Regulation of placenta-specific expression of the aromatase cytochrome P-450 gene. Involvement of the trophoblast-specific element binding protein.

The aromatase (cytochrome P-450AROM) gene contains multiple untranslated exons I that are differentially transcribed in a tissue-specific manner. DNA sequences within the initial -301 upstream of placenta-specific exon I (exon Ia) are sufficient for placenta-specific expression of aromatase. In gel mobility shift assay, three separate domains in this region form specific binding complexes with proteins extracted from choriocarcinoma JEG-3 nuclei. A fragment containing these domains activates transcription driven by a heterologous promoter in a cell type-specific manner. Two of the binding domains that form major complexes in gel shift assay compete with each other and with a DNA fragment containing the trophoblast-specific element (TSE), which is derived from the enhancer region of the human chorionic gonadotropin alpha-subunit gene and is believed to confer placenta-specific expression of the gene. The core sequence RNCCTNNRG is sufficient for recognition of the TSE-binding protein, which is detected only in nuclear extracts prepared from placenta and choriocarcinoma. A mutation introduced in the distal TSE core in aromatase promoter resulted in marked reduction of transcriptional activity, although TSE region by itself did not show enhancer activity as that in human chorionic gonadotropin alpha-subunit gene.

Aromatase (cytochrome P-450 AROM ) is a unique member of the cytochrome P-450 superfamily. It catalyzes the conversion of androgen to estrogen, which is a rate-limiting step in estrogen biosynthesis (1). Aromatase is expressed in various cells and tissues including ovarian cells (2)(3)(4), testicular Sertoli and Leydig cells (5,6), placenta (1,7), adipose tissue (8 -10), skin fibroblasts (11), and various parts in the brain including the amygdala and hypothalamus (12,13). Recent investigations revealed that aromatase mRNAs expressed in adipose tissue, ovary, fetal liver, and brain have differences in the untranslated exon I from that reported previously for the placenta (14 -18). The use of alternative transcription start sites that occur as a consequence of employing tissue specific promoters is a unique feature of the aromatase gene expression and seems to be the underlying molecular basis for the complexity of the expression of this gene. However, how individual promoters exert their tissue specificities is not known.
The placenta is the primary site of estrogen synthesis in pregnant women and 16␣-hydroxydehydroisoandrosterone sulfate derived from the fetal adrenal and liver is the major precursor for placental estrogen production. Human aromatase mRNA expressed in placenta has a unique untranslated exon I, which is designated as exon Ia in this laboratory (16,17) or I-1 by Mahendroo and others (14,15). This type of aromatase mRNA is detected only in placenta and in cultured cells of trophoblast origin. Although cDNAs containing other types of untranslated 5Ј-region have been isolated from placental cDNA libraries (19,20), the 5Ј-end of exon Ia is the major initiation site in placenta. A choriocarcinoma cell line JEG-3 expresses aromatase mRNA of placental type (14,21). This cell line expresses aromatase in unstimulated condition as well as in response to phorbol ester and various reagents that raise intracellular cAMP levels (21,22). The aromatase activity in the cells was closely related to the content of aromatase mRNA that was very labile, suggesting transcriptional control is the main regulatory point for the enzyme activity. Thus, this cell line seems to provide a good system to investigate transcriptional regulation of placental type aromatase.
In the present paper, we describe analysis of the 5Ј-flanking region of aromatase exon Ia for the placenta-specific expression of this gene. We show that the initial Ϫ301 upstream of the exon Ia is sufficient for the basal expression of the exon. This region, conferring cell specificity, contains at least three binding domains. Two of them are recognized by the same transacting factor that binds to the trophoblast-specific element previously located in the enhancer region of human glycoprotein hormone ␣-subunit (23). A common transcription factor appears to be involved in the expression of two hormone-related genes that are characteristic of human placenta.
