Human Activin βA Gene IDENTIFICATION OF NOVEL 5′

On the basis of cDNA cloning, primer extension, and transfection experiments, we identified a novel 5′ exon of the human activin βA subunit gene, and found its enhancer and promoter regions as well as multiple transcription start sites. A series of deletion and mutation analyses of the enhancer sequences defined the 45-base pair core region (DR-1 core) containing two short elements with similarity to AP-1 (12-O-tetradecanoylphorbol-13-acetate response element; TRE) and CREB/ATF (cyclic AMP response element; CRE) binding sites, both of which were necessary for full enhancer activity. Gel shift and antibody supershift assays using DR-1 core region revealed the formation of two specific DNA-protein complexes, one of which could be partially dissociated by a competing oligonucleotide containing a single copy of the consensus TRE, but the other of which contained neither CREB/ATF nor AP-1 as major components. Although 12-O-tetradecanoylphorbol-13-acetate and cAMP induced the activin enhancer/promoter-driven CAT activity, such drug induction was obscured when either the TRE- or CRE-like elements were mutated in the native promoter context. Our results demonstrate that the promoter and enhancer regions identified here are essential for maintaining the efficient promoter activity of the human activin βA subunit gene.

Activins, originally characterized based on their ability to stimulate follicle-stimulating hormone secretion by cultures of rat anterior pituitary cells (1,2), are members of an extensive family of growth and differentiation factors that include inhibin, transforming growth factor-␤, Mü llerian inhibitory substance, the fly decapentaplegic gene complex, and the product of Xenopus Vg-1 mRNA (3). Further investigations revealed multiple functions of activin, such as stimulation of hematopoiesis (4,5), paracrine regulation of ovarian and testicular functions (6 -8), modulation of nerve cell differentiation (9,10), and mesoderm induction in early embryos (11)(12)(13)(14). Inhibins were initially described as an inhibitory activity of follicle-stimulating hormone production, and for the most part they act in opposition to the activities of activins (15). Although functionally antagonistic, both are structurally related, i.e. they are disulfide-bonded dimeric peptides sharing a common ␤-subunit. Inhibin was later shown to be composed of a specific ␣-subunit complexed with one of the two closely related ␤-subunits, whereas activin was found to be formed by a homo-or heterodimer of the two ␤-subunits. These three subunits are encoded by three separate genes, and the S1 nuclease analysis in adult rat tissues has shown that the expression of mRNAs encoding these subunits (␣, ␤A, and ␤B) varies by severalfold in a tissue-specific manner (16). Thus, differential subunit association results in the formation of dimers with opposing biological activities in a manner dependent on their sites of production.
Since expression of the activin and inhibin subunit genes must each be precisely regulated, the identification of cis-acting elements that control the activin gene expression and the elucidation of how these elements are involved in the response to extracellular stimuli are of key importance. Recently, we reported that expression of the human activin ␤A gene is stimulated by the addition of 12-O-tetradecanoylphorbol 13-acetate (TPA) 1 (17) or 8-bromo-cyclic AMP (8-Br-cAMP) (18) and that the effects of these two drugs were synergistic in human fibrosarcoma HT1080 cells but not in either HeLa or HepG2 cells (Ref. 19 and data not shown). However, the DNA sequences that regulate expression of the gene have not been defined. Although several groups including ours have reported the cloning and sequence of the coding portion of the human activin ␤A gene (20 -22), neither the complete structural organization of the activin ␤A gene nor its functional promoter or any other regulatory elements have been previously defined. We therefore report the complete structure for the activin ␤A subunit gene, as well as studies identifying crucial regulatory elements and transcription factors that are involved in its transcriptional control.
Isolation of cDNA Clones and Nucleotide Sequence Determination-Total RNA was extracted from HT1080 cells (that were treated with TPA at 100 ng/ml for 4 h) by using ISOGEN (Nippon Gene). Poly(A) ϩ RNA was selected by oligo(dT)-cellulose chromatography. cDNA was synthesized using random primers (6-mer; Pharmacia Biotech Inc.), followed by ligation to adapters containing EcoRI and NotI sites, digested with EcoRI, and finally ligated with EcoRI-digested, dephosphorylated ZAPII DNA (Stratagene). Screening of this library (7.5 ϫ 10 4 plaques) with 32 P-labeled human activin ␤A subunit probe (221-bp DraI fragment in Fig. 1A; bases 810-1824) (22) yielded five positive plaques. All five were purified, and the nucleotide sequence of their 5Ј-ends was determined.
Plasmid Constructions-All of the fragments used in the constructions were excised from the genomic clone (22) containing the human activin 5Ј-upstream region. ph␤ACAT19 was constructed as follows. A 1015-bp DraI fragment (bases 810-1824; Ref. 22) was ligated with HindIII linker, digested with EcoT14I and HindIII to generate an 852-bp fragment (bases 973-1824). This fragment and a 972-bp BglII-EcoT14I fragment (bases 1-972) were inserted in the BglII/HindIII sites of pUCSV0CAT (23). To make ph␤ACAT35, a 3.5-kb BamHI fragment (bases ϳ847) was blunt-ended by the fill-in reaction using the Klenow fragment (TaKaRa), ligated with HindIII linker, digested with HindIII, and inserted in the HindIII site of pUC0CAT (24). A 2.7-kb HindIII-BglII fragment from the ph␤ACAT35 and the BglII-HindIII fragment (bases 1-1824) from the ph␤ACAT19 were inserted in the HindIII site of pUC0CAT to make ph␤ACAT45. Construction of pUCSV0CAT and pUCSV3CAT was described previously (23).
