Hoxd13 and Hoxa13 Directly Control the Expression of the EphA7 Ephrin Tyrosine Kinase Receptor in Developing Limbs*

Hoxa and Hoxd genes, related to the Drosophila Abd-B gene, display regionally restricted expression patterns and are necessary for the formation of the limb skeletal elements. Hox genes encode transcription factors, which are supposed to control the expression of a series of downstream target genes, whose nature has remained largely elusive. Several genes were identified that are differentially expressed in relation to Hox gene activity; few studies, however, explored their direct regulation by Hox proteins. Ephrin tyrosine kinase receptors and ephrins have been proposed as Hox targets, and recently, evidence was gained for their role in limb development. The expression of the EphA7 gene in developing limbs was shown to correlate with the expression of Hoxa13 and Hoxd13; however, its direct regulation by these genes has never been assessed. We have characterized the EphA7 promoter region and show that it contains multiple binding sites for paralog group 13 Hox proteins. We found that one of these sites is bound in vivo by HOXA13 and HOXD13 and by endogenous Hoxd13 in developing mouse limbs. Moreover, we show that HOXD13 and HOXA13 activate transcription from the EphA7 promoter and that a mutation of the HOXA13/HOXD13 binding site was sufficient to abolish activation. Conversely, the HOXD13(147L) mutation, identified in patients displaying a novel brachydactyly-polydactyly syndrome, does not bind to in vivo, and fails to transactivate the EphA7 promoter. These results establish that EphA7 is a direct downstream target of Hoxd13 and Hoxa13 during limb development, thus providing further insight into the regulatory networks that control limb patterning.

Transcription factors belonging to the HOX family of homeodomain-containing proteins control cell fates and regional identities along the primary body and limb axes (1,2). Hox genes that are related to the Drosophila Abd-B gene and are located at the 5Ј-end of the Hoxa and Hoxd clusters (paralogous groups 9 -13) display regionally restricted expression patterns in the developing limbs (3)(4)(5). Targeted mutagenesis and overexpression of Abd-B-related Hoxa and Hoxd genes showed that their function in limb development is to control the size, shape, and number of specific bones by regulating processes such as mesenchymal cell aggregation, chondrification, and ossification (6 -8) (reviewed in Refs. 5 and 9). The genetic pathways in which they act, however, are still poorly characterized. In particular, the identity of their downstream effector genes remains still elusive (10). Indeed, whereas HOX proteins have been shown to bind specific DNA sequences and are supposed to regulate overlapping sets of target genes, only a few of them have been isolated. Target genes for HOX proteins have been identified using various approaches, including microarray hybridization screenings and candidate gene analysis (11)(12)(13)(14)(15)(16). Few studies, however, have shown direct regulation by HOX proteins on the promoters of their putative downstream target genes.
Ephrins and ephrin tyrosine kinase receptors (Eph) 2 have been recurrently indicated as HOX downstream target genes (12,(17)(18)(19). Eph and ephrins are expressed in various regions of the vertebrate embryo in dynamic patterns, and they were found to play crucial roles in the control of cell shape, cell migration, cell sorting, wiring of neurons in the nervous system, and the formation of boundaries between structures (reviewed in Refs. 20 and 21). Recently, experimental evidence has been gained for a role of Eph-ephrin signaling in limb development. Overexpression in developing chick limbs of ephrin A2 has been shown to disrupt limb cartilage morphogenesis causing digit bifurcations and syndactyly (22). Similarly, heterozygous ephrin B1 null female mice have been reported to display preaxial polydactyly and syndactyly (23). Finally, the EphA7 gene, which is expressed at embryonic day 13.5 (E13.5) in the perichondrium of the mesenchymal condensations of the phalanges, was shown to be significantly down-regulated in the forelimbs of Hoxa13 Ϫ/Ϫ mice. Its expression, however, was not completely absent, suggesting that the transcription of this ephrin receptor gene might be under the control of more than one paralogous group 13 Hox protein (19). Indeed, we could show that the misexpression of Hoxd13 by retroviral infection of developing chick limbs leads to a marked increase of EphA7 expression in the phalangeal mesenchymal condensations, indicating that Hoxd13 might regulate EphA7 transcription as well. Interestingly, in the same set of experiments, we found that the misexpression of a mutant HOXD13 protein, HOXD13(I47L), did not result in an increase of EphA7 expression (24). The HOXD13(I47L) mutation, identified in patients showing a novel brachydactyly-polydactyly syndrome, represents a single amino acid substitution, involving residue 47 of the homeodomain that is located within the recognition helix. This substitution was found to alter rather than abolish DNA binding, since it selectively impaired the ability of HOXD13 to recognize one of its two different classes of consensus binding sequences (24). This finding prompted us to speculate that the I47L mutation, via a reduction of the repertoire of potential sites recognized by the HOXD13 protein, would cause a partial or complete failure to regulate a subset of the genes normally controlled by HOXD13. Since EphA7 turned out to be one of these genes, we decided to further investigate its regulation by HOXD13. In particular, we wanted to ascertain whether EphA7 is indeed a direct target of HOXD13 and/or HOXA13 and whether the nature of the regulatory sequences mediating HOXD13 regulation would explain the lack of control by the mutated HOXD13(I47L) protein.
