A Highly Conserved Enhancer in the Dlx5/Dlx6Intergenic Region is the Site of Cross-Regulatory Interactions betweenDlx Genes in the Embryonic Forebrain

Identification of highly conserved elements in the zebrafish dlx4/dlx6 and mouse Dlx5/Dlx6 intergenic regions

We identified two highly conserved sequences in the region between the zebrafish dlx4 and dlx6 genes and their mouse orthologs Dlx5/Dlx6. A 1.4 kbXhoI–EcoRI fragment from the zebrafishdlx4/dlx6 intergenic region (Fig. 1) was found to hybridize to a pair of restriction fragments in the mouseDlx5/Dlx6 intergenic region. Nucleotide sequence analysis revealed two sequences, of ∼400 and 300 bp, respectively, that are highly similar between the two species (Fig.2A,B). The orientation and the relative position of the two sequences relative to dlx4 (Dlx5) anddlx6 (Dlx6) are identical (Fig. 1). The 400 bp sequence, named zI46i, is closer to the dlx6 gene than is the 300 bp sequence, hereafter named zI46ii. The orthologous mammalian elements are hereafter called mI56i and mI56ii, respectively. We have identified highly similar sequences at the humanDLX5/DLX6 locus by searching the GenBank database (Fig.2A,B).

Nucleotide sequence comparisons indicate the human and mouse mI56i elements to be identical except for a 3 bp insertion in the human sequence (Fig. 2A). The zI46i sequence is 83% identical to its mammalian counterparts over 384 bp (Fig.2A), including a central 131 bp with 94% sequence identity. The human and mouse mI56ii sequences are 98% identical, and zI46ii shares ∼85% identity over 275 bp with its mammalian counterparts (Fig. 2B). The zI46ii sequence lacks a stretch of adenines, ∼20 nucleotides in length, that is found in the two mammalian sequences.

A zebrafish dlx4/dlx6 intergenic fragment _targets reporter gene expression to the forebrain and olfactory placodes in transgenic mice

A construct containing the lacZ reporter gene under the control of a β-globin minimal promoter and the entire zebrafish dlx4/dlx6 intergenic region (plus a short segment of the transcription unit of dlx4 and a few base pairs of the transcription unit of dlx6;zfdlx4/6lacZ transgene; Fig. 1) was injected into fertilized mouse eggs to produce transgenic animals. We obtained seven lines of transgenic mice. All seven lines showed lacZexpression, beginning at approximately embryonic day 10 (E10), in two groups of forebrain cells, one in the ventral thalamus/hypothalamus and one in the basal telencephalon (Fig.3A; I and II, respectively). Examination of whole-mount embryos stained for β-galactosidase activity indicated that the patterns of reporter transgene expression are strikingly similar to patterns of endogenous mouse Dlxexpression in the forebrain (Shimamura et al., 1997). Mouse embryos express Dlx5 and Dlx6 in two separate domains within the forebrain. Domain I is a longitudinal alar plate stripe that begins at the zona limitans intrathalamica and extends rostrally through the ventral thalamus (VT) and several hypothalamic areas (Hy) to the rostral midline. Domain II is a longitudinal region in the basal telencephalon that extends rostrally from part of the caudal ganglionic eminence (CGE; amygdala primordium), through the lateral ganglionic eminence (LGE) and medial ganglionic eminence (MGE), and into the septal and preoptic (POA) primordia (Bulfone et al., 1993b; L. Puelles, E. Kuwana, A. Bulfone , K. Shimamura, J. Keleher, S. Smiga, E. Puelles, and J. Rubenstein, unpublished observations).

Similarly to endogenous Dlx5 and Dlx6, expression of the reporter transgene decreased after E14.5, and lacZ transcripts were virtually undetectable at postnatal day 0 (P0). Yet, β-galactosidase activity persisted much longer and, in some areas, remained strong even in the adult (P120) brain (data not shown). A higher sensitivity of the enzymatic assay for β-galactosidase activity compared to in situ hybridization or the sheltering of the β-galactosidase protein from metabolism could explain the apparent persistence of β-galactosidase compared to thelacZ transcripts. In addition to the developing forebrain, the reporter transgene under the control of zebrafish sequences was expressed in the olfactory placodes in five of seven mouse lines (Figs.1, 3A,B). There were very few additional sites of transgene expression: one line had a few labeled eye cells; one line had expression in the developing shoulder area, and one showed expression in the trunk somites (data not shown). Additional sites of endogenous mouse Dlx5 and Dlx6 expression were negative, including the branchial arches, the otic vesicle, and the limb apical ectodermal ridge (AER).