For construction of 5Ј-deletion mutants of the activin 5Ј-upstream region, the 3.5-kb BamHI fragment (the same as that used for construction of ph␤ACAT35) was inserted in the BamHI site of pBluescriptII KS(ϩ) (Stratagene) in the reverse orientation. This plasmid was linearized with XbaI and PstI restriction enzymes, partially digested with exonuclease BAL31 (TaKaRa) for various lengths of time, treated with Klenow fragment to form blunt ends, and ligated in the presence of BglII linker. DNAs were transferred into Escherichia coli HB101. After determination of the 5Ј-deleted ends by DNA sequencing, the fragments were excised with BglII (sites were provided by linker sequence and the activin gene at position 1 (22)). The deleted fragments were inserted in the BglII site, located upstream of the 1.9-kb BglII-HindIII fragment, of ph␤ACAT19 to make ph␤ACAT30, -29, -28, -26, and -24.
The human c-Jun expression plasmid, pRSV-c-Jun (25), was kindly provided by Professor Michael Karin (University of California, San Diego).
Transient Transfections-Twenty-four hours prior to DNA transfection, HT1080 cells were plated at 2 ϫ 10 5 cells/60-mm diameter dish in medium supplemented with 10% fetal bovine serum. Three hours before transfection, the dishes received fresh medium. Cells were transfected with the different constructs by the calcium phosphate coprecipitation method, exposed to the precipitate for 9 h, and kept in fresh medium (supplemented with drugs, if necessary) for 24 h. Cells were collected, and cell extracts were prepared by freezing and thawing and analyzed for CAT activity.
Stable Transfections-HT1080 cells were co-transfected with a 5:1:1 molar ratio of the activin 5Ј-flanking region-CAT chimeric constructs, pSV2neo (26), and pCH110, a plasmid that contains the E. coli ␤-galactosidase gene controlled by the SV40 promoter (27) (21 g of total DNA/100-mm diameter dish; 1 ϫ 10 6 logarithmically growing cells) by the calcium phosphate coprecipitation method. After selection with 800 g/ml G418 for 8 -12 days, the resistant cells were pooled and analyzed for CAT activity. One-fifth of the cell extract was assayed for ␤-galactosidase activity to monitor the transfection efficiency.
CAT Assay-For CAT assay (28), cell extracts (20 -50 g for transient transfection and 20 g for stable transfection assay) were incubated at 37°C for 1.5 h in 140 mM Tris-Cl (pH 7.8), 0.68 Ci/ml [ 14 C]chloramphenicol (DuPont), and 580 M acetyl coenzyme A (Wako Pure Chemical Industries, Ltd.). [ 14 C]Chloramphenicol and its acetylated products were separated by TLC. The TLC plate was exposed to RX film (Fuji). The conversion of chloramphenicol to its acetylated form was measured using a FUJIX BIO-imaging analyzer BAS2000 (Fuji).
Primer Extension Analysis-Primer extension experiments were performed by the method of Bodner and Karin (29) with some modifications. Two oligonucleotide primers complementary to regions within exons 1 and 2 of the human activin ␤A gene were synthesized (PR3PX and HBA-6 in Fig. 1A, respectively). These primers were end-labeled with [␥-32 P]ATP (Amersham Corp.) and T4 polynucleotide kinase (Toyobo) and were purified on Sephadex G-50 chromatography (Pharmacia). The labeled primers (5 ϫ 10 5 cpm) were mixed with 20 g of poly(A) ϩ RNA from HT1080 cells or yeast tRNA (as a control) and precipitated with ethanol. The dried pellet was directly dissolved in 10 l of hybridization buffer (10 mM Tris-HCl (pH 7.9), 1 mM EDTA, 0.25 M KCl), and hybridization was carried out at 55°C for 60 min. After the addition of 24 l of a reaction mix (10 mM Tris-HCl (pH 8.7), 10 mM MgCl 2 , 5 mM DTT, 100 g/ml actinomycin D, 0.4 mM dNTPs, 10 units of Rouse-associated virus (RAV-2) reverse transcriptase (TaKaRa)), extension was performed at 37°C for 45 min. Free nucleotides were removed by Sephadex G-25 chromatography (Pharmacia) followed by ethanol precipitation. The extended products were denatured and subjected to electrophoresis on an 8 M Urea, 5% polyacrylamide gel or on a standard sequence gel with appropriate molecular weight markers.
Gel Shift Assay-Nuclear extracts were prepared using a modification of the protocol of Lee et al. (30). The final protein concentration was 8 -13 mg/ml. Probes were synthesized by PCR, purified by Ultrafree C3 filter (Millipore), and phosphorylated at both ends using T4 polynucleotide kinase and [␥-32 P]ATP. Nuclear extracts (5-10 g) were preincubated for 20 min at 4°C in an 18-l reaction mixture containing 20 mM Tris-HCl (pH 7.6), 100 mM NaCl, 2 mM EDTA, 4 mM MgCl 2 , 2 mM dithiothreitol, 20% glycerol, and 1 g of double-stranded poly(dI-dC) (Pharmacia) in the presence or absence of a 50 -200-fold molar excess of a specific double-stranded competitor DNA. 0.32 ng (10,000 cpm) of a radiolabeled DNA probe were added and the incubation continued for 30 min at 4°C. The incubation mixture was loaded on a 3.5% nondenaturing polyacrylamide gel in 1 ϫ TBE (1 ϫ TBE: 89 mM Tris-HCl, 89 mM boric acid, and 2 mM EDTA) buffer, and electrophoresed at 4°C with a field strength of 10 V/cm with buffer circulation. The gels were dried and exposed to x-ray film at Ϫ70°C overnight. For competition experiments, competitor DNA was incubated in the mixture prior to the addition of nuclear extract. Oligonucleotides containing the consensus binding sequences for AP-1(s) and CREB/ATF (31) were synthesized on a DNA synthesizer. The AP-1(w) fragment was simply made by ligating the AP-1(s) fragment on the plasmid vector. For supershift assays, 1-2 g of antibody (Santa Cruz Biotechnology) were included in the reaction mixture.

