Retinoic Acid Repression of Bone Morphogenetic Protein 4 in Inner Ear Development

Abstract

Bone morphogenetic protein 4 (BMP4) and retinoic acid are important for normal development of the inner ear, but whether they are linked mechanistically is not known. BMP4 antagonists disrupt semicircular canal formation, as does exposure to retinoic acid. We demonstrate that retinoic acid directly down-regulates BMP4 transcription in a mouse inner ear-derived cell line, and we identify a novel promoter in the second intron of the BMP4 gene that is a target of this regulation both in the cell line and in the mouse embryonic inner ear in vivo. The importance of this down-regulation is demonstrated in chicken embryos by showing that the retinoic acid effect on semicircular canal development can be overcome by exogenous BMP4.

Organogenesis requires the concerted actions of hormones such as retinoic acid and growth factors such as the bone morphogenetic proteins (BMPs). The BMPs, which are members of the transforming growth factor β (TGF-β) superfamily, were first identified by their ability to induce ectopic bone formation (42). Subsequently, BMPs have been shown to play important roles in axis determination at stages as early as gastrulation and in the development of numerous organs (15).

Morphologically visible inner ear development begins with the otic placode, a thickening of the surface ectoderm on each side of rhombomere 5 of the hindbrain. In mammals and birds, the placode invaginates to form an otic pit and then closes to form the otocyst, from which all components of the inner ear develop, except for a small contribution from the neural crest (8). The three semicircular canals (SCCs) emerge as pouch-like protrusions from the otocyst (3). BMP4 is expressed early in the developing otocyst, at embryonic day 9 in the mouse (26) and at stage 11 (day 2) in the chicken (28, 43), and becomes restricted to cell foci in the anterior and posterior otic epithelium at the time of otic pit closure (43). BMP4 also is expressed in regions of the otocyst that have been correlated with the later development of sensory epithelia, including those located at the bases of the SCC ampullae in chickens (28, 43), mice (26), and Xenopus frogs (17). Noggin, a potent antagonist of BMP4, is expressed in the periotic mesenchyme adjacent to the otocyst (10), suggesting that it helps establish a BMP4 gradient important for otocyst development. Indeed, malformations of the SCCs occur with the strategic placement of beads carrying noggin-producing cells around the otocyst, and the simultaneous addition of beads carrying BMP4-expressing cells rescues SCC development (5, 10). These data indicate that BMP4 plays an important role in the patterning of the SCCs.

All-trans retinoic acid (at-RA), the major biologically active metabolite of vitamin A, is necessary for normal embryonic development. In vertebrates, excess vitamin A during critical periods of development leads to widespread malformations of the face, limbs, heart, central nervous system, and skeleton (25). In the early 1980s, a correlation was made between pregnant women's use of isotretinoin (13cis-RA) to treat acne and abnormalities in their infants' ears, hearts, brains, and thymuses (19). Pharmacokinetic studies showed that 13cis-RA is rapidly isomerized to at-RA (18). Pregnant mice and hamsters exposed to excess at-RA or 13cis-RA exhibited similar widespread malformations. Two types of inner ear defects were observed: (i) failure to develop beyond the otocyst stage and (ii) a reduction in the number of SCCs, in conjunction with a smaller and flattened cochlea (4). These abnormalities closely resemble the human ear anomalies known as Michel aplasia and the Mondini-Alexander defect (4). Chicken otocysts implanted with at-RA-saturated beads exhibited several types of malformations, ranging from the absence of SCCs to a phenotype exhibiting only a rudimentary inner ear mass (6). Thus, the malformations due to excess at-RA resemble the malformations observed when BMP4 action is antagonized by noggin, suggesting that at-RA and BMP4 function through the same pathway to regulate SCC development. Furthermore, when chicken otocysts were implanted with at-RA-saturated beads, BMP4 expression decreased in the developing anterior crista (6). However, it is not known whether this was a direct effect of at-RA or whether it was causally related to the subsequent SCC abnormalities.

The biological actions of at-RA are mediated by retinoic acid receptors (RARs), which are members of the nuclear receptor superfamily and closely related to thyroid hormone receptors. Ligand-activated RARs bind to specific retinoic acid response elements (RAREs) typically located in the promoter regions of target genes (11). Three genes encode different RAR isotypes (α, β, and γ). The RARs are expressed in the mouse developing inner ear (30), and mice with combined null mutations of RARα and RARγ have abnormal otocyst development (22, 39), further supporting an important role for at-RA in the development of this organ.

These studies indicate that both BMP4 and at-RA play important roles in the morphogenesis of the SCCs and suggest that they function through the same pathway or intersecting pathways. In support of this hypothesis, we demonstrate that at-RA down-regulates BMP4 transcription in otocyst cells and that this is a direct effect, mediated by RARs, on a novel promoter within the second intron of the BMP4 gene. This repression stands in contrast to the previously described induction of BMP4 by at-RA in bone (13). The in vivo significance of this negative regulation is demonstrated by showing in the chicken embryonic otocyst that the inhibitory effect of at-RA on SCC formation is overcome by exogenous BMP4. Thus, these data indicate that repression of BMP4 expression is the downstream mediator by which at-RA regulates SCC formation and suggest the use of a novel BMP4 promoter to restrict this regulation to specific developing organs.

MATERIALS AND METHODS

Cell culture.

2B1 cells (1) were cultured at 32°C in Ham's F-12 nutrient mixture with 15% fetal bovine serum and 10⁵ U of gamma interferon per liter. Cells treated with hormone or carrier were incubated in similar media with 10% charcoal-stripped fetal bovine serum for at least 24 h prior to hormone addition.

Western analysis.

