Transcriptional Regulation of the Bmp2 Gene

Bmp2, a highly conserved member of the transforming growth factor-β gene family, is crucial for normal development. Retinoic acid, combined with cAMP analogs, sharply induces the Bmp2 mRNA during the differentiation of F9 embryonal carcinoma cells into parietal endoderm. Retinoic acid (RA) also induces the Bmp2 gene in chick limb buds. Since normalBmp2 expression may require an endogenous retinoid signal and aberrant Bmp2 expression may cause some aspects of RA-induced teratogenesis, we studied the mechanism underlying the induction of Bmp2. Measurements of the Bmp2mRNA half-life and nuclear run-on assays indicated that RA stimulated the transcription rate of the Bmp2 gene. The results of ribonuclease protection and primer extension assays indicated that Bmp2 transcription started 2,127 nucleotides upstream of the translation start site in F9 cells. To identify genetic elements controlling this transcription rate increase, upstream and downstream genomic sequences flanking the Bmp2 gene were screened using chloramphenicol acetyltransferase reporter genes in F9 cells and β-galactosidase reporter genes in Saccharomyces cerevisiae that were cotransformed with retinoic acid receptor and retinoid X receptor expression plasmids. RA-dependent transcriptional activation was detected between base pairs −2,373 and −2,316 relative to the translation start site. We also identified a required Sp1 binding site between −2,308 and −2,298. The data indicate that Bmp2 is directly regulated by retinoic acid-bound receptors and Sp1.

Bmp2, a highly conserved member of the transforming growth factor-␤ gene family, is crucial for normal development. Retinoic acid, combined with cAMP analogs, sharply induces the Bmp2 mRNA during the differentiation of F9 embryonal carcinoma cells into parietal endoderm. Retinoic acid (RA) also induces the Bmp2 gene in chick limb buds. Since normal Bmp2 expression may require an endogenous retinoid signal and aberrant Bmp2 expression may cause some aspects of RAinduced teratogenesis, we studied the mechanism underlying the induction of Bmp2. Measurements of the Bmp2 mRNA half-life and nuclear run-on assays indicated that RA stimulated the transcription rate of the Bmp2 gene. The results of ribonuclease protection and primer extension assays indicated that Bmp2 transcription started 2,127 nucleotides upstream of the translation start site in F9 cells. To identify genetic elements controlling this transcription rate increase, upstream and downstream genomic sequences flanking the Bmp2 gene were screened using chloramphenicol acetyltransferase reporter genes in F9 cells and ␤-galactosidase reporter genes in Saccharomyces cerevisiae that were cotransformed with retinoic acid receptor and retinoid X receptor expression plasmids. RA-dependent transcriptional activation was detected between base pairs ؊2,373 and ؊2,316 relative to the translation start site. We also identified a required Sp1 binding site between ؊2,308 and ؊2,298. The data indicate that Bmp2 is directly regulated by retinoic acid-bound receptors and Sp1.
Bone morphogenetic proteins (BMPs) 1 are developmentally critical growth factors of the transforming growth factor-␤ family that were first described as having osteogenic activity in rats (1)(2)(3). Bmp2 and Bmp4 transcripts are widely expressed in vertebrate embryonic structures undergoing induction and morphogenesis (4 -6). BMP signaling is involved in key embryonic processes such as epithelio-mesenchymal interactions (5), interdigital apoptosis in the developing limb (7), and dorsalventral axis specification (8). Mice having null mutations in the Bmp2 or Bmp4 (9,10) or the Bmp receptor IA (11) genes die during early embryogenesis. The phenotypes of these mutants prove that BMP signaling is required for numerous extraembryonic and embryonic developmental processes.