RESULTS
Cloning and Sequence Analysis of the 5Ј-untranslated Region of cDNA for Human Activin ␤A-The complete genomic structure (including a description of the number and organization of exons and introns) of the activin ␤A subunit gene has not been reported. Although more extended sequences lying 5Ј to the translated region of the ␤A gene have been reported for the mouse (32), rat (33), and human (34) cDNAs, the corresponding genomic sequences have not been defined. We previously cloned the gene for the human activin ␤A subunit and determined the nucleotide sequence extending from 1857 bp 5Ј to the translation start codon (22). However, a computer-assisted homology search did not identify significant identities between either rodent 5Ј-untranslated region and this region of the human genomic sequences (not shown). We therefore considered the possibility that a previously unidentified 5Ј exon might exist and that this hypothetical exon lay even further upstream of the activin ␤A-coding sequence. To test this hy-pothesis, we cloned activin ␤A cDNAs from a cDNA library of TPA-treated HT1080 cells, since we had previously found that TPA markedly induced the accumulation of mRNA encoding activin ␤A (17)(18)(19). Several cDNA clones for human activin A were isolated that extended 5Ј to the start of translation, and all of these clones contained the same untranslated sequences, although of varying length. Comparison of these untranslated sequences with our previously reported genomic sequences (22) defined the position of the first exon as lying beyond 145 bp upstream from the first nucleotide of the ATG initiation codon, since the cDNA sequences began to diverge from the genomic sequences at this point.
Structure of Activin ␤A Gene and Its 5Ј-Flanking Sequences-The results presented here, in comparison with the sequences of the cDNA and genomic clones (22) for the activin ␤A subunit, allowed us to derive the complete structure of this gene (Fig. 1B). The gene consists of three exons interrupted by two introns, apparently differing from the two-exon structure reported for the human ␣ (21) and ␤B subunit (35) genes. The first exon of the activin ␤A gene contains the 5Ј-noncoding region, while exon 2 includes both 143 bp of noncoding and the initiation of coding sequences. Exon 3 contains the sequences encoding the mature, post-translationally processed ␤A subunit protein and the entire 3Ј-noncoding region, including several polyadenylation signals (17). Intron 1 contains typical 5Ј-donor (gt(a/g)) and 3Ј-acceptor (cag) splice consensus se-quences ( Fig. 1A) (36).
To determine the transcription start site of the human activin A gene, we conducted primer extension analysis using two different primers, complimentary to sequences within exon 1 or 2 ( Fig. 2A), and these yielded two major and several minor extension products that were in good agreement with each other (lanes 4 and 7 in Fig. 2B) and with the RNase protection assays (data not shown). No extension product was formed when yeast tRNA (lane 5) or RNA from noninduced HT1080 cells (lanes 3 and 6) were used. When the size of the products (Fig. 2B, lane 4) was compared with sequence ladders primed using the same primer on a high resolution sequence gel, multiple transcription start sites were detected (Fig. 2C, lane 2). Comparing the data obtained by both primer extension and RNase protection analyses, the band having the strongest intensity (an A located 143 bases 5Ј to the initiation codon) was assigned position ϩ1 (Figs. 1A and 2C).
We determined the sequences of the putative 5Ј-flanking regions and found several nucleotide consensus motifs that are involved in transcriptional regulation of many genes (Fig. 1A). Possible TATA boxes (TATAAA and TACAAA) were found 29 and 47 bp, respectively, upstream of the major transcription start sites determined by primer extension analysis, a typical promoter feature of many eukaryotic genes. In addition, CCAAT boxes, another element often found in RNA polymerase II-governed promoters, are present at positions Ϫ270, Ϫ103, and Ϫ60. GATA and CACCC box binding sites, which are common structural and functional features of the promoters of many genes, are also present (37). Two putative AP-1 (c-Jun homodimer or c-Jun⅐c-Fos heterodimer) binding sites (TPA response element; TRE) (T1 site, Ϫ201 to Ϫ195; T2 site, Ϫ153 to Ϫ147) (38,39), one lymphokine-specific sequence (40,41), one PEA-3 binding sequence (42)(43)(44), one cyclic AMP response element (CRE)-like site (C site, Ϫ123 to Ϫ116) (45), and a consensus AP-2 binding site (46) were also present.
Identification of Sequences Required for Activin ␤A Transcription-To identify the sequences responsible for controlling the expression of the human activin ␤A gene, CAT constructs that contained various lengths of the 5Ј-flanking sequences of this gene were generated (Fig. 3A). A DraI restriction site, lying 36 bp 5Ј to the translation start site and within the mature mRNA (Fig. 1A), was used as the 3Ј-end of each con-struct. A series of the 5Ј-deletion mutants were transiently transfected into HT-cJun cells (Fig. 3B), a HT1080 cell line that had been stably transformed with the pRSV-cJun expression vector (17), because the levels of activity expressed by these transiently transfected constructs were too low to be determined accurately. The promoterless plasmid pUCSV0CAT was used as a background control, and pUCSV3CAT (a reporter directed by the SV40 enhancer-promoter region) was used as a positive control. As shown in Fig. 3B, deletion of activin ␤A sequences up to position Ϫ3.0 kb to the gene (in ph␤ACAT30) resulted in a 2-fold increase in CAT activity relative to the parental full-length plasmid (ph␤ACAT45). This suggests the presence of negative regulatory region(s) within Ϫ4.5 and Ϫ3.0 kb of the gene. Deletion of the next (approximately) 100 bp (in ph␤ACAT29) led to a dramatic reduction in activity (18% of ph␤ACAT30). Further deletion of the next 100 bp (ph␤ACAT28) reduced activity to background levels.