To assess the degree to which expression of thezfdlx4/6lacZ transgene matches endogenousDlx5 and Dlx6 expression in the forebrain, we have compared their expression patterns, using X-gal staining and radioactive in situ hybridization, respectively, on transverse brain sections. We analyzed sequential sections from E10.5, E12.5, E14.5, E17.5, and P0 mice.

Dlx5 and 6 are expressed in domains I and II in slightly different, but overlapping patterns: Dlx5 is expressed strongest in the subventricular zone (SVZ), whereasDlx6 is expressed strongest in the mantle zone (MZ) (Fig.3E,F). Neither gene is expressed appreciably in the ventricular zone (VZ) (Liu et al., 1997).

Remarkably, the zfdlx4/6lacZ transgene is expressed in a pattern extremely similar to that of the mouseDlx5 and Dlx6 genes. It is apparent, however, that despite the degree of overlap between thezfdlx4/6lacZ transgene and Dlx5 andDlx6 genes, there is a greater similarity between thezfdlx4/6-enhancer-driven lacZ and Dlx5expression patterns. β-Galactosidase activity and Dlx5transcripts can first be detected in the forebrain at ∼E10, as a thin layer of cells overlying parts of the ventricular zones in the basal telencephalon and diencephalon, respectively. On E10.5, E12.5, and E14.5, zebrafish dlx4/6-enhancer driven β-galactosidase expression in the mouse telencephalon matches mouseDlx5 expression more closely than that of Dlx6(Fig. 3D–I; data not shown). In domain I, thezfdlx4/6lacZ transgene has a pattern that also appears to be more similar to that of Dlx5 thanDlx6 (Fig. 3G–I).

The zebrafish dlx4/dlx6 intergenic enhancer is active in the forebrain of zebrafish embryos

To determine if the intergenic sequences that _target reporter gene expression to the forebrain of transgenic mice can reproducedlx expression in zebrafish embryos, we microinjected, into one-cell stage embryos, a construct that contained the 1.4 kb zebrafishEcoRI–XhoI intergenic fragment fromdlx4/dlx6 (Fig. 1), the same β-globin minimal promoter fragment as for the transgenic mouse experiments, and the gene coding for the GFP as a reporter. Primary transgenic zebrafish embryos carrying this construct expressed GFP specifically in forebrain cells forming two domains (Fig. 3J) with patterns strikingly similar to endogenous dlx4/dlx6 expression (Fig.3K,L). Of the 750 embryos that survived microinjection until the second day of embryonic development, four had very high levels of GFP expression in the forebrain, 25 had 5–10 GFP-positive forebrain cells, and 30 had one or two positive forebrain cells. Fifteen embryos showed one or a few GFP-positive cells in ectopic locations. The onset of GFP expression in the forebrain was ∼17–19 hr after fertilization (hpf), shortly after the onset of dlx4 expression as detected by in situ hybridization. GFP expression persisted in the forebrain until at least 36 hpf.

The two domains of dlx expression in the forebrain of zebrafish embryos (Akimenko et al., 1994) are reminiscent of the two domains observed in the mouse embryonic forebrain. To compare expression patterns of dlx genes in the zebrafish forebrain, we made sections of 48-hr-old embryos hybridized with dlxprobes. Interestingly, the patterns of expression of dlx1and dlx2 in both the telencephalon and the diencephalon indicate that the two genes are expressed in more immature cells, as reflected by their position closer to the ventricular walls than the cells that express dlx4 and dlx6 (Fig.4). A similar observation had been made previously for the mouse orthologs of these four genes (Liu et al., 1997).

Most of the forebrain activity of the dlx4/dlx6intergenic enhancer is located in zI46i

The two conserved sequences located in the zebrafishdlx4/dlx6 intergenic region (zI46i and zI46ii) were inserted separately into reporter constructs and used to produce transgenic mouse embryos. At E11, forebrain lacZ expression _targeted by the zI46i enhancer construct was indistinguishable from that _targeted by the full-length I4/6 enhancer (compare Figs.5A, 3B), although none of the embryos showed expression in the olfactory epithelium (n = 12; Fig. 1).