Reporter Gene Construction-To characterize the promoter regulatory region of the placental aromatase gene, chimeric constructs were prepared for transient expression. Initially, a 5Ј-flanking fragment (Ϫ2743 to ϩ22) was inserted into the HindIII site of pSV00CAT (24) (2.8KACAT). Deletion mutants in the 5Ј-flanking region of the fragment were generated with exonuclease III. The lengths of 5Ј-flanking region in deletion mutants were confirmed by nucleotide sequencing analysis. FB and HB fragments were generated by PCR. The obtained fragments were cloned in pBlueScript KS, cut out from the vectors, and inserted in pBLCAT2 (25). The orientation and the positions of the insertions are shown in the figures. The C2nB fragment, generated by PCR, was cloned in pBlueScript KS and then subcloned to pBLCAT2. The StuI-XhoI fragment of the clone that carries C2n mutation was replaced to wild type StuI-XhoI fragment of pFBCAT2.
Transfection of Cells-Choriocarcinoma JEG-3 cells (ATCC) were maintained in minimum essential medium supplemented with 10% fetal bovine serum (Whittaker Bioproducts), 0.1 mM non-essential amino acids, and 1 mM pyruvate. HeLa cells (ATCC) were maintained in minimum essential medium containing 0.1 mM non-essential amino acids and 10% fetal calf serum. HepG2 cells (ATCC), obtained from Riken Cell Bank (Tsukuba, Japan), were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. All of the cells were cultured at 37°C in a 95% air, 5% CO 2 humidified atmosphere.
For the transfection experiment, cells were plated at 1 ϫ 10 6 cells/ 90-mm plate 2 days before transfection. The medium was changed 4 h before transfection. 8 g of CAT plasmid and 1 g of reference plasmid RSV/Luci (26) with 0.1 mg of O,OЈ-didodecyl-N-[p-(2-trimethylammonioalkyloxy)benzoyl]-L-glutamate bromide (27) were suspended in 1 ml of culture medium without serum and kept for 30 min at room temperature. Culture dishes were washed once with 8 ml of the culture medium without serum and refed with 4 ml of the culture medium containing 5% fetal calf serum. Transfection was started by dropwise addition of the liposome-DNA suspension. After 12-16 h, the culture medium was changed to medium containing 10% fetal calf serum and gentamicin. Cells were cultured for 48 h before cell lysis.
Luciferase and Chloramphenicol Acetyltransferase Assay-The reference plasmid RSV/Luci in CAT assay expresses the luciferase protein under control of the Rous sarcoma virus promoter. Luciferase assays were performed as described by Brasier et al. (28). 10 l of lysate from each plate was added to 100 l of 25 mM glycylglycine buffer (pH 7.5), 5 mM MgCl 2 , and 5 mM ATP, and the reaction was started by injection of 100 l of 25 mM glycylglycine buffer (pH 7.5) containing 0.1 mol of luciferin and 1 mM dithiothreitol into the sample. Light output was measured for 30 s at 25°C using a Luminescence Reader BLR-301 (Aloka). A CAT assay was conducted as described (29) with samples containing equal amounts of luciferase units. The amounts of acetylated chloramphenicol on TLC plates were quantitated with a Bioimage Analyzer BA100 (Fuji Film Co.).
Preparation of Nuclear Extracts-Nuclear extracts from JEG-3, HeLa, and HepG2 cells were prepared as described by Uchiumi et al. (30) and from human placenta according to the procedure described by Gorski et al. (31). Protein concentrations were determined with a BCA protein assay kit using bovine serum albumin as a standard.