In this report, we show that the EphA7 promoter region contains multiple potential binding sites for HOX paralog group 13 proteins. We found, however, that only one of these sites is bound in vivo by the HOXA13 and HOXD13 proteins. A mutation of this evolutionarily conserved site was sufficient to abolish the transcriptional activation of the EphA7 promoter by HOXA13 and HOXD13. We moreover found that this EphA7 Hox group 13 site is not bound by the HOXD13(I47L) mutated protein both in vivo and in vitro, thus providing a molecular basis for the lack of up-regulation of EphA7 expression by HOXD13(I47L) in developing limbs. Our results thus establish that EphA7 is a direct downstream target of the Hoxd13 and Hoxa13 proteins during limb development, providing additional evidence that Hox gene products directly control the aggregation properties of limb mesenchymal cells by regulating the ephrin receptorephrin signaling system.
Electrophoretic Mobility Shift Assay-The GST-HOXD13HD and GST-HOXD13(I47L)HD fusion proteins were produced in Escherichia coli, purified according to established methods, and analyzed by SDS-PAGE and Coomassie staining. The purified proteins were diluted in 13 l of ␦-buffer (20% glycerol, 20 mM KCl, 2 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol) and preincubated with 100 ng poly(dI-dC) in a total volume of 20 l of 1ϫ binding buffer (0.1 M KCl, 2 mM MgCl 2 , 4 mM spermidine, 0.1 mg/ml bovine serum albumin) for 15 min on ice. 32 P-Labeled fragments containing a consensus binding site for HOXD13 or the EphA7 HOXD13 binding site 3 were obtained by annealing of the following oligonucleotides: HOXBSA/B (24); EphA7BSA, 5Ј-TCGAC-CCTAATTGATATTATTGGAGTGTGGAGCAC-3Ј; EphA7BSB, 5Ј-TCGCGTGCTCCACACTCCAATATCAATTAGGG-3Ј. 30,000 cpm of the labeled probes were added to the samples and incubated for 30 min on ice. Reactions were separated on a 6% polyacrylamide gel in 0.5% TBE, dried, and exposed to Eastman Kodak Co. X-OmatR film at Ϫ80°C.