We also generated transgenic mice with a reporter construct that contained the second conserved sequence from the zebrafishdlx4/dlx6 intergenic region zI46ii. Two lines of transgenic mice and 10 primary transgenic mouse embryos were produced. Embryos from one transgenic line showed lacZ expression in the olfactory epithelium (data not shown) resembling that obtained with the full-length I4/6. This was also observed in one primary transgenic embryo, which, in addition, had expression in the AER of the limb buds, where Dlx genes are expressed (Dollé et al., 1992;Bulfone et al., 1993a). Finally, one primary zI46ii transgenic embryo showed correct lacZ expression in the forebrain (data not shown). Thus, zI46ii was much less efficient at _targetinglacZ to the forebrain (0 of 2 lines; 1 of 10 primary transgenic embryos) compared to full-length I4/6 (seven of seven lines) or to zI46i (12 of 12 primary transgenic embryos; Fig. 1).

We next tested whether the orthologous mouse mI56i could regulate correct Dlx expression in the forebrain. Four stable lines and four primary transgenic embryos were produced with the mI56i construct, and nearly all showed forebrain expression (Fig.5B); one transgenic line did not express lacZanywhere, possibly because of an integration effect. Reversing the orientation of mI56i had no effect on its expression (nine of nine primary transgenic embryos; Fig. 1; data not shown). Thus, both the orthologous mI56i and the zI46i fragments are enhancers that efficiently replicate the correct pattern of Dlx expression in the forebrain.

Unlike the zI46i enhancer, mI56i in either orientation frequently reproduced correct Dlx expression in the branchial arches (two of four stable lines and seven of 13 primary embryos; sum of both orientations, Figs. 1, 5B; data not shown), olfactory placode (one line and one primary embryo) and AER (one line; Fig. 1). No expression in the otic vesicle was observed in any embryos, but some expression was detected in the middle ear, which is consistent with the branchial arch expression.

To begin to identify the essential sequences within these enhancers, we used a deletion fragment of the zI46i enhancer, corresponding to positions 187–316 (Fig. 2A) and examined its activity in transgenic mouse embryos at E11. Of five primary transgenic embryos, all appeared to have correct expression in domain II in the forebrain (Fig. 5C). However, β-galactosidase expression in domain I (ventral thalamus and hypothalamus) was occasionally weaker or not detectable (data not shown). A similar construct also _targeted expression of GFP principally to the forebrain of zebrafish embryos (data not shown).

Activity of the zebrafish intergenic enhancer is reduced in mice lacking Dlx1 and Dlx2

Mutant mice that lack both Dlx1 and Dlx2function have a time-dependent block in basal telencephalon differentiation (Anderson et al., 1997b). Although early neurogenesis appears to be normal, later neurogenesis is not. This phenotype seems to be caused by a defect in the production and/or function of the subventricular zone. Accordingly, in Dlx1/2 mutantsDlx5 and Dlx6 expression is not detectable in the subventricular zone of the LGE and MGE, but is maintained in early born mantle cells at E12.5 (Anderson et al., 1997b). As described above, the zebrafish and mouse intergenic enhancers are highly active in the SVZ of the basal telencephalon. Therefore, it is possible that the Dlx1 or Dlx2 proteins might be, at least in part, responsible for the activity of this enhancer. To test this hypothesis, we bred mice containing the zebrafish dlx4/dlx6 full intergenic reporter construct with mice heterozygous for a mutation that inactivates both Dlx1and Dlx2. We then inbred mice that are heterozygous for both the mutation and the transgene to generate Dlx1/2−/− homozygotes that also had at least one zfdlx4/dlx6lacZallele.

In embryos that are homozygous for the Dlx1/Dlx2 mutation, β-galactosidase activity is strikingly reduced in the subventricular zone of the developing striatum (Fig.6A–D). These results parallel the changes in endogenous Dlx5 and Dlx6expression in the Dlx1/2 mutant mice (Fig.6E–L). Based on these results, we propose thatDlx1 and/or Dlx2 function is required, directly or indirectly, to regulate Dlx5 and Dlx6expression via their intergenic enhancer.

