Retinoic acid-responsive enhancers located 3' of the Hox A and Hox B homeobox gene clusters. Functional analysis.

Homeobox genes control the spatial identity and differentiation of tissues in the developing vertebrate embryo. Retinoids are signaling molecules involved in the regulation of Hox genes. We previously identified a 3′ enhancer called the RAIDR5, which contained a DR5 retinoic acid response element (RARE) and was responsible for the retinoic acid (RA)-associated expression of the murine Hoxa-1 gene in teratocarcinoma cells. We demonstrate that a similar enhancer, which contains a DR5 RARE, is located at a DNase I-hypersensitive site 3′ of the murine Hoxb-1 gene. This enhancer, the Hoxb-1 RAIDR5, regulates the RA responsiveness of the Hoxb-1 gene and is different in location and sequence from the RA-regulated 3′ Hoxb-1 enhancers previously described. Several DNA elements within the murine Hoxa-1 RA-inducible RAIDR5 enhancer, including the DR5 RARE, conserved element (CE) 1, and CE2, are conserved in the murine Hoxb-1 RAIDR5 enhancer, the human homolog of Hoxa-1, and in the chicken Hoxb-1 gene. Gel shifts show that the CE2 sequence TATTTACTCA binds an RA-inducible factor, while UV cross-linking indicates that a 170-kDa protein binds to this sequence. Thus, the Hoxa-1 and Hoxb-1 genes possess 3′ enhancers with similar sequences through which their expression and responsiveness to endogenous retinoids are controlled.

Genetic analysis of Drosophila melanogaster led to the identification of the homeotic genes, which specify the identity of body parts during embryonic development (for reviews, see Refs. [1][2][3][4]. The vertebrate homologs of these genes, the homeobox or Hox 1 genes, appear in structurally related chromosomal gene clusters known as the Hox A, B, C, and D clusters, which are thought to have arisen from a single ancestral clus-ter by repeated duplication. The expression domains of Hox genes during embryogenesis have been suggested to create a "code" that specifies the positional fate of cells in the embryo (5,6). Consistent with this function, experiments have demonstrated that changes in the expression of Hox genes lead to corresponding changes in tissue identity (7)(8)(9). Hox genes encode homeodomain-containing transcription factors (10), and expression of a particular Hox protein, or a combination of proteins, is thought to specify a positional identity by creating a unique pattern of gene expression (for reviews, see Refs. 2 and 11).
There is considerable evidence that retinoids such as retinoic acid (RA) play an important role in regulating Hox genes. Retinoids are required for normal vertebrate embryogenesis; the offspring of animals deficient in vitamin A have numerous developmental defects (12,13). The addition of exogenous retinoids during vertebrate embryogenesis also leads to developmental abnormalities in zebrafish (14), Xenopus (15)(16)(17)(18), chickens (19), hamsters (20); mice (21)(22)(23)(24), and humans (25). Most striking is the ability of locally applied retinoids to respecify positional identity during development or regeneration of the vertebrate limb (for reviews, see Refs. 9 and 26 -28). Such observations have led to the hypothesis that retinoids function as morphogens in vertebrates, specifying positional identities in a concentration-dependent fashion.
The retinoic acid receptors (RARs ␣, ␤, and ␥) and retinoid X receptors (RXRs ␣, ␤, and ␥) are members of the steroid/thyroid/ retinoid gene family, which includes the steroid receptors, thyroid hormone receptors, the RARs, the RXRs, and the vitamin D receptor, and are ligand-inducible regulators of transcription (for review, see Ref. 29 -31). In addition to all-transretinoic acid, 9-cis-retinoic acid (32,33), 4-oxoretinoic acid (34), 3,4-didehydro-RA (35), 4-oxoretinaldehyde (36), and 4-oxoretinol (37) are activating ligands for the RARs and can affect pattern formation in vertebrates (for review, see Ref. 38). RAR-RXR heterodimers are thought to be the species which control the expression of RA-responsive genes (for review, see Refs. 30 and 39). There is evidence that these heterodimers bind either DR 2 or DR 5 retinoic acid-response elements (RAREs). These elements consist of two "direct repeat" (DR) sequences of AGGTCA separated by either two or five bases (for review, see Ref. 30). Animals that lack more than one normal RAR show developmental abnormalities that correspond to those seen in vitamin A deficiencies, definitely demonstrating the importance of signal transduction by the retinoic acid receptors in vertebrate development (40,41).
