CCAAT/Enhancer-binding Proteins Are Mediators in the Protein Kinase A-dependent Activation of the Decidual Prolactin Promoter*

In the course of decidualization, human endometrial stromal cells (ESC) activate the alternative upstream promoter of the decidual prolactin (dPRL) gene. The dPRL promoter is induced by the protein kinase A pathway in a delayed fashion via the region −332/−270 which contains two overlapping consensus binding sequences, B and D, for CCAAT/enhancer-binding proteins (C/EBP). Here we show that sites B and D both bind C/EBPβ and -δ from ESC nuclear extracts. When decidualization of cultured ESC was induced by treatment with 8-Br-cAMP, complex formation on sites B and D was enhanced. Western blot analysis revealed an elevation of both C/EBPβ isoforms, liver-enriched activator protein and liver-enriched inhibitory protein, with a delayed onset between 8 and 24 h of cAMP treatment, while C/EBPδ expression remained unaffected. Cyclic AMP-mediated activation of dPRL promoter construct dPRL-332/luc3 was abrogated by mutation of sites B and D at −310/−285. An expression vector for liver-enriched activator protein potently induced transcription of dPRL-332/luc3 and further enhanced cAMP-mediated induction, while liver-enriched inhibitory protein expression vector abolished the cAMP response, implying that C/EBPs serve as mediators in the delayed cAMP signal transduction to the dPRL promoter. The ratio between activating and repressing isoforms is likely to dictate the transcriptional output.

Decidualization is a differentiation process of the endometrium in preparation for blastocyst implantation. In humans this process is independent of the presence of a blastocyst and is first apparent in the stromal cells surrounding the spiral arteries in the second half of the luteal phase. The key stimulus for decidualization in vivo is progesterone acting on an estrogen-primed uterus, however, in vitro in cultured endometrial stromal cells (ESC), 1 progesterone only acts as a weak inducer of decidualization (1). Its action is synergistically enhanced by a number of factors such as prostaglandin E 2 , corticotropin releasing factor, and gonadotropin free ␣-subunit (2)(3)(4). In fact, in vitro decidualization of ESC can be triggered in the absence of progesterone by agents that elevate intracellular cAMP levels, including gonadotropins, relaxin, or cAMP analogs (3,(5)(6)(7)(8)(9). Induction of prolactin (PRL) gene expression serves as a decidualization marker in human ESC (8,10). PRL expressed in the human endometrium is referred to as decidual PRL (dPRL) to distinguish it from pituitary-derived PRL. Transcription of the human PRL gene is driven by two alternative tissue-specific promoters, the dPRL promoter being located approximately 5.7 kilobases upstream of the pituitary promoter at an additional non-coding exon 1A (11)(12)(13). Utilization of the dPRL promoter has been detected in decidualized endometrial stroma, in myometrial smooth muscle cells, and in hematopoietic cells (12)(13)(14)(15), and is specific to humans and primates (16).
Little is known about the molecular mechanisms governing dPRL promoter control. Cyclic AMP, which is a major decidualization stimulus in vitro, also controls dPRL gene transcription (12). We have shown previously that the cAMP response of the transfected dPRL promoter in ESC occurs in two phases. An early weak induction, mediated by an imperfect cAMP response element (CRE-L) at position Ϫ12 relative to the major transcriptional start site, is detectable within 6 h of treatment with 8-Br-cAMP. This is followed by a delayed strong induction which sets in after 12-18 h of stimulation and is dependent on the dPRL promoter region Ϫ332/Ϫ270 (17). Mutation of the CRE-L abolishes the early, but not the delayed cAMP-mediated induction of promoter activity (17). Computerized search for transcription factor binding sequences revealed two consensus sites for members of the CCAAT/enhancer-binding protein (C/ EBP) family in the dPRL-332/-270 promoter fragment.
C/EBP factors belong to the superfamily of basic region/ leucine zipper DNA-binding proteins (18). So far, six members of the C/EBP family have been described: C/EBP␣, -␤, -␦, -⑀, -␥ and -(reviewed in Ref. 19). C/EBP␣, -␤, and -␦ are found in liver, adipose tissue, intestine, lung, cells of the inflammatory system, and in reproductive tissues, while C/EBP⑀ is restricted to myeloid and lymphoid lineages (19 -27). Specificity of gene control by C/EBPs is ensured through their ability to homoand heterodimerize and to interact with other transcription factors, together with cell-specific and temporal expression patterns and different transactivation potentials (28 -37).
C/EBP␤ was originally identified as a mediator of interleukin-6 (IL6) signaling and is therefore also known as NF-IL6 (38,39). The related factor C/EBP␦ is occasionally designated NF-IL6␤ (29). From the single C/EBP␤ mRNA, protein isoforms with different functions can be generated by a leaky ribosomal scanning mechanism involving three methionine residues in the intronless C/EBP␤ gene, Met 1 , Met 24 ,and Met 199 . Proteins initiated at Met 1 or Met 24 have a calculated molecular mass of about 36 and 33.5 kDa and are referred to as LAP (liver-enriched activator protein) whereas translation ini-tiation at Met 199 results in the 16-kDa isoform LIP (liverenriched inhibitory protein) which lacks the transactivation domain of the longer forms. LIP readily heterodimerizes with LAP, counteracting its activation potential in substoichiometric amounts and therefore acts as a potent repressor (40).
