AP-2γ and the Homeodomain Protein Distal-less 3 Are Required for Placental-specific Expression of the Murine 3β-Hydroxysteroid Dehydrogenase VI Gene, Hsd3b6

The enzyme 3β-hydroxysteroid dehydrogenase/isomerase (3β-HSD) is essential for the biosynthesis of all active steroid hormones. It exists as multiple isoforms in humans and rodents, each the product of a distinct gene. Human 3β-HSD I in placenta is essential for placental progesterone biosynthesis and thus is essential for the maintenance of pregnancy. The murine ortholog, 3β-HSD VI, is the only isoform expressed in giant trophoblast cells during the first half of mouse pregnancy. This study was designed to identify the cis-acting element(s) and the associated transcription factors required for trophoblast-specific expression of 3β-HSD VI. Transfection studies in placental and nonplacental cells identified a novel 66-bp trophoblast-specific enhancer element located between −2896 and −2831 of the 3β-HSD VI promoter. DNase protection analysis of the enhancer element identified three trophoblast-specific binding sites, FPI, FPII, and FPIII. Electrophoretic mobility shift assays with oligonucleotides representing the protected sequences, FPI and FPIII, and nuclear extracts isolated from human JEG-3 cells and from mouse trophoblast cells, demonstrated the same binding pattern that was distinct from the binding pattern with mouse Leydig cell nuclear proteins. Further electrophoretic mobility shift assays identified AP-2γ and the homeodomain protein, Dlx 3, as the transcription factors that specifically bind to FPI and FPIII, respectively. Site-specific mutations in each of the binding sites eliminated enhancer activity indicating that AP-2γ and Dlx 3, together with an additional transcription factor(s) that are conserved between humans and mice, are required for trophoblast-specific expression of 3β-HSD VI.

(3␤-HSD) 1 is essential for the biosynthesis of all active steroid hormones: the adrenal steroid hormones, cortisol, corticosterone, and aldosterone; the testicular steroid hormone, testosterone; and the ovarian and placental hormones, progesterone and estradiol. The 3␤-HSD enzyme exists as multiple isoforms in humans and rodents, each the product of a distinct gene (1). To date, six isoforms have been identified in mouse and two in human. These isoforms are expressed in a tissue-and temporal-specific manner (1)(2)(3)(4)(5). In the mouse, the two major isoforms involved in steroid hormone biosynthesis are 3␤-HSD I and 3␤-HSD VI. 3␤-HSD I is the major or only isoform expressed in gonads and adrenal glands, whereas 3␤-HSD VI is the only isoform expressed in giant trophoblast cells during mid-pregnancy (3). The orthologous isoforms in human are human 3␤-HSD II, the isoform expressed in the gonads and adrenal glands (6), and human 3␤-HSD I (7,8), the only isoform expressed in placenta throughout pregnancy. The expression of human 3␤-HSD I in placenta is essential for placental progesterone biosynthesis and, thus, is vital for maintenance of pregnancy (9). Consistent with the role of human 3␤-HSD I in placental progesterone biosynthesis, we have shown that mouse 3␤-HSD VI is required for progesterone biosynthesis in giant trophoblast cells between embryonic day (E) 9.5 and E10.5 (10).
Progesterone biosynthesis from cholesterol requires the activity of two enzymes, cholesterol side chain cleavage cytochrome P450 (P450scc) which catalyzes the conversion of cholesterol to pregnenolone and 3␤-HSD which catalyzes the conversion of pregnenolone to progesterone. This latter step is brought about by distinct tissue-specific isoforms of 3␤-HSD. Previous studies designed to identify placental-specific regulatory elements in the human placental-specific 3␤-HSD promoter were unsuccessful (8). Moreover, identity of transcription factors essential for placental-specific expression of P450scc remains to be resolved (11). Therefore, the question is whether there is a unique tissue-specific transcription factor or factors required for the expression of the trophoblast-specific isoform of 3␤-HSD as well as the other enzyme required for progesterone biosynthesis, P450scc, in human placenta and mouse giant trophoblast cells.
In the present study, we identify a 66-bp trophoblast-specific enhancer element located between Ϫ2896 and Ϫ2831 5Ј of the transcription start site of the murine 3␤-HSD VI promoter and demonstrate the requirement for two transcription factors, AP-2␥, and the homeodomain protein, Distal-less 3 (Dlx 3), in determining trophoblast-specific expression of the 3␤-HSD VI gene. Dlx 3 is a member of a family comprising at least six Dlx genes sharing a homeobox sequence similar to that found in the Drosophila Distal-less gene (12). Dlx proteins are transcriptional activators which play an essential role during vertebrate development (12). AP-2␥ is a member of a family of three closely related and evolutionary conserved sequence-specific DNA-binding proteins which include AP-2␣, AP-2␤, and AP-2␥ (13). Dlx 3 and AP-2␥ are expressed in both murine and human placental trophoblast cells and are required for the placentalspecific expression of a number of genes (14 -18). Our studies demonstrate that these two transcription factors regulating the trophoblast-specific expression of the steroidogenic enzyme, 3␤-HSD VI, are conserved in both murine and human trophoblast cells. This is the first report on the identification of at least two of the transcription factors required for placentalspecific expression of 3␤-HSD in humans and mice.