Evidence that retinoids were involved in the regulation of homeobox genes initially came from cell culture studies (42)(43)(44)(45). In both human and murine teratocarcinoma cells, retinoic acid can induce the expression of many homeobox genes. This retinoic acid inducibility, like the expression pattern of genes in the embryo, corresponds to the order of the genes in the cluster (46,47). Thus, the 3Ј most genes in each chromosomal cluster, e.g. HOXA1 and HOXB1, are the first to be induced by RA. This induction is transient in that it is observed over a 12-24-h period, which is followed by a decline in expression (44 -47). Co-linear RA responsiveness has also been demonstrated in mouse and Xenopus embryos (17, 48 -50). These observations indicate that the expression of Hox genes is controlled in part by retinoic acid or by other retinoids that activate the retinoic acid receptors.
The murine Hoxa-1 gene is the 3Ј most homeobox gene in the "Hox a" chromosomal cluster on mouse chromosome 6 (51). Previously, we described the identification of a retinoic acidresponsive enhancer located approximately 4.5 kilobases 3Ј of the murine Hoxa-1 gene (formerly Hox 1.6; see Scott (52) for nomenclature). This enhancer, called the Hoxa-1 RAIDR 5 , was shown to be required for retinoic acid inducibility of Hoxa-1 in cultured F9 and P19 teratocarcinoma cells, and the ϳ300-bp enhancer functioned in either orientation when placed upstream of the basal TK promoter (53). Furthermore, this enhancer function was shown to require a DR 5 RARE since point mutations within the RARE abolished RA inducibility (53). We now report the identification of very similar enhancers located 3Ј of the murine and chicken Hoxb-1 (formerly called Hox 2.9) genes. The murine Hoxb-1 gene is the 3Ј most homeobox gene in the "Hox b" chromosome cluster on mouse chromosome 11 (51). We also report the identification of a highly conserved 3Ј enhancer in the human HOXA1 gene. We demonstrate that these related RAIDR 5 enhancers all mediate retinoic acid responsiveness in cultured cells and share common sequence elements, including a DR 5 RARE.

EXPERIMENTAL PROCEDURES
Isolation of Genomic Clones-The human HOXA1 clones were isolated from a genomic library from human lymphocyte DNA in EMBL3 obtained from Dr. Winnie Wong. Standard hybridization conditions were modified by reduction of formamide concentration to 25%, and the probe used was the PstI-EcoRI fragment containing the 3Ј RA-responsive enhancer from the murine Hoxa-1 gene (53). Hoxb-1 and Hoxd-1 genomic clones were isolated from FIX (Stratagene) and EMBL (constructed by Anton Berns) libraries containing 129SV genomic DNA. The Hoxb-1 cDNA probe was provided by Dr. Joseph Grippo. The chicken Hoxb-1 phage (54) was obtained from Dr. Gregor Eichele. Library screening, phage purification, and DNA isolation were by standard methods (55).
CAT Transfection Assays and ␤-Galactosidase Assays-Standard methods were used to subclone DNA fragments from genomic phage into Bluescript, pBLCAT2 (56), or pGOTKCAT. The latter is a derivative of pGL2 (Promega) in which the luciferase gene is replaced by the TK minimal promoter and the CAT reporter from pBLCAT2 and includes a transcriptional terminator upstream of the polylinker to reduce background. Transient transfections, CAT assays, and ␤-galactosidase assays were performed as described previously (53). Quantitation of CAT assays was carried out using a PhosphorImager (Molecular Dynamics). ␤-Galactosidase units were calculated by the formula: ␤-gal units ϭ (A 420 ) (100)/(volume of lysate per reaction) (time).
Mutagenesis of Conserved Elements-Conserved elements 1 and 2 were mutated by polymerase chain reaction amplification of subcloned enhancer fragments with the following oligonucleotides, paired with primers to flanking plasmid sequences: CE1, top strand: 5Ј-CTGAGGATCCCGGAGGCTATTCAGATGC-3Ј; CE1, bottom strand: 5Ј-GGGATCCTCAGCTCAAAGGTGAACCCAAAGC-3Ј; CE2, top strand: 5Ј-CGCTCTGCAGAACAGTCTGAAAGGGTTG3-Ј; and CE2, bottom strand: 5Ј-TGTTCTGCAGTTGATCGGTTTGAAT-3Ј. These oligomers are complementary to sequences adjacent to conserved elements 1 and 2, but their 3Ј tails are noncomplementary and contain unique restriction sites. Following polymerase chain reaction, products were cut with appropriate enzymes, purified, and ligated. Plasmid primers were then used to amplify the fragments containing disrupted CE1 or CE2. To generate an enhancer with mutations in both CE1 and CE2, the process was repeated with a fragment containing mutations in CE1, using the oligonucleotides to CE2. Products were sequenced to confirm that the mutations in the conserved elements were the only changes introduced. Mutated DNA fragments were then cloned into the Hoxa-1 (1.6) lacZ minigenes (see Fig. 5). These minigenes differ from the previously described constructs (53) in that the 3Ј end of the lacZ gene is joined to the PvuII site at bp 203 of the Hoxa-1 (1.6) cDNA. The resulting constructs include the constitutive intron of Hoxa-1, unlike the minigenes used in previous work (53) and lack only 44 base pairs of exon 1.