To test whether these same regions are also active without c-Jun expression, some representatives of this deletion series  (17) were transiently transfected with 6 g of the fusion genes. pUCSV3CAT, which contains both SV40 enhancer and promoter, and pUCSV0CAT were used as positive and negative controls, respectively. Following a 36-h incubation, cell lysates were prepared and analyzed for CAT activity as described under "Experimental Procedures." Relative CAT activities were determined by averaging eight independent experiments and comparing them with the activity obtained with pUCSV3CAT (assigned a value of 100). C, basal expression of the deletion constructs. HT1080 cells were stably transfected with a series of activin/CAT fusion genes (15 g) and pSV2neo (3 g). After selection with 0.8 mg/ml of G418 for 2 weeks, cell lysates were prepared and analyzed for CAT activity. Results are expressed as relative CAT activities compared with those obtained with ph␤ACAT30 (assigned a value of 100), and the values are the averages of four independent experiments.
FIG. 4. Activation of heterologous TK promoter by DR-1 enhancer fragment. Enhancer activity of the DR-1 element is shown. One or two copies of the DR-1 fragment were inserted in the sense or antisense orientations, upstream of the p0/tk-CAT, bearing the herpes simplex virus thymidine kinase gene promoter (17). HT-cJun cells were transiently transfected with 6 g of plasmids. After 36 h, cell lysates were prepared and analyzed for CAT activity as described under "Experimental Procedures." Enhancer activities are expressed relative to the CAT activity obtained with p0/tk-CAT (assigned a value of 1). The values represent the average of at least four independent experiments.
were stably transfected into the HT1080 cells and analyzed for CAT activity (Fig. 3C). Deletion of the activin ␤A gene sequences from Ϫ4.5 to Ϫ3.0 kb resulted in an overall 4.8-fold increase in CAT activity, again suggesting the existence of negative regulatory elements within this region that affect both basal and c-Jun-induced transcription. Further deletion of 5Ј sequences (e.g. in ph␤ACAT29) severely reduced reporter gene activity (13% of the ph␤ACAT30 activity). ph␤ACAT28 exhibited no activity above background. These results showed that a positive regulatory element, located between the region described by constructs ph␤ACAT30 and ph␤ACAT28, is required for expression of the activin A gene. We designated this region DR (distal region)-1 (Fig. 3A).
To ascertain whether or not DR-1 fulfilled the requirements of an enhancer, a DNA fragment corresponding to sequences lying between Ϫ275 and Ϫ89 (designated DR-1; nucleotide numbering as shown in Fig. 1A) was synthesized by PCR and then fused to a CAT reporter gene construct that was transcriptionally directed by the herpes simplex virus TK promoter (Fig.  4A). These test constructs were then transiently transfected into HT-cJun cells. CAT activities were normalized to that obtained with the enhancerless vector, p0/tk-CAT, which was arbitrarily assigned the value 1. The addition of the DR-1 fragment in the sense and antisense orientations produced a 19-and 17-fold stimulation, respectively, of CAT activity (Fig.  4A). Furthermore, duplicated copies of the DR-1 fragment enhanced the activation even more dramatically (69 -82-fold; Fig.  4A). We have also examined the activity of the DR-1 fragment at a position downstream of the CAT gene and obtained 5-fold activation of the TK promoter (data not shown). These results show that the 187-bp DR-1 fragment functions as a powerful enhancer in HT-cJun cells.
Identification of Activin ␤A Promoter-Since the DR-1 fragment alone was unable to confer transcription to a cis-linked reporter gene (data not shown), further deletion analysis of more proximal regions to the transcription initiation sites, placed under the control of the DR-1 enhancer, was performed to determine whether any additional segments of this 5Ј region contributed to gene activity. Removing DR-1 from ph␤ACAT30 completely abolished CAT activity (ph␤ACAT28), and as anticipated, the addition of DR-1 (in the reverse orientation) to ph␤ACAT28 (pDR1(Ϫ)/CAT28) partially restored this activity (Fig. 5A). However, DR-1 was unable to restore the activity of constructs containing more significant deletions (e.g. in constructs pDR1(Ϫ)/CAT26, pDR1(Ϫ)/CAT24, and pDR1(Ϫ)/ CAT19; Fig. 5A), suggesting that the region of the activin A gene described by the end points lying between ph␤ACAT28 and ph␤ACAT26 is also required for transcription. We designated this region PR (proximal region)-2 (Ϫ88 and ϩ103, nucleotide numbering as in Fig. 1A; Fig. 5, top).
To further characterize the PR-2, this DNA region was prepared by PCR and analyzed for activity after transfection into HT-cJun cells. As shown in Fig. 5B, PR-2 alone had little CAT activity (PR2(ϩ)/CAT) in comparison with the negative control. When the DR-1 enhancer was included in the construct with the sense-oriented PR-2 fragment (pDR1(Ϫ)/PR2(ϩ)/CAT), a significant stimulation in activity was observed. However, the reverse-oriented PR-2 fragment in the same construct (pDR1(Ϫ)/PR2(Ϫ)/CAT) gave background levels of activity. Transcriptional enhancement was also observed using the heterologous SV40 enhancer (pSVE(ϩ)/PR2(ϩ)/CAT). These results thus demonstrate that PR-2 contains the activin ␤A gene promoter.