Nuclear extracts were obtained from 2B1 cells by using N-PER (Pierce). Ten micrograms was subjected to sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis and Western blotting. The membranes were probed with anti-RARα (C-20; Santa Cruz) or anti-RARγ (C-19; Santa Cruz) polyclonal antibodies at 1:1,000, followed by horseradish peroxidase-conjugated anti-rabbit immunoglobulin G (sc-2301; Santa Cruz) at 1:20,000. The bands were visualized with SuperSignal West Dura Extended Duration Substrate (Pierce).

RAR cDNAs.

2B1 cell RNA was subjected to 5′ rapid amplification of cDNA ends (5′ RACE; Gene Racer Kit; Invitrogen) to determine which RARα isoforms are expressed. The reverse transcriptase (RT) primer used, 5′-ACCGACTCCTTGGAC, was specific for the hinge domain of mouse RARα, and a nested primer specific for the DNA binding domain of RARα, 5′-CGGAAGAAGCCCTTACAGCCCTCA, was used for the PCR. The PCR products were cloned into pcTOPO4 (Invitrogen). Sixteen colonies were sequenced, and all were RARα1 or RARα2. The full-length coding sequences for these RARs were then obtained by RT-PCR and ligated into the HindIII and XbaI sites of pcDNA3.1+ (Invitrogen). To determine which RARγ isoforms are expressed, 2B1 cell RNA was subjected to RT-PCR. The reverse primer was specific for RARγ bp 1724 to 1747 (GenBank accession number X15848), and forward primers were used that are specific for each of RARγ isoforms 1 to 6 (16). With this approach, RARγ isoforms 1, 2, and 4 gave products. The full-length cDNAs for these isoforms were obtained by RT-PCR and ligated into the NheI and HindIII sites of pcDNA3.1+.

RNase protection analysis (RPA).

With the Ambion MAXIscript Kit, antisense [α-³²P]UTP-labeled cRNA probes were transcribed from templates containing either mouse cyclophilin (Ambion catalog no. 7675) or a specific BMP4 sequence (GenBank accession no. L47480) as follows: (i) exon 4 (nucleotides 8417 to 8700), (ii) exon 1A plus intron 1 (nucleotides 2437 to 2733), (iii) exon 1B long plus intron 2 (nucleotides 4599 to 4865), (iv) the short 5′ end of exon 1B spliced to full-length exon 2, followed by the 5′ end of exon 3 (nucleotides 4519 to 4539, followed by nucleotides 5589 to 5711 and nucleotides 6747 to 6776), or (v) the 3′ end of intron 2, followed by exon 2 and the 5′ end of exon 3 (nucleotides 5473 to 5588, followed by nucleotides 5589 to 5711 and nucleotides 6747 to 6785). 2B1 cells were cultured with the RAR-specific ligand 4-[(E)-2-(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthalenyl)-1-propenyl]benzoic acid (TTNPB; Sigma) at 10 μM or the vehicle (ethanol) for the times indicated. RNA was isolated with Trizol (Gibco). Hybridizations with the radiolabeled probes were accomplished with the Ambion RPA III kit with 15 to 20 μg of RNA. After digestion with RNase, the protected fragments were separated on 5% acrylamide-8 M urea denaturing gels. The bands were analyzed by phosphorimager.

Nuclear run-on analysis.

The protocols used for nuclear isolation by Dounce homogenization and preparation of the nitrocellulose membrane with denatured target DNAs were previously described (12). Additional details are as follows. 2B1 cells were cultured with or without 10 μM TTNPB for 6 h prior to isolation of nuclei. Target DNAs were linearized, and denatured plasmids containing BMP4 exon 4, the empty vector (pDP18), or mouse cyclophilin were loaded at 5 μg per slot in triplicate slots. The membranes were prehybridized overnight at 65°C in prehybridization-hybridization buffer [10 mM N-tris(hydroxymethyl)methyl-2-aminoethanesulfonic acid (TES; pH 7.5), 2% SDS, 10 mM EDTA, 0.25 mg of Escherichia coli RNA per ml, 0.3 M NaCl, 1× Denhardt's solution]. Radiolabeled RNA probes were synthesized from 5 × 10⁷ nuclei incubated with 0.3 mCi of [α-³²P]UTP (3,000 Ci/mmol)-1.25 mM each ATP, CTP, and GTP in 25 mM Tris (pH 8)-12.5 mM MgCl₂-750 mM KCl at 30°C for 30 min. The labeled RNAs were isolated with a Qiagen RNeasy Midi kit. The membranes were incubated with the probes (10⁷ cpm/ml) for 48 h at 65°C. The hybridized membranes were washed at 65°C three times in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.1% SDS and two times in 0.1× SSC-0.1% SDS and then analyzed by phosphorimager.

Actinomycin D analysis.

In preliminary experiments, addition of 40 μg of actinomycin D (from an 8-mg/ml stock in dimethyl sulfoxide [DMSO]) per ml to culture media inhibited the incorporation of [³H]uridine into 2B1 cell RNA by >99%. Therefore, 2B1 cells were treated with 10 μM TTNPB or the vehicle for 3 h, at which time actinomycin D (or DMSO) was added to 40 μg/ml; this was followed by further incubation for 0, 2, 4, or 7 h. RNA was isolated and subjected to RPA with the cyclophilin and BMP4 exon 4 probes as described previously.

Cycloheximide analysis.

In preliminary experiments, addition of 35 μM cycloheximide to culture media inhibited the incorporation of [³H]leucine into 2B1 cell trichloroacetic acid-precipitable protein by >95%. Therefore, 2B1 cells were cultured with or without 35 μM cycloheximide for 30 min, at which time 10 μM TTNPB or vehicle (ethanol) was added and the incubations were continued for another 3 or 6 h. RNA was isolated with Trizol and subjected to RPA with cyclophilin and BMP4 exon 4 probes as described previously.