EXPERIMENTAL PROCEDURES
Isolation of Genomic Clones-A phage 129/SvJ mouse genomic library from Stratagene (La Jolla, CA) was screened with an 180-bp probe labeled with [␣-32 P]dCTP, which includes 18 bp of coding region and 167 bp of 3Ј-untranslated region of the 3␤-HSD VI cDNA (3). Prehybridization and hydribization were performed as described (19). Two clones (402 and 602) were identified to represent 3␤-HSD VI. The clones were subjected to restriction enzyme and Southern blot analysis using exon-specific oligonucleotide probes (exon 2, 5Ј-CCAGAGGATT-GTCCAGTTG-3Ј; exon 3, 5Ј-GACATCTAGGATGGTCTG-3Ј; exon 4, 5Ј-AGGAAGCTCACAGTTTCCA-3Ј). Restriction fragments from clone 402 were subcloned into the pBluescript-KS vector and subjected to further analysis to identify the location of exon 1 and to establish the start site of transcription.
Preparation of Giant Trophoblast Cells-Timed pregnant C57BL/6J mice (Jackson Laboratory, Bar Harbor, ME) were killed by CO 2 followed by cervical dislocation at E9.5 and E10.5, and the uterine horns were removed and placed in phosphate-buffered saline for further isolation of giant trophoblast cells as described (5). E9.5 and E10.5 implantation sites were separated from the myometrium and the embryos were removed. Giant trophoblast cells were gently removed by scraping the inner face of the hemi-implantation site and collected for preparation of RNA or proteins.
Primer Extension-The transcription initiation site of 3␤-HSD VI was determined using the primer extension kit from Promega (Madison, MI). Messenger RNA was isolated from E10.5 mouse giant trophoblast cells and adult mouse testes using the QuickPrep Micro mRNA Purification Kit (Pharmacia, Piscataway, NJ). Primer GSP3 (5Ј-GGT-TCTGATCTCTGCAAAGGAACCAG-3Ј, 132 bp 5Ј of exon 1 end), which is specific for 3␤-HSD VI, was end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase. Approximately 100 fmol of labeled primer and 3-5 g of mRNA were hybridized at 65°C for 20 min, followed by gradual cooling to room temperature. The reactions were extended by avian myeloblastosis virus reverse transcriptase at 42°C for 1 h and 15 min and followed by heating at 99°C for 5 min. The extended DNA fragments were precipitated with ethanol and subjected to electrophoresis in a 6% denaturing polyacrylamide gel.
Construction of Plasmids-To characterize the promoter region of the mouse 3␤-HSD VI gene, a series of 5Ј deletions spanning between Ϫ4700 and Ϫ40 (Fig. 3) were subcloned into a promoterless luciferase reporter vector pA3LUC at the SmaI-HindIII sites (21). To make heterologous constructs, different fragments, including Ϫ3004/Ϫ1989, Ϫ3004/Ϫ2500, Ϫ3004/Ϫ2723, and Ϫ2722/Ϫ2500, were subcloned into a heterologous thymidine kinase promoter-driven luciferase reporter vector TK164LUC at the SmaI site. Further deletions within the Ϫ3004/ Ϫ2723 fragment, including sequences Ϫ2896/Ϫ2725, Ϫ2830/Ϫ2725, and Ϫ2896/Ϫ2831, were generated by PCR using primers with SmaI site added at the 5Ј end and subcloned into TK164LUC/SmaI. The sequences Ϫ2896/Ϫ2857 and Ϫ2867/Ϫ2831 were synthesized with a half SmaI site added in both 5Ј and 3Ј ends and subcloned into TK164LUC/SmaI. The PCR fidelity and orientations of inserts of each construct were verified by restriction enzyme digestion and sequencing.
Mutagenesis in the Enhancer Region-Mutations in the potential AP-2, TEF, or Dlx 3-binding sites in each footprint (FPI, FPII, and FPIII) ( Table I) that were identified by DNase I footprinting were introduced by the ExSite PCR-based site-directed mutagenesis kit (Stratagene) following the instructions given in the manual. Positive clones were identified and verified by sequencing in the DNA Sequencing Facility at Stanford University.
Preparation of Nuclear Extracts-Crude nuclear extracts from JEG-3 and MA-10 cells were prepared as described (24). Nuclear extracts from E10.5 giant trophoblast cells and E15.5 placental tissues were isolated using the 1-h minipreparation techniques (25). Total protein was quantitated by Bradford assay and normalized against extraction buffer. The extracts were aliquoted and stored at Ϫ70°C until use for footprinting or EMSA.
DNase I Footprinting Assay-DNase I footprinting was carried out using Promega's core footprinting kit with modifications. To generate the probe, the 120-bp fragment (Ϫ2916/Ϫ2800) amplified from PCR was subcloned into the pBluescript-KS vector at the EcoRI-HindIII sites. The insert was sequenced for verification. A 167-bp/XbaI-XhoI fragment released from the vector was labeled with T4 polynucleotide kinase and [␥-32 P]ATP. Digestion with AccI or SpeI resulted in formation of a sense or antisense probe, respectively. Increasing amounts of crude nuclear extract were incubated with the labeled probe (10,000 cpm) at room temperature for 10 min, followed by digestion with 6 units of DNase I at room temperature for 3 min. The digested products were extracted with phenol/chloroform (1:1), ethanol-precipitated, and analyzed using 5% denaturing polyacrylamide gel electrophoresis.