Electrophoretic Mobility Shift Assays-The sequences of complementary CE2 oligonucleotides used as probe in the binding assays are: sense CE2, GGCAAACCGATCAATATTTACTCAGTCTGA; and antisense CE2, GGTCAGACTGAGTAAATATTGATCGGTTTG. The mutated CE2 oligonucleotides are identical to those above except they contain the dinucleotide substitutions as indicated in Fig. 6. The sequence of the AP-1 consensus oligonucleotides are: sense AP-1, AACTTTGACTCAG; and antisense AP-1, CTGAGTCAAAGTT. Complementary nucleotides were annealed by heating at 85°C for 2 min, followed by successive 15 min incubations at 65, 37, 22, and 4°C. The double-stranded oligonucleotide to be used as probe was end-labeled by filling in 5Ј overhangs with an [␣-32 P]dCTP using Klenow enzyme. Binding reactions were performed in a total volume of 20 l containing: 0.5 ng (50,000 cpm) of probe, 5 g of nuclear extract, 1 ϫ binding buffer (10 mM Hepes, pH 7.9, 1 mM MgCl 2 , 60 mM KCl, 0.5 mM EDTA, 1 mM DTT, 10% glycerol), and 2 g of poly(dI-dC). Samples were incubated for 20 min at 22°C and separated on a 4% nondenaturing polyacrylamide gel, which was dried and visualized by autoradiography. Supershift experiments were performed similarly to the standard binding assay with the following modifications. Following the 20-min 22°C incubation, 2 l of either anti-Fos or anti-Jun antibody (Santa Cruz Biotechnology) was added to the binding reaction and incubated another 2 h at 4°C. Samples were separated on a 6% nondenaturing polyacrylamide gel at 4°C.
Ultraviolet Cross-linking of Protein to DNA-Binding reactions were carried out in a 20-l volume as described in the previous section. Following a 20-min incubation, the reactions were exposed to 254-nm wavelength ultraviolet light from a hand-held luminometer for 10 min at a distance of 5 cm. 20 l of 2 ϫ SDS-sample buffer (100 mM Tris-HCl, pH 6.8, 100 mM DTT, 4% (w/v) SDS, 0.2% (w/v) bromphenol blue, 20% (w/v) glycerol) were added to each reaction, and the samples were boiled for 5 min to denature proteins. The samples were subjected to SDSpolyacrylamide gel electrophoresis utilizing a 5% stacking gel and an 8% separating gel. A 10-kDa protein ladder (Life Technologies, Inc.) was run as a size standard. To visualize the protein standards, the gel was stained overnight in Coomassie Blue staining solution (50% (v/v), methanol 0.05% (v/v) Coomassie Brilliant Blue R-250, 10% (v/v) acetic acid, and 40% H 2 O) and destained for several hours in 5% (v/v) methanol, 7% (v/v) acetic acid, and 88% H 2 O. The gel was then dried and visualized by autoradiography.

RESULTS
Identification of a Retinoic Acid-responsive Enhancer 3Ј of the Human HOXA1 Locus-Since we had previously identified a retinoic acid-responsive enhancer located about 4.5 kb 3Ј of the start site of transcription of the murine Hoxa-1 gene (53), we wanted to determine whether or not this enhancer sequence was conserved in other Hox genes located in the most 3Ј position of Hox chromosomal clusters such as the murine Hoxb-1 gene, the chicken Hoxb-1 gene, and the human HOXA1 gene. Genomic clones for the murine Hoxa-1 and Hoxb-1 genes, the human HOXA1 gene, and the chicken Hoxb-1 gene were isolated or obtained as described under "Experimental Proce-dures." (Murine Hox genes are labeled using small letters, human Hox genes with capital letters (52).) DNA hybridization experiments demonstrated that the human and murine Hoxa-1 genes were homologous not only in their transcribed regions, but also in portions of their 3Ј-flanking DNA. DNA fragments containing the ϳ300-bp murine Hoxa-1 retinoic acid-responsive enhancer, called RAIDR 5 (53), were used to identify homologous sequences in the human HOXA1 gene. This fragment of human DNA was tested for the presence of a retinoic acidresponsive enhancer by subcloning cross-hybridizing DNA fragments into a minimal promoter-CAT vector and transfecting these constructs into F9 or P19 teratocarcinoma cells. The structure of the human HOXA1 gene, and the location of the 3Ј RA-responsive enhancer are shown in Fig. 1. One NsiI-SacI restriction fragment mediates retinoic acid responsiveness in transient transfection assays (Fig. 1). Thus, a retinoic acidresponsive enhancer is present about 4.5 kb 3Ј of the human HOXA1 gene, in a location very similar to that which we previously described for the murine Hoxa-1 gene (53).