Transient transfections.

2B1 cells were transfected in 24-well clusters with Lipofectamine Plus (Invitrogen) in accordance with the manufacturer's instructions. Each well received 10 ng of an RAR expression vector or the empty vector (pcDNA3.1+), 200 ng of a BMP4 promoter firefly luciferase vector, and 1 ng of the pRLSV40 (Promega) Renilla luciferase (Rluc) vector as an internal control. The BMP4 1B promoter used in these studies includes nucleotides 2686 to 4539 of GenBank accession no. L47480, which are bp −1833 to +21 relative to the 1B transcription start site. The intron 2 promoter used in these studies includes nucleotides 4631 to 5616 of GenBank accession no. L47480, which are bp −888 to +97 relative to the major transcription start site. Both promoter sequences were ligated into the KpnI and BglII sites of pGL3-Basic (Promega). The transfected cells were treated with 10 μM TTNPB or the vehicle (ethanol) on the day after the transfection; this was followed by continued incubation for 48 h. The harvested cells were analyzed with the Dual Luciferase Plus Kit (Promega) in accordance with the manufacturer's instructions.

RT-PCR.

RT-PCR analyses were accomplished with a one-step RT-PCR kit (Qiagen) in accordance with the manufacturer's instructions. A touchdown RT-PCR program was used as follows: step 1, 50°C for 30 min; step 2, 95°C for 15 min; step 3, 94°C for 1 min; step 4, 70°C for 1 min; step 5, −1°C per cycle; step 6, 72°C for 1 min (14 repetitions of steps 3 to 6); step 7, 94°C for 1 min; step 8, 56°C for 1 min; step 9, 72°C for 1 min (24 repetitions of steps 7 to 9); step 10, 72°C for 10 min. The following forward PCR primers were used: (i) exon 1A, 5′-GAAGGCAAGAGCGCGAGG; (ii) exon 1B, 5′-CAGGCCGAAAGCTGTTC; (iii) intron 2, 5′-GAGCCTGTCTGCTCCAGAGTCTC; (iv) exons 3 and 4, 5′-CGGATTACATGAGGGATCTTTACCG. The following reverse PCR primers were used: 5′-ACGACCATCAGCATTCGGTTAC for the exon 1A, 1B, and intron 2 RT-PCR products and 5′-ATGGCGACGGCAGTTCTTATTC for the exon 3 and 4 RT-PCR product.

Isolation and real-time RT-PCR analysis of embryonic otocyst RNA.

ICR embryonic day 10.5 (E10.5) timed pregnant mice from Taconic received at-RA (25 or 100 mg/kg) or the vehicle by gavage. The at-RA was dissolved in DMSO at 150 mg/ml and then diluted with sesame oil to achieve a volume of administration of 0.2 ml per mouse. Six hours after gavage, the embryos were removed and the otocysts were microdissected and placed in Trizol for isolation of RNA. The RT-PCR protocol was as described above, with the addition of SYBR Green I diluted 100,000-fold from the stock (Molecular Probes catalog no. S-7563) for detection of the PCR products. The primers used for the β-actin RT-PCR were as follows: forward, 5′-TGTGATGGTGGGAATGGGTCAG; reverse, 5′-TTTGATGTCACGCACGATTTCC. The samples were run in and analyzed with a DNA Engine Opticon System (MJR). The protocol was approved by the University Committee on the Use and Care of Animals.

Bead and pellet preparation and implantation into chicken ears.

Fertilized White Leghorn chicken eggs were obtained from Dave Bilbie Aviaries, Ann Arbor, Mich. The protocols used for bead implantation and inner ear analysis by paint fill were previously described (10). All implants were performed at stages 16 to 17 and harvested at stages 34 to 35. A full-length mouse BMP4 cDNA was inserted with a hemagglutinin epitope tag into pcDNA3.1+. The plasmid was transfected into CHO cells with a Transfast Transfection Kit (Promega). Stable clones were selected with G418 and analyzed by Western blot assay for BMP4-hemagglutinin protein expression (data not shown). Control CHO cells or BMP4 CHO cells were placed on Cibacron beads (Sigma) and implanted into the center of stage 16 to 17 otocysts as previously described (10). at-RA beads were prepared as previously described (6). Briefly, a 100-μl volume of AG1X-8 beads resuspended in DMSO was pelleted and then resuspended in 0.125 mg of at-RA per ml of DMSO (or in plain DMSO as a control). After 20 min of incubation, the beads were rinsed twice with phosphate-buffered saline and then implanted into the centers of stage 16 to 17 otocysts.

RESULTS

RARα and RARγ are expressed in an otocyst-derived cell line.

To study the regulation of BMP4 by at-RA in the developing inner ear, we began by employing the 2B1 cell line, which is derived from the otocyst of an E9.5 Immortomouse and expresses BMP4 (1). This otocyst cell line was derived from the earliest stage of otic development that can easily be isolated from the mouse. The 2B1 cell line is a multipotent cell line resembling the precursor common to both hair cells and supporting cells. Since mice with combined null mutations of RARα and RARγ show defects in their otocysts (22, 39), we tested whether 2B1 cells express these RAR isotypes. Western analysis revealed that both are present (Fig. 1A). RARs contain a centrally located DNA binding domain and a carboxyl-terminal ligand binding domain (Fig. 1B), both of which are highly conserved between isotypes. The RARs primarily differ in their amino-terminal A and B domains, which are unique for each isotype (11). In addition, each isotype is subdivided into a number of isoforms that differ only in their A domains. We used a combination of 5′ RACE and RT-PCR to determine that 2B1 cells express RARα isoforms 1 and 2 and RARγ isoforms 1, 2, and 4 (data not shown). Each of these isoforms was cloned from 2B1 cells. Their sequences matched those reported by Zelent et al. (44) and GenBank accession numbers X56572 (RARα1), X56565 (RARα2), X15848 (RARγ1), M32069 (RARγ2), and M32071 (RARγ4). Because of the lack of isoform-specific antibodies, we were not able to distinguish the RARα or RARγ isoforms by Western blotting. The expression of RARα and RARγ in 2B1 cells suggests that at-RA could potentially regulate BMP4 expression in these cells.