Several lines of evidence indicate an important role for C/EBPs as mediators of hormonal signals in reproductive tissues. C/EBP␤ is regulated by gonadotropins in the ovary and testis (20, 24 -26, 41). Deletion of the C/EBP␤ gene in female mice leads to sterility, caused by the inability to form corpora lutea (20). C/EBP␤ is also an essential factor for mammary gland differentiation and proliferation (21,22).
In this study we demonstrate for the first time the presence of C/EBP transcription factors in human endometrial stroma, an interaction between C/EBP family members and the dPRL promoter, and their involvement in cAMP-induced dPRL gene expression.

EXPERIMENTAL PROCEDURES
Cell Culture-Primary cultures of purified human ESC were prepared and maintained as described previously (12,17). Cells of the first passage were used for transfections and for extraction of nuclear and cytoplasmic protein and RNA. Primary cultures of myometrial smooth muscle cells were prepared as detailed elsewhere (42) and maintained in Dulbecco's modified Eagle's medium/Ham's F-12, 10% fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 10 Ϫ9 M 17␤estradiol. COS-7 cells were kept in the same medium but without estradiol.
Transient Transfection, Protein Extraction, and RNA Isolation-Transient transfections were performed by the calcium phosphate precipitation method for ESC and myometrial cells and with DOTAP reagent (Roche Molecular Biochemicals) for COS-7 cells overnight as described previously (43) in triplicates using 12-well dishes if not indicated otherwise, or 24-well dishes. Medium was replaced the next morning and 0.5 mM 8-Br-cAMP (Biolog, Bremen, Germany) added for stimulation experiments. Cell harvest was performed for ESC and myometrial cultures 48 h and for COS-7 cells 24 h after medium replacement. Luciferase activity was measured with the luciferase reagent kit (Promega) and expressed as relative light units. Transfections were repeated at least three times, and representative experiments are shown (mean Ϯ S.D.). Nuclear and cytoplasmic protein extracts were prepared as described by Schreiber et al. (43,44) with minor modifications and protein concentrations were determined using the DC protein detection kit (Bio-Rad). For RNA isolation the method developed by Gough (45) was applied, using the cytosolic lysate generated during protein extraction as the source.
Antibodies-Rabbit antisera against rat C/EBP␣ (numbers 6 and 247) were kindly provided by Dr. Steve McKnight (University of Texas Southwestern Medical Center, Dallas, TX) and diluted 1:10 for use as a working stock solution. Rabbit antibodies raised against rat C/EBP␤ and human C/EBP␦ (0.1 g/l) were purchased from Santa Cruz Biotechnology.
SDS-Polyacrylamide Gel Electrophoresis, Western Blotting, and Immunodetection-A modified method of Rittenhouse and Marcus (46) was used for protein analysis. Nuclear and cytosolic proteins were loaded on 12% SDS-polyacrylamide gels and electrophoresed for 1 h at 150 V. Gels were transferred for 1 h at 1.2 mA/cm 2 in a semi-dry chamber with a three buffer system (cathode buffer: 25 mM Tris-HCl, pH 9.4, 40 mM ⑀-aminocaproic acid, 20% methanol; anode buffer I: 30 mM Tris-HCl, pH 10.4, 20% methanol; anode buffer II: 300 mM Tris-HCl, pH 10.4, 20% methanol) onto PVDF Immobilon membrane (Millipore) and stained with Fount India Ink (Pelikan) to control for even loading and transfer efficiency. Blots were blocked overnight at 4°C with Blotto (5% nonfat dry milk in 15 mM Tris-HCl, pH 7.6, 136 mM NaCl), exposed to primary antibodies for 1 h at room temperature (dilution 1:1000 in Blotto), and then incubated with secondary antibody (horseradish peroxidase-conjugated anti-rabbit IgG, Sigma), diluted 1:1000 in Blotto, for 1 h at room temperature. Detection was performed with the ECL system (Pierce).
RT-PCR Analysis-Total RNA was used for oligo(dT)-primed cDNA synthesis with SuperScript RNase H Ϫ reverse transcriptase (Life Technologies Inc.). PCR for dPRL and GAPDH cDNAs was performed as described previously (42,43). For amplification of C/EBP␤ cDNA (38) Taq DNA polymerase and solution Q (Qiagen) were used. The sense primer TCTCCGACCTCTTCTCCGACGA spans cDNA positions 353-374 relative to the first start codon, and the antisense primer CAGCT-GCTTGAASAASTKCCG anneals to the region 982-1002. After transfer of the electrophoresed PCR products to positively charged nylon membrane (Roche Molecular Biochemicals), Southern blot hybridization was performed with internal oligonucleotides labeled with terminal deoxynucleotidyl transferase (Life Technologies) and digoxigenin-11-dUTP (Roche Molecular Biochemicals), and detected with the DIG luminescent detection kit (Roche Molecular Biochemicals). The probe sequences were for GAPDH: TCGTCATGGGTGTGAACCATG; for hPRL: CAAGGGGGCCACGCTCTGGCA; for C/EBP␤: TTGCGCACG-GCGATGTTGTTG (antisense to positions 840 -860).