Gel Mobility Shift Assays-DNA fragments FB, FD, and HB were created by polymerase chain reaction and size selected by electrophoresis through 2% SeaPlaque GTG-agarose gels. They were end labeled with [␥-32 P]ATP (6000 Ci/mmol) by T4 polynucleotide kinase (Takara Shuzo Co.). A 32 P-C2 probe was prepared by labeling the C2 fragment with [␣-32 P]dCTP by the Klenow fragment of Escherichia coli DNA polymerase I. The nuclear extract was incubated with radioactive probe (approximately 10 kcpm) at room temperature for 30 min in a final volume of 22 l of 20 mM HEPES KOH (pH 7.9) buffer containing 2 g of poly(dI-dC) (Pharmacia Biotech Inc.), 50 mM KCl, 0.5 mM EDTA, 0.5 mM dithiothreitol, and 10% glycerol. For the competition analysis, nuclear extract was incubated for 10 min with the indicated molar excess of cold oligonucleotide prior to addition of the radiolabeled probe. The samples were then loaded on a 5% non-denaturing acrylamide gel (acrylamide:bisacrylamide ratio, 29:1) and run for 2.5 h at 4°C at 175 V in TAE buffer (25 mM Tris, 3.3 mM sodium acetate, and 1 mM EDTA). Gels were dried and exposed to Kodak-Omat AR film at Ϫ70°C in the presence of intensifying screens for 16 -48 h.

Specificity of the Promoter/Regulator Region of Exon Ia-
When various lengths of truncated 5Ј-flanking region of aromatase exon Ia up to 12 kilobases were transiently expressed in JEG-3 cells, a choriocarcinoma cell line, the shortest construct that showed the full transcription activity was a CAT construct containing Ϫ301 to ϩ22 of aromatase exon Ia (p-301ACAT). A shorter construct, p-212ACAT, exhibited 14% of full activity, and p-115ACAT had minimal transcription activity (Fig. 2). Thus, the region downstream of Ϫ301 is likely to contain the regulatory domain for basal expression of exon Ia, namely, placental type aromatase mRNA.
These three constructs and a CAT construct with the tk promoter were transfected into two other cell lines: a hepatic cell line, HepG2, and HeLa cells. In HepG2, exon Ib-type aromatase mRNA, a major form expressed in fibroblasts and fetal liver, was found. 2 HeLa cells did not express detectable amounts of aromatase mRNA under our culture and assay conditions including 30 cycles of reverse transcription-PCR. When the same luciferase units were applied to CAT assay, CAT constructs with truncated aromatase exon Ia upstream poorly activated transcription in HepG2 and HeLa cells, compared to the nonspecific promoter tk. Therefore, the regulator/ promoter activity of the 5Ј-upstream region of aromatase exon Ia was found specific to trophoblasts.
Determination of Binding Domains-For further analysis of this regulator/promoter region, labeled DNA fragments encompassing this region and nuclear extracts from JEG 3 cells were prepared for gel mobility shift assay. In Fig. 1, a schematic diagram of the promoter region of aromatase Ia is illustrated with the size and location of tested fragments in the gel mobility shift assay (start points of the CAT constructs listed in Fig.  2 are shown with arrows).
In experiments where the HB fragment was used as a radiolabeled probe (Fig. 3, panel C), two specific bands appeared as shown by the arrow and arrow head. The major band was inhibited by excess amounts of unlabeled HB, FC1, C2, and C4 FIG. 2. Transcriptional activities of truncated aromatase exon Ia 5upstream fragments. Derivatives of pSV00CAT fused to three different lengths of the 5Ј-upstream region of aromatase Ia were transfected into JEG-3, HepG2, or HeLa cells as described under "Experimental Procedures." p-301ACAT carries the Ϫ301 to ϩ22-bp fragment of the upstream region of aromatase Ia. p-212A CAT and p-115ACAT contain the Ϫ212 to ϩ22-bp fragment and the Ϫ115 to ϩ22-bp fragment, respectively. Transfection efficiency was standardized by cotransfection of the luciferase expression vector pSV/Luci. The average of the duplicates is shown. (Ϫ177/Ϫ153) (Fig. 3C, lanes 2-4 and 8). The restriction enzyme MvaI cuts the HB fragment at Ϫ166/Ϫ165, and the digested HB probe completely lost the ability to form the major band (data not shown). Both of the resulting fragments of digested HB failed to effectively compete for the binding of major bands (Fig. 3C, lanes 6 and 7). The formation of the minor band was not inhibited with C2 or C4 fragments but was inhibited by a segment of HB (Ϫ212/Ϫ167) (lane 6) and a 25-bp fragment encompassing Ϫ199/Ϫ176 (C3) (lane 5). In panel D, mutual competition between C2 and C4 is illustrated. The results of these gel mobility shift assays can be summarized as follows; two sites in the FB region, C2 and C4, bind to the same nuclear factor to form a major band and another factor, which forms minor complex binds to a distinct region between the two domains.