The evolutionary conservation of the Bmp2 and Bmp4 genes and their Drosophila homolog dpp is remarkable. Conservation exists at both the functional and sequence levels (8,(12)(13)(14). Both BMP2 in mouse and DPP in Drosophila have pleiotropic functions and are expressed in a highly tissue-and stagespecific manner. Multiple promoters and alternative splicing produce three major and several minor dpp transcripts (15). Our work in murine cells and the Drosophila studies indicate that the multiple core promoters are closely involved in Bmp2 and dpp tissue-specific regulation. It is likely that the regulation of this essential growth factor in mammals equals the complexity of dpp regulation in Drosophila.
Many Bmp2-expressing tissues develop abnormally in vitamin A-deficient embryos or after exposure to the potent teratogen retinoic acid (RA). These include the heart and cardiovasculature, limbs, central nervous system, craniofacial structures, and vertebrae (see Refs. 3 and 16 and references therein). The first indication that the Bmp2 gene was regulated by RA was the discovery that it was strongly induced in F9 embryonal carcinoma cells stimulated to differentiate with RA (17). Subsequently, the Bmp2 gene was found to be induced by RA in the developing chick limb (18). Since retinoid signaling may contribute to the normal pattern of embryonic Bmp2 expression and since the aberrant induction of Bmp2 by excess RA may cause some RA-associated deformities, elucidating the genetic regulatory elements controlling the RA inducibility of Bmp2 will increase our understanding of normal development and teratogenesis.
F9 cells, a widely used model of cellular differentiation and early embryonic development, are an excellent biochemical system for investigating RA-inducible genes. F9 embryonal carcinoma cells differentiate rapidly and synchronously into primitive endoderm upon treatment with RA and into parietal endoderm upon treatment with RA and cAMP analogs (19). This model system has been used to identify retinoic acid response elements (RAREs) controlling the expression of several important developmental genes, such as Hoxa1 (20), laminin B1 (21), and now Bmp2.
Here we describe the first genetic regulatory elements controlling the RA-regulated induction of the Bmp2 gene in embryonic cells. RA-regulated gene expression is mediated by nuclear receptors, which act as retinoid-dependent transcrip-tion factors (22). Encoded by six different genes, RAR␣, -␤, and -␥ and RXR␣, -␤, and -␥, the receptors can act as homodimers and heterodimers, often with unique DNA binding and transactivation specificities (see Refs. 23 and 24). RARs bind to and are activated by all-trans-RA and 9-cis-RA, while the RXRs are activated only by 9-cis-RA. Our experiments in yeast suggest that, like Hoxa1 and RAR␤, ligands that bind both RARs and RXRs synergistically activate the Bmp2 promoter. The work also suggests that, as in Drosophila, multiple transcription start sites are utilized in different mammalian tissues.

EXPERIMENTAL PROCEDURES
F9 Cell Culture and Differentiation-F9 embryonal carcinoma cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated calf serum and 2 mM glutamine. The cells were induced to differentiate into primitive endoderm by adding RA alone and into parietal endoderm by adding RA, 250 M dibutyryl cAMP, and 500 M theophylline (RACT).
Nuclear Run-on Assays-F9 cells were untreated or treated 72 h with 0.5 M RA, dibutyryl cAMP, and theophylline as described above. Nuclei were isolated and nuclear run-on assays were performed as de-2 D. Israel, personal communication, A, nuclear run-on assays. 32 P-Labeled probes generated from transcripts initiated in these cells at the time of RNA extraction were hybridized to identical nitrocellulose strips spotted with a vector control (pGEM3Z) or plasmids containing the cDNA encoding a constitutive ribosomal protein (36B4) or BMP2 (pBMP2-452). B, ribonuclease protection assays. Antisense probes extending from nucleotide Ϫ1,541 to Ϫ1,233 and nucleotide Ϫ2,287 to Ϫ1,663 were hybridized to F9 cell RNA or to yeast tRNA, and ribonuclease protection assays were performed.  (36). C, primer extension assays. An antisense oligonucleotide corresponding to nucleotides Ϫ2,065 to Ϫ2,046 was hybridized to F9 cell RNA or to yeast tRNA, and primer extension was performed. One band was visible in the RACT lane only. A sequence generated from this oligonucleotide and pBMP2-NX is shown to the left of the primer extension lanes. scribed previously (25,28). Plasmids 36B4 (29), pGEM3Z (Promega), and pBMP2-452 were used for hybridization.