Cell Culture, Transfection, and Transduction-NIH3T3 mouse fibroblasts and C3H10T1/2 mouse embryonic fibroblasts were cultured in Dulbecco's modified Eagle's medium (Celbio) supplemented with 10% fetal calf serum (Celbio), 2 mM L-glutamine (Invitrogen), 100 units/ml penicillin, and 100 g/ml streptomycin. Transfections in NIH3T3 cells were carried out by CaPO 4 precipitation (25). In a typical experiment, 12 g of reporter plasmid, 2.5-5.0 g of expression construct, and 0.1 g of CMV-␤-galactosidase (Clontech) as internal control were used per 6-cm dish. Forty-eight hours after transfection, cells were washed, lysed, and assayed for luciferase and ␤-galactosidase expression (26). Each transfection was done in duplicate in the same experiment, and the plotted luciferase activities represent the average of 3-6 different experiments. To transiently express HA-HOXD13, HA-HOXA13, or HA-HOXD13(I47L) in NIH3T3 cells, we transfected 5 g of the corresponding expression vector per 10-cm dish. To transduce C3H10T1/2 cells, viral stocks of the LXI⌬N and LHOXD13I⌬N retroviral expression constructs were produced by transient transfection of Eco-Phoenix cells as described previously (27). The viral supernatant was added to a subconfluent culture of C3H10 T1/2 cells in the presence of 0.8 g/ml Polybrene; incubation was performed twice, initially for 6 h and then overnight. Cells were washed, and medium was replaced. HOXD13 transient expression in NIH3T3 and stable expression in C3H10T1/2 were detected by Western blot analysis using anti-HA (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) antibodies. An antibody against the NFY-B nuclear transcription factor (kind gift of C. Imbriano) was used as a loading control. Immunostained bands were detected with a chemiluminescence system (Amersham Biosciences).
RT-PCR Analysis-RNA from NIH3T3 and C3H10T1/2 cells was extracted using the RNeasy kit (Qiagen) according to the manufacturer's protocol. Synthesis of cDNA was done starting from 3 g of RNA using the SSII reverse transcriptase kit (Invitrogen). Semiquantitative PCR was performed with the following oligonucleotides: EphA7RTfor, 5Ј-TCTACACCACGACTGGTGGAAAAA-3Ј; EphA7RTrev, 5Ј-C-CGCTCGAGCTTGGGTTTCGAATCATTTTGTCT-3Ј. Glyceraldehyde-3-phosphate dehydrogenase control RT-PCR was done using standard oligonucleotides.

RESULTS
The 5Ј Region of the Mouse EphA7 Gene Contains Several Potential Binding Sites for the HOXD13 Protein-To identify possible HOXD13dependent regulatory elements within the transcriptional control regions of the EphA7 gene, we analyzed genomic sequences upstream to the putative transcription start site of the EphA7 gene. An interspecies comparison using the UCSC Genome Browser (29) was made, with the idea that sequence conservation might highlight relevant regulatory regions. We found that at distances greater than 5 kb 5Ј to the putative transcription start site of EphA7, the degree of sequence conservation drops significantly, whereas the highest degree of interspecies similarity is found within a region of ϳ2 kb upstream of the start site (Fig. 1A), suggesting that regulatory elements crucial for EphA7 expression might be confined to this region. The MatInspector software (30,31) was subsequently employed to scan the 5 kb upstream region of EphA7 for possible HOXD13 binding sites. We previously showed, through optimal DNA binding site selection experiments, that the HOXD13 protein has an equal preference for two types of sites, one having TTAT and the other having TTAC as core consensus sequence (24). Based on our previous results, a binding site matrix was generated, using the MatInd program (30). We identified five putative HOXD13 binding sites within the 5Ј region of the EphA7 gene having a matrix similarity of 1.0 (Fig.  1A). Four of them have TTAT as a core consensus sequence, and one has TTAC. Moreover, four of the identified sites are located within the proximal region of the promoter (sites 2-5; Fig. 1A), and three of these (sites 3-5) are located downstream from the putative transcription start site (Fig. 1, A and B). A sequence alignment of these putative HOXD13 DNA binding sites with the corresponding available genomic sequences from different species revealed that of the five sites identified within the mouse genome, only site 3 displays a high degree of conservation (Fig.  1A, bottom part). These results indicated that the sequences upstream to and in the vicinity of the presumed transcription start site of EphA7 contain several putative HOXD13 binding sites that match the optimal consensus binding sequence determined in vitro, suggesting that Hoxd13 and possibly also Hoxa13 may directly associate with these sites to regulate EphA7 expression.