Transcriptional Activity of the Fragments-The two DNA fragments that contain regions for specific binding with JEG-3 nuclear extracts were linked to the CAT reporter gene upstream of a heterologous tk promoter, and transient transfection assays were carried out in JEG-3, HeLa, and HepG2 cells (Fig. 4A). FB fragment, which contains C2, C3, and C4 domains, stimulated transcription driven by tk by nearly 100-fold in JEG-3 cells. Whereas in HeLa and HepG2 cells, the extent of the stimulation was small, showing this fragment carries the tissue specificity that was observed in truncated promoter p301ACAT. The smaller fragment HB included in the FB had about 12% of the activity of pFBCAT in JEG-3 cells. The ratio of the activities of the two fragments in JEG-3 cells was comparable to that of the activity ratio of two truncated promoters, p-212ACAT and p-301ACAT (14%). The HB fragment that contains C3 and C4 domains still carries tissue specificity, being the stimulation greater in trophoblast cells, but the extent was smaller compared with that of pFBCAT2. In accordance with the cell specificity of the stimulative activity, the specific binding complex with C2 was observed only with nuclear extracts prepared from JEG-3 cells. Nuclear extract prepared from the placenta also showed specific binding complex with C2 (Fig.  4B).
DNA Sequence Required for Binding in the C2 Domain-The core DNA sequence necessary for binding of the C2 fragment was determined by competition analysis. Sequential mutations were introduced into the C2 fragment, and their competitive activities were examined with C2 or HB as radiolabeled probes. As shown in Fig. 5, four pairs of nucleotides between Ϫ290 to Ϫ282 (GG CC TC GG) were important for the binding to C2. At least one base in each of the four pairs was necessary for the recognition. These mutated C2 fragments also showed similar competitive effects on formation of the major binding complex with HB probe, which contains proximal binding domain C4. This result further shows that these two binding domains are recognized by a same trans-factor.
As we found that this sequence requirement for C2 and C4 binding was not conflicting with that of TSE, which resides in the upstream region of human chorionic gonadotropin ␣-subunit gene (23,35,36), the 24-bp segment (Ϫ182/Ϫ159), designated as TSE in the promoter of the ␣-subunit gene (37), was synthesized and examined for competition with formation of the binding of C2, C4, and HB. As shown in Fig. 6A, TSE effectively inhibited formation of the binding complex of C2, C4 (lanes 2 and 5), and the major complex with HB (lane 9). The binding factor(s) for TSE is referred to as the TSE-binding protein (TSEB) (23). Little is known about TSEB except its cell line-specific occurrence and that it cannot bind to a TSE homologue that has a C to T mutation at Ϫ172 (TSE-172) (37). The failure of this mutated TSE to compete for the binding of C2, C4, and HB (lanes 1, 6, and 10) further confirms that the binding activity to the C2 and C4 regions arises from the same factor with TSEB. As C2 and C4 are recognized by the same factor(s) TSEB, their core binding domains are designated as TSE-C arom 1 and 2, respectively. Interestingly, the minor complex formed with labeled HB was competed by both TSE and TSE-172 (indicated by an arrow) (lanes 9 and 10). In panel B, three sequences are aligned to give a consensus sequence for FIG. 4. Specificity of activation of tk-driven transcription by the fragments containing regions for specific binding and C2 binding. A, JEG-3, HepG2, and HeLa cells were transfected with tkCAT(pBLCAT2), pHBCAT, or pF-BCAT. Transfections were performed, and the activity was calculated as described under "Experimental Procedures." The average of the duplicates was shown relative to the activity of the pBLCAT2 construct of each cell. B, gel mobility shift assays were carried out with labeled C2 fragment as the probe. Nuclear extracts from either JEG-3 cells (2.7 mg, lanes 1, 2, 5, 6), HepG2 cells (3  mg, lane 3), HeLa cells (3 mg, lane 4), or placenta (3.5 mg, lanes 7, 8) were incubated for 30 min at room temperature and subjected to electrophoresis. Samples in lanes 1, 5, and 8 contained a 400-fold molar excess of unlabeled C2 fragment as a competitor.