To determine whether the loss of Dlx5, Dlx6, andzfdlx4/6lacZ expression in the SVZ of theDlx1/2 mutants is attributable to a loss of those cells or to a change in gene regulation in SVZ cells, we studied the expression of Dlx1 and Dlx2 mRNAs. In theDlx1/2 mutants, the 5′ end of these genes was not deleted (Qiu et al., 1997). Thus, if the Dlx1/2 cis-acting regulatory sequences responsible for Dlx1/2expression are intact, and the truncated Dlx1 and/orDlx2 transcripts are stable, we should be able to detect the cells that normally express Dlx1/2 in these mutants. In fact, in situ hybridization demonstrates that both truncated genes are still expressed in the proliferative zones of the Dlx1/2 mutants (Fig. 6M–P;data not shown). This strongly supports the model that cells expressingDlx1/2 are maintained in the mutants and that there is molecular dysregulation within these cells leading to the loss of Dlx5, Dlx6, andzfdlx4/6lacZ expression.

Dlx proteins can upregulate transcription from conserved intergenic sequences

Dlx proteins bind DNA and can regulate transcription (Liu et al., 1997; Zhang et al., 1997). Because analysis of Dlx1/2 mutant mice suggests that Dlx1 and/or Dlx2 function is necessary for proper expression of Dlx5/Dlx6, one possibility is that this interaction is directly mediated by transcriptional activation of the intergenic enhancer(s) by Dlx1 or Dlx2. To test this model, we performed transient cotransfection assays in cultured cells. Reporter plasmids were constructed to contain either the zebrafish 1.4 kb I4/6 fragment, which contains both zI46i and zI46ii, or to contain only one of these elements. Effector plasmids were constructed to express full-length zebrafish Dlx1, Dlx2, Dlx3, Dlx4, or Dlx6 proteins under the control of the SV40 early promoter, or full-length mouse Dlx1, Dlx2, or Dlx5 proteins under the control of the cytomegalovirus promoter.

Cotransfection, into mouse P19 murine embryonic carcinoma cells, of a construct expressing the zebrafish Dlx2 protein resulted in a 20-fold increase in the activity of the CAT reporter gene placed under the control of the 1.4 kb dlx4/dlx6 intergenic fragment (Fig.7A). All of the zebrafishDlx expression vectors were able to activate expression of the reporter construct to a similar extent (data not shown). The mouse Dlx1, Dlx2, and Dlx5 expression vectors were also able to activate transcription of the same reporter construct in a neuroepithelial cell line (MNS-71; G. Yu, T. Zerucha, M. Ekker, and J. L. R. Rubenstein, unpublished observations). This indicates that Dlx proteins from either zebrafish or mouse are able to recognize similarly the zebrafish intergenic enhancer sequences in at least two different cell types. Not all homeodomain proteins could activate transcription from the 1.4 kbdlx4/dlx6 intergenic fragment; the products of severalsine oculis-related genes (six genes) were unable to activate transcription from this sequence (data not shown).

To determine if both zI46i and zI46ii contain _targets for Dlx proteins, the same Dlx expression constructs were cotransfected with reporter constructs containing either zI46i or zI46ii. All Dlx proteins examined activated expression through the zI46i element (Fig.7A; data not shown) but none activated transcription from a reporter containing zI46ii (Fig. 7A; data not shown), except perhaps for a weak (less than twofold activation) by the mouse Dlx5 protein. Furthermore, the degrees of activation produced by Dlx proteins on zI46i reporter constructs were comparable to those obtained with the 1.4 kb I4/6 fragment.

In an attempt to narrow down the region of zI46i required for activation by Dlx proteins, a series of deletion fragments of zI46i were prepared and subcloned into the reporter plasmid. The orientation of each of the deletion fragments was maintained relative to the orientation of the full-length zI46i. In transient cotransfection experiments, Dlx2 activated transcription of constructs containing the 187–316 deletion fragment to an extent similar to that observed with the full-length zI46i (Fig. 7A). However, Dlx2 did not activate transcription of constructs containing either the 1–204 or 305–473 fragments (Fig. 7A).