The EphA7 Promoter Is Bound in Vivo by HOXD13-We then set out to verify which of the putative sites was actually bound in vivo by HOXD13. For this purpose, we exogenously expressed a HA-tagged HOXD13 protein both stably and transiently in two different mouse cell lines of mesenchymal derivation, C3H10T1/2 and NIH3T3 fibroblasts. In C3H10T1/2 cells, stable expression was achieved by retrovirus-mediated gene transfer ( Fig. 2A, left), whereas transient expression in NIH3T3 cells was obtained using a plasmid expression vector ( Fig. 2A,  right). We found that the exogenous expression of HOXD13 could efficiently activate transcription of the endogenous EphA7 gene both in C3H10T1/2 and in NIH3T3 cells (Fig. 2B), indicating that the expressed HOXD13 protein is functional and that the chosen cell backgrounds are permissive for activation of EphA7 transcription by HOXD13 and therefore represent suitable models for studying the interactions of HOXD13 with the EphA7 promoter in vivo.
To verify the binding in vivo of HOXD13 to the candidate binding sites within the EphA7 promoter, we used the chromatin formaldehyde cross-linking and immunoprecipitation (ChIP) technique (32). Cross-linked and sonicated chromatin from HA-HOXD13-expressing C3H10T1/2 cells was immunoprecipitated using an anti-HA antibody (␣HA) and was analyzed by PCR for the presence of the five putative HOXD13 binding sites. In addition, as a control, we tested for the presence of a region of 263 bp located at Ϫ6000 bp (control site; Fig. 3A). As shown in Fig. 3B, the immunoprecipitated chromatin showed a substantial enrichment only of the sequence including site 3, indicating that only site 3 was efficiently bound in vivo by HOXD13. No enrichment was detected for the remaining sites, as well as for the control site (Fig.  3B). Thus, of the five sites matching the optimal DNA-binding consensus sequence of HOXD13, which could hence all be potentially bound by HOXD13, only site 3 turned out to be actually contacted in vivo by HOXD13.
We then wanted to test whether also endogenous Hoxd13 is bound in vivo, in the developing limb, to site 3 of the EphA7 promoter. To this end, chromatin was prepared from E13.5 mouse fore and hind limbs and immunoprecipitated using a ␣Hoxd13 antibody. As a control, immunoprecipitation experiments were performed using chromatin extracted from E13.5 mouse brain, a tissue that does not express Hoxd13 (33). The immunoprecipitated limb chromatin showed a significant enrichment of the fragment containing site 3 (Fig. 3C). No enrichment of the site 3-containing sequence was observed in the control brain chromatin (Fig. 3C), as well as no enrichment of the control sequence (C site; Fig.  3C) was detected. Thus, these data confirm that site 3 is bound in vivo during mouse limb development at a stage when both Hoxd13 and EphA7 are coexpressed in the perichondrium of the digit condensations.
The HOXD13(I47L)-mutated Protein Does Not Bind to the EphA7 Promoter-We previously reported that the HOXD13(I47L) mutation, identified in patients showing a novel brachydactyly-polydactyly syndrome, caused a selective impairment of the DNA binding potential of HOXD13 (24). Indeed, the HOXD13(I47L) mutant protein, unlike its wild type counterpart, proved to be unable to up-regulate the expression of the endogenous EphA7 gene, if overexpressed in developing chick limbs, suggesting that it is unable to bind to the EphA7 promoter. To test this assumption, we verified by chromatin immunoprecipitation whether HOXD13(I47L) was able to bind site 3 of the EphA7 promoter in vivo. HA-HOXD13, HA-HOXA13, or HA-HOXD13(I47L) was transiently expressed in NIH3T3 cells, and their chromatin was tested by immunoprecipitation using an ␣HA antibody. As in the case of HA-HOXD13-expressing C3H10T1/2 cells, the immunoprecipitated chromatin from NIH3T3 cells expressing HA-HOXD13, showed a significant enrichment only of the sequence containing site 3 ( Fig. 4A; data not shown), indicating that also in this cell background only site 3 of the EphA7 promoter was occupied in vivo by HOXD13. Similarly, HOXA13, which was previously reported to be a candidate regulator of EphA7 in limb development (19), was found to be bound in vivo to EphA7 site 3 (Fig. 4A). In contrast, chromatin immunoprecipitated from NIH3T3 cells expressing HA-HOXD13(I47L) showed no enrichment for the sequence containing site 3 (Fig. 4A), indicating that HOXD13(I47L) does not bind to this site in vivo.