Analysis of DR-1 Enhancer-A search of the DR-1 region for known consensus binding sites for transcription factors identified several potential sequences, such as two TRE (AP-1 binding sequences, T1 and T2)-like and CRE (CREB/ATF binding sequences, C)-like sequences (Fig. 1A). It is reported that the CRE is efficiently trans-activated by cJun/AP-1 and that some of the CREB/ATF transcription factors heterodimerize with c-Jun to bind to the CRE (47). Deletion mutations were targeted to determine whether the three potential T1, T2, and C sites (Figs. 1A and 6A) individually or combinatorially contributed to DR-1 enhancer function. A series of 5Ј-and/or 3Јdeletion mutants from the DR-1 enhancer fragment were placed in the sense orientation 5Ј to the TK gene promoter (in p0/tk-CAT), and their contribution to transcriptional activity was tested in transiently transfected HT-cJun cells (Fig. 6A). Deletion of site T1 (pDR1(Ϫ180:Ϫ89)/tk-CAT) from the parental construct (pDR1(Ϫ275:Ϫ89)/tk-CAT) moderately reduced the enhancer activity of DR-1. Nevertheless, the same deletion had a statistically insignificant effect on enhancer activity when introduced into pDR1(Ϫ275:Ϫ114)/tk-CAT (compared with pDR1(Ϫ180:Ϫ114)/tk-CAT). Further deletion of sequences at the 5Ј-end (in pDR1(Ϫ151:Ϫ89)/tk-CAT) reduced the enhancer activity to 16% of the original activity, which is probably attributable to destruction of sequence T2, 5Ј-TGTCTCA-3Ј, lying between Ϫ153 and Ϫ147. Results examining the activity of the 3Ј-deletion mutants indicated that sequences up to nucleotide Ϫ114 had no significant role in enhancer activity of DR-1 (constructs pDR1(Ϫ275:Ϫ114)/tk-CAT and pDR1(Ϫ180: Ϫ114)/tk-CAT). However, deletion of 26 nucleotides 5Ј from position C reduced enhancer activity to less than 18% of the original level (pDR1(Ϫ275:Ϫ140)/tk-CAT). A DNA fragment that contains both T2 and C sites (pDR1(Ϫ158:Ϫ114)/tk-CAT) exhibited significant enhancer activity. In summary, this detailed mutational analysis identified the core enhancer sequences of DR-1 as lying between nucleotide positions Ϫ158 and Ϫ114, and we subsequently refer to this region as the DR-1 core.
To directly demonstrate a functional role for the two putative sequence motifs in the activity of the DR-1 core enhancer, we deleted three and seven nucleotides from each of the putative binding sites for transcription factors (as well as from the region between the sites) using PCR site-directed mutagenesis (Fig. 6B). Four different deletion mutants for DR-1 core enhancer were examined: one for the T2 site (⌬T), one for the C site (⌬C), one for the double mutant containing both of them (⌬CT), and one for the intervening sequence (⌬i). Each of these mutants was cloned 5Ј to the TK promoter (in the p0/tk-CAT reporter plasmid), and their individual effects on transcriptional activity were tested after transfection into HT-cJun cells. As shown in Fig. 6C, mutation of sequences lying between these two elements (⌬i; pDR1(core⌬i)/tk-CAT) modestly increased enhancer activity, whereas the ⌬T mutant (pDR1(core⌬T)/tk-CAT) reduced CAT activity to about 22% of the wild type enhancer level, but still well above the activity of the TK promoter alone. In contrast, the activities of the ⌬C and ⌬CT mutants (pDR1(core⌬C)/tk-CAT and pDR1(core⌬CT)/tk-CAT) virtually completely abolished enhancer activity. These results suggest that a combined action of T2 and C sequences is important for enhancer activity of the DR-1 core fragment, although the T2 function may be dependent on the presence of site C.
Detection of Specific DNA/Protein Interactions within DR-1-To examine the ability of the T2 and C elements to interact with trans-acting factors, the DR-1 core fragment was examined in gel shift assays with nuclear extracts prepared from HT1080 cells (Fig. 7). After electrophoresis of the products of binding reactions in polyacrylamide gels, two complexes, designated C1 and C2, were identified (lanes 2 and 11). Formation of both complexes was efficiently inhibited after adding an excess of unlabeled DR-1 core fragment (lanes 3 and 4). To test whether the T2 and C sites were also critical for protein-DNA complex formation, the ⌬T and ⌬C mutations were individually introduced into the DR-1 core fragment. When these mutant fragments were used in competition experiments, only the DR-1 core fragment with ⌬T mutation (core-⌬T) was able to efficiently compete for protein binding in both complexes (lane 5), indicating that the core-⌬T fragment retains binding activity to both complexes. On the other hand, the DR-1 core fragment harboring the ⌬C mutation failed to compete with C2 complex formation (lane 6).