Recently a consensus DNA-binding site was identified for theXenopus Dlx3 ortholog, Xdll2 using a binding site selection procedure from a random oligonucleotide pool (Feledy et al., 1999). The consensus site identified is (A/C/G)TAATT(G/A)(C/G). Because of the similarity of the homeodomains of the Dlx family, it is likely that Dlx proteins other than those of the Dlx3 paralogous group will recognize a similar sequence. The 187–316 fragment of zI46i that is activated by Dlx2 contains two sites consistent with this consensus sequence. These two sites correspond to positions 207–214 and 263–270 of the zI46i sequence (Fig. 2). We mutagenized these two sites, individually or in combination. Mutagenesis of either site or of both sites almost entirely abolishes the activation of transcription by Dlx2 (Fig.7B). When tested in transgenic mice, a construct that contained mutations in both putative binding sites had little if any activity in the forebrain. Of seven primary transgenic mouse embryos, three had no detectable lacZ expression in the forebrain, and two had a few weakly positive cells at the anterior end of domain II (data not shown). These positive cells represented only a very small fraction of the endogenous pattern. Combined with our observation that activity of the forebrain enhancer is dramatically decreased in mice lacking Dlx1 and Dx2 function (Fig. 6), these results indicate that activation by Dlx proteins, presumably by Dlx1 and/or Dlx2, is essential for the activity of the I46 enhancer in the forebrain.

To determine if the Dlx proteins are able to directly interact with the zI46i element, electrophoretic mobility shift assays were performed. Nuclear extracts from SF7 SCID fibroblasts expressing a fusion of a c-myc fragment with full-length zebrafish Dlx2 (MTG-Dlx2) produced a lower mobility complex with the 187–316 deletion fragment of zI46i (Fig.8A). Migration of the lower mobility complex obtained with MTG-Dlx2 was further retarded in the presence of the 9E-10 anti-c-myc antibody, indicating the lower mobility complex contains MTG-Dlx2. No complexes of lower mobility were obtained with a control SF7 nuclear extract (Fig.8A). Furthermore, neither of the other two deletion fragments of zI46i (1–204 and 305–473) produced a complex of lower mobility in the presence of MTG-Dlx2-containing SF7 extract (Fig.8A), a result consistent with the absence of activation, by Dlx2, of reporter constructs containing these intergenic fragments in cotransfection experiments.

We next examined the effects of mutagenesis of the putative binding sites. As seen in Figure 8B, the Δ210 and Δ266 mutagenized fragments both produced a shift of lower mobility that migrated to the same place as that obtained with the wild-type 187–316 fragment. These shifts could be supershifted with the 9E-10 antibody. In contrast, we observed no supershift when the double mutant fragment was incubated with the MTG-Dlx2 and 9E-10 (Fig. 8B). Mutation at the 210 site seems to increase background binding, which is also observed in the control Sf7 extract and overlaps with the expected position for the Dlx2–DNA complex. This is especially evident for the double mutant but can also be seen in the Δ210 lanes.

The similar migrations of the lower mobility complexes obtained with the wild-type and single mutant fragments suggest that we can only observe the wild-type fragment bound by one Dlx2 molecule. This is consistent with the observation that only a small proportion of the labeled fragment is bound and suggests that Dlx2 binds to the two sites independently instead of cooperatively, which is not surprising considering the relatively large distance that separates the two binding sites (∼56 bp). The fragment bound by two Dlx2 molecules would be proportionally too weak to be observed in this assay.

Taken together with the results of our transient expression assays (Fig. 7), our observations suggest that two Dlx2 protein molecules bind to the zI46i enhancer independently but that optimal function of the enhancer requires occupancy of the two binding sites.

One intergenic enhancer is sufficient to recapitulate forebrainDlx expression

A zebrafish sequence from the intergenic region between thedlx4 and dlx6 genes is sufficient, once combined to a minimal promoter, to direct expression in cells of the telencephalon and diencephalon, of either zebrafish or mice, that normally express Dlx genes (Fig. 3). This strongly suggests that the regulatory mechanisms controlling Dlx expression in the forebrain have been conserved during vertebrate evolution, and it lends support to the idea that Dlx function during forebrain development has also been conserved. Additional evidence for conserved function of Dlx genes in forebrain development comes from the differential expression of Dlx genes in the telencephalon and diencephalon where more immature cells expressDlx1 and Dlx2 compared to Dlx5 andDlx6, as seen both for the mouse genes (Liu et al., 1997) and for their zebrafish orthologs (Fig. 4; Akimenko et al., 1994).

Functional conservation of enhancer sequences between mammals and teleost fish has been previously observed for the Otx2(Kimura et al., 1997), hoxb1 (Marshall et al., 1994), and the Hoxd-11 genes (Beckers et al., 1996; Gerard et al., 1997), although in the latter case, temporal, spatial, and mechanistic differences could be observed between the fish enhancer and its mammalian counterpart.