Identification of a Retinoic Acid-responsive Enhancer 3Ј of the Murine Hoxb-1 Gene-DNase I hypersensitivity assays were used to map a putative retinoic acid-responsive enhancer near the murine Hoxb-1 gene using the same strategy employed to identify the RA-responsive enhancer 3Ј of the Hoxa-1 gene (53). Following RA treatment of F9 murine teratocarcinoma cells, a DNase I hypersensitive site was detected approximately 6.5 kb 3Ј of the Hoxb-1 gene (data not shown; position of the hypersensitive site is indicated by the open arrow in Fig.  2). Minimal promoter-CAT vectors containing a series of DNA fragments including and surrounding the location of the RAinducible hypersensitive site were shown to mediate responsiveness to retinoic acid, as shown in Fig. 2. More specifically, an NsiI-SacI fragment which includes the region of the hypersensitive site was shown to be sufficient for RA inducibility in the transient transfection assays in either F9 (Fig. 2) or P19 cells (data not shown).
Identification of a Retinoic Acid-responsive Enhancer from the Chicken Hoxb-1 Gene-The chicken Hoxb-1 (2.9) genomic phage DNA was analyzed for 3Ј enhancer activity by subcloning restriction fragments of the 3Ј-flanking DNA and testing various fragments for RA inducibility. In transient transfection experiments, an approximately 2-kb BamHI-EcoRI DNA fragment located 3Ј of the chicken Hoxb-1 gene was shown to confer RA responsiveness on a minimal promoter (Fig. 3). Other 3Ј chicken DNA fragments lacked the ability to confer RA responsiveness (Fig. 3, and data not shown).
Comparison of the DNA Sequences of the 3Ј Hoxa-1 and Hoxb-1 Retinoic Acid-responsive Enhancers: Conservation of a DR 5 RARE and Elements CE1 and CE2-The aligned sequences of the RA-responsive enhancers from the murine Hoxa-1, the human HOXA1, and the murine and chicken Hoxb-1 genes are presented in Fig. 4. First, it is striking that two different homeobox genes, Hoxa-1 and Hoxb-1, located on different chromosomes, have such similar sequences in their RA-inducible enhancers. Second, with respect to the mouse Hoxa-1 and human HOXA1 3Ј retinoid-responsive enhancers, the region of near identity shown in Fig. 4 is part of a large block of conserved sequence. Furthermore, homology of the Hoxa-1 and b-1 enhancers among different species is particularly noteworthy, given that these DNA sequences come from species separated by an estimated 300 million years of evolution (57).
As expected from the ability of these enhancers to mediate RA responsiveness, all four enhancers include a DR 5 (direct repeat 5) RARE. The half sites of these RAREs are precisely conserved, as is their separation by 5 base pairs. Thus, these ϳ300-bp enhancers are called the Hoxa-1 RAIDR 5 and Hoxb-1 RAIDR 5 . Two other conserved elements (CE1 and CE2; boxed in Fig. 4) are also present in all four enhancers. The conservation of these sequences is equivalent to that of the RARE, suggesting that CE1 and CE2 play equally important roles in controlling the expression of the vertebrate homeobox genes. Therefore, we examined the functions of the CE1 and CE2 elements in more detail using electrophoretic mobility shift assays and transient transfection experiments.
Functional Analysis of the CE1 and CE2 in RAIDR 5 Enhancer-To examine the functional importance of CE1 and CE2, mutations were introduced that altered each nucleotide of CE1 and CE2, as described under "Experimental Procedures." In addition, a double mutation in which both CE1 and CE2 elements were disrupted and another mutation in which the conserved RARE was specifically mutated were also characterized. These mutations were made in the context of a Hoxa-1 minigene/lacZ reporter construct that contains the lacZ gene inserted into the coding sequence of Hoxa-1 as well as 6.5 kb of 5Ј-flanking region and about 3 kb of 3Ј-flanking DNA (53) (Fig.  5). P19 teratocarcinoma cells were first transiently transfected with these constructs, and cells were either untreated (stem) or treated for 24 h with RA.