Plasmid dPRL(-332/-191)/-32/luc3 was constructed by PCR with Pwo polymerase (Peqlab, Erlangen, Germany), using dPRL-332-S (see above) as upstream primer and the downstream primer GTCAA-GATCTCTCCCAGGAGACATTTGG (antisense), corresponding to positions Ϫ208/Ϫ191 and carrying an overhang with BglII recognition sequence (underlined). The PCR product was digested with BglII and inserted into the Ecl136II (AGS, Heidelberg, Germany) and BglII sites of dPRL-32/luc3. Plasmid dPRL(-332/-134)/-32/luc3 was generated with the sense primer ATTATGTTCTGAGGGCTGCTTGTGTTGT (positions Ϫ332/Ϫ302) and the antisense primer AGGATCCCATCAATCTAAAT-GAGTG (positions Ϫ154/Ϫ135) with BamHI overhang (underlined). The PCR product was digested with BamHI and inserted into Ecl136II and BglII sites of dPRL-32/luc3. Plasmid dPRL-270/luc3 was generated by PCR, using dPRL-332/luc3 as template. The upstream primer AGGATCCACCAGATGCCAGCAG-CAC spans positions Ϫ270 to Ϫ253 of the dPRL promoter and contains a BamHI overhang (underlined); the antisense primer is anchored in exon 1A: CAAGAAGAATCGGAtCcTACAGGCTTT, covering positions ϩ54 to ϩ80 relative to the transcription start site, and containing two mismatches (lowercase letters) to introduce a BamHI restriction site (underlined) at position ϩ65. The PCR product was subjected to incomplete digestion with BamHI to prevent restriction at the BamHI site close to the 5Ј end of the upstream primer. This created a dPRLϪ270/ ϩ65 fragment with 5Ј blunt and 3Ј BamHI ends which was ligated into pGL3-Basic digested with Ecl136II and BglII.
Computerized searches for transcription factor consensus binding sites were performed with TFSEARCH v1.3 by Yutaka Akiyama using the TRANSFAC data bases (47). Mutations of consensus sequences were introduced into the dPRL promoter by site-directed mutagenesis of dPRL-332/luc3 using the QuikChange system and Pfu polymerase (Stratagene). PCR products were transformed and plasmid DNA prepared to retrieve the mutated insert by Acc65I/NcoI digestion and ligate it into native pGL3-Basic cleaved with the same enzyme combination. The sense sequences of the complementary oligonucleotides used for site-directed mutagenesis (D-B-mut, F-mut, G-mut) and the resultant constructs (dPRL-332/D-Bmut/luc3, dPRL-332/F-mut/luc3, dPRL-332/ G-mut/luc3) are illustrated in Fig. 6.
As a reporter control for C/EBP expression vectors NFIL6RE/luc3 was constructed. Complementary oligonucleotides with Acc65I overhangs were annealed and inserted into the Acc65I site of the dPRL-32/luc3 minimal promoter construct. The resultant insert sequence GGTACCCAGATCTGGCAGGATCCAGATTGCGCAATCTGCGGT-ACC carries a palindromic NF-IL6 response element consensus sequence (bold faced) flanked by Acc65I sites (underlined).
Expression vectors for transient transfections were created in pSG5 (Stratagene) under the SV40 promoter. The human C/EBP␣ cDNA insert (48) was excised from hCMV-C/EBP␣ (kindly provided by Dr. G. Darlington, Baylor College of Medicine, Houston, TX) with BamHI and ligated into BamHI cleaved pSG5 to give pSG-C/EBP␣. Expression vectors for human C/EBP␤ and C/EBP␦ (pCMV/NF-IL6, pCMV/NF-IL6␤) were a gift from Dr. S. Akira (Hyogo College of Medicine, Hyogo, Japan) (29,38). Inserts were excised with SalI, blunt-ended and cloned into pSG5 which had been linearized with BamHI and blunt-ended. This yielded pSG-C/EBP␤, pSG-C/EBP␦, and pSG-C/EBP␤rev, which carries the C/EBP␤ cDNA in the reverse orientation and was used as a negative control.
To generate pSG/LIP, pSG-C/EBP␤ plasmid was cut with EcoRI (in the polylinker region of pSG5 upstream of the cDNA insert) and SacII (in the insert). Released cDNA fragments were removed and the remaining plasmid polished with mung bean nuclease (Promega) and recircularized, retaining as the insert the C-terminal portion of CEBP␤ cDNA extending from the SacII site 49-base pair upstream of the Met 199 codon to beyond the stop codon.
To generate the expression vector pSG/LAP, Met 199 was changed to Leu by PCR-mediated mutagenesis using the following primers: LAP-H-S (CGCAGGCtTGGCGGCaaGCTTCCCGTAC, corresponding to positions 588 -615 in the human C/EBP␤ cDNA relative to the first Met codon; lowercase letters represent nucleotide changes, italicized triplets indicate codon mutations from Met 199 to Leu and Gly 202 to Ser, and underlined is a HindIII site introduced by mutation), LAP-H-AS (antisense to LAP-H-S), and NFIL6-H (TTGCGCACGGCGATGTTGTTG, antisense to positions 840 -860 of human C/EBP␤ cDNA). Two separate PCR reactions were performed with Pfu polymerase on template pSG-C/EBP␤. Primer pairs T7 (anchored in the polylinker of pSG5 5Ј to the cDNA insert) and LAP-H-AS, and LAP-H-S and NFIL6-H were used to amplify overlapping portions of the C/EBP␤ cDNA, introducing the mutations Met 199 to Leu, Gly 202 to Ser, and a HindIII site within the overlap. Both PCR products were digested with HindIII and ligated to one another to span the entire amplified region from the T7 to the NFIL6-H sequences. The purified ligation product was cleaved with BsiWI to isolate an internal 495-base pair fragment containing the mutations. This BsiWI fragment was used to replace the wild type BsiWI fragment in pSG-C/EBP␤ thus yielding pSG/LAP.