We then verified the binding of HOXD13 and HOXD13(I47L) to EphA7 site 3 in vitro in electrophoretic mobility shift assays. As shown in Fig. 4B, HOXD13 bound efficiently the site 3 sequence, whereas HOXD13(I47L) bound site 3 weakly, only at higher concentrations. A control site, having a TTAC core sequence, was bound comparably well by HOXD13 and HOXD13(I47L) (Fig. 4B). Taken together, these results show that HOXD13(I47L), unlike its wild type counterpart, does not bind to EphA7 site 3 both in vivo and in vitro, providing a molecular basis for the lack of regulation of EphA7 expression by HOXD13(I47L).
HOXD13 and HOXA13 Activate Transcription from the EphA7 Promoter, whereas HOXD13(I47L) Does Not-We next wanted to assess whether the EphA7 promoter region could mediate transcriptional activation by HOXD13 or HOXA13. For this purpose, we generated a lucif-erase reporter construct containing a fragment ranging from Ϫ2000 to ϩ55 of the EphA7 genomic sequence (pXP-EphA7(Ϫ2000); Fig. 5A), which comprises the promoter region with the highest degree of evolutionary conservation and includes site 3 (Fig. 1A). NIH3T3 cells were FIGURE 1. The 5 region of the mouse EphA7 gene contains potential HOXD13 binding sites. A, interspecies comparison of the EphA7 genomic sequences spanning from Ϫ6000 bp upstream to ϩ500 bp downstream to the putative transcription start site. Alignments to the Chr. 4:28992000 -28998500 mouse genomic sequence were performed using the UCSC Genome Browser (29). The highest density of interspecies conservation is found within a region of 2 kb upstream to the transcription start site (indicated by an arrow in the diagram below). The degree of sequence conservation drops significantly for distances greater than 5 kb upstream to the start site of transcription. A schematic representation of the EphA7 genomic region is shown below. An arrow indicates the transcription start site. Five different putative HOX binding sites (sites 1-5, indicated by black boxes) were found within the analyzed genomic region using the MatInspector software and a HOXD13 binding site matrix generated by the MatInd program (30). The interspecies conservation of the five sites is shown below the schematic diagram. Gray boxes highlight the consensus core binding sequences. B, nucleotide sequence of the EphA7 proximal promoter region (Ϫ497 to ϩ286). The putative transcription start site (ϩ1) is indicated by an arrow. The gray boxes highlight HOXD13 binding sites. The conserved HOXD13 binding site 3 is underlined. DPE, downstream promoter element. SP1, binding sites for the Sp1 transcription factor. The EphA7 protein translation start site (ATG) is indicated.
transiently co-transfected with pXP-EphA7(Ϫ2000) and increasing amounts of SV40-driven constructs expressing HA-HOXD13, HA-HOXA13, or HA-HOXD13(I47L). As shown in Fig. 4B, both HOXD13 and HOXA13 significantly increased the basal reporter activity, whereas the expression of HOXD13(I47L) led only to a modest activation of the reporter. We then generated a deletion construct (pXP-EphA7(Ϫ497); Fig. 5A) containing a fragment from Ϫ497 to ϩ55 of the EphA7 promoter, comprising site 3. As shown in Fig. 4C, both HOXD13 and HOXA13 could efficiently activate the pXP-EphA7(Ϫ497) reporter basal activity to levels comparable with those obtained with the pXP-EphA7(Ϫ2000) reporter. Conversely, the coexpression of HOXD13(I47L) induced only a minor activation of the pXP-EphA7(Ϫ497) reporter (Fig. 5B). Finally, we mutated the sequence of site 3 within the context of the pXP-EphA7(Ϫ497) reporter (pXP-EphA7(Ϫ497m); Fig. 5A), changing it from ATATTATTGG into ATAGGCGTGG, to alter the core consensus sequence for HOXD13 binding. As shown in Fig. 5C, the expression of HOXD13, HOXA13, or HOXD13(I47L) had virtually no effect on the (pXP-EphA7(Ϫ497m) reporter basal activity, indicating that a mutation of site 3 was sufficient to abolish transactivation by HOXD13 and HOXA13 on the EphA7 promoter.