Electrophoretic Mobility Shift Assay (EMSA)-EMSAs were performed as described previously (26). The 20-l binding reaction containing 5-10 g of crude nuclear extracts and 0.05-0.25 ng of radiolabeled probes (50,000 cpm/reaction) was incubated for 20 min at room temperature. The sequences of the double-stranded oligonucleotides used for EMSAs are listed in Table I. For competition assays, a 50 -500-fold molar excess of unlabeled oligonucleotides was added to the binding reaction mixture. For supershift assays, 2 g of polyclonal antisera to AP-2␣, AP-2␥, Nkx2-5 (Santa Cruz Biotechnology, Santa Cruz, CA), Dlx 3 (a gift of Dr. Mark Roberson), or preimmune sera were added to the mixture prior to the addition of the probe. The binding reactions were resolved on a 5% nondenaturing polyacrylamide gel.
Western Blot Analysis-Extracts of MA-10 cells, transfected COS-7 GTACTTGGTGTAATTACCAT AP2␥ and Dlx3 Determine Placenta-specific Expression of 3␤-HSD cells, E9.5 or E10.5 giant trophoblast cells, or adrenal glands or testes from 50-day-old mice were prepared by homogenization in extraction buffer followed by centrifugation as described previously (3). The supernatant was subjected to SDS-PAGE and Western blot analysis. Membranes were first incubated with antiserum generated against human placental 3␤-HSD and then with the horseradish peroxidaselabeled secondary antibody and exposed using the Enhanced ChemiLuminescence kit (Amersham Biosciences, Inc., Arlington Heights, IL).
In Situ Hybridization-Cryosections (8 m) of implantation sites from E9.5 and E10.5 were subjected to in situ hybridization as described (22) using a 359-bp 35 S-labeled sense or antisense cRNA probe representing 45 bp from the 3Ј end of the coding region and 314 bp from the 3Ј-untranslated region of 3␤-HSD VI cDNA (3). Slides were exposed at 4°C and developed after 4 days. The slides were stained with hematoxylin and eosin and examined under light microscopy with bright-and dark-field illumination.

RESULTS
Genomic Structure of the 3␤-HSD VI Gene-To characterize the 3␤-HSD VI gene, a phage genomic DNA library was screened with a 3␤-HSD VI cDNA probe. Two phage clones, 402 and 602, were identified as encompassing the entire 3␤-HSD VI gene. Restriction mapping and Southern blot analysis revealed that clone 402 contains ϳ10 kb of 5Ј-flanking region, and exon 2 through exon 4 ending at the SacI site in the 3Ј-untranslated region (3), whereas clone 602 contains a portion of intron 2, exon 3, exon 4, and 9 kb 3Ј-flanking region (Fig. 1A).
Identification of Exon 1 Sequence and Mapping of the Transcription Start Site of the 3␤-HSD VI Gene-5Ј-RACE analysis was performed to identify the exon 1 sequence. The 3Ј 340-bp sequences from the 5Ј-RACE PCR product matched with known cDNA sequences including 206 bp of exon 2 and 134 bp of exon 3 (3). To confirm that the new sequence from the 5Ј-RACE represents the exon 1 sequence, reverse transcriptase-PCR was performed using RNA isolated from E10.5 mouse giant trophoblast cells and from mouse testicular tissues with a forward primer designed from the 5Ј-RACE exon 1 sequence and a backward primer designed from the known sequence of exon 2 (3). A fragment of the expected size was obtained. Thus, exon 1 was identified and the size of intron 1 was established to be 3.1 kb, which differs from human 3␤-HSD I or mouse 3␤-HSD I (1).
The transcription start site for 3␤-HSD VI was confirmed by primer extension (Fig. 1B). Using mRNA from adult mouse testes and from E10.5 mouse giant trophoblast cells, a single transcription start site was identified at a guanosine residue 315 bp 5Ј of the translation initiation codon in exon 2 which defines the size of exon 1 as 252 bp.
Mouse Trophoblast-specific Expression of 3␤-HSD VI mRNA and Protein-We previously reported the expression of 3␤-HSD mRNA and protein in mouse giant trophoblast cells at E9.5 (5). However, the earlier study did not use isoform-specific probes for the identification of 3␤-HSD. To determine the isoformspecific expression of 3␤-HSD VI, in situ hybridization of sections of E9.5 and E10.5 implantation sites were analyzed using a 3␤-HSD VI-specific antisense probe (see "Experimental Procedures"). Fig. 2A shows exclusive expression of 3␤-HSD VI mRNA in giant trophoblast cells at E9.5 and E10.5 with considerably greater expression at E10.5. No expression of 3␤-HSD VI mRNA was observed in decidua or embryo. The increase in 3␤-HSD VI mRNA in giant trophoblast cells at E10.5 is accompanied by a parallel increase in 3␤-HSD VI protein as determined by Western blot analysis (Fig. 2B).
SF-1 Is Not Required for Expression of 3␤-HSD VI in JEG-3 Cells-In gonads and adrenal glands expression of steroidogenic enzymes is dependent on steroidogenic factor 1 (SF-1) (27,28). SF-1 null mice develop normal placental trophoblast cells that express steroidogenic enzymes (29) indicating that this factor is not involved in placental-specific expression of steroidogenic enzymes. Our studies (Table II) demonstrate that the transcriptional activity of the murine 3␤-HSD VI proximal promoter transfected into JEG-3 cells that do not contain SF-1, is 22-fold greater than the transcriptional activity of the gonadal-and adrenal-specific 3␤-HSD I promoter. Furthermore, co-transfection with an SF-1 expression vector resulted in a dose-dependent increase in 3␤-HSD I transcriptional activity, but had no effect on 3␤-HSD VI transcriptional activity. Thus, SF-1 is not required for expression of the 3␤-HSD VI enzyme in the placenta.