The results from the untreated P19 stem cells showed that mutation of the CE1 element did not significantly change ␤-galactosidase activity as compared with that observed with the wild type construct (Fig. 5). Mutation of either the CE2 element or the RARE also had no effect on the activity of the reporter (Fig. 5). These results indicate that CE1 and CE2 do not appear to be functional elements in P19 stem cells.  -1 (2.9), human HOXA1 (1.6), and murine Hoxa-1 (1.6). Regions of identity of greater than three sequential nucleotides are shaded; blocks of homology common to all four sequences are: conserved elements 1 and 2 (CE1 and CE2, boxed) and the direct repeats of the RAREs (arrows). Gaps are introduced to maximize sequence alignment. This region corresponds to the region between the XhoI site (ϩ1) and SacI site (ϩ467) of the murine Hoxa-1 gene (Fig. 1) from nucleotides 238 to 455. The activities of all of the reporter constructs were increased by the addition of RA to P19 cells, with the exception of the construct containing a mutated RARE (Fig. 5). Mutation of CE1 resulted in an increase of approximately 3.0-fold and mutation of CE2 resulted in a 4.9-fold increase in activity, as compared to the 3.1-fold RA-associated increase in reporter activity observed with the wild type enhancer/reporter construct (Fig. 5). Mutation of both elements CE1 and CE2 resulted in an RA-associated increase in reporter activity of 3.7fold (Fig. 5). These results indicate that the CE2 element plays a negative role in the RA responsive of the Hoxa-1 gene, since mutation of the CE2 element leads to a greater increase in the response to RA. The CE1 element does not function as a negative element in RA-treated P19 cells.
CE2 of the Hoxa-1 3Ј RAIDR 5 Is a Retinoic Acid-inducible Factor Binding Site-Electrophoretic mobility shift assays were performed using CE1 and CE2 sequences from the mouse Hoxa-1 3Ј RAIDR 5 enhancer as probes and nuclear extracts prepared from P19 or F9 teratocarcinoma stem cells versus P19 or F9 cells that had been treated for 48 h with RA. No binding to the CE1 element was detected in nuclear extracts from either stem cells or RA-treated cells (data not shown). Binding to the CE2 element was also absent in P19 stem cells (Fig. 6,  lanes 2-4). In contrast, binding to the CE2 element was detected in RA-treated P19 cells (Fig. 6, lane 5) and F9 cells (data not shown). CE2 binding was shown to be specific since it could be effectively competed with a 100-fold molar excess of unlabeled CE2 oligonucleotide but could not be competed by an unrelated, nonspecific double-stranded oligonucleotide (Fig. 6,  lane 6). When the CE2 element of the murine Hoxb-1 enhancer was used as a probe, an apparently identical binding complex was observed (data not shown). The CE2 element of the Hoxb-1 enhancer differs from that of the Hoxa-1 CE2 element by 1 base out of 14 (Fig. 4).
Analysis of the CE2 element revealed the presence of an AP-1 site (58 -60). The AP-1 consensus site is TKANTCA where K is a G or T and N represents any nucleotide. To determine whether AP-1 was involved in recognition of the CE2 element, supershift assays were performed using antibodies that recognized conserved regions of both Fos and Jun family members. Under a variety of binding and gel buffer conditions, no shifting of the CE2 complex was observed (data not shown), suggesting that Fos and Jun are not present in the CE2 binding complex. To confirm this, a consensus AP-1 recognition site was examined for its ability to act as a competitor for the CE2 binding complex in an electrophoretic mobility shift assay. A 100-fold molar excess of the unlabeled AP-1 consensus element did not compete for CE2 binding activity (Fig. 7, lane 10), indicating that the CE2 complex did not contain Fos or Jun.