These results confirm the importance of site 3 in mediating transcriptional activation of the EphA7 promoter by paralog group 13 Hox proteins. These data moreover show that, unlike HOXD13 and HOXA13, HOXD13(I47L) is unable to bind to EphA7 site 3 and thus to regulate the EphA7 promoter. This result is in accordance with the inability of HOXD13(I47L) to up-regulate endogenous EphA7 in developing limbs (24).

DISCUSSION
The nature of the downstream target genes for Hox proteins has remained largely elusive. Whereas various approaches have led to the identification of a number of genes that are differentially expressed in relation to Hox gene activity, few studies have addressed the direct regulation of target gene promoters by Hox transcription factors. In this work, we have analyzed the promoter of the mouse EphA7 gene for direct regulation by the HOXD13 and HOXA13 proteins. Recent reports have pointed to the EphA7 ephrin tyrosine kinase receptor gene as a possible downstream target of Hoxa13 and Hoxd13 protein function. Indeed, EphA7 is co-expressed in developing limbs with Hoxd13 and Hoxa13 in the perichondrium of the mesenchymal condensations of the phalanges, and the expression of EphA7 was found to be significantly reduced in Hoxa13 Ϫ/Ϫ mutant mice (19). Additionally, the overexpression of Hoxd13 by retroviral infection of developing chick limbs was shown to induce a marked increase of EphA7 expression in the phalangeal mesenchymal condensations, indicating that EphA7 transcription might be regulated by Hoxd13 as well (24).
The Genomic Region Upstream to EphA7 Contains Multiple Binding Sites for Group 13 HOX Proteins-Since the EphA7 promoter had never been previously characterized, we analyzed the genomic sequence upstream to the transcription start site of EphA7 for the presence of putative Hoxd13 binding sites using an in silico approach. Binding site selection studies have shown that the Drosophila AbdB protein has a preference for sites with a TTAT core sequence and that Abd-B-related vertebrate Hox proteins preferentially bind to sites with a TTAT or TTAC core sequence (34,35). In accordance, we previously reported that HOXD13 binds equally well to two distinct sites, one TTTTAT-TGG with TTAT as core binding sequence and the other TTTACGAG with TTAC as core element (24). Similarly, Shen et al. (35) reported a Hoxd13 consensus binding site with a TTAC core sequence. Based on these data, we generated a position weight matrix (36) describing the Hoxd13 binding site and exploited it to identify Hoxd13 binding sites within the EphA7 promoter region. Five different putative Hoxd13 binding sites were thus identified, matching with the reported Hoxd13 consensus binding sequences. Four of the putative Hoxd13 binding sites (sites 2-5) were found to match with the TTAT-based consensus site, whereas only one (site 1) matched with the TTAC-based consensus. Interestingly, four of the identified sites are localized to the most conserved part of the EphA7 promoter region (Ϫ2000 bp to ϩ500 from the transcription start site), and three of them (sites 3-5) map 3Ј and in close proximity to the transcription start site.
HOXD13 and HOXA13 Bind in Vivo to a Single Evolutionarily Conserved Site within the EphA7 Promoter-The presence of multiple sites matching the optimal DNA binding sequence of Hoxd13 would in theory imply that Hoxd13 and possibly Hoxa13 occupy concomitantly all or most of these sites to regulate EphA7 expression. We found, however, by ChIP that of the five putative Hoxd13 binding sites within the EphA7 promoter, only site 3 (TTATTG) is actually bound in vivo by HOXD13 as well as by HOXA13, indicating that site 3 is the only functionally relevant paralog group 13 Hox binding site within the EphA7 promoter. In accordance with this result, a mutation of site 3 in the context of the EphA7 promoter was sufficient to abolish the transcriptional activation of EphA7 by HOXA13 and HOXD13.
Significantly, site 3 is the only HOX group 13 binding sequence within the EphA7 regulatory region that shows a substantial degree of evolutionary conservation, being identical in six of seven species compared. The remaining sites, conversely, display a considerably lower degree of conservation, with site 1 being present only in the mouse genome. Thus, evolutionary conservation within the EphA7 promoter, of the site 3 paralog group 13 regulatory element is consistent with its functional relevance in the regulation of EphA7 expression.