FIG. 5. The effects of mutations introduced into the C2 fragment on competition for binding to C2 or HB. Gel mobility shift assays were carried out as described in the text with nuclear extracts from JEG-3 and radiolabeled C2 or HB. In the designated lanes, the nuclear extract was incubated for 10 min with a 200-fold molar excess of the competitor before the addition of the probe. TSEB recognition. The consensus sequence RNCCTNNRG also suffices requirement for the binding shown in Fig. 5. The observation that the C4 fragment with a mutation at Ϫ166 and Ϫ167 (CC to GG) failed to compete for the binding of C4 (data not shown) also supports this alignment.
In the last experiment transcriptional activity of this region was further examined (Fig. 7). The HB fragment that contains the TSE-C arom -2 and C3 enhanced transcription when linked to the tk promoter in both directions and placed downstream of tk. The full activity of pFBCAT was not retained when the fragment was reversely placed upstream or downstream of tk promoter. The C2 fragment only marginally activated transcription when single, double, or four copies of the 26-bp fragment were linked to the tk promoter. As the distal TSE core sequence did not enhance transcription by itself, transcriptional activity of a pFBCAT that has two G to T mutations at Ϫ283 and Ϫ282 (corresponding C2n-type mutation in Fig. 5) was examined. This mutation resulted in marked reduction of the transcription activity. DISCUSSION In the present study, we have shown that DNA sequences within the initial Ϫ301 upstream of placenta-specific exon I (exon Ia) are sufficient for basal transcription of placental type aromatase mRNA. Transcription activity directed by this promoter/regulator is cell type specific. Further deletion resulted in gradual loss of the transcriptional activity, suggesting involvement of multiple regulatory elements as shown in other systems (38). The fragment encompassing the distal half of the region has an activity to enhance transcription driven by a heterologous tk promoter, which was also cell type specific.
Gel mobility shift assay demonstrated three distinct binding domains in this region. Two separate binding domains at Ϫ300/ Ϫ274 (C2) and Ϫ177/Ϫ153 (C4) form a major complex. The mutual trans-acting factor for C2 and C4 was further found to be the same factor that recognizes the TSE, which resides in the upstream region of the human glycoprotein hormone ␣-subunit gene (Fig. 6) (23,(35)(36)(37). In human, this gene is expressed in the pituitary as subunits of glycoprotein hormones as well as ␣-subunit of chorionic gonadotropin in placenta. The placental expression of this gene has been extensively studied; it requires a multi-component enhancer composed of tandem cAMP-responsive elements and an adjacent upstream regulatory element. TSE is described as a subdomain of upstream regulatory element and is considered to be a regulatory element that confers tissue-specific expression. Although little is known about its binding protein, TSEB (23,37), the cytosine at Ϫ172 in TSE is critical for recognition of TSE sequence by TSEB. The observation that TSE with a C to T substitution at Ϫ172 also failed to compete for binding to C2 or C4 gives support to the conclusion that TSEB is the factor. The two new TSE-like sequences in the aromatase promoter and the competition analysis of their homologues with sequentially introduced mutations now reveal the sequence required for TSEB recognition, namely, the core sequence of TSE (TSE-C) as RNCCTNNRG.