Sequencing-Sequencing was performed manually (25), by the Molecular Biology Core Facility at the H. Lee Moffitt Cancer Center (Tampa, FL) or by the DNA Sequencing Laboratory at the Interdisciplinary Center for Biotech Research (Gainesville, FL), using primers from the vector and internal sequences. Analysis of the RA-responsive upstream sequence for putative transcription factor binding sites was performed using TFSEARCH 3 versus the TFMATRIX transcription factor binding site profile data base (30) and by visual inspection. Promoterscan II 4 was used to located putative promoter sequences (31).
Primer Extension-Primer extension was performed as described previously (25). The primer 5Ј-GTGGGAAGCGCAGCGGCGGC-3Ј, corresponding to the complement of the sequence extending from Ϫ2,064 to Ϫ2,045, was labeled and hybridized to 29.7 g of RNA and extended with avian myeloblastosis virus reverse transcriptase (Life Technologies, Inc.).
Ribonuclease Protection Assays-Ribonuclease protection assays were performed as described by Zinn et al. (32) with the following modifications. 32 P-Labeled RNA probes were made from pBMP2-H linearized with BglII and transcribed with T3 RNA polymerase and pBMP2-NX linearized with NotI and transcribed with T7 RNA polymerase. 10 6 cpm of each probe were hybridized to 10 g of RNA overnight at 45°C in 80% deionized formamide, 40 mM PIPES, pH 8.5, 400 mM NaCl, and 1 mM EDTA. After treatment with ribonuclease A, the product was electrophoresed on a sequencing gel.
F9 Cell Transfections and CAT Assays-All methods were essentially as described by Vasios et al. (21). F9 cells were cultured for 48 h without drugs or in the presence of CT, RA, or RACT, transfected by calcium phosphate precipitation, and then cultured an additional 24 or 48 h with drugs. All cells were cotransfected with the reporter gene and with p␤AclacZ (21), which contains the ␤-galactosidase coding region driven by the constitutive ␤-actin promoter. Cell extracts were normalized for transfection efficiency as determined by ␤-galactosidase expression. Equivalent amounts of extract were incubated at 37°C for 7 h with 250 mM Tris, pH 7.8, 5.3 mM acetyl coenzyme A, and 32.4 M 14 C-chloramphenicol (51.5 Ci/mol; NEN Life Science Products). After separation by thin layer chromatography (Whatman No. 4410221), chloramphenicol acetylation was quantified with a Molecular Dynamics Phospho-rImager or a Beckman 60001C liquid scintillation counter.

Bmp2 Transcription in F9 Cells Increases in Response to
RACT Treatment-RNA abundance may be regulated by alterations in transcription rate and in message stability. Previous experiments using the transcriptional inhibitors actinomycin D and 5,6-dichloro-1-␤-D-ribofuranosylbenzimidazole showed that Bmp2 mRNA stability does not change with RA treatment (35). This observation suggested, but did not prove, that RA increased Bmp2 transcription rates. To test this hypothesis, we isolated nuclei from untreated cells or cells treated with RACT for 72 h and performed nuclear run-on assays (Fig. 1A). Bmp2 gene transcription increased 3.7-fold in RACT-treated cells relative to untreated cells. In contrast, the transcription of 36B4, a constitutively expressed ribosomal protein, did not vary. This directly demonstrated transcriptional induction of the Bmp2 3 TFSEARCH is available on the World Wide Web at http://pdap1. trc.rwcp.or.jp/research/db/TFSEARCH.html. 4 Promoterscan II is available on the World Wide Web at http://biosci. cbs.umn.edu/software/promoterscan.htm.