FIG. 1.

RAR expression. (A) Analysis of RARα and RARγ expression in 2B1 cells by Western blot analysis. (B) Schematic diagram of a prototypical RAR. Three RAR genes (α, β, and γ) encode receptors that are unrelated in their amino-terminal A and B domains but highly similar in their DNA and ligand binding domains (DBD and LBD, respectively).

Ligand-occupied RARs decrease BMP4 expression in otocyst cells.

To investigate if retinoic acid has an effect on BMP4 expression, RNA was isolated from 2B1 cells that had been incubated with the vehicle or TTNPB for various amounts of time. RPA revealed that TTNPB decreased the amount of BMP4 mRNA (Fig. 2A). The effect was maximal at 6 h (expression decreased to 29% of that in control cells, normalized to cyclophilin expression) and persisted for at least 48 h, at which time normalized expression was at 50% of that in control cells. Similar data were obtained with at-RA (data not shown). However, because at-RA can be metabolized to 9cis-RA, which activates retinoid X receptors (14, 21), TTNPB was used as the RAR-specific ligand in subsequent studies.

FIG. 2.

TTNPB down-regulates BMP4 transcription in 2B1 cells. (A) RPA of BMP4 expression. 2B1 cells were cultured in the presence of the RAR-specific ligand TTNPB or the vehicle for up to 48 h. RNA was isolated, and expression of BMP4 RNA was quantified by RPA and phosphorimager analysis (normalized to cyclophilin as a control). Incubation with yeast RNA yielded no protected bands. The full-length (unprotected) probes are in the far right lane. Similar results were obtained in a second experiment. (B) Nuclear run-on assay of BMP4 transcription. 2B1 cells were cultured in the presence of the RAR-specific ligand TTNPB or the vehicle for 6 h. Nuclei were isolated, and the rate of BMP4 transcription was quantified in triplicate by nuclear run-on assay and phosphorimager analysis. Similar results were obtained in a second experiment. (C) TTNPB does not decrease the half-life of BMP4 RNA. 2B1 cells were cultured in the presence of the RAR-specific ligand TTNPB or the vehicle for 3 h, at which point actinomycin D was added. Cells were harvested at various time points (time zero represents the point of actinomycin D addition). The amount of BMP4 RNA was assessed by RPA with cyclophilin as a control. Incubation with yeast RNA yielded no protected bands. The full-length (unprotected) probes are in the far right lanes. (D) Semilog plot of the data in panel C (BMP4 normalized to cyclophilin) quantified by phosphorimager analysis. nt, nucleotides.

The decrease in BMP4 expression could reflect either a decrease in production or an increase in degradation. Since RARs are transcription factors, a nuclear run-on analysis was employed to test whether TTNPB down-regulates the rate of BMP4 transcription. The data indicate that TTNPB reduced the rate of BMP4 transcription to 20% of that seen in control cells (Fig. 2B).

We also examined whether TTNPB decreases the half-life of BMP4 RNA. 2B1 cells were treated with TTNPB or the vehicle for 3 h, at which point actinomycin D was added to block RNA synthesis. The cells were harvested at subsequent time points, and the amount of BMP4 RNA was quantified by RPA (Fig. 2C). The results demonstrate that the half-life of BMP4 RNA was 2.4 h in the presence of TTNPB and 2.1 h in the presence of the vehicle (Fig. 2D). Thus, TTNPB does not increase the rate of degradation of BMP4 RNA. We concluded that the TTNPB-induced decrease in BMP4 RNA is accounted for by a decrease in BMP4 transcription.

The effect of TTNPB on BMP4 transcription could be direct or could be an indirect consequence of TTNPB regulating another transcription factor that has the BMP4 gene as a target. For example, TTNPB may induce the expression of a transcriptional repressor. This could explain how TTNPB, which generally induces the transcription of target genes, could actually repress BMP4 transcription. If the effect of TTNPB on BMP4 transcription is indirect, then it should be blocked by inhibition of protein synthesis. Therefore, 2B1 cells were treated with cycloheximide (or the vehicle) for 30 min, at which point TTNPB (or the vehicle) was added. The cells were harvested 3 and 6 h later, and BMP4 RNA was quantified by RPA. At both time points, the ability of TTNPB to decrease BMP4 RNA was unaffected by inhibition of protein synthesis (Fig. 3). These data indicate that TTNPB inhibits BMP4 transcription directly; i.e., TTNPB does not regulate an intermediary gene whose product targets BMP4.

FIG. 3.

The TTNPB-mediated decrease in BMP4 expression does not require new protein synthesis. (A) Cycloheximide or the vehicle was added to cultures of 2B1 cells. Thirty minutes later, TTNPB or the vehicle was added and the cells were cultured for an additional 3 or 6 h. The amount of BMP4 RNA was assessed by RPA with cyclophilin as a control. (B) Graph of the data in panel A (BMP4 normalized to cyclophilin) quantified by phosphorimager analysis.

Promoter 1B and a novel intronic promoter drive BMP4 transcription in mouse otocyst cells.