Since the T2 and C sites were similar to the authentic TRE (25) and CRE (47), respectively, we investigated whether the DR-1 core binding complexes contained these factors. As shown in Fig. 7, a 200-fold molar excess of a single copy of consensus TRE-containing oligonucleotide (AP-1(s)) (lanes 8 and 13) was able to compete for C1 complex formation only weakly, indicating that this complex includes members of the AP-1 family. However, the DR-1 core fragment appears to compete much more effectively than does TRE oligonucleotide for the formation of complex C1 (compare lanes 3, 4, and 8), suggesting that this complex does not consist principally of AP-1. This was also confirmed in supershift experiments (lane 9) using a polyclonal antibody that recognizes Jun family proteins (c-Jun, JunB, and JunD). Only the formation of complex C1 was perturbed by the inclusion of antibody to the DNA association reaction, but only to a very limited extent. The formation of both complexes was also unaffected by the addition of a CRE competitor (lanes 7 and 12); similarly, the addition of a monoclonal antibody against CREB/ATF family proteins (ATF-1, CREB-1, and CREM-1) left the two complexes essentially unaffected (lane 15). Parallel experiments showed that anti-Jun family and anti-CREB/ATF family antibodies effectively supershifted the FIG. 6. Identification and functional analysis of the DR-1 core enhancer element. A, identification of core element in the DR-1 enhancer fragment. Solid boxes represent sequence motifs suggesting potential protein binding sites for AP-1 (T1 and T2) and CREB/ATF (C). Subfragments of the DR-1 enhancer (striped bars) were synthesized using PCR and inserted upstream of the TK promoter in p0/tk-CAT. The vertical lines indicate the positions of the 5Ј termini of the primers used to synthesize the fragments. HT-cJun cells were transfected with the 6 g of the indicated constructs. After 36 h, cell lysates were prepared and analyzed for CAT activity as described under "Experimental Procedures." Results are expressed as -fold activation compared with that obtained with p0/tk-CAT (TK/CAT), and the values shown are the mean Ϯ S.E. of 12 independent experiments. B, nucleotide sequence of the DR-1 core enhancer fragment and the positions of deleted nucleotides (⌬T, ⌬C, and ⌬i with brackets). Potential binding sites for AP-1 (T2) and CREB/ATF (C) are boxed. Consensus Dorsal binding sequence (54,55) is also shown. C, mutational analysis of the DR-1 core enhancer activity. The wild-type (core) fragment and its mutated derivatives, in which either T2 (core⌬T) or C (core⌬C) and a part of their intervening sequence (core⌬i) or a combination of them (core⌬CT) were disrupted, were fused to the p0/tk-CAT (TK). CAT assays were performed as described above. The level of activation is depicted as a percentage of the level obtained with the wild-type construct (core, 100%). The mean Ϯ S.E. of at least six independent experiments is shown. complexes formed on legitimate CRE (lanes 16 -19) and TRE (lanes 20 -23) probes, respectively. Interestingly, a tandem duplicated copy of the TRE oligonucleotide (200-fold molar excess of AP-1(w), lane 14) effectively eradicated C2 complex formation, although an 800-fold molar excess of AP-1(s) did not (data not shown). No competition with either complex was observed when a 200-fold molar excess of unlabeled oligonucleotides containing consensus binding sites for Sp-1, NF-1, and C/EBP were added to the reactions (data not shown).
Induction of Activin ␤A Gene Transcription by TPA and cAMP-We previously reported that expression of the human activin ␤A gene in HT1080 cells was induced by both TPA and 8-Br-cAMP and that the effects of these two drugs were synergistic (17)(18)(19). We postulated that this induction might be mediated at the transcriptional level, possibly through a DR-1 element. To investigate this possibility, both pDR1(ϩ)/PR2(ϩ)/ CAT and pPR2(ϩ)/CAT were transiently transfected into HT1080 cells, followed by treatment with one or both of the inducing agents. An autoradiogram of one representative experiment is shown in Fig. 8A, and in support of the hypothesis that DR-1 mediates the inductive responsiveness of the activin ␤A gene, the expression of pDR1(ϩ)/PR2(ϩ)/CAT, but not PR2(ϩ)/CAT (data not shown), was stimulated by TPA (2.3fold) or 8-Br-cAMP (2.2-fold) or both (5.5-fold). Furthermore, a duplicated DR-1 fragment directing expression from a heterologous (TK gene) promoter showed a synergistic responsiveness to induction by both drugs (Fig. 8B). These results demonstrated that the DR-1, including T2 and C sites, and the PR-2 are both necessary and sufficient for induction of the activin ␤A gene by TPA and/or 8-Br-cAMP in HT1080 cells.
Site-directed mutagenesis was performed to determine whether T2 and C sites were functionally relevant to the induction of activin ␤A gene transcription by TPA and/or 8-Br-cAMP (Fig. 9). The pDR1(ϩ)⌬T/PR2(ϩ)/CAT construct containing the 3-bp T2 site deletion produced 80% lower basal activity in comparison with the wild-type response (pDR1(ϩ)/PR2(ϩ)/ CAT). However, this mutant also responded significantly (albeit slightly weaker) to TPA induction (1.4-fold), 8-Br-cAMP induction (1.9-fold), and the additive effects mediated by the addition of both TPA and 8-Br-cAMP (3.4-fold) in the native promoter context. Basal expression produced from the construct containing the 3-bp C site deletion (pDR1(ϩ)⌬C/PR2(ϩ)/ CAT) was reduced by 95% in comparison with the wild-type construct; as with the ⌬T mutant, however, a modest inductive response to drug treatment was still retained. The three constructs were also found to respond positively to elevated AP-1 levels mediated by co-transfection with a c-Jun expression vector. DISCUSSION On the basis of the results presented here, the structural organization of the human activin ␤A gene was shown to consist of three exons interrupted by two introns (Fig. 1B), which is totally different from the composition of the human ␣ or ␤B subunits, since both have only two exons (20,21,34,42). When we sequenced and compared the 5Ј-flanking regions of both the human (Fig. 1) and mouse 2 activin ␤A genes, the DR-1 and PR-2 regions were found to be highly similar in their nucleotide sequence. This evolutionary conservation suggests that these conserved regions play an important role in the regulation of activin A expression in both species. In addition, the T2 and C sites in the DR-1 core (Fig. 6) are also conserved in both the human and mouse activin ␤A promoters at the same positions. In contrast, the sequences of the 5Ј-flanking region of the human ␤A gene is different from that of ␣ (rat (48 -50) and mouse (51)) and ␤B (human (35), rat (49,52), and sheep (53)) subunit genes. The human, rat, and ovine activin ␤B subunit genes have no consensus TATA and CAAT boxes but instead consist of a series of GC boxes, a typical promoter for housekeeping genes.