Promoter Activity and Cell-specific Expression of the 5Ј-Flanking Region of the 3␤-HSD VI Gene-Because of the lack of a mouse trophoblast cell line and the difficulty in obtaining mouse primary giant trophoblast cells for transfection studies, the human placental choriocarcinoma cell line, JEG-3, was chosen for the transfection studies described herein. JEG-3 cells express human 3␤-HSD I (30), the tissue-specific ortholog to mouse 3␤-HSD VI. To establish trophoblast-specific expression, nonplacental cell lines, both mouse Leydig tumor cells (MA-10) and monkey kidney cells (COS-7), were also used for the promoter and enhancer analyses of the mouse 3␤-HSD VI gene.
To characterize the promoter sequences involved in transcriptional regulation of the 3␤-HSD VI gene in trophoblast cells, a series of 5Ј deletions of the 3␤-HSD VI gene, spanning from Ϫ4700 to Ϫ40 5Ј of exon 1 (Fig. 3), were subcloned into a promoterless luciferase reporter vector, pA3LUC, and transiently transfected into JEG-3, MA-10, or COS-7 cells. As shown in Fig. 3, a marked increase in basal transcription activity was observed in all three cell lines when the promoter sequence was increased from Ϫ40 to Ϫ91, suggesting that a common positive cis-acting element is located between Ϫ40 and Ϫ91, although the luciferase activity was ϳ10-fold greater in JEG-3 and MA-10 cells compared with COS-7 cells. A database analysis of this sequence shows that there are several potential binding sites for transcription factors CREB and Sp1 in the segment between Ϫ40 and Ϫ91. No further analysis was performed with this proximal sequence.
As larger fragments from Ϫ91 to Ϫ4700 were included, the basal transcription activities remained essentially the same in both MA-10 and COS-7 cells. In contrast, a large increase in basal activity was observed in JEG-3 cells with fragments greater than Ϫ1989. The results suggest that the sequence between Ϫ3004 and Ϫ1989 may contain a trophoblast-specific enhancer element.
Analysis of Enhancer Activity and Characterization of the Enhancer Region-To determine whether the sequence between Ϫ3004 and Ϫ1989 has the properties of an enhancer, this region was subcloned 5Ј of the heterologous thymidine kinase (tk) promoter either in the sense or antisense direction and transfected into JEG-3, MA-10, and COS-7 cells. Results of the transfections are shown in Fig. 4. The sequence between Ϫ3004 and Ϫ1989 increased tk promoter activity over 8-fold in either the sense or antisense directions in JEG-3 cells, but not in COS-7 or MA-10 cells. These data indicate that the sequence between Ϫ3004 and Ϫ1989 has the characteristics of a trophoblast-specific enhancer element. Because a further increase in transcriptional activity was observed between Ϫ4700 and Luciferase activity was normalized to ␤-galactosidase activity and expressed relative to the promoterless vector pA3LUC, whose activity is set as 1. Each value represents the mean Ϯ S.E. of three separate transfections, each performed in triplicate. The inset illustrates the relative luciferase activity between Ϫ40 and Ϫ91 in the three cell lines (ϫ20).
Ϫ3004 (Fig. 3), this 1700-bp fragment was also subcloned 5Ј of the tk promoter and transfected into JEG-3 cells. No increase in the promoter activity was observed with this fragment (data not shown).
To further define the regulatory domains within the Ϫ3004/ Ϫ1989 fragment of the 3␤-HSD VI promoter, subsequent deletions were made and subcloned 5Ј of the tk promoter. Fig. 5A shows the results of the deletion constructs transfected into JEG-3, MA-10, and COS-7 cells. No enhancer activity with any of the deletion constructs was observed in MA-10 or COS-7 cells. In contrast, in JEG-3 cells, deletions of the 3Ј sequence from Ϫ1989 to Ϫ2500, did not change enhancer activity compared with the Ϫ3004/Ϫ1989 fragment. Further deletion from Ϫ2500 to Ϫ2723 displayed similar enhancer activity. The data suggest that the 282-bp fragment between Ϫ3004 and Ϫ2723 contains the trophoblast-specific enhancer element.
To characterize the minimal enhancer region, additional 5Ј deletions within Ϫ3004/Ϫ2723 were made and subcloned 5Ј of the tk heterologous promoter. The transfection results are shown in Fig. 5B. When the 5Ј sequence was deleted from Ϫ3004 to Ϫ2896, the activity remained the same as the Ϫ3004/ Ϫ2723 construct. However, further 5Ј deletion to Ϫ2830 reduced transcriptional activity by 73%, suggesting that the enhancer element is located between Ϫ2896 and Ϫ2830. To confirm that this 66-bp fragment between Ϫ2896 and Ϫ2831 has trophoblast-specific enhancer activity, it was directly subcloned 5Ј of the tk heterologous promoter. Surprisingly, this 66-bp fragment increased tk promoter activity ϳ3-fold relative to the Ϫ3004/Ϫ2723 fragment (Fig. 5B). This finding not only provides conclusive evidence that the sequence (Ϫ2896/Ϫ2831) contains the enhancer element, but also suggests that an inhibitory element is located between Ϫ2830 and Ϫ2723 as removal of this 108-bp fragment from the Ϫ3004/Ϫ2723 region resulted in a significant increase in enhancer activity. Further 3Ј or 5Ј deletions of the Ϫ2896/Ϫ2831 fragment resulted in complete loss of the enhancer activity. The data suggest that the entire sequence between Ϫ2896 and Ϫ2831 is required for trophoblast-specific enhancer activity.