FIG. 2. Bmp2 gene structure and CAT reporter constructs.
A schematic representation of the Bmp2 genomic sequence is shown. The sequence is numbered with respect to the translation start site (36). Filled boxes represent exons (36). D (distal) indicates transcription initiation from the RA-dependent promoter in F9 cells, whereas P (proximal) indicates the additional transcription initiation site utilized in osteoblasts (37). The bars below indicate the location and relative sizes of the sequences cloned into CAT reporter vectors pBLCAT2 and pBLCAT3. For pCAT5ЈNN6.8, pCAT5Ј-XX4.5, and pCAT5ЈNN6.3, fragments were inserted 5Ј of the TK promoter in pBLCAT2. For pCAT3ЈBB3.4 and pCAT-3ЈBN3.0, fragments were inserted 3Ј of the CAT coding region. For pCAT5Ј-NB6.3B, pCAT4.5X, and pCAT4.5X⌬Not, fragments were inserted 5Ј of the CAT coding region in the promoterless reporter vector pBLCAT3. gene by RA.
Location of the Transcription Start Site-Two Bmp2 promoters have been described in osteoblast cells (36,37). We used ribonuclease protection assays to determine if these transcription start sites or others were RA-inducible in F9 cells (Fig. 1B). An antisense RNA probe that extended from nucleotide Ϫ1,541 to Ϫ1,233 relative to the translation initiation site was generated from pBMP2-H digested with BglII. The entire Bmp2 probe was protected by RNA isolated from RACT-treated cells (Fig. 1B, thick arrow), indicating that Bmp2 transcription initiated upstream of nucleotide Ϫ1,541 in F9 cells. No protected fragments were observed in the reactions containing RNA from untreated cells, yeast tRNA (Fig. 1B), or a sense probe extending from Ϫ2,230 to Ϫ1,233 (data not shown). The open arrow indicates the predicted location of a fragment generated by a transcript originating at the proximal promoter (nucleotide Ϫ1,344) described by Feng et al. (36). The absence of a fragment at this location indicates that, in contrast to osteoblast cells, this promoter is not used in F9 cells. Using an antisense probe that extended from Ϫ2,287 to Ϫ1,663 generated from pBMP2-NX, we observed a fragment of approximately 493 nucleotides. This suggests that transcription starts at approximately Ϫ2,156, near the distal promoter used in osteoblasts.
To confirm the transcription start site, primer extension was performed utilizing reverse transcriptase and a primer complementary to base pairs Ϫ2,064 to Ϫ2,045 relative to the translation start site (Fig. 1C). An extended product was detected only in the reaction containing RNA from cells treated with 1 M RA and CT for 96 h and not in the reactions containing RNA No other RACT-dependent extended products were observed, suggesting utilization of one major transcriptional start site in F9 cells. The differences in mobility between RNA and DNA molecular weight markers explain the small discrepancy in start site position as determined by primer extension or RNase protection assays.
Regulation of Bmp2 Promoter Activity by RA in F9 Cells-CAT assays were used to detect RA-or RACT-dependent increases in CAT reporter activity driven by Bmp2 genomic DNA in F9 cells. The reporter constructs containing sequence flanking the Bmp2 gene are shown in Fig. 2. Since two transcriptional start sites have been described (37), all nucleotide positions are indicated relative to the translational start site (Fig.  2). Several large regions of the Bmp2 upstream flanking region were inserted upstream of the Herpes simplex virus TK minimal promoter in pBLCAT2. These fragments, extending from Ϫ8,583 to Ϫ2,320 (pCAT5ЈNN6.8), Ϫ3,367 to ϩ1,206 (pCATXX4.5), and Ϫ2,287 to ϩ3,360 (pCAT5ЈNN6.3), failed to drive CAT expression in F9 cells treated for 96 h with 1 M RA or 1 M RA and CT (Fig. 3A). A fragment that included the 3Ј-end of the transcribed region (9,316 -12,700; pCAT3ЈBB3.4) did not affect CAT activity. In contrast, a fragment distal to the 3Ј-end (12,700 -15,700; pCAT3ЈBN3.0) caused a 3.5-4-fold CAT activation relative to the pBLCAT2 vector alone (Fig. 3A). Since activation occurred in cells treated with CT, RA, or RACT, this sequence must contain a non-RA-dependent regulatory element. Considering the highly tissue-and stage-specific expression of Bmp2, many regulatory elements are likely to control Bmp2 expression.