Comparisons of the enhancer activities of zI46i (mI56i) and zI46ii suggest the former plays a more important role in forebrain expression. zI46ii may still be necessary for optimal Dlx expression in the ventral forebrain, but this enhancer may require, to function efficiently, the presence of other regulatory sites, either from zI46i or from the promoters of one or both flanking genes.

Detailed analysis of reporter transgene activity in the mouse forebrain suggests that both the full dlx4/dlx6 intergenic construct and the zI46i (mI56i) sequences reproduce the endogenousDlx5 expression pattern more faithfully than theDlx6 expression pattern. This was observed principally by comparison of β-galactosidase expression and endogenous transcript levels in the LGE and MGE of the telencephalon (Fig. 3). Identical results were obtained with constructs from either zebrafish or mouse origin. Therefore, the observed differences cannot be attributed solely to the inability of a zebrafish enhancer to precisely recapitulateDlx expression in a mouse embryo. One possible explanation for this result is that sequences, necessary for maximal expression in cells of the mantle, are absent from our constructs, and, therefore, located outside the intergenic region. It is also possible that, although the intergenic enhancer is sufficient to direct expression to the ventral forebrain, its activity is modulated by specific interactions with other cis-acting regulatory elements, such as the promoters of each of the two flanking genes, Dlx5 andDlx6. An overall distinct set of transcriptional activators binding to upstream and intergenic regulatory sequences would be responsible for the differences in Dlx5 and Dlx6expression patterns. Experiments in zebrafish designed to examine the interactions between the intergenic forebrain enhancer and thedlx4 and dlx6 promoters are presently under way to address this issue. In summary, although it is possible that the intergenic forebrain enhancer is shared between Dlx5 andDlx6, which would explain their partially overlapping patterns of expression, additional mechanisms must account for the overall distinct expression of the two genes in the forebrain.

Transgenic animals carrying constructs containing both zI46i and zI46ii always exhibit expression of the reporter gene in the ventral forebrain. Reporter expression is often seen in the olfactory placodes, but never in regions of the embryo, such as the branchial arches, the inner ear, and the AER of the limb buds where Dlx5 andDlx6 or their zebrafish orthologs are also expressed. These results suggest that elements necessary for proper expression in the latter areas are located outside the conserved Dlx5/Dlx6intergenic region. On the other hand, several transgenic animals with the zI46i or mI56i constructs showed expression in the ectomesenchyme of the branchial arches reminiscent of endogenous Dlxexpression. The mechanisms that underlie such results are, at present, unclear but may involve integration effects. It is also possible that intergenic sequences, outside zI46i (mI56i) are necessary to restrict the activity of this enhancer to the ventral forebrain.

Dlx proteins interact with the forebrain-specific regulatory sequences

Expression of Dlx5 and Dlx6 is affected in the ventral forebrain of Dlx1/2 null mutants (Anderson et al., 1997b). Thus, Dlx5 and Dlx6transcripts are not detectable in the SVZ of the LGE at E12.5 and E14.5. Like endogenous Dlx5 and Dlx6, the activity of the zebrafish dlx4/dlx6 intergenic transgene (I4/6) is drastically reduced in the Dlx1/2 null mutants (Fig. 6). The mostly normal expression of the truncated Dlx1and Dlx2 transcripts in forebrain cells of the mutant (Fig.6; data not shown) suggests that the SVZ cells that normally expressDlx5 and Dlx6 are still present. Therefore, the reductions in Dlx5 and Dlx6 expression and in transgene activity strongly suggest that Dlx1 and/orDlx2 are required for the induction and/or maintenance ofDlx5 and Dlx6 expression and that this is mediated by the intergenic enhancer sequences. This might be achieved by direct regulation of Dlx5/Dlx6 expression by the Dlx1 or Dlx2 proteins which can function as transcriptional activators (Liu et al., 1997; Zhang et al., 1997). Such cross-regulatory interactions involving homeobox genes have been described, for example, for members of the Hox clusters (Gould et al., 1997; Nonchev et al., 1997; Studer et al., 1998) and for the zebrafish dlx genes (Zerucha et al., 1997). Alternatively, Dlx1 or Dlx2 may activate a yet unknown factor that is an essential regulator of Dlx5 andDlx6 expression. The above two mechanisms are not mutually exclusive. Our finding that the Dlx2 (Fig. 7A) and Dlx1 (data not shown) proteins, of either zebrafish or mouse origin are able to upregulate transcription of reporter constructs containing the conserved I4/6 intergenic sequences in cotransfection experiments supports the view that these sequences are the site of cross-regulatory interactions in vivo. Upregulation by Dlx2 was almost completely abolished when the putative binding sites were mutagenized (Fig. 7B). Furthermore, we were able to demonstrate binding of Dlx2 to the 187–316 fragment of zI46i in gel mobility shift assays (Fig. 8A). Mutagenesis of both sites in 187–316 abolished binding by the Dlx2 protein (Fig. 8B). The 187–316 fragment of zI46i is responsible for most if not all of the activation by Dlx proteins in transfection experiments and is able to _target transgene expression to the ventral forebrain in mice and zebrafish.