The construct pABVP16 was kindly provided by Drs. Sergio Onate and Sophia Tsai (Baylor College of Medicine, Houston, TX). It contains coding region for the transactivation domain (AD) of herpes simplex virus VP16 under control of the Rous sarcoma virus promoter and was used to generate pVP16/LAP-DBD. The sequence encoding the DNAbinding domain (DBD) was excised from pSG-C/EBP␤ by digestion with BsiWI (immediately downstream of the Met 199 codon), polishing with mung bean nuclease, and digestion with BglII in the polylinker 3Ј to the cDNA insert. The fragment was ligated into pABVP16 which had been prepared as follows: the plasmid was cut with XhoI in the polylinker 3Ј to the VP16 sequence, filled in with the Klenow fragment of DNA polymerase I and dTTP/dCTP only, polished with mung bean nuclease, and finally cleaved with BamHI. The ligation resulted in an in-frame fusion of VP16-AD and LAP-DBD. All mutations and preservation of open reading frames were verified by sequencing.

Specific Complexes Formed between the dPRL Promoter Fragment dPRL-332/-270 and Nuclear Proteins from Decidualized ESC Contain Primarily C/EBP␤-The region Ϫ332/
Ϫ270 is required for the delayed cAMP response of the dPRL promoter in ESC. We therefore used this promoter fragment as a probe in gel shift assays in an attempt to identify nuclear proteins induced during cAMP-stimulated decidualization of ESC. Nuclear extracts from untreated, undifferentiated ESC cultures were compared with those from cultures which had been treated with 8-Br-cAMP for 6 days to decidualize (Fig. 1). No specific DNA-protein complexes were obtained with extracts from undifferentiated cells. However, a number of specific complexes appeared with extracts from decidualized cells: a faster migrating distinct doublet and a more diffuse range of bands migrating more slowly (compare lanes 6 and 7). In order to define the binding sequences in dPRL-332/-270 more closely, we used five subfragments (A-E, illustrated in Fig. 1) as competitors. Two of these subfragments, B and D, were able to compete for binding, eliminating all complexes specific to decidualized cells. Fragment B spans the sequence dPRL-301/-270 and overlaps with fragment D, which covers dPRL-311/-291. We therefore concluded that the sequence involved in  6 and 7). Binding is competed by subfragments B and D both of which contain consensus binding sequences for C/EBPs indicated by horizontal brackets: at positions Ϫ310 to Ϫ297 for C/EBP␤, and at Ϫ298 to Ϫ285 for C/EBP␣ and -␤.
binding with decidual proteins must be localized to the overlap. However, a double-stranded oligonucleotide containing the overlap Ϫ301 to Ϫ291 was not able to compete with the fulllength probe dPRL-332/-270 for binding of decidual proteins (data not shown). This led us to postulate that sequences with functional homology must be present in both subfragments, not or not entirely contained within the overlap, to explain their ability to compete for the same proteins. Computerized search revealed consensus binding sites for members of the C/EBP family in both subfragments. A binding site for C/EBP␤ is found at Ϫ310/Ϫ297 in fragment D, and for C/EBP␣ and -␤ at Ϫ298/Ϫ285 in fragment B ( Fig. 1; see also Fig. 6B). These clustered sites with coordinates Ϫ310/Ϫ285 will be referred to as the D-B site.
We then determined by supershift analysis if the DNA/protein adducts formed with extracts from decidualized ESC on dPRL-332/-270 contained C/EBPs (Fig. 2). Specific complexes were quantitatively supershifted by antibody to C/EBP␤; a faint supershift was obtained with C/EBP␦ antibody while addition of C/EBP␣ antiserum was without effect (Fig. 2A). This indicates that the multiple and diffuse complexes all contain C/EBP␤ primarily in homodimeric form and partially het-erodimerized with C/EBP␦ or potentially other basic region/ leucine zipper proteins. In the majority of endometrial preparations, no C/EBP␣ was detectable by supershift analysis. However, when we used a particular specimen that contained C/EBP␣ and employed subfragments B and D as probes, different affinities for C/EBP isoforms became apparent. While fragment B was able to bind C/EBP␣, -␤, and -␦, fragment D only interacted with C/EBP␤ and -␦ (Fig. 2B). The banding pattern was again diffuse, and binding intensity was increased in decidualized compared with undifferentiated ESC (compare lanes 1 and 6). It is noteworthy that specific complexes formed on fragment D were completely supershifted with C/EBP␤ antibody, whereas fragment B gave rise to a very abundant complex which was specific, i.e. could be competed with unlabeled probe (lane 2, left panel), but was unaffected by addition of antibodies to C/EBPs. The nature of this complex is presently unknown.