Potential Transcription Factor-binding Sites-To determine the corresponding protein-binding nucleotides within the 66-bp enhancer, DNase I footprinting assay was performed using an end-labeled fragment containing the sequence Ϫ2916/Ϫ2872 (including the 66-bp fragment identified above as the trophoblast-specific enhancer element). Three protected regions, FPI, FPII, and FPIII, were observed on both antisense (data not shown) and sense strands in JEG-3 cells (Fig. 6A), but not in MA-10 cells (data not shown). A database search (TRANSFAC) for potential transcription factors revealed that the first region (FPI) protected a sequence between Ϫ2885 and Ϫ2876 which contains a potential Ker1-binding site (GGCTGTAGCC) (31) (Fig. 6B). Footprint FPII between Ϫ2869 and Ϫ2857 encompassed a predicted TEF site (CGCATTCTA) (32). A perfect binding site for GATA (WGATAR) between Ϫ2853 and Ϫ2841 (33) or a potential binding site for Nkx2-5 (TAATT) between Ϫ2850 and Ϫ2846 (34) was protected in the third footprint (FPIII).
To determine whether the sequence representing each footprint interacts with the predicted factor present in trophoblast cells, EMSAs were performed using double-stranded oligonucleotides corresponding to each of these footprints as probes FIG. 4. The sequence between ؊3004 and ؊1989 of the 3␤-HSD VI promoter contains the JEG cell-specific transcriptional element. The fragment comprising the sequence between Ϫ3004 and Ϫ1989 was subcloned in either sense or antisense orientation 5Ј of the heterologous tk promoter-driven luciferase reporter vector (TK164LUC). The reporter constructs were transfected into the three cell lines and luciferase activity of each construct was expressed relative to the vector TK164LUC. Each value represents the average plus the range of two separate transfections, each performed in triplicate.

FIG. 5. Identification of the trophoblast-specific minimal enhancer element.
A, 5Ј or 3Ј deletions within the sequence between Ϫ3004 and Ϫ1989 (as illustrated on the left) were subcloned 5Ј of the tk promoter in the TK164LUC vector and transfected into the three cell lines. Luciferase activity of each construct is expressed relative to the vector TK164LUC. B, further 5Ј or 3Ј deletion mutants were constructed using the 282-bp fragment (Ϫ3004/Ϫ2723) identified in A as a backbone. The constructs were transfected into JEG-3 cells. Luciferase activity of each construct is expressed relative to the Ϫ3004/Ϫ2723 construct which displayed full enhancer activity. The activity of the Ϫ3004/Ϫ2723 construct was set at 1. Each value represents the average plus the range of two separate transfections, each performed in triplicate. and nuclear extracts isolated from JEG-3 cells, E10.5 giant trophoblast cells, or MA-10 cells. The probe representing FPI or the probe representing FPIII, each gave rise to a specific DNAprotein complex. No binding activity was observed with the probe representing FPII (data not shown).
Identification of AP-2␥ as the Trophoblast-specific Transcription Factor That Binds to FPI-The FPI element forms a specific DNA-protein complex (complex I) with identical mobility with nuclear extracts of either JEG-3 or giant trophoblast cells (Fig. 7B, lanes 2 and 5) which was not observed with the nuclear extract of MA-10 cells (lanes 8 -10). Five hundred-fold molar excess of unlabeled FPI competed for binding of the labeled FPI probe (Fig. 7B, lanes 3 and 6), whereas an unlabeled FPI oligonucleotide containing mutations within the potential Ker1 site (mFPI) could not compete (lanes 4 and 7). This observation suggests that the FPI element may interact with the Ker1-binding protein. To test this hypothesis, labeled probes representing FPI and Ker1 were subjected to EMSAs using JEG-3 cell nuclear extract. Fig. 7C demonstrates that each of these probes yielded a DNA-protein complex with identical mobility and, furthermore, 500-fold molar excess of either unlabeled FPI or unlabeled Ker1 competed for binding of each of the probes. Binding of the JEG-3 nuclear protein to the Ker1 probe is greater than to the FPI probe (Fig. 7C, lanes 1 and 4) and competition with unlabeled FPI for binding to the Ker1 probe is somewhat less than the competition observed with unlabeled Ker1 (Fig. 7C, lanes 5 and 6). These results suggest that FPI and Ker1 bind the same trophoblast nuclear protein with different affinities.