To study the TTNPB effect in more detail, it was necessary to identify the BMP4 promoter used in 2B1 cells. There are two known BMP4 promoters, 1A and 1B, with 1A being predominant in the cell types evaluated to date (7) (Fig. 4A). The two promoters transcribe different first exons (exons 1A and 1B), but these are spliced to the same exon, exon 2, which encodes the translation start site. Exon 1B is either 279 or 21 bp, depending on its splice donor site. We used 5′ RACE to determine which promoter drives BMP4 transcription in 2B1 cells. As indicated in Fig. 4A, the BMP4 gene-specific primers were directed toward sequences in exon 3. Eight clones were sequenced, all of which contained a perfect splice junction between exons 2 and 3. Unexpectedly, in seven of these clones, the sequence extended 5′ from exon 2 into intron 2. These clones contained the most distal 70 to 79 bp of intron 2, with their 5′ ends beginning at bp 5510, 5513 (two clones), 5514, or 5519 (three clones); the start of exon 2 is at bp 5589 (numbering is from GenBank accession no. L47480). In addition to these clones, one clone began with exon 1B, which was perfectly spliced to exon 2. No exon 1A-containing clones were identified. Similar results were obtained in a second independent 5′ RACE assay. To confirm that the 5′ RACE assay was capable of detecting exon 1A-containing transcripts, whole-mouse embryo RNA was subjected to the same analysis. The products contained exon 1A spliced to exon 2, as expected. Overall, these results suggest that otocyst cells largely utilize a novel BMP4 promoter located within intron 2 and that they also utilize the 1B promoter but not the 1A promoter that predominates in other cell types.

FIG. 4.

BMP4 transcription is driven by the 1B promoter and a novel promoter in intron 2. (A) Schematic diagram of the mouse BMP4 gene. Exon 1B is shown as a box with a dashed line near the 5′ end because, in 2B1 cells, this exon can be 21 or 279 bp. The diagram also indicates the locations of nested 5′ RACE primers complementary to exon 3 (primers 3.1R and 3.2R), as well as PCR primers used to detect transcripts driven by the 1A or 1B promoter (1AF plus 3.2R and 1BF plus 3.2R, respectively), a novel promoter located within intron 2 (i2F plus 3.2R), or all transcripts containing common coding exons 3 and 4 (3F plus 4R). (B) BMP4 promoter usage in 2B1 cells and whole-mouse embryo RNA analyzed by RT-PCR. RNA was isolated from 2B1 cells or an E9.5 whole-mouse embryo (WME). RT-PCR was performed with the intron-spanning primer pairs shown in panel A. No products were formed when the RT step was omitted (data not shown). (C) BMP4 promoter usage in 2B1 cells analyzed by RPA. 2B1 cells were cultured in the presence of TTNPB or the vehicle for 6 h. RNA was isolated, and expression of transcripts driven by the 1A, 1B, or intron 2 promoter was analyzed. Probes also were used to distinguish between the long and short versions of exon 1B-containing transcripts, and a probe for common coding exon 4 also was used. Cyclophilin served as a control. RNA from vehicle-treated cells was analyzed in lanes 1 to 5, and RNA from TTNPB-treated cells was analyzed in lanes 6 to 10. The full-length (unprotected) probes are in lanes 11 to 16.

To confirm the 5′ RACE data, RNAs from 2B1 cells and day 9.5 whole-mouse embryos were analyzed by RT-PCR with intron-spanning primers specific for transcripts originating from the 1A, 1B, or putative intron 2 promoter, as well as the common coding region (exons 3 and 4) that would be contained in transcripts driven by all BMP4 promoters. The whole-mouse embryo and 2B1 RNA samples produced similar amounts of RT-PCR product for the BMP4 common coding region, showing that both samples strongly express BMP4 (Fig. 4B, lanes 10 and 5). The exon 1A product was strongly expressed in a whole-mouse embryo sample (lane 7) but was barely detected in 2B1 cells (lane 2), supporting the 5′ RACE data that failed to identify 1A promoter transcripts in 2B1 cells. Exon 1B-containing transcripts were barely detectable in the whole-mouse embryo sample (lane 8), whereas 2B1 cells more clearly expressed primarily the shorter splice product (lane 3). The intron 2 promoter transcript was expressed strongly in the otocyst cell line (lane 4). Importantly, this product also was weakly present in the whole-mouse embryo sample (lane 9), indicating that transcription from the intron 2 promoter is not unique to 2B1 cells. These results support the 5′ RACE data, indicating that otocyst cells primarily drive BMP4 transcription from a novel intron 2 promoter and the 1B promoter instead of the 1A promoter.

To further substantiate this conclusion, an RPA was performed with radiolabeled probes that are specific for exon 1A, long exon 1B, short exon 1B, or intron 2-containing transcripts that are properly spliced to exons 2 and 3 (Fig. 4C). Prior to RNA harvesting, 2B1 cells were treated with TTNPB or the vehicle for 6 h to analyze which promoter(s) is down-regulated by at-RA. Mature RNAs containing exon 1A or the longer version of exon 1B were not detectable (lanes 1 and 2, vehicle; lanes 6 and 7, TTNPB). Importantly, this analysis confirmed the existence of mature RNA transcripts containing the 3′ end of intron 2 (lanes 4 and 9), as well as a mature transcript containing the short version of exon 1B (lanes 3 and 8). The presence of multiple protected fragments in the intron 2 lane is likely explained by the absence of a TATA box in the intron 2 promoter, resulting in multiple transcription start sites. These data are consistent with the variation in the 5′ ends of the intron 2 products obtained by 5′ RACE. Overall, the data confirm that BMP4 transcription in 2B1 otocyst-derived cells is driven by a novel promoter contained within intron 2, as well as by the 1B promoter, rather than the conventionally predominant 1A promoter. Furthermore, the RPA data also demonstrate that both the 1B and intron 2 transcripts are down-regulated by TTNPB (lane 3 versus lane 8 and lane 4 versus lane 9).