C/EBP␤ Isoforms LAP and LIP, but Not C/EBP␦ Proteins Are Induced by 8-Br-cAMP Treatment in ESC-The rat C/EBP␤ gene has been reported to be transcriptionally controlled by the cAMP regulatory element-binding protein, CREB (49). We therefore investigated whether cAMP treatment up- regulated C/EBP␤ expression in ESC. Three individual ESC cultures were prepared and either left untreated or treated with 8-Br-cAMP for 6 days before harvesting cytoplasmic RNA and nuclear proteins. RT-PCR analysis showed massive induction of dPRL mRNA in response to the cAMP analog, confirming at the molecular level the morphologically apparent decidualization in all three specimens (Fig. 3A). C/EBP␤ mRNA was detectable both in undifferentiated and in decidualized cells and appeared moderately increased in the latter. We then prepared Western blots with nuclear proteins from the same individual cultures. Nuclear extracts from COS-7 cells which had been transfected with expression vectors for C/EBP␤ and C/EBP␦ were run in parallel. Immunodecoration with C/EBP␤ antibody resulted in the appearance of multiple bands, three of which are likely due to alternative translation initiation ( C/EBP␦ protein levels, in contrast, varied widely between individual specimens but were never affected by 8-Br-cAMP treatment (Fig. 3B). Only a single band of 32 kDa apparent molecular mass was detected. C/EBP␣ protein could not be demonstrated in ESC by Western blot analysis, but a single band was immunodecorated with C/EBP␣ antibody in control nuclear extracts from COS-7 cells transfected with C/EBP␣ expression vector (not shown).
If C/EBP␤ in ESC was in fact transcriptionally controlled by the PKA pathway, up-regulation might be rapid and transient, having already passed after 6 days of stimulation with 8-Br-cAMP. We therefore performed a kinetic study including earlier time points of 1, 2, 4, 8, and 24 h and 2 and 4 days. We also compared C/EBP␤ levels in the cytosolic versus the nuclear fraction to detect a potential cAMP-induced change in subcellular distribution (Fig. 3C). In nuclear extracts, LAP and LIP were detectable in control cells. An increase in protein levels for both isoforms, but more distinctly for LIP, became apparent PCR was performed for detection of dPRL, C/EBP␤, and GAPDH mRNA. PRL cDNA was amplified with primers anchored in the decidua-specific exon 1A and in exon 5 of the human PRL gene. GAPDH amplification was performed for standardization. PCR products were subjected to nonradioactive Southern blot hybridization. B, Western blot analysis of C/EBP isoforms in ESC. Nuclear proteins were harvested from the same individual ESC cultures as in panel A. As positive controls, nuclear extracts were prepared from COS-7 cells which had been transiently transfected with pSG-C/EBP␤ or pSG-C/EBP␦ (lanes 7 or 8, respectively). The amounts of proteins loaded were: 10 g (lanes 1 and 2), 15 g (lanes 3 and 4), 30 g (lanes 5 and 6), 1 g (lane 7), and 2 g (lane 8). Immunodetection was carried out with antibodies to C/EBP␤ (upper panel) and C/EBP␦ (lower panel). Migration of molecular mass markers is indicated on the left in kilodaltons. C, time course of C/EBP␤ protein induction in ESC in response to 8-Br-cAMP. Nine flasks of an ESC preparation were plated in parallel and left untreated or treated for the indicated time periods with 8-Br-cAMP before simultaneous harvesting of protein extracts. Western blot analysis was carried out on nuclear (upper panel) and cytosolic (lower panel) protein fractions with C/EBP␤ antibody to study changes in isoform expression and subcellular distribution. Twenty g of nuclear protein and 15 g of cytosolic protein were loaded per lane. only between 8 and 24 h of 8-Br-cAMP treatment, reached a maximum between 2 and 4 days, and declined after 6 days. Parallel kinetics were observed in cytosolic extracts for LAP levels. However, LIP was not detectable in the cytosolic fraction. As LAP levels in both subcellular compartments changed in parallel, no cAMP-induced nuclear translocation appears to take place in ESC.

Fragment dPRL-332/-270 Confers C/EBP Responsiveness to a Minimal Promoter in Co-transfection Experiments-
Since the dPRL promoter fragment Ϫ332/Ϫ270 was able to bind C/EBP␣, -␤, and -␦ in vitro, we next inserted this fragment in front of the minimal homologous promoter element dPRLϪ32/ϩ65 in a luciferase reporter vector (Fig. 4A). This construct was then co-transfected with expression vectors for C/EBP␣, -␤, or -␦ to investigate their ability to interact with the dPRL promoter in vivo. COS-7 cells were chosen as the recipient cells because they express very low endogenous levels of these transcription factors as determined by Western blot and EMSA/supershift analyses (see Figs. 3B and 5A, and data not shown). The reporter fusion 2x(dPRL-332/-270)/-32/luc3 was strongly induced by C/EBP␦ and moderately by C/EBP␣, while C/EBP␤ displayed only a very small activation potential (Fig. 4B). Surprisingly, the intact dPRL promoter extending from Ϫ332 to ϩ65 in the reporter construct dPRL-332/luc3 was unresponsive to any of the C/EBP factors. The results obtained with the isolated fragment Ϫ332/Ϫ270 clearly show that C/EBPs bind to it and can transactivate in vivo, C/EBP␦ being the most potent factor. The very minimal transactivating ability of C/EBP␤ can be explained by the fact that COS-7 cells transfected with the C/EBP␤ cDNA translate not only LAP but also substantial amounts of LIP from this template (see Fig. 3B). No explanation for the unresponsiveness of the wild type promoter construct dPRL-332/luc3 to C/EBP could be provided at this point.