To determine which nucleotides in the murine Hoxa-1 CE2 element are essential for binding, a panel of CE2 oligonucleotides containing dinucleotide base mutations were examined for their ability to compete the CE2 complex binding in an electrophoretic mobility shift assay (Fig. 7). Mutations M1 and M7 were effective competitors and therefore still retained the ability to bind the CE2 complex (Fig. 7, lanes 3 and 9). Mutations M2 through M6 could not compete for the binding complex indicating that the 10 nucleotides, TATTTACTCA, are critical to CE2 binding (Fig. 7, lanes 4 -8). This sequence constitutes a novel binding site since, although it overlaps with the AP-1 site, it contains additional residues that are not essential for AP-1 binding. Importantly, these ten bases are the nucleotides that are specifically conserved in the homeobox 3Ј RAIDR 5 enhancers of the murine Hoxb-1, human HOXA1, and chicken Hoxb-1 genes (Fig. 4). Thus, the CE2 element is likely to play an important role in the transcriptional regulation of The sequence GCATAAATAAT has been identified as a high affinity Brn 3.0 binding site. This sequence overlaps with a complementary region, TATTTA, of the CE2 element. To examine whether Brn 3.0 was involved in binding the CE2 element, a mobility shift assay was performed using nuclear extracts prepared from the pituitary ATt-20 cell line. High levels of both Brn 3.0 and the CE2 factor(s) were detected in these ATt-20 cells, but the Brn 3.0 and CE2 binding complexes migrated with different mobilities on the gel (data not shown). In addition, the Brn 3.0 site could not compete the CE2 complex and the CE2 site could not compete the Brn 3.0 complex, indicating that Brn 3.0 is not involved in binding to the CE2 element (data not shown).
To analyze further the CE2 binding complex, UV cross-linking was performed. Binding reactions were carried out as described under "Experimental Procedures," using nuclear extracts from P19 cells that had been treated for 48 h with RA. Following this incubation, the samples were exposed to ultraviolet light at a wavelength of 254 nm and at a distance of 5 cm for 10 min. The samples were then separated on an 8% SDSpolyacrylamide gel. The probe alone is shown in Fig. 8, lane 1.
In the absence of any UV treatment, no bands are observed (lane 2). Following UV treatment a band is observed at approximately 180 kDa (lane 3), and a smeared band is observed below 60 kDa. The 180-kDa band is specifically competed away by a 100-fold molar excess of unlabeled CE2 competitor (lane 4), but is unaffected by unlabeled competitor that contains a mutation of the CE2 site (lane 5). The smear below 60 kDa is not significantly affected by specific competitor indicating that these are nonspecific interactions. The 180-kDa band observed includes the molecular mass of the 32 bases of one strand of the labeled 32-base pair probe which has a molecular mass of approximately 10 kDa (depending on which strand of the CE2 binding site is effectively cross-linked to the protein). The cross-linked protein therefore has an approximate molecular mass of 170 kDa.
The 3Ј RAIDR 5 Enhancer of the Murine Hoxb-1 Gene Is Distinct from the Previously Identified 3Ј RAIDR 2 Enhancer-In previous work (62)(63)(64)(65)(66), several enhancers surrounding the murine and human Hoxb-1 genes were analyzed, but none of these 5Ј or 3Ј enhancers was reported to contain a DR 5 RARE. However, a functional RARE 3Ј of the murine, chicken, and pufferfish Hoxb-1 (2.9) genes was shown to mediate a portion of the RA responsiveness of the Hoxb-1 gene in transgenic animals. Marshall et al. (62) have reported that this RARE is a functional DR 2 -type RARE in the murine and pufferfish Hoxb-1 genes, while Ogura and Evans (64) have reported similar results for the human HOXB1 gene. We have confirmed that the mouse Hoxb-1 gene has this second RA-responsive enhancer, located 3Ј of this gene but much closer to the transcribed sequence (referred to as the DR 2 RARE, see Fig. 9) than the RAIDR 5 enhancer described above.
This RAIDR 2 enhancer, which contains a DR 2 RARE, exhibits different properties in cultured teratocarcinoma cells when compared to the RAIDR 5 enhancer we have identified in this report. Using P19 teratocarcinoma cells, RA responsiveness of the murine Hoxb-1 gene from this 3Ј RAIDR 2 enhancer could be detected (Fig. 9). However, in F9 teratocarcinoma cells these same constructs did not mediate a response to RA, consistent with the observed absence of a DNase I-hypersensitive site at FIG. 8. UV treatment cross-links a 170-kDa protein to the labeled CE2 probe. Binding reactions were carried out as described previously. All samples were incubated with 5 g of nuclear extract from P19 cells that had been treated for 48 h with RA and the 32 -Plabeled Hoxa-1 enhancer CE2 element with the exception of lane 1 which contains probe only and no nuclear extract. Lane 2 contains extract but was not subjected to UV irradiation. Lane 3 is identical to lane 2 but was subjected to 10 min of UV irradiation. Lane 4 is identical to lane 3 but contains a 100-fold molar excess of unlabeled wild type CE2 competitor, and lane 5 is identical to lane 3 but contains a 100-fold molar excess of the unlabeled mutant CE2 competitor site M3, which was described previously. Samples were separated by SDS-polyacrylamide gel electrophoresis utilizing a 5% stacking gel and an 8% separating gel. The size and migration of known protein standards is indicated (in kilodaltons) on the left. The specific 180-kDa binding complex (including the size of the cross-linked DNA probe) is indicated by the arrow on the right. This experiment was performed three times with essentially identical results obtained in all three experiments. this location in F9 cells (Fig. 2). Thus, the murine Hoxb-1 RAIDR 2 enhancer functions to induce RA responsiveness only in some teratocarcinoma cell lines. This is a markedly different result from that obtained with the RAIDR 5 enhancers; these function in all teratocarcinoma and embryonic stem cell lines we have tested, including F9 and P19 teratocarcinoma cells, and CCE embryonic stem cells (Figs. 1-3 and 5, and data not shown).