TTNPB down-regulates the 1B and intron 2 promoters through RAR-mediated action.

A BMP4 promoter luciferase assay was established to further investigate the involvement of RARs in the transcriptional regulation of BMP4 in otocyst cells. 2B1 cells were transiently cotransfected with the individual RARα or RARγ clones described earlier (or with the empty vector), a luciferase reporter driven by either 0.9 kb of the intron 2 promoter or 1.8 kb of the 1B promoter, and an Rluc plasmid as an internal control. In the absence of exogenous RAR, TTNPB caused a very modest but reproducible ∼25% decrease in luciferase activity from both promoters (Fig. 5, top and middle panels). This effect was enhanced by all three RARγ isoforms, achieving 2.5- to 2.8-fold repression with RARγ1 and RARγ4. The RARα isoforms also repressed luciferase expression, but the magnitude of repression was less than that obtained with RARγ1 and RARγ4. For all of the RARs, the fold repression was due entirely to a TTNPB-mediated decrease in luciferase expression; the unliganded RARs did not increase luciferase expression relative to the empty vector (data not shown).

FIG. 5.

TTNPB down-regulates the intron 2 and 1B promoters in transfected 2B1 cells. 2B1 cells were transfected with reporter plasmids in which luciferase is driven by 0.9 kb of the intron 2 promoter (top panel) or 1.8 kb of the 1B promoter (middle panel), along with an Rluc plasmid (as an internal control). Cells also received an RAR expression vector (or the empty vector [v]), as indicated. The cells were cultured with or without TTNPB for 2 days and then harvested for luciferase assays. Rluc was unaffected by TTNPB. Data are expressed as fold repression, which is defined as luciferase/Rluc for cells cultured without TTNPB divided by luciferase/Rluc for cells cultured with TTNPB. Results are the mean ± the standard error of the mean (n = 3). In the bottom panel, the transfection was as described for the upper two panels except that the reporter plasmid was 2xPal, in which the luciferase reporter is driven by a positive RARE. Data are expressed as fold induction, which is defined as luciferase/Rluc for cells cultured with TTNPB divided by luciferase/Rluc for cells cultured without TTNPB. Results are the mean ± the standard error of the mean (n = 3).

The effects of the RARα and RARγ isoforms also were tested on a positive RARE luciferase vector, 2xPal (Fig. 5, bottom panel). Each of these RARs supported TTNPB induction of luciferase from this vector, demonstrating that ligand-occupied RARs can induce or repress different promoters within the same 2B1 cells. In addition, the receptor potency for promoter induction was opposite that for repression: the RARα isoforms induced expression from 2xPal more strongly than the RARγ isoforms. Thus, RARα and RARγ appear to differ fundamentally in their abilities to repress or activate different target genes.

RA down-regulates BMP4 expression in developing mouse otocysts in vivo.

To determine whether the regulation of BMP4 expression in 2B1 cells mimics the regulation of BMP4 in the developing inner ear, pregnant mice at 10.5 days postcoitus were treated with at-RA or the vehicle by gavage. Six hours later, the embryos were removed, the otocysts were isolated under a dissecting microscope, and RNA was prepared. A real-time RT-PCR was used to measure BMP4 and β-actin expression in the otocyst RNAs. As demonstrated in Fig. 6A, actin was expressed at the same level in the vehicle and at-RA-exposed samples. However, RT-PCR for the BMP4 intron 2 promoter's transcript (the identity of which was verified by sequencing) revealed decreased expression following exposure to at-RA. These samples were analyzed at two doses of input RNA (1.25 and 5 ng) to demonstrate the quantitative nature of the assay. The difference between the at-RA- and vehicle-treated mice in the number of PCR cycles needed to reach the inflection point of the fluorescence curve is consistent with an approximately 2.2-fold decrease in BMP4 expression following treatment with at-RA (Fig. 6B). We were not able to perform similar studies of the 1B promoter transcript, in part because of limiting amounts of RNA and in part because we had difficulty developing an RT-PCR assay for this transcript that had the appropriate dose-response relationship with increasing amounts of input RNA.

FIG. 6.

Retinoic acid (RA) down-regulates BMP4 transcripts driven by the intron 2 promoter in E10.5 mouse embryo otocysts. Day 10.5 postcoitus pregnant mice received at-RA (100 mg/kg) or the vehicle by gavage. Six hours later, the embryos were removed and the otocysts were microdissected. (A) Otocyst RNA was isolated, and a real-time RT-PCR was used to study BMP4 expression driven by the intron 2 promoter. β-Actin served as a control. The assay was run in triplicate with the higher dose (5 ng) of input RNA and in quadruplicate with the lower dose (1.25 ng). The graph shows the mean fluorescence ± the standard error of the mean at each cycle number. (B) Quantification of the data in panel A expressed as BMP4 normalized to β-actin per nanogram of input RNA, with the mean expression level from vehicle-treated mice defined as 1.0. Similar results were obtained in a second experiment with 25 mg of at-RA per kg.

Retinoic acid effects on developing inner ears can be rescued by BMP4.