The protein that specifically binds to Ker1 has previously been identified as AP-2 (31). The AP-2 family consists of three distinct proteins, AP-2␣, AP-2␤, and AP-2␥. AP-2␥ is the predominant AP-2 family member expressed in mouse placenta and giant trophoblast cells (15). Both AP-2␣ and AP-2␥ are expressed in JEG-3 cells and in human placenta (16,17). To identify whether the protein that specifically binds to FPI is a member of the AP-2 transcription factor family and, furthermore, to distinguish between AP-2␣ and AP-2␥, EMSAs were performed with the FPI probe and nuclear extracts of JEG-3 cells, of E10.5 giant trophoblast cells, or of E15.5 mouse placentas in the presence or absence of polyclonal antisera to AP-2␣ or AP-2␥. As shown in Fig. 8A, mouse placental nuclear proteins formed a DNA-protein complex (lane 4) with identical mobility to complex I as shown in Fig. 7 for JEG-3 cell and mouse giant trophoblast cell nuclear extracts. This complex was displaced by the addition of antiserum to AP-2␥ (lanes 3, 6, and 10) but not by the addition of AP-2␣ antiserum (lanes 7 and 11) or by normal rabbit serum (NRS, lane 8). The results shown in Fig. 8B illustrate that the JEG-3 cell nuclear protein that specifically binds to the Ker1 probe also is AP-2␥ and not AP-2␣. The data presented in Fig. 8 establish that the human and murine trophoblast-specific protein that interacts with FPI is the transcription factor AP-2␥.
Identification of Distal-less 3 (Dlx 3) as the Trophoblastspecific Transcription Factor That Binds to FPIII-As described above, a database search for transcription factors involved in producing FPIII in the DNase I protection assay (Fig.  6) indicated two potential binding sites, GATA or Nkx2-5. These two potential sites overlapped. To characterize binding of JEG-3 and giant trophoblast nuclear proteins to the FPIII sequence, EMSAs were performed using a radiolabeled probe comprising the entire protected sequence as well as an FPIII mutant as competitor (Fig. 9A). A specific DNA-protein complex (complex III) with identical mobility was observed with nuclear extracts of both JEG-3 and giant trophoblast cells (Fig.  9A, lanes 2 and 5), which was distinct from the complex formed with MA-10 cell nuclear extract (Fig. 9A, lane 8). This finding indicates that the FPIII-binding nuclear protein is trophoblastspecific. The competition assay with mFPIII which contains mutations comprising the entire GATA site, as well as the 5Ј nucleotides of the Nkx2-5-binding site, resulted in loss of competition with the FPIII probe (Fig. 9A, lanes 4 and 7). To delineate the recognition sequence for the trophoblast nuclear protein within the FPIII element, a series of FPIII oligonucleotides with sequential double mutations were tested for their ability to compete with the FPIII probe in an EMSA (Fig. 9B). As illustrated in Fig. 9B, oligonucleotides containing mutations involving the nucleotides as represented by m3, m4, m5, m7, m8, and m9 lost or markedly reduced the capacity of these mutants to compete with the wild type FPIII probe indicating that the sequence TAATTG from Ϫ2848 to Ϫ2843 was the critical binding site for the FPIII-binding protein. The TAATT sequence is identical to the binding site for the murine homeodomain protein, Nkx-2.5, or the current nomenclature, Nkx2-5 (34,35), and to the binding site of another homeodomain protein Dlx 3. Roberson et al. (18) recently demonstrated that Dlx 3 is the transcription factor that binds to the junctional regulatory element (JRE) of the human glycoprotein hormone ␣ (hCG␣) subunit gene and is required for basal placental-specific expression of this gene (18) (Fig. 10A). This observation suggests that Dlx 3 may be the human and murine trophoblast-specific nuclear protein that binds to FPIII. To examine whether Dlx 3 is the FPIII-binding protein, EMSAs were carried out with radiolabeled probes FPIII and JRE (18) and JEG-3 cell nuclear extract (Fig. 10B). The FPIII-protein complex (lane 1) displayed identical mobility to the JRE-protein complex with JEG-3 cell nuclear extract (lane 8). In addition, both unlabeled FPIII and JRE oligonucleotides had similar self-and cross-competition with each of the probes for binding to the trophoblast nuclear protein (lanes 2-7 and  9 -14). These results are consistent with Dlx 3 being the transcription factor that forms complex III. To establish that the trophoblast-specific protein that binds to FPIII is Dlx 3, EM-SAs were performed with the FPIII probe and the JER probe, nuclear extracts of JEG-3 cells (Fig. 10C) or of mouse placentas, FIG. 6. DNase I protection analysis of the trophoblast-specific minimal enhancer element. A, a fragment of the 3␤-HSD VI gene from Ϫ2916 to Ϫ2800 was end-labeled on the sense strand, incubated with increasing amounts of crude nuclear extract of JEG-3 cells (from left to right, 30-120 g) and subjected to DNase I digestion. The extent of each footprint is indicated on the right. B, nucleotide sequences of the minimal enhancer region with the nucleotides comprising each footprint are indicated by the bar above the sequence. The putative transcription factor binding sites found within each footprint are marked by a line above or below the appropriate sequence. The letters in italic indicate mismatches from the consensus sequences. (Fig. 10D), specific antisera to Dlx 3 or to Nkx2-5 or preimmune antiserum. The results illustrated in Fig. 10, C and D, demonstrate that Dlx 3 is the nuclear protein in both human trophoblast cells (JEG-3) and mouse placentas that forms a DNAprotein complex with the FPIII or the JER element. Although, incubation with JEG-3 cell nuclear extract and a radiolabeled oligonucleotide containing an Nkx2-5-binding site yielded a DNA-protein complex with identical mobility to the DNA-protein complex formed with the FPIII probe (data not shown), the addition of antiserum to Nkx2-5 did not result in either a supershift or disruption of the DNA-protein complex (Fig. 10, C  and D, lanes 2 and 6).