The chicken homolog of the Hoxb-1 3Ј proximal enhancer does not show the same cell type variation as the cognate enhancer from the mouse. However, while it functions in both P19 and F9 cells, in F9 cells the fold induction by RA is considerably less (Fig. 9). After examination of the sequence of this DNA fragment from the chicken Hoxb-1, we were unable to identify a DR 2 RARE identical to that previously identified by Marshall et al. (62) or any other blocks of homology between the chicken and its cognate fragment from the mouse Hoxb-1 gene in this region (data not shown). Thus, this DR 2 RARE appears to be less conserved than the DR 5 RARE within the RAIDR 5 enhancers identified in this report. Creation of chimeric constructs between the 3Ј RAIDR 2 enhancer of the mouse Hoxb-1 and similar portions of the chicken Hoxb-1 gene would allow us to map the sequences involved in the cell type-specific function (P19 versus F9 cells) of the murine Hoxb-1 RAIDR 2 .
Further evidence that, in cultured teratocarcinoma cells, the RAIDR 5 and RAIDR 2 enhancers are functionally distinct comes from the fact that the DR 2 RARE from the murine Hoxb-1 gene, without additional genomic sequences, is unable to confer significant RA responsiveness on a heterologous promoter (63,64). 2 In contrast, the DR 5 RARE (a 21-mer with 2 bases on each side of the DR 5 sequence) mediates RA responsiveness in all teratocarcinoma cell lines tested when it is placed upstream of a basal thymidine kinase promoter/reporter construct, even in the absence of co-transfected receptor (53). 2 Thus, the RAIDR 5 enhancer 3Ј of the murine and chicken Hoxb-1 genes, which we describe here, is unlikely to be redundant to other previously described Hoxb-1 regulatory elements (62)(63)(64)66), given the evolutionary conservation of this enhancer and the apparent functional differences between these DR 5 and DR 2 RAREs (for review, see Ref. 30). DISCUSSION Experiments in which sequences from Hox loci have been used to control lacZ expression in transgenic animals have demonstrated the complexity of Hox gene regulation (for review, see Refs. 6 and 27). Multiple, distinct enhancer-like elements, each functioning in a stage-and tissue-specific manner, are thought to act in combination to create the temporal and spatial expression patterns of Hox genes. Similar combinatory action of enhancer elements have been well documented for the homeotic genes of Drosophila (67,68). We previously described the identification of one such enhancer, located at about 4.5 kb 3Ј of the murine Hoxa-1 gene transcription start site, and demonstrated that this enhancer, and specifically the DR 5 RARE within it, was required for the RA responsiveness of the Hoxa-1 gene in teratocarcinoma cells (53).
In this study we show that this RA-responsive enhancer is also present in other homeobox genes similar in chromosomal position to Hoxa-1 (see Fig. 10). Conservation of this RAIDR 5 enhancer sequence suggests that the DR 5 RARE and the two conserved elements CE1 and CE2 serve regulatory functions which are required for expression of both the Hoxa-1 and Hoxb-1 genes. To date, the only RAREs identified in the Hox clusters are the DR 5 RARE of Hoxa-1 (53, 69) (this report for the human HOXA1 gene), the DR 5 RAREs of the murine and chicken Hoxb-1 genes (this report), the DR 2 RAREs in the murine and human Hoxb-1 genes (62)(63)(64)66), and a divergent DR 5 RARE conserved in the promoters of the human and murine D4 genes (70,71). Thus, while it has been shown that most Hox genes are inducible by RA in cultured cells (46,47), how extensive a role direct regulation by activated retinoic acid receptors plays in the creation of the expression patterns of other Hox genes remains to be determined (for review, see Ref. 72).