To identify cis-acting DNA elements responsible for transcription of the human activin ␤A gene, fusion genes were constructed with progressive 5Ј-deletions linked to a reporter FIG. 7. Identification of nuclear proteins that bind to the DR-1 core enhancer element. Gel shift assay and competition analyses are shown of complexes formed by factors in HT1080 nuclear extract with the DR-1 core element (Ϫ158 to Ϫ114), CREB/ATF, or AP-1. Nuclear extracts (lanes 1, 10, 16, 20, negative controls; lanes 2-9, 11-15, 17-19, 21-23, 10 g) were incubated with 1.5 ϫ 10 4 cpm of 32 P-labeled probe. In a competition assay, a 50 -200-fold molar excess of unlabeled DNA fragments (DR1 core, core-⌬T, or core-⌬C) or a 200-fold molar excess of double-stranded oligonucleotides containing the consensus binding sites for CREB/ATF (lanes 7, 12, and 18) or AP-1 (single site (s) for lanes 8, 13, and 22, and double sites (w) for lane 14) was added to the reaction mixture. Arrows point to specific DNA-protein complexes. In supershift assays, nuclear extracts were preincubated with the indicated antibodies prior to incubation with the labeled probes for DR-1 core (lanes 9 and 15), gene (Fig. 3A), and their promoter activities were examined in transiently transfected HT-cJun cells (Fig. 3B). The presence of a positive regulatory element for the activin ␤A gene was suggested by the observation that deletion of this region resulted in a greater than 90% decrease of activity (Fig. 3B). Further stable transfection experiments (Fig. 3C) confirmed this observation. A DNA segment corresponding to this sequence, designated DR-1, was able to activate the heterologous TK promoter in a position-and orientation-independent manner (Fig. 4A), indicating that this element possesses the normal properties of an enhancer. Another critical finding was the identification of the promoter region (designated PR-2) that directs expression of the activin ␤A gene under the control of either DR-1 or the heterologous SV40 enhancer (Fig. 5).
Transfection studies using a combination of deletion and mutation approaches, in the context of heterologous TK promoter, were used to define the DR-1 core enhancer region (Fig.  6). The moderate induction of DR-1 core enhancer activity by deletion of sequences between the T2 and C sites (⌬i mutation) whose mutation partially overlaps with the putative binding sequences for Drosophila protein Dorsal, a member of the Rel family of transcription factors, that normally represses gene expression (e.g. dpp) (54,55) might be attributable to the negative regulatory function of this putative Dorsal binding site. While mutation of the C site (⌬C) completely abolished core enhancer activity, mutation of the T2 site (⌬T) reduced, but did not abolish, enhancer activity. Although Dorsal element has a minor contribution to regulate enhancer activity, these results suggest that C and T2 sites are major components of DR-1 core enhancer.
To determine whether proteins in nuclear extracts prepared from HT1080 cells could specifically interact with DR-1 core sequences, gel shift assays were conducted. These studies de-fined two specific protein-DNA complexes (Fig. 7). However, these binding activities were not cell type-specific, since they were found in all cell lines tested (HeLa, HepG2, T98G, and A172 cells; data not shown). These results are entirely consistent with those from stable transfection experiments, showing that the basal transcriptional activity expressed from the largest reporter construct is not cell type-specific and that the DR-1 enhancer can also mediate a transcriptional activation in HeLa and HepG2 cells (data not shown). Factor(s) that confers celltype specificity to the activin ␤A gene promoter likely interacts with a region outside the sequences defined in the present studies, or the gene could be regulated post-transcriptionally.
In DNA binding competition experiments (Fig. 7), complex C1 formation was greatly diminished by including either core-⌬T or core-⌬C competitor oligonucleotides. These results show that neither the T2 nor C sites are required for C1 complex formation on the DR-1 core. Unlabeled consensus AP-1 oligonucleotides (both single and double sites) displaced the binding of factor(s) to DR-1 core (C1 complex) only partially. An antibody that recognizes multiple members of Jun family proteins (c-Jun, JunB, and JunD) moderately interfered with C1 complex formation. Furthermore, bacterially expressed c-Jun protein, which is a component of the AP-1, also reacted with the DR-1 core probe (data not shown). Taken together, these results provide significant evidence that the DR-1 core enhancer is recognized by proteins of the AP-1 family of transcription factor(s) in HT1080 cells and suggest that the complex C1 is actively involved in the regulation of activin ␤A gene expression. However, the data also show that the Jun family and ATF proteins themselves cannot account for most of the DNA complex formation with the DR-1 core elements.