If down-regulation of BMP4 expression is indeed the mechanism by which at-RA affects SCC development, then forced expression of BMP4 should block the at-RA effect. This was tested in chicken embryos because of the ease with which they can be manipulated. An at-RA-saturated bead was placed in one otocyst of a stage 16 to 17 chicken embryo along with a bead containing BMP4-secreting CHO cells or control CHO cells. Inner ear morphology was analyzed at stages 34 to 35 (6) and categorized as type 0, 1, 2, or 3 to signify malformations of 0, 1, 2, or all three SCCs (Fig. 7A). All otocysts exposed to at-RA plus control CHO cell beads developed abnormally, and most had all three SCCs affected (type 3) (Fig. 7B, black bars). Beads saturated with the vehicle instead of at-RA never caused developmental abnormalities (n = 9; data not shown), nor did CHO cell beads in the absence of at-RA (n = 13; data not shown). These data are in agreement with those of Choo et al. (6) and are remarkably similar to the defects resulting from inhibition of BMP4 by implantation of beads bearing cells that make noggin, a BMP4 antagonist (5, 10).

FIG. 7.

Retinoic acid (RA) inhibits the development of SCCs in chicken embryo otocysts, and exogenous BMP4 overcomes this effect. (A) A small window was cut into White Leghorn chicken egg shells, and two beads (one bead soaked in at-RA or the vehicle and one bead adsorbed with either BMP4-secreting CHO cells or control CHO cells) were placed in the center of the otic vesicle of stage 16 to 17 embryos. Embryos were harvested at stages 34 to 35. Shown is a paint fill analysis of chicken inner ears illustrating representative abnormal phenotypes following exposure to at-RA. Inner ears were categorized by phenotype as follows: A, normal inner ear representing a type 0 phenotype (D, dorsal; A, anterior; L, lateral); B, type 1 inner ear with a single canal missing (ssc); C, type 2 inner ear with two missing SCCs (lsc and ssc); D, type 3 inner ear with no SCCs. Bar, 100 μm. Abbreviations: cd, cochlea; cc, common crus; es, endolymphatic sac; la, lateral ampulla; lsc, lateral SCC; pa, posterior ampulla; psc, posterior SCC; sa, superior ampulla; ssc, superior SCC. (B) Graph summarizing phenotypic abnormalities in ears treated with at-RA beads plus control CHO cell beads or at-RA beads plus BMP4 CHO cell beads.

Importantly, the presence of beads adsorbed with BMP4-secreting CHO cells greatly reduced the at-RA effect. Although all of the inner ears exposed to at-RA plus control CHO beads developed abnormalities, with 90% being type 2 or 3 (Fig. 7B, black bars), the presence of BMP4 beads plus at-RA resulted in only 69% of the ears developing abnormalities and only 25% being type 2 or 3 (Fig. 7B, hatched bars; P = 0.0007 for at-RA plus CHO versus at-RA plus BMP4 [Mann-Whitney rank sum test]). No abnormalities were caused by implantation of BMP4 beads in the absence of at-RA (n = 16; data not shown). The results are consistent with the at-RA effect on SCC development being mediated by inhibition of BMP4, and they suggest that at-RA may be a key regulator of BMP4 expression in the developing inner ear.

DISCUSSION

BMP4 is a signaling molecule related to TGF-β that binds to cell surface receptors leading to SMAD phosphorylation and activation (38). BMP4 is expressed early in the developing otocyst and is known to be important in the patterning of the SCCs. Studies of SCC development in chicken embryos reveal malformed canals when the BMP4 antagonist noggin, chordin, or DAN is introduced into the developing otocyst (5, 9, 10). Similar SCC malformations are observed in chicken otocysts that have been exposed to exogenous at-RA (6). These data suggest that BMP4 induces and at-RA inhibits SCC formation. BMP4 null mutations in mice are lethal at embryonic stages prior to otocyst development (41), thus preventing evaluation of the role of BMP4 in inner ear development. A better understanding of the roles of BMP4 and at-RA in inner ear development requires a model system in which these two morphogens can be studied. By using a mouse otocyst-derived cell line (2B1), we found that BMP4 expression is driven by two promoters that are distinct from the major promoter used in bone. In contrast to the major bone promoter, we show that these promoters in otocyst cells are down-regulated by ligand-occupied RARs. The significance of this unusual regulation was demonstrated in chicken embryos, where the ability of at-RA to disrupt SCC formation was overcome by exogenous BMP4, indicating that this at-RA effect is mediated by repression of BMP4.

Promoter usage in the 2B1 otocyst cell line.

The BMP4 gene has two previously characterized promoters, 1A and 1B, with 1A being predominant in the cell lines and tissues studied to date (7, 13). We found that the 1B promoter and a novel intron 2 promoter are used to drive BMP4 expression in 2B1 otocyst-derived cells. The function of the intron 2 promoter was substantiated in vivo by showing that its transcription product is present in mouse embryo otocysts and that it is down-regulated by at-RA. A Pustell DNA matrix analysis revealed that the proximal ∼600 bp of the mouse intron 2 promoter are well conserved with human BMP4, similar to the conservation observed with the 1A and 1B promoters (data not shown). The protein products of the three BMP4 transcripts are identical. Hence, we speculate that the 1B and intron 2 promoters have been selected to permit at-RA repression in restricted locations and/or developmental stages, as the more commonly used 1A promoter is up-regulated by at-RA in osteoblasts (13). The intron 2 and 1B promoters may confine BMP4 expression appropriately in space and time and/or prevent excess BMP4 expression during inner ear development. The inner ear and the temporal bone are forming in the same region of the head at the same time by using a common pool of periotic mesenchyme cells to contribute to both inner ear cartilage and temporal bone formation. The use of different BMP4 promoters could contribute to cell lineage sorting into temporal bone or ear, allowing these two processes to proceed simultaneously and independently.