The sequence of EphA7 site 3 (TTATTGG) matches with the optimal DNA binding sequence determined in site selection experiments (TTT-TATTGG), indicating that the binding specificity of HOXD13 in vitro corresponds to that observed in vivo. It can be therefore concluded that paralog group 13 Hox proteins apparently do not require co-factors, which alter and/or increase their intrinsic DNA binding specificity, to bind in vivo to their regulatory element within the EphA7 promoter. Similar results were reported for the Bmp2 and Bmp7 regulatory elements, which were shown to be bound by Hoxa13 in vivo at discrete sequences that are closely related to the optimal DNA-binding consensus sequence for paralog group 13 Hox proteins (14).
Our finding that of the multiple potential binding sites for group 13 Hox proteins within the mouse EphA7 promoter only a single site is actually bound in vivo is in contrast with a previously proposed model for Q50 homeodomain protein binding in vivo (37). This model was suggested on the basis of the in vivo binding behavior of the Drosophila Eve and Ftz proteins. According to it, Q50 homeodomain proteins, unlike the majority of transcription factors, bind, in the absence of cofactors, at uniform levels to many sites throughout the length of their target genes (38,39). Our results and the results reported by Knosp et al. (14) for the Hoxa13 protein conversely indicate that vertebrate group 13 Hox proteins bind to discrete sites within specific regulatory regions of their direct target genes, behaving in this respect as "conventional" transcription factors. The discrepancy between the binding behavior of Eve/Ftz and group 13 Hox proteins may derive from the possible differences in the intrinsic properties of these proteins or, more likely, from the different approaches used to detect in vivo binding. Indeed, whereas the in vivo binding by Eve and Ftz was examined by UV photocross-linking, a technique that allows efficient mapping of binding in vivo only to relatively large DNA fragments, possibly including many different binding sites, in this and in the study by Knosp et al. (14), formaldehyde cross-linking ChIP was used, which allows the detection of binding at a higher resolution (40). In fact, we were able to clearly discriminate between bound (site 3) and unbound (sites 4/5) sites located only 165 bp apart.
Several mechanisms could be envisaged to explain the lack of binding in vivo by Hoxd13 and Hoxa13 to sites matching the optimal DNA-binding sequence for paralog group 13 Hox proteins. However, chromatin configuration or the binding of other transcription factors at or in close proximity to these binding sites are the most likely causes for their lack of accessibility and thus of their functional ineffectiveness. Indeed, the DNA context was shown to play an important role in differentiating functional binding sites from mere physical binding sites (41,42).
The HOXD13(I47L) Mutated Protein Displays a Selective Impairment of Its DNA Binding Ability in Vivo-We previously reported that a missense mutation that substitutes leucine for isoleucine at position 47 The arrows indicate the primers used in ChIP assays to PCR-amplify genomic regions containing the HOXD13 binding sites or a control region (C site). B, ChIP on C3H10T1/2 cells expressing HOXD13 using ␣HA antibodies. The enrichment was analyzed by PCR using oligonucleotides that amplify genomic regions of 174 bp (site 1), 297 bp (site 2), 245 bp (site 3), or 221 bp (site 4/5) within the mouse EphA7 promoter. No enrichment of the genomic regions containing sites 1, 2, and 4/5 or of the control site was detected even after 41 cycles of amplification. PCRs were performed in triplicate; a representative set is shown. C, ChIP on pooled fore and hind limbs or on brain tissue, dissected from E13.5 mouse embryos, using an anti-Hoxd13 (␣Hoxd13) antiserum. PCRs were performed in triplicate; representative experiments are shown. No enrichment was observed with immunoprecipitated brain chromatin even after 42 cycles of amplification, as well as no enrichment was detected for the control site (C site). I, input chromatin; CA, non-specific control antibody.  1-3, 7, and 8), or GST-HOXD13(I47L) (lanes 4 -6, 9, and 10) proteins were used. Retarded complexes are indicated by an arrow.