Activation of the activin ␤A promoter by c-Jun co-expression in the transfection experiments reported here (Fig. 8) may be, at least in part, achieved through C1 complex formation. This is consistent with the observation that pDR1(ϩ)⌬T/PR2(ϩ)/ CAT and pDR1(ϩ)⌬C/PR2(ϩ)/CAT could be activated by c-Jun (Fig. 9), because enhancer fragments (core-⌬T and core-⌬C, respectively) used in these constructs still retain the ability to form complex C1. The c-Jun/AP-1 is a primary nuclear target of receptor-mediated signal transduction pathways activated by extracellular ligands (56). Although the precise location is still not determined, the identification of functional c-Jun/AP-1responsive site(s) within the activin enhancer and promoter suggests that certain extracellular ligands may regulate the expression of the activin ␤A subunit gene. pDR1(ϩ)⌬C/PR2(ϩ)/CAT, which contains a mutation in the C site but an intact T2 site, directed almost background levels of expression (Figs. 9 and 10A), indicating that site C appears to be indispensable for enhancer function. In other words, T2 functions as a cis-acting DNA element only when site C is present nearby. On the other hand, pDR1(ϩ)⌬T/PR2(ϩ)/CAT, which contains a mutation in T2 but an intact C site, directed low, but nonetheless significant, levels of expression. This demonstrates that a nuclear protein binding to the C site even in the absence of a functional T2 site weakly trans-activates the activin ␤A promoter. Thus, we can speculate that T2 site binding protein(s) may take part in C2 complex formation, utilizing the C site binding factor(s) as a scaffold. There is no discrepancy between this model and the experimental observation that the core-⌬T competitor effectively disrupted the C2 complex (Fig. 7). Since the site C binding factor(s) would hold a dominant position hierarchically in complex formation in this model, a C2 complex would never form in the absence of the C site binding factor(s) (Fig. 10B). The gel shift assay result using monomeric versus dimeric AP-1 sites as competitors also supports this model (Fig. 7); tandem duplicated and properly spaced consensus AP-1 binding sites may substitute for the binding sites in C2 complex formation in the DR-1 core (Fig. 10B).
Previous investigations of the regulation of the human activin ␤A gene by various drugs, including phorbol ester (TPA) and 8-Br-cAMP, were based on assays for the biological activity or mRNA levels of activin (4,18,57). In this report, we describe the first identification of the enhancer and promoter for this gene, as well as the possible basis for the effects of these drugs on transcription from this locus. Transcriptional activity of the activin ␤A enhancer-promoter-CAT fusion constructs in HT1080 cells significantly increased after treatment with TPA or 8-Br-cAMP (Fig. 8). Hence, the increase in activin ␤A mRNA level was due, at least in part, to an increase in activin ␤A gene template activity. The discrepancy between the relative increase in levels of mRNA (19) and gene transcription rates implies that the stability of mRNA may be affected by the drug treatment.
TPA and cAMP responsiveness of several genes is mediated via the common AP-1 binding site (referred to as a TRE, or TPA-responsive element) and/or cAMP response element (CRE), respectively (47). The CREs identified in the promoters of several c-AMP-responsive genes (43,58,59) have been shown to interact with the nuclear factors CREB/ATF (60,61). A CRE-like sequence identified within the DR-1 core region (5Ј-TGATGTCA-3Ј) differs from the consensus sequence at one nucleotide (5Ј-TGACGTCA-3Ј) (Fig. 1A). However, this same sequence has been reported to be a functional CRE in the human retinoic acid receptor type ␤ gene (62). Although transfection analyses disclosed the importance of this CRE-like sequence in maintaining DR-1 enhancer activity (Figs. 6 and 9), neither competitor oligonucleotides nor monoclonal antibodies specific for CREB/ATFs interfered with the formation of complexes on the DR-1 core probe in the gel shift assay, suggesting that these complexes do not contain CREB/ATF transcription factors.
In the transfection experiments reported here, enhancerpromoter-CAT fusion constructs with mutations at TRE-like (pDR1(ϩ)⌬T/PR2(ϩ)/CAT) and CRE-like (pDR1(ϩ)⌬C/PR2(ϩ)/ CAT) sites lowered basal expression by 80 and 95%, respectively, and yet TPA and/or 8-Br-cAMP responsiveness was somewhat retained in both constructs (Fig. 9). There can be no doubt that these particular cis-acting elements play a role in maintaining normal, basal activin ␤A gene expression and are probably responsible in part for the inducibility. However, in the absence of TRE-like and CRE-like sites in each construct, other sequences must function as drug-responsive elements. One possibility is that another TRE-like site (T1), which was left intact in the site-directed mutant T2 and C constructs, may provide an alternative site through which TPA and/or cAMP inducibility might be mediated. Because AP-1 recognizes not only TRE, but also CRE, and expression of the protooncogenes c-fos and c-jun/AP-1 is induced via the cAMP pathway (63), it is possible that the expression of the activin ␤A gene by cAMP may also require the c-Jun/AP-1 system. Of special note within the context of the DR-1 enhancer, it has also been reported that a PEA-3 binding site in conjunction with the AP-1 element within an enhancer and promoter has been shown to be involved in phorbol ester induction, and that PEA-3 and AP-1 motifs function synergistically and cooperatively for transcriptional activation (42)(43)(44). The cis-acting DNA elements that are known to mediate cAMP activation also include the AP-2 element (46). Although this motif was identified in the PR-2 region (Fig. 1A), the pPR2(ϩ)/CAT construct was not responsive to 8-Br-cAMP induction (data not shown).
In this report, we show experiments that allow us to completely describe the primary structure of the human activin ␤A gene. We also detail and delineate two critical regulatory ele- The putative action of both regulatory elements on transcription (A) and binding (B) is shown. A, site C appears to be indispensable for enhancer function, because the activin ␤A promoter is weakly transactivated even in the absence of a functional T2 site, but it functions only when the site C is present nearby. B, a C2 complex would never form in the absence of the C site binding factor(s). The DR-1 core fragment with an asterisk indicates the 32 P-labeled probe. ments, the promoter and upstream enhancer, that are necessary for basal and efficient expression of the gene. Finally, we provide functional evidence that transcriptional regulation of the human activin ␤A gene by TPA and 8-Br-cAMP involves both of these 5Ј-flanking DNA sequence elements. Further study of the functional significance of the T2 site, C site, and other putative regulatory elements within and beyond the DR-1 region promises to provide insight into the exquisite control of inhibin and activin production in vitro and in vivo.