Of relevance to this study is the fruit fly decapentaplegic (dpp) gene, which is homologous to mammalian BMP4, with the mature proteins being over 75% similar (33). dpp has two coding exons and at least three 5′-noncoding exons that are driven by three alternate promoters. Functional analysis in the fruit fly showed that the different dpp promoters are used at different developmental stages and in different tissues. Thus, the promoter usage of BMP4 appears to have parallels with that of dpp. Of particular interest regarding the BMP4 intron 2 promoter is the sequence (GT)₂₂ located 442 bp 5′ to the transcription start site. GT stretches have a propensity to adopt a Z-DNA conformation, and analysis of human genes has revealed many GT stretches with a bias toward locations near sites of transcription initiation (37). It has been hypothesized that the Z-DNA conformation plays a role in regulating gene transcription.

Down-regulation of BMP4 by at-RA.

In contrast to the much more common situation of gene induction by at-RA, the general mechanism underlying at-RA-mediated repression of target genes is not known. In fact, a consensus cis-acting DNA sequence to mediate repression has yet to be defined. The intron 2 and 1B promoters do not contain classical direct repeat or palindromic RAREs; however, this is not surprising since these elements are characteristic of genes induced by at-RA. One way in which at-RA could repress gene transcription is by inducing a gene-specific repressor. For example, the nuclear receptor FXR represses cholesterol 7α-hydroxylase by inducing expression of the transcriptional repressor SHP (23). We excluded a similar mechanism by showing that TTNPB-mediated repression of BMP4 transcription does not require protein synthesis. Perhaps the most thoroughly studied example of at-RA-mediated repression is inhibition of AP-1-mediated gene activation (32). This regulation does not appear to require RAR-DNA interactions (27), although the actual mechanism is controversial. Proposed mechanisms include a direct interaction between the ligand-occupied RAR and AP-1, competition for transcriptional coactivators, inhibition of c-Jun N-terminal kinase, functional interference with JunB/Fra1 dimers (reference 34 and references therein), and indirect effects due to at-RA induction of glycogen synthase kinase 3 (2). Thyroid hormone receptors (TRs) are closely related to RARs and are known to repress the thyrotropin beta gene (TSHβ) in a ligand (T3)-dependent manner. Again, despite numerous studies, the mechanism is unclear. However, suggested mechanisms include the recruitment of histone deacetylases to the T3/TR-DNA complex (31) or off-the-DNA effects (35). Opposite to repression by T3-occupied TRs, unliganded TRs appear to enhance TSHβ gene expression (35). However, our transfection data do not reveal an effect of unliganded RARs on BMP4 promoter activity. Thus, there may be mechanistic differences in the negative regulation of TSHβ by TRs and BMP4 by RARs. Further studies are required to understand the mechanism(s) of transcriptional repression by ligand-occupied nuclear receptors, including repression of BMP4 by at-RA.

RAR isotype function in otocysts.

As with other members of the nuclear receptor superfamily, RARs have a modular structure made up of six functional domains designated A to F (11). The A and B domains of nuclear receptors are involved in trans activation functions that are both cell type and promoter specific (36). The three types of RARs (α, β, and γ) each have several isoforms that vary in their N-terminal A domains (16, 20, 44). We cloned five RAR isoforms from 2B1 cells: α1, α2, γ1, γ2, and γ4. All of these isoforms were capable of repressing luciferase expression driven by either the BMP4 1B or intron 2 promoter in this cell line. Importantly, in the same cells, these RARs supported induction of luciferase from the RARE 2xPal, indicating that the promoter must dictate the direction of the response.

Unlike the other RARs, RARγ4 does not have an A domain. Since RARγ4, along with RARγ1, was the most potent repressor of luciferase expression from both BMP4 promoters, it can be inferred that BMP4 repression does not require the A domain. Overall, the RARγ isoforms were stronger repressors at the BMP4 promoters while the RARα isoforms were stronger inducers at the RARE 2xPal. The structural basis for this difference remains to be defined, but the difference presumably reflects differences in protein interactions between RARα and RARγ.

Functional implications of BMP4 and at-RA for inner ear development.

The SCC malformations obtained by antagonizing BMP4 (5, 10) or exposing otocysts to at-RA (6) demonstrate that BMP4 has an inductive role and at-RA has a repressive role in SCC formation. The ectopic otocysts observed in mouse embryos with combined RARα and RARγ null mutations (22, 39) or vitamin A deficiency (40) further suggest a repressive role for at-RA in otocyst development. These observations, combined with the knowledge that RARs are transcription factors, led us to postulate that the effect of at-RA on SCC development is mediated by down-regulation of BMP4. Here, we demonstrate that at-RA down-regulates BMP4 transcription in otocyst-derived cells and that a novel promoter within intron 2 is a target of this down-regulation both in the cell line and in vivo in mouse embryo otocysts. Importantly, by using chicken embryos, we show that the inhibitory effect of at-RA on SCC formation is overcome by exogenous BMP4, indicating that repression of BMP4 lies downstream of at-RA in regulating SCC development. Given that the otocyst expresses RARs (30) and the dehydrogenases necessary to convert vitamin A to at-RA (24, 29), we speculate that at-RA plays a key role in otocyst development via its ability to regulate BMP4 expression directly. at-RA also can affect the otocyst through indirect mechanisms. The hindbrain, especially rhombomere 5, provides signals that are critical to otocyst development, and hence, factors that disturb rhombomere 5 development also disturb otocyst development (8). Vitamin A deficiency (40) or combined null mutations of RARα and RARγ (22, 39) impair the development of rhombomere 5, providing a second, indirect mechanism by which at-RA can affect otocyst development. Thus, at-RA regulates otocyst development by multiple mechanisms, including indirect effects through alterations in hindbrain development, as well as the direct effects on otocyst BMP4 expression, as demonstrated herein.