Since developmental regulation of the dpp gene is mediated by the core promoter (38), we hypothesized that the RA responsiveness of the Bmp2 gene was similarly controlled. If so, then the TK minimal promoter in pBLCAT2 might have interfered with RA-induced transcription. Therefore, Bmp2 sequences were inserted into pBLCAT3, which lacks a minimal promoter. A 1,709-bp fragment, containing nucleotides Ϫ3,367 to Ϫ1,658 (pCAT4.5X), induced CAT activity 2.8-fold in cells treated for 96 h with 1 M RA or 1 M RA and CT relative to the activity observed in CT-treated cells (Fig. 3B). This fragment included 1,240 base pairs upstream of the transcription start site at Ϫ2,127. Finally, a fragment containing only 161 nucleotides upstream of the transcriptional start site (Ϫ2,288 to Ϫ1,537; pCAT5ЈNB6.3B) failed to induce CAT activity (Fig. 3B). These results are consistent with the presence of elements required for the RA response and promoter activity between nucleotides Ϫ3,367 and Ϫ2,288. As will be discussed below, we used a yeast reporter system to further delineate this RARE.
A Bmp2 RARE Drives RA-dependent ␤-Galactosidase Expression in Yeast-It is difficult to distinguish genes regulated directly by retinoid-activated receptors from those indirectly activated by other transcription factors induced by RA in mammalian cells. To avoid the complications associated with endogenous receptors and other transcription factors in F9 cells, we co-transformed yeast with mammalian receptor expression vectors and reporter genes driven by Bmp2 genomic sequences. Although yeast do not normally express retinoid receptors, yeast transformed with receptor genes synthesize functional receptors. These can stimulate the RA-dependent expression of reporter genes controlled by mammalian RAREs (33,39,40). We inserted a fragment containing base pairs Ϫ3,367 to Ϫ1,658 of the Bmp2 gene in front of the cyc1 promoter and the ␤-galactosidase coding region of the yeast vector, p⌬SS (27). The yeast strain BJ5409 was transformed with this plasmid and various combinations of RAR␤, RAR␥ or RXR␥ yeast expression vectors. Treatment of yeast expressing RAR␤ or RAR␥ and RXR␥ with 1 M all-trans-RA or 9-cis-RA induced ␤-galactosidase activity 1.7-and 2.3-fold, respectively, relative to untreated yeast (Fig. 4A). Yeast transfected with RAR␤ or RAR␥ alone and treated with 9-cis-RA also induced ␤-galactosidase activity 1.6-fold, indicating that the RAR homodimers could activate Bmp2 nearly as efficiently as the RAR/RXR heterodimer (Fig.  4B). In contrast, yeast expressing RXR alone or yeast lacking receptors failed to express ␤-galactosidase in response to RA treatment (Fig. 4B). These experiments indicate that the RA responsiveness of this Bmp2 sequence in yeast requires activation of RAR homodimers or RAR/RXR heterodimers.
In addition to the naturally occurring all-trans-RA, which activates only RARs, and 9-cis-RA, which activates both RARs and RXRs, several synthetic receptor-selective retinoids are available. TTNPB is often used to demonstrate RAR selectivity in mammalian cells because, unlike all-trans-RA, it cannot be converted to 9-cis-RA.