The loss of Dlx1 and Dlx2 expression only eliminates Dlx5, Dlx6, andzfdlx4/6lacZ transgene expression in the SVZ, whereas their expression is maintained in the early-born mantle cells of the rostral telencephalon (Fig. 6). This observation indicates that other transcription factors regulate Dlx5, Dlx6,and I4/6lacZ expression in a subset of early forebrain cells. Furthermore, the loss of Dlx5 and Dlx6expression in the SVZ of the Dlx1/Dlx2 null mutants also raises the possibility that the mutant phenotype may be attributable to the loss of function of all four genes, implying some functional redundancy between them. This is also supported by the recent observation that loss of Dlx5 function alone does not produce any obvious forebrain phenotype (Acampora et al., 1999; Depew et al., 1999). A better understanding of any differences in biochemical activities of Dlx proteins, such as involvement in specific protein–protein interactions, would help elucidate the functional consequences of the partially overlapping expression of these genes during development.

Intergenic region and Dlx gene evolution

The high degree of sequence similarity that we observed in the intergenic region between a pair of Dlx genes of mouse, human, and teleost fish (Fig. 2) is remarkable, considering that these sequences are outside the coding regions of either genes. High degrees of sequence similarity outside gene coding regions have been observed previously between human and mouse sequences (for example, see Becker et al., 1996; Williams et al., 1998), and some sequence conservation has also been found with sequences of distantly related vertebrates such as the pufferfish, Fugu rubripes, and zebrafish (Marshall et al., 1994; Morrison et al., 1995; Beckers et al., 1996;Kimura et al., 1997). However, none of the above examples compare in length and/or in percentage identity with the elements we report in the present study.

The convergently transcribed configuration of pairs ofdistal-less-related genes is ancient because it has been reported for the ascidian Ciona intestinalis (Di Gregorio et al., 1995). The distance that separates the two genes is relatively small (2–10 kb) for all cases reported thus far. It is likely that the paired organization arose after the divergence of arthropods from the lineage that would give rise to vertebrates because insects are thought to have only one distal-less gene. It is possible that one or a few regulatory sequences found downstream of thedistal-less gene in the common ancestor to modern day invertebrates and vertebrates were preserved after the first duplication and inversion event that produced the first pair ofdistal-less/Dlx genes. Enhancer sequences have been described downstream of the Drosophila distal-less gene (Vachon et al., 1992; O'Hara et al., 1993) and it will be interesting to determine if there is any degree of functional conservation in these enhancers and those described in the current study. A potential evolutionary advantage of enhancer-sharing by the two linked genes would be consistent with the conservation of the paired, convergently transcribed configuration and in particular the maintenance of a relatively short intergenic distance. Enhancer sharing has been previously demonstrated for some of the clustered Hox genes (van der Hoeven et al., 1996; Gould et al., 1997; Sharpe et al., 1998).

Vertebrates have at least three pairs of linked Dlx genes (Simeone et al., 1994; McGuinness et al., 1996; Ellies et al., 1997;Liu et al., 1997), the Dlx1/Dlx2, Dlx5/Dlx6(dlx4/dlx6 in zebrafish), and Dlx3/Dlx7 pairs. The presence of conserved regulatory sequences may not be unique to theDlx5/Dlx6 gene pair because the Dlx1/Dlx2intergenic region also contains highly conserved sequences (T. Zerucha, M. Qiu, J. K. Liu, J. L. R. Rubenstein, and M. Ekker, unpublished observations), although the roles of such sequences in Dlxgene regulation are, at present, unclear. The function and evolution of intergenic enhancer sequences, combined with studies of the functional specificity of Dlx genes, will enable us to understand the mechanistic basis for the concerted action of Dlx proteins in embryonic cells and the position of the Dlx genes in regulatory cascades during development.