The PKA-mediated activation of the dPRL promoter is cellspecific and cannot be elicited in COS-7 cells (data not shown). All following transfection experiments were therefore performed in cell types that are capable of activating the endogenous dPRL promoter, i.e. primary cultures of ESC and of human myometrial smooth muscle cells. The wild type promoter in dPRL-332/luc3 is strongly activated by 8-Br-cAMP in transfected ESC (Fig. 4C). Deletion of the region Ϫ332/Ϫ270 (construct dPRL-270/luc3) significantly reduced the cAMP response. However, the isolated fragment Ϫ332/Ϫ270 alone, inserted in front of the minimal dPRL promoter, was unresponsive to cAMP. Further deletion analyses revealed that a fusion construct carrying element Ϫ332/Ϫ191 was still unresponsive while a construct containing element Ϫ332/Ϫ134 partially regained cAMP responsiveness. This indicates that neither the proximal CRE-L (see Fig. 4A) nor region Ϫ332/Ϫ270 alone or in combination are sufficient but that an additional cis-acting element located between Ϫ191/Ϫ134 contributes to full cAMP inducibility of the dPRL promoter.

FIG. 4. Activation of dPRL promoter constructs by C/EBP expression vectors in COS-7 cells and by cAMP in ESC.
A, schematic depiction of the minimal dPRL promoter/reporter gene construct Ϫ32/luc3 carrying sequence Ϫ32/ϩ65 of the dPRL gene fused to the luciferase gene (luc), the intact dPRL-332/luc3 promoter construct containing sequence Ϫ332/ϩ65, deletion constructs in which regions Ϫ332/Ϫ270, Ϫ191/Ϫ32, or Ϫ134/Ϫ32 had been removed, and the fusion constructs 1x

The cAMP Response of dPRLϪ332/ϩ65 in ESC Is Diminished by Mutation of the Composite C/EBP-binding Site D-B and Is
Abolished by Coexpressed LIP-In order to dissect the contributions of alternative translation products to the overall regulatory capacity of expression vector for C/EBP␤, we created expression vectors for LIP and LAP. The latter was generated by mutating the methionine residue Met 199 to a leucine to prevent downstream translation initiation. To verify correct translation initiation, COS-7 cells were transfected with these expression vectors, and nuclear extracts subjected to Western blot analysis using antibody to C/EBP␤ (Fig. 5A). LIP expression vector yielded a single band of 16 kDa (lane 3) comigrating with the smallest product generated by C/EBP␤ expression vector (lane 1). LAP expression vector (lane 2) yielded a doublet of bands around 40 kDa comigrating with the major products generated by the C/EBP␤ expression vector (lane 1). Both expression vectors for C/EBP␤ and LAP yielded an additional minor product of about 30 kDa which is also visible in nuclear extracts from cAMP-stimulated ESC cultures (see Fig. 3C) and may represent a cleavage product.
At first, the newly created expression vectors were tested in co-transfection experiments in ESC using the fusion construct 2x(dPRL-332/-270)/-32/luc3 as the reporter (Fig. 5B, left panel). The transactivation potential of LAP was very pronounced and comparable to that of C/EBP␦. In contrast, LIP did not activate the reporter construct. C/EBP␤ displayed an intermediate effect, most likely reflecting the co-translation of LAP and LIP in ESC as shown before in COS-7 cells.
We then investigated the effect of these expression vectors on the wild type dPRL promoter in ESC, alone or in combination with 8-Br-cAMP treatment (Fig. 5B, right panel). Basal activity of dPRL-332/luc3 was increased approximately 3-fold by 8-Br-cAMP treatment. C/EBP␦, C/EBP␤, and LAP all elicited a significant induction of dPRL promoter activity which was further enhanced by the addition of 8-Br-cAMP. The strongest activation (13-fold) was obtained by a combination of LAP and 8-Br-cAMP. In contrast, coexpression of LIP abrogated the cAMP-mediated induction of the dPRL promoter. Even more pronounced effects of C/EBPs were obtained in myometrial smooth muscle cells which display a very low basal activity of the dPRL promoter (Fig. 5C). C/EBP␦ caused a significant 23-fold induction which was further enhanced by 8-Br-cAMP. Introduction of C/EBP␤ only resulted in a minimal activation, still allowing enhancement by 8-Br-cAMP, whereas expression of LIP completely abrogated cAMP induction. Most interestingly, LAP expression vector was as potent as C/EBP␦ in transcriptional activation, effecting an 87-fold induction when combined with 8-Br-cAMP.
In order to investigate whether the C/EBP consensus binding sequences were instrumental in transducing the cAMP signal to the dPRL promoter, we mutated these sites in the context of the cAMP-responsive promoter construct dPRL-332/luc3. In addition to the composite D-B site (coordinates Ϫ310/Ϫ285) we also mutated a downstream C/EBP consensus site (site F, coordinates Ϫ214/Ϫ201) and a consensus sequence for binding of Ets factors (site G, coordinates Ϫ248/Ϫ239) (Fig. 6). The latter two sequences overlap with regions that have recently been described as decidual-specific footprints formed with extracts from ESC decidualized by progesterone treatment (Ϫ209/Ϫ201 and Ϫ257/Ϫ244) (50). In co-transfection experiments in ESC, the cAMP responsiveness of promoter constructs was tested by a 48-h incubation with 8-Br-cAMP. In addition, we co-transfected the expression vector for the repressor isoform LIP (Fig.  7A). The intact promoter in dPRL-332/luc3 displayed the highest basal activity and was induced 4.1-fold by 8-Br-cAMP. Deletion of the major cAMP-responsive region -332/-270 in dPRL-270/luc3 had the same effect as mutation of the D-B sites in dPRL-332/D-Bmut/luc3; basal activity was reduced to about 30%, and cAMP inducibility was reduced to about 50% of that seen with dPRL-332/luc3. The other two promoter mutations dPRL-332/F-mut/luc3 and dPRL-332/G-mut/luc3 also displayed lowered basal activities but, in contrast to dPRL-332/D-Bmut/luc3, their cAMP inducibility was not impaired (4.5-and 4.7-fold, respectively). This indicates that the D-B sites are essential for cAMP responsiveness of the dPRL promoter, whereas the F and G sites are dispensable for the cAMP response but contribute to basal promoter activity. Co-transfection of LIP completely abolished cAMP induction of all reporter constructs. It is noteworthy that LIP also reduced basal activity of dPRL-332/luc3, indicating that the elevated basal transcription from dPRL-332/luc3, as compared with dPRL-270/luc3, is dependent on endogenous activating C/EBP isoforms.