All Three FP Elements within the 66-bp Enhancer Are Essential for 3␤-HSD VI Transcriptional Activity in JEG-3 Trophoblast Cells-To determine whether the binding sites within each FP are functional, site-directed mutations were introduced into the heterologous enhancer-tk construct Ϫ2896/ Ϫ2831 (Fig. 5B). To determine the requirement for the AP-2␥binding site identified in FPI, the mutation, TAGGCA 3 GGTACC (mFPI), which had been shown to disrupt binding of AP-2␥, was introduced (Fig. 7A). For disruption of the Dlx 3-binding site in FPIII, TGATAA 3 AAGCTT (mFPIII) was introduced into the enhancer construct and the potential TEFbinding site in FPII was mutated (GCATTC 3 AAGCTT) (Table I). Transfection into JEG-3 cells of each of the mutated 66-bp enhancer-tk constructs resulted in complete loss of enhancer activity (Fig. 11). These results indicate that all three proteins, AP-2␥, FPII-binding protein, and Dlx 3 homeodomain protein, are essential for determining trophoblast-specific enhancer activity. The results do not eliminate the possible requirement for additional proteins. DISCUSSION Progesterone biosynthesis in the human placenta is essential for maintenance of pregnancy (9). Although progesterone is required for maintenance of pregnancy in the mouse, the source has not been unequivocally established. The biosynthesis of progesterone from cholesterol requires two steroidogenic enzymes, P450scc and 3␤-HSD. P450scc is the product of a single gene that is expressed in human gonads and adrenal glands (36), the human placenta (37), mouse decidua, and giant trophoblast cells (5). Unlike P450scc, 3␤-HSD exists in multiple isoforms and the isoform expressed in human and mouse trophoblast cells is distinct from the isoform expressed in the gonads and adrenal glands, with one exception, mature mouse Leydig cells express 3␤-HSD VI in addition to the major gonadal isoform, murine 3␤-HSD I (Fig. 2B and Refs. 3 and 4). Both P450scc and 3␤-HSD VI are expressed in a tissue-and temporal-specific manner during the first half of mouse pregnancy (Fig. 2) (5). Several investigations have been undertaken during the past few years in the search for a factor or factors which determine placental-specific expression of steroidogenic enzymes. Guerin et al. (8) were unsuccessful in their attempt to identify a placental-specific element in the promoter of human 3␤-HSD I. They did identify a strong positive regulatory element in the first intron which functioned in a ubiquitous man- FIG. 7. Specific binding activity of FPI with trophoblast cell nuclear proteins. A, sequences in the oligonucleotides representing FPI, Ker1, an AP-2 binding element in the human k14 keratin gene (31) and the mutated FPI (mFPI) which represents a five-nucleotide substitution mutation in the potential Ker1 site within the FPI sequence. The consensus nucleotides for AP-2 binding are illustrated in bold. B, EMSA with 10 g of nuclear protein isolated from JEG-3 cells, or E10.5 mouse giant trophoblast cells (GTC), or MA-10 cells and radiolabeled probe FPI. Lane 1 is free of nuclear extract. C, EMSA with JEG-3 nuclear extract and with probes FPI or Ker1. ner. This element had sequence similarities to the positive nontissue-specific element identified in the murine 3␤-HSD VI promoter between Ϫ40 and Ϫ90. The inability by Guerin et al. (8) to identify a placental-specific element in human 3␤-HSD I, the ortholog of murine 3␤-HSD VI, may have been due to the limited length of the promoter examined. A recent report by Huang and Miller (11) described two transcription factors related to human immunodeficiency virus-inducible LBP proteins, LBP-1B and LBP-9, that had the capacity to modulate placental-specific expression of human P450scc but were not involved in determining the placental-specific expression of this enzyme. Thus, the identification of the factors that regulate the tissue-specific expression of the steroidogenic enzymes necessary for progesterone biosynthesis in placenta and giant trophoblast cells remains to be established.
In the current study, we identify a 66-bp placental/trophoblast-specific enhancer element located between Ϫ2896 and Ϫ2831 of the murine 3␤-HSD VI promoter. DNase I protection analysis of the enhancer element identified three trophoblastspecific binding sites, referred to as FPI, FPII, and FPIII. The protein that binds to FPI was established to be a member of the AP-2 family of transcription factors. EMSA studies using specific antisera to AP-2␣ and AP-2␥ demonstrated that the protein binding to FPI is AP-2␥ and not AP-2␣. Furthermore, identical results were obtained with nuclear extracts of human placental JEG-3 cells, of mouse placentas or of mouse giant trophoblast cells, which demonstrate that the transcription factor that binds to FPI and is involved in trophoblast-specific expression of murine 3␤-HSD VI is the same in humans and mice. Human placenta and JEG-3 cells express both AP-2␣ and AP-2␥ (15,16), while AP-2␥ is the predominant member of the AP-2 family expressed in mouse trophoblast cells and its expression is restricted to the trophoblast lineage (15). Shi and Kellems (15) suggest that AP-2␥ may be one of the key transcription factors regulating gene expression in trophoblast cells. Thus, the finding in this study demonstrating that AP-2␥ is required for trophoblast-specific expression of 3␤-HSD identifies one of the important target genes whose product is required for progesterone biosynthesis and thus essential for maintenance of pregnancy.