The potential roles that the conserved elements CE1 and CE2 in the RAIDR 5 enhancer play in the regulation of Hoxa-1 and Hoxb-1 in various cell types are currently being examined in greater depth. The data presented in this report indicates that the CE2 binding factor(s) is a novel, previously unidentified protein(s). The transactivation data indicate that the CE2 element plays a negative role in the RA responsiveness of the Hoxa-1 gene. Although the level of inhibition of expression observed is small, the repression correlates with the induction of CE2 binding activity in P19 cells following RA treatment. In addition, CE2 binding is detected in F9 cells, another teratocarcinoma line, and is up-regulated by RA with the same kinetics as observed for P19 cells (data not shown). Since the Hoxa-1 and Hoxb-1 genes are only transiently activated by RA in murine and human teratocarcinoma cells (44 -47), the RA induction of CE2 binding activity most likely plays a role in the repression of the RAIDR 5 enhancer at later times after RA addition. The relatively weak inhibitory effect observed (Fig. 5) may reflect experimental limitations in using transient transfections to perform this analysis. For instance, previous experiments (53) demonstrated that, following RA treatment, a DNase I hypersensitive site is found at the Hoxa-1 RAIDR 5 3Ј enhancer, indicating a reorganization of chromatin at this site. This change in chromatin may be essential to the function of the CE2 element. The generation of stable cell lines containing the Hoxa-1 minigene reporter constructs should help elucidate any role of chromatin structure on the ability of the CE2 site to function in regulating Hoxa-1 expression.
UV treatment of binding reactions cross-links a single protein band of approximately 170 kDa to the labeled CE2 binding site. UV cross-linking in Fig. 8 was performed for 10 min. In experiments not shown, samples were also UV-treated for time periods of 30, 60, and 90 min. In these cases, the 170-kDa band was still the only band observed. These results indicate that the lack of detection of additional proteins binding to the CE2 element was not due to an insufficient amount of UV crosslinking treatment. UV treatment results in the cross-linking of only a small percentage of protein (73)  to the probe. Therefore, only a single protein of a multimeric protein complex binding to the CE2 element will likely acquire label. Since only one band is observed, these results indicate that the CE2 binding protein(s) recognize the CE2 site as: 1) a monomer, 2) a homodimer, or 3) a heterodimer of two proteins of identical molecular mass. It is also possible that additional protein(s) can bind to this element in other cell types.
The consensus binding site recognized by the CE2 factor(s) is TATTTACTCA. These 10 nucleotides are identically conserved among the 3Ј enhancers of the murine Hoxa-1, murine Hoxb-1, human HOXA1, and chicken Hoxb-1 genes, suggesting that the CE2 element may play a role in regulating all of these genes. The CE2 element overlaps with an AP-1 site, but experiments show that neither Fos nor Jun is present in the binding complex. Although not identical, CE2 does show some homology to other known factor binding sites. Three elements which are similar in sequence to the CE2 binding site are those for the transcription factors Brn 3.0 and MEF-2, and the DE-2 element of the proopiomelanocortin promoter (61, 74 -76).
The 3Ј enhancers identified in this study are likely to be important not only in the regulation of Hoxa-1 and Hoxb-1 genes themselves, but also in the activation of entire Hox loci. Most of the genes in the Hoxa and Hoxb chromosomal clusters (not all have been studied) have been shown to be activated in response to RA in a co-linear manner with respect to time and RA concentration ( Fig. 10) (see Refs. 39 and 72 for review). There are several models which could explain this temporal Hox cluster activation. In one model, the Hox genes would be activated via an RA-responsive "locus control region," similar to that which controls the globin genes (77,78); in this model the 3Ј RAIDR 5 enhancer would function to regulate the RA responsiveness of the entire chromosomal Hox gene cluster. In a second model, the Hoxa-1 and Hoxb-1 genes would be activated directly via the RAIDR 5 enhancers analyzed in this report, and then the protein products of these genes would activate the next 5Ј genes in the clusters, i.e. Hoxa-2 and b-2, respectively. In a third model, multiple RAREs would control the responses of Hox genes in a manner similar to that in which stripes of gene expression are generated in Drosophila embryos by binding sites with differing affinities for the morphogens bicoid and hunchback (79,80). Although there is no direct evidence to support the validity of any one of these three mechanisms, in the case of the first two proposed models, the 3Ј RAIDR 5 enhancers, identified here, are critical to the activation of the entire Hox cluster. Aspects of these models can be tested using cultured teratocarcinoma and embryonic stem cells, and such experiments are currently in progress.