To further support the notion that the loss of cAMP inducibility in dPRL-332/D-Bmut/luc3 was a consequence of the inability of C/EBP DBD to bind to the mutated sequence, we created a fusion of the DBD of C/EBP␤ with the strong transactivation domain of VP16. The resultant expression vector pVP16/LAP-DBD was used in a co-transfection experiment in ESC (Fig. 7B). The fusion protein VP16/LAP-DBD activated the reporter construct 2x(dPRL-332/-270)/-32/luc3 more than 100-fold (not shown), strongly activated a positive control construct carrying a palindromic C/EBP response element linked to a minimal promoter (NFIL6RE/-32/luc3) and was also able to transactivate the wild type dPRL promoter in dPRL-332/luc3. Mutation of the D-B sites in dPRL-332/D-Bmut/luc3 resulted in a complete loss of inducibility. Three conclusions can be drawn from this experiment. The DBD of C/EBP␤ binds to the wild type dPRL promoter in vivo; the mutations introduced by us disable the D-B sites; and lastly, no other sequence in the dPRL promoter, including site F, binds the C/EBP␤ DBD. DISCUSSION The observation that prolonged treatment with cAMP analogs causes decidualization and induction of dPRL expression in isolated ESC in vitro prompted us to search for cAMPinducible factors that interact with the dPRL promoter. We focussed our interest on the region Ϫ332/Ϫ270 because it is essential for the pronounced but delayed transcriptional activation of the dPRL promoter in response to elevated cAMP (17). Gel shift analyses revealed that dPRL-332/-270 formed specific complexes with nuclear proteins from ESC, and that the intensity of complex formation was increased in decidualized cells which had been exposed to 8-Br-cAMP for 6 days. The pattern of complex formation on probe dPRLϪ332/Ϫ270 was always diffuse in the upper range and included two distinct sharper bands in the lower range of the gel. In spite of this multiplicity of DNA/protein adducts, they were quantitatively supershifted by antibody to C/EBP␤. This indicates the engagement of FIG. 6. Map of dPRL promoter constructs and mutations of C/EBP and Ets factor-binding sites. A, consensus binding sequences for transcription factors of the C/EBP (B, D, and F) (ovals) and Ets family (G) (rhombus), and the CRElike element (CRE-L) (hexagon) are depicted in the minimal dPRL promoter construct Ϫ32/luc3, the 5Ј deletion construct dPRL-270/luc3, and the intact construct dPRL-332/luc3. Mutated consensus sequences in the context of dPRL-332/luc3 are crossed-out. Sites G and F are labeled with question marks because their functionality could not be confirmed in the course of this study. B, sequence of the dPRL promoter region from Ϫ332 to Ϫ187 with the consensus binding sequences for C/EBP (D, B, and F) and Ets factors (G) highlighted by boxes. The oligonucleotides used for site-directed mutagenesis (D-Bmut, G-mut, and F-mut) are shown below the sequence. Lowercase letters indicate mutations; new restrictions sites, integrated for diagnostic purposes, are underlined.
We have shown by Western blot analysis that ESC express both LAP and LIP, and that protein levels increase in response to cAMP treatment. It is therefore conceivable that the distinct high mobility complexes seen in EMSA, which are specific to extracts from decidualized cells and supershifted by C/EBP␤ antibody, contain LIP homodimers or heterodimers with another small molecule. Low mobility complexes may, at least in part, result from heterodimerization of various C/EBP␤ isoforms. Even when using in vitro transcription/translation mixture primed with pSG-C/EBP␤ only, we observed a broad range of bands in EMSA on probe dPRL-332/-270 (not shown). In addition, low mobility complexes may contain heterodimers with C/EBP␦ because we could demonstrate both by EMSA/ supershift and by Western blotting that this isoform is present in ESC, albeit at much lower levels than C/EBP␤. Whether other nuclear proteins from ESC interact with C/EBP␤ on dPRL-332/-270 remains to be determined.
In addition to elevation of C/EBP␤ protein levels, stimulation of the PKA pathway may contribute to the enhanced formation of complexes between the dPRL promoter and C/EBP␤ in other ways. (i) The level or binding affinity of another unknown nuclear factor, engaged in complex formation with C/EBP␤, may increase and serve to recruit C/EBP␤ to the dPRL promoter. (ii) The binding affinity of C/EBP␤ may be enhanced as a consequence of post-translational modification. Several serine residues in C/EBP␤ have been described as targets for PKA phosphorylation, however, the role of phosphorylation for transactivating function of the protein has not fully been determined. While in vitro binding studies seem to indicate that phosphorylation at Ser 240 in the DBD of rat C/EBP␤ reduces DNA binding affinity and at Ser 105 in the transactivation domain is without effect (51), phosphorylation in vivo at Ser 288 in the human C/EBP␤ appears essential for nuclear translocation (52,53). This serine residue and its contiguous sequence are highly conserved in the rat and human proteins, human Ser 288 corresponding to rat Ser 240 (38,54). Our data do not support the notion that PKA-mediated phosphorylation triggers nuclear translocation of C/EBP␤ in ESC. In a kinetic study including short term stimulations with 8-Br-cAMP we could not detect a change in subcellular distribution of C/EBP␤ in response to the stimulus.