(I47L) of the HOXD13 homeodomain, identified in patients showing a novel brachydactyly-polydactyly syndrome, causes a selective impairment of the ability of HOXD13 to bind DNA. HOXD13(I47L) was found to be unable to recognize binding sites having TTAT or TAAT as the core sequence, whereas sites with a TTAC or TAAC core were bound with the same efficiency as wild type HOXD13 (24). The EphA7 promoter contains one potential Hoxd13 binding site with a TTAC core (site 1, TTTACG) and could therefore in theory be regulated by HOXD13(I47L) as well. We previously observed, however, that misexpression of the HOXD13(I47L) mutant protein, as opposed to wild type HOXD13, did not result in the up-regulation of endogenous EphA7 in developing chick limbs (24), suggesting that HOXD13(I47L) is unable to bind the EphA7 promoter. Indeed, our results indicate that the only functionally relevant HOX group 13 binding site within the EphA7 promoter is site 3 (TTATTG), with a TTAT core sequence, which in principle should not be recognized by HOXD13(I47L). Using ChIP analysis, we could confirm that EphA7 site 3 is not bound in vivo by HOXD13(I47L). HOXD13(I47L) proved furthermore to be unable to activate transcription from the EphA7 promoter. Finally, the only potential binding site with a TTAC core, site 1, turned out to be the least evolutionarily conserved group 13 Hox binding site within the EphA7 promoter, thus further supporting the assumption that it is functionally irrelevant. In conclusion, our results showing that EphA7 is a direct target of Hoxd13 allow us to confirm that the selective loss of its DNA binding ability leads to a failure of HOXD13(I47L) to regulate some of the direct target genes that are normally controlled by HOXD13, providing a molecular basis for the pathogenesis of the novel brachydactyly-syndactyly syndrome produced by this mutation.
EphA7 Is a Direct Downstream Target of Hoxd13 and Hoxa13 in Limb Development-Signaling between Eph and their ephrin ligands plays a fundamental role in controlling several developmental processes and has been proposed to be a downstream effector of a number of Hox genes in some of these (reviewed in Ref. 43). Evidence for a role of Eph-ephrin signaling also in limb patterning has been recently accumulating (19,22,23). During limb development, segmentation and bifurcation processes shape blocks of condensed mesenchyme to generate the prepattern of the limb skeletal elements. These processes are particularly important in the autopod, which is composed by numerous skeletal elements such as the phalanges of the digits and the carpal/tarsal bones. Eph-ephrin signaling in the limbs was proposed to be involved in the regulation of intercellular affinity, which is a prerequisite for the correct condensation of limb mesenchyme (22,23). Thus, positional identity of the mesenchymal cells, as provided by the expression of Hox genes, may be implemented by differential variations in cellular affinities, which affect the pattern of limb mesenchymal condensations. Hoxa13 and Hoxd13 probably play crucial and partially overlapping roles in this process. Indeed, in Hoxa13 Ϫ/Ϫ mutant mice, the fore and hind limb most anterior digits are absent and the prechondrogenic condensations that give rise to the carpal, metacarpal, tarsal, and metatarsal bones are not properly formed. Analogously, Hoxd13 Ϫ/Ϫ mutant mice display a significant growth retardation of several autopodal cartilage elements and a selective lack of bony elements that develop late in wild type mice (44). In Hoxa13 and Hoxd13 double homozygous mutant mice, the mesenchymal condensations of the autopod are not formed at all (45). Moreover, the misexpression of Hoxa13 in developing chick limbs was shown to alter mesenchymal cell-cell interactions, further underscoring the role of group 13 Hox genes in the regulation of genetic pathways controlling aggregation and sorting properties of the distal autopodal mesenchyme (6). Our data show that both Hoxd13 and Hoxa13 bind to the EphA7 promoter in vivo and activate transcription from it. They moreover show that endogenous Hoxd13 is bound to the EphA7 promoter in developing mouse limbs at E13.5, a stage when both Hoxd13 and EphA7 are co-expressed in the perichondrium of the digit condensations. Thus, taken together, our results establish that Hoxd13 and Hoxa13 directly control the expression of EphA7 in the developing autopod, providing further evidence that Hox gene products directly control the aggregation adhesion properties of the mesenchymal cells within the limbs by regulating the ephrin/ephrin receptor signaling system.