In the ␣-subunit promoter, TSE is associated with another element, ␣ACT, adjacent to two tandemly repeated cAMPresponsive elements to compose a tissue-specific enhancer. TSE in the ␣-subunit promoter has little independent transcriptional activity when linked to homologous or heterologous promoters, and only when linked to the cAMP-responsive elements does it stimulate basal and cAMP-dependent expression. Likewise, the trophoblast-specific enhancer in the aromatase gene may be composed of multiple cis-elements including two FIG. 6. Consensus sequence for recognition by TSEB. A, inhibition of specific binding to C2 and C4 fragments by TSE sequences. A gel mobility shift assay was carried out as described in the text with nuclear extracts from JEG-3. The arrowhead and the arrow indicate the position of the major and minor binding complexes, respectively. B, consensus sequence for recognition by TSEB. Underlining of the C2 and C4 sequences indicates the base pairs where introduction of mutation reduced the competitor activity of the fragment. The asterisk in the TSE sequence indicates the Ϫ172 C, where mutation to T abolished the binding activity. TSE-Cs, although neither of them is associated with cAMPresponsive element or other known cis-elements. The 27-bp DNA fragment containing TSE-C arom 1 was inactive to stimulate tk-driven transcription by itself even when two or four tandem copies were introduced. The binding to TSE-C arom 1, however, seems to make an indispensable part of the aromatase enhancer. A pFBCAT2 analogue that has a mutation which eliminates TSE-C arom -1 binding loses most of the transcriptional activity that the parent pFBCAT2 shows. At present we do not have the direct evidence of the contribution of the proximal TSE-C or the C3 region to the enhancer activity. The two TSE-Cs seem to be equivalent in terms of recognition by TSEB, but their function may be different because of their position. Indeed, as shown in Fig. 7, the transcriptional activities of FB and HB fragments seem to be affected by their orientation and position in the CAT constructs.
Importance of the region between Ϫ242 and Ϫ183, which contains the C3 domain, was shown by Toda et al. (39,40) in transient transfection assays in BeWo cells. They reported that the 89-bp fragment had an enhancer activity when multiple copies were placed with a heterologous promoter. Although the binding to the C3 domain was not competed by the two TSE-C arom sequences, the TSE derived from the ␣-subunit gene competed for binding to the C3 domain, and the TSE Ϫ172 that failed to compete for TSE and TSE-C arom s is competitive for binding to the C3 region. Pittmann et al. (37) recently described a new binding protein that recognizes the region partly overlapping TSE in the ␣-subunit promoter. This transacting element binds to a distinct recognition sequence from that of TSEB and seems to be involved in placenta-specific expression of the ␣-subunit gene as well. It is therefore of particular interest whether this second upstream regulatory element binding protein is also involved in the expression of placental aromatase.
In this study, we demonstrated that TSE-like elements are located in the promoter region of placenta-specific exon I of aromatase and that at least the distal TSE-like element is transcriptionally functional. TSE-like elements in aromatase gene seem to compose a placenta-specific enhancer with other factor(s), as shown in the ␣-subunit gene, though their compositions seem to be quite different. The observation that the element recognized by TSEB, originally found in the ␣-subunit gene, also functions in the promoter regions of unrelated genes to confer placenta specificity emphasizes its functional importance as a placenta-specific transcription element. Although the ␣-subunit and aromatase genes are not closely linked in evolution, both are involved in producing hormones characteristic of the human placenta. Interestingly, the placental expressions of these genes are observed only in some mammals. Nevertheless, the fact that the two hormones are secreted from an early stage of pregnancy and play important roles in development and maintenance of the human placenta highlights the significance of a common trans-acting factor governing their expression. A potential TSE-C sequence is seen in the immediate upstream of human chorionic gonadotropin ␤-subunit genes (41), though their functional significance has not been tested.
Although various types of transcription factor have been found in trophoblasts so far, only a few are specific to this cell type (42). Recently, transcription factors of the basic helix-loophelix family that are important cell lineage determinants in many cell types have been identified with limited occurrence in extra-embryonic structures (43,44). At present, however, their immediate target genes have not been shown. TSEB is likely to be a immediate regulator of the genes in the later stage of placental differentiation. Purification of TSEB and cDNA clon-ing should clarify its identity as a trophoblast-specific transcription factor and provide the basis for further investigations of its roles and temporal and causal relationships with other factors that determine the differentiation to trophoblast lineage.