The protein that specifically binds to FP III was identified as the homeodomain protein, Dlx 3. Dlx 3 was recently identified as a placental-specific transcriptional activator of the hCG␣ subunit gene (18). This transcription factor is expressed in human placenta as well as in JEG-3 cells (18) and in the trophoblast cell lineage that forms the placenta in mice (14). Targeted deletion of the Dlx 3 gene resulted in embryonic death between E9.5 and E10.5 due to placental defects (14). Whether mutation in the human Dlx 3 gene would lead to loss of the fetus during human pregnancy is not known at present. Immunocytochemical studies on sections of first trimester human chorionic villus samples showed that expression of Dlx 3 was found primarily in the trophoblast cell layer surrounding the villus and expression was restricted to the nucleus (18).
EMSA studies using nuclear extracts of either JEG-3 or giant trophoblast cells and an oligonucleotide representing the FPII sequence identified by DNase I protection analysis did not result in the formation of a DNA-protein complex, suggesting that the binding of a nuclear protein to FPII requires the prior binding of either AP-2␥ or Dlx 3 or both. Although we were unable to show binding of a nuclear protein to the FPII oligonuclotide within the 66-bp enhancer element, binding of a specific protein is essential for trophoblast-specific transcriptional activity as demonstrated by the loss of transcriptional activity when site-directed mutations were introduced separately into each of the FP-binding sites of the heterologous enhancer-tk construct (Fig. 11). This study also demonstrates that AP-2␥, FPII-binding protein, and Dlx 3 are required for trophoblast-specific expression of the murine 3␤-HSD VI gene.
A GenBank TM search of the human 3␤-HSD I promoter sequence (accession number AL121995) identified an AP-2-binding site at Ϫ2857/Ϫ2848 identical to the FPI core-binding site of murine 3␤-HSD VI and a binding site for Dlx 3 at Ϫ2495/ Ϫ2489 identical to the FPIII-binding site of murine 3␤-HSD VI (Table III). The placental-specific function of these potential binding sites for AP-2␥ and Dlx 3 in the human 3␤-HSD I promoter needs to be established.
The identification of a trophoblast-specific enhancer in the murine 3␤-HSD VI promoter and the demonstration that the two transcription factors, AP-2␥ and Dlx 3, which are required for the cell-specific expression of this gene, are expressed in both human placenta and mouse trophoblast cells, and the identification of binding sites for these two transcription factors in the human 3␤-HSD I promoter, suggests that AP-2␥ and Dlx 3 are the placental nuclear proteins that determine the cell-specific expression of human 3␤-HSD I whose product is required for placental progesterone production. It is of interest to note that these two transcription factors have been shown to be required for placental-specific expression of the hCG␣ subunit gene (16,18). Another similarity between the cell-specific expression of the hCG␣ subunit and the expression of 3␤-HSD is the fact that the nuclear factor SF-1 determines pituitaryspecific expression of the LH/CG␣ subunit and adrenal-and gonad-specific expression of 3␤-HSD (27). It is intriguing to FIG. 9. Characterization of recognition sequence for trophoblast nuclear protein within the FPIII element. A, specific binding of trophoblast cell nuclear proteins to FPIII. EMSA was performed with 4 g of nuclear extract of JEG-3 cells, or GTC or MA-10 cells and FPIII as a probe. Lane 1 is absence of nuclear extract. Indicated competitors were added at 500-fold molar excess. B, identification of nucleotides within the FPIII element recognized by the trophoblast protein. A series of FPIII oligonucleotides with sequential double mutations as underlined (m1-m5) and as illustrated on top of the figure (m6 -m10) were incubated with JEG-3 cell nuclear extract and subjected to EMSA. Lane 1 represents the absence of nuclear extract. Indicated competitors were added at 500-fold molar excess. speculate that AP-2␥ and Dlx 3 are the SF-1 analogous nuclear transcription factors that coordinate trophoblast-specific expression of the placental hormones and enzymes required for maintenance of pregnancy.  -less 3 (Dlx 3). A, similarity between the core binding nucleotides (bold) of the FPIII element and the JRE element, which contains a Dlx 3-binding site characterized in the hCG␣ promoter (18). B, EMSA was performed with nuclear extract of JEG-3 cells (5 g) and radiolabeled probe FPIII (lanes 1-7) or JRE (lanes 8 -14). Indicated unlabeled competitors were added in increasing amounts (50-, 200-, and 500-fold molar excess). C, radiolabeled FPIII (lanes 1-4) or JRE (lanes 5-8) probes were incubated with 5 g of JEG-3 cell nuclear extract in the absence or presence of specific antisera against Nkx2-5 (lanes 2 and 6) or Dlx 3 (lanes 3 and 7) or preimmune rabbit sera (lanes 4 and 8). D, identical to C except the probes were incubated with 10 g of mouse placental nuclear extract.
FIG. 11. Effect of mutations in core nucleotides within each footprint on trophoblast-specific enhancer activity. The binding sites for AP-2␥, Dlx 3, or the potential TEF protein within the sequence representing each footprint, FPI, FPIII, or FPII, respectively, were mutated as described under "Experimental Procedures." The black symbols signify mutations within the indicated FP. Both mutated and wild-type Ϫ2896/Ϫ2831/TK164LUC constructs were transiently transfected into JEG-3 and MA-10 cells. Luciferase activity of each construct is expressed relative to the vector TK164LUC. Each value represents the average plus the range of two separate transfections, each performed in triplicate. AP2␥ and Dlx3 Determine Placenta-specific Expression of 3␤-HSD