LG100268 is an RXR-selective retinoid (41). We treated RAR␤-and RXR␥-expressing yeast with 1 M TTNPB, LG100268, or 9-cis-RA or the combination of TTNPB and LG100268. Like all-trans-RA, TTNPB activated transcription slightly (1.5-fold) but less effectively than 9-cis-RA (2.1fold, Fig. 4C). Interestingly, the RXR agonist induced activity as effectively as the panagonist 9-cis-RA, which can activate both the RARs and the RXRs. Combined exposure to these retinoids stimulated activity by 3.6-fold (Fig. 4C). The synergistic activation of several RA-responsive genes by simultaneous ligand binding of each receptor subunit within a heterodimer has also been observed in mammalian cells (42,43). These results are the first to demonstrate that the developmentally crucial Bmp2 gene is activated directly by retinoid-bound receptors.
Having proven that this sequence could drive the yeast ␤-galactosidase reporter gene in a ligand-and receptor-dependent manner, we localized this element more precisely using a series of deletion constructs (Fig. 5). Deletions of 3Ј-flanking sequences up to position Ϫ2,316 and 5Ј-flanking sequences up to Ϫ2,373 do Yeast strain BJ5409 was transformed with expression vectors encoding RAR␤ and RAR␥. These yeasts were subsequently transformed with various portions of Bmp2 genomic DNA driving the ␤-galactosidase gene as indicated. Transcriptional induction by 1 M 9-cis-RA is presented by -fold induction as indicated in the histogram. "1-Fold" induction indicates no difference in ␤-galactosidase activity between untreated and RA-treated cells. n ϭ 3. Bars show S.E. not alter the induction by 9-cis-RA (bars 1-4). The reporter constructs containing only base pairs Ϫ3,195 to Ϫ2,369 or Ϫ3,367 to Ϫ3,191 were not induced by 9-cis-RA (bars 5-6). These results indicate that a 57-bp Bmp2 promoter sequence located between Ϫ2,373 and Ϫ2,316 bp contains a RARE that is necessary and sufficient to induce RA-mediated transcription.
Sequencing and Analysis of the Upstream Region of the Bmp2 Gene-We sequenced base pairs Ϫ3,367 to Ϫ1,658 of the Bmp2 gene to identify consensus sequences for other known regulatory proteins. Sequences consistent with a TATA-containing promoter sequence and a transcription start site at nucleotide Ϫ2,127 are depicted in Fig. 6. As shown in Fig. 1C, the primer extension assay confirmed the activity of this promoter in RACT-treated F9 cells. These features are consistent with a promoter at this site.
The sequence was also scanned for putative regulatory protein binding sites. A putative Sp1 site was identified between Ϫ2,308 and Ϫ2,298 (Fig. 6). Since others have demonstrated the importance of Sp1 sites for RA responsiveness (44 -46), we deleted a 32-bp fragment containing the site (Fig. 2). This deletion failed to alter the magnitude of RA inducibility in either yeast (data not shown) or F9 cells (Fig. 7A). However, in F9 cells, both the basal and the induced transcription activity of the CAT reporter gene declined by 30% (Fig. 7A). We also demonstrated that recombinant Sp1 protein bound this sequence (Fig. 7B). These observations suggest that Sp1 influences the transcription activity of the Bmp2 gene but does not play a role in RA responsiveness. DISCUSSION Approximately 200 genes have been shown to be RA-responsive in one cell or another. Some genes are regulated directly by RA-bound receptors, e.g. Hoxa1 (20), while others are secondarily regulated by other transcription factors modulated in RA-treated cells, e.g. Fgf4 (47). Since both retinoid deficiencies and overdoses can cause embryonic malformations via the aberrant expression of key proteins controlling differentiation, proliferation, apoptosis, and morphogenesis, it is important to understand which genes are directly regulated. We now present evidence that the gene encoding the essential growth and differentiation factor, BMP2, is a direct target of RA.
Bmp2 is transcriptionally induced by RA in F9 embryonal carcinoma cells (Fig. 1A). Several pieces of evidence suggest that F9 cells utilize a Bmp2 promoter initiating transcription at nucleotide Ϫ2,127 relative to the translation initiation site. First, ribonuclease protection assays indicate that the longest Bmp2 transcript initiates near this site (Fig. 1B). Second, this is the end of an RA-inducible primer extension product (Fig.  1C). Third, the predicted size of a transcript starting at Ϫ2,127 is consistent with the mRNA size of 3.8 kilobases (17,36). Finally, sequences resembling a TATA-containing promoter are near nucleotide Ϫ2,127 (31).
In mouse osteoblasts, distal and proximal transcription start sites were observed at nucleotides Ϫ2,127 and Ϫ1,344 (36, 37). We did not observe a ribonuclease protection fragment corresponding to the proximal start site in untreated or in RACTtreated F9 cells (Fig. 2B). Thus, the proximal promoter does not mediate the RA-induced transcription of Bmp2 in F9 cells. The existence of multiple Bmp2 transcription start sites in different cell types is not surprising, because the expression of the Bmp2 gene is highly dynamic. In addition, dpp, the gene encoding the Drosophila homolog of Bmp2, has three major and several minor transcripts produced from several promoters (15). Like the dpp transcript, the Bmp2 transcript has an unusually long 5Ј-untranslated region of 1,125 nucleotides that might contain as yet uncharacterized regulatory elements. Since Bmp2 and dpp are pivotal developmental genes, their spatial and temporal expression must be tightly regulated. We have demonstrated here that tissue-specific promoters are one mechanism involved in this tight regulation.
We have shown that this promoter and 1,709 base pairs of flanking region drive RA-dependent reporter gene expression in F9 cells. RA, which induces primitive endoderm differentiation, and RA and CT (dibutyryl cyclic AMP and theophylline), which induce parietal endoderm differentiation, caused equal activation of the CAT reporter gene (Fig. 3B). The endogenous Bmp2 RNA is undetectable in undifferentiated cells and is induced modestly in RA-treated cells. In contrast, although CT does not induce differentiation and has no effect on Bmp2 mRNA abundance, the combination of RA and CT induces the message abundance strikingly. Thus, sequence outside of Ϫ3,367 to Ϫ1,658 must contain the elements responsible for the synergistic activity of RA and cAMP.
Our demonstration that a 57-bp fragment of Bmp2 genomic DNA can drive the expression of a ␤-galactosidase reporter gene in yeast transformed with retinoid receptors strongly suggests that this gene is directly regulated by receptor binding. The RAR/RXR heterodimer in the presence of ligands that activate both subunits activated the Bmp2-driven reporter most efficiently (Fig. 4C). Similar synergy has been observed for many genes in mammalian cells, including Hoxa1 and the RAR␤ gene (42,43). A requirement for specific receptor combinations and specific ligand activities may mediate the tight regulation of developmentally crucial genes such as Bmp2.
Known retinoic acid-responsive elements are highly polymorphic and conform loosely to the form of two repeated halfsites separated by nonconserved "spacer" DNA: RG(G/T)-TCAN 5 RG(G/T)TCA (22). Although the most frequent forms are direct repeats separated by 5 base pairs (N), some RAREs consist of inverted repeats, much wider spacing, and diverse arrangements of half-sites. Since the 57-bp Bmp2 RARE lacks identity to any previously described RAREs, point mutational analyses will be necessary to identify the precise sequences bound by retinoid receptors. Considering that the numerous combinations of the six retinoid receptors and their various isozymes have distinct ligand-and DNA-binding specificities and that over 200 genes are known to be modulated in retinoidtreated cells (48), many more types of functional RAREs are likely to be found.