A random site selection approach to determine DNA binding specificity of the C/EBP family had revealed similar, but not identical sequence preferences (55). We also observed differences in binding of C/EBP␣, C/EBP␤, and C/EBP␦ to two sites in the dPRL-332/-270 promoter fragment. Whereas the C/EBP site in subfragment B was able to bind all three C/EBPs, C/EBP␣ failed to bind to the C/EBP site in subfragment D. This is in agreement with the binding specificities of sites B and D predicted by the TFSEARCH program.
We performed a series of transfection experiments to determine the role of C/EBPs in the transcriptional control of the dPRL promoter. C/EBP␣ and C/EBP␦ were clearly able to bind to the region dPRL-332/-270 and to transactivate the promoter fusion construct 2x(dPRL-332/-270)/-32/luc3, C/EBP␣ being less potent than C/EBP␦. This may be due to the presence of only one binding site for C/EBP␣ in the region dPRL-332/-270 as compared with two sites for C/EBP␦ binding, or it may reflect different properties of the transactivation domains and different requirements for interacting partners of these two proteins. Although there is a certain extent of redundancy in that C/EBP␣, -␤, and -␦ can compensate for one another in some systems (56), the pronounced phenotypes observed in mice lacking one of the three genes argue for specific physiological roles of each factor (20)(21)(22)57,58;reviewed in Ref. 19). In addition to slightly different DNA sequence preferences, distinctive functions are likely to reside in different proteinprotein interaction interfaces. For example, C/EBP␤ but not C/EBP␣ is able to synergize with Sp1 by physical interaction (33). Surprisingly, in numerous transfection studies in COS-7 cells, none of the C/EBP expression vectors was capable of transactivating the wild type intact dPRL promoter in dPRL- 332/luc3. This may in part be explained by different spacing between the cognate DNA sequence and the transcriptional start site in the wild type promoter compared with the fusion construct 2x(dPRL-332/-270)/-32/luc3. When the C/EBP-binding sites are located in close proximity to the start site, as is the case in 2x(dPRL-332/-270)/-32/luc3, the transactivation domains of C/EBPs may be capable of directly interacting with the basal transcription machinery. In the context of the wild type promoter, additional cofactors may be required to establish contact between the C/EBP transactivation domains and the basal transcription machinery. Such endogenous factors are possibly present in limiting amounts when C/EBPs are overexpressed by transfection, or they are cell-specific and lacking in COS-7 cells which are not capable of activating the endogenous dPRL promoter.
In contrast, in cell types capable of activating the endogenous dPRL promoter, we observed substantial activation of dPRL-332/luc3 by C/EBP␦ and LAP. This was the case in ESC and, more dramatically, in myometrial smooth muscle cells which display a very low basal activity of the dPRL promoter. These cells therefore show very clearly that by supplying C/EBP␦ or LAP in combination with an activated PKA pathway, highly efficient transcription of the dPRL promoter can be elicited. Introduction of LIP resulted in a complete abrogation of cAMP induction, both in myometrial and endometrial stromal cells. These observations strongly argue for the importance of endogenous C/EBPs as mediators of the cAMP signaling pathway to the dPRL promoter in PRL-expressing cells and for the requirement of cell-specific cofactors integrating C/EBPand cAMP-mediated signals, such factors lacking in COS-7 cells.
A physiological role for C/EBP␤ in mediating hormonally induced PKA signaling is supported by a number of reports. In rat granulosa cells, treated with an ovulatory dose of hCG, an increase in LAP and LIP isoforms is observed after 2 h (25). Luteinizing hormone induces C/EBP␤ protein in preovulatory ovaries maintained in an ex vivo perfusion system; when C/EBP␤ levels are reduced by application of antisense oligonucleotides, ovulation is inhibited (24). C/EBP␤-deficient female mice fail to down-regulate COX-2 and P450 arom in response to luteinizing hormone and do not develop corpora lutea (20). In cultured rat Leydig cells, C/EBP␤ expression is elevated both at the mRNA and at the protein level within 4 h of hCG stimulation, returning to control levels thereafter (41). Treatment with follicle stimulating hormone rapidly and transiently increases C/EBP␤ expression in cultured rat Sertoli cells as well as in whole testes of hypophysectomized animals (26). The molecular basis, at least in the rat, for gonadotropin-or cAMPmediated up-regulation of C/EBP␤ expression is the presence of two CREs in the C/EBP␤ promoter. PKA stimulates C/EBP␤ transcription both from the transfected and the endogenous promoter in rat hepatocytes (49). In addition, autoregulation has been demonstrated for the mouse C/EBP␤ promoter (59). Such a mechanism may come into play as a secondary event to the cAMP-mediated up-regulation to further enhance signaling. In contrast to all above reports, we observed a complete absence of a rapid and transient up-regulation of C/EBP␤ protein in cAMP-stimulated ESC. An increase in LAP and LIP protein only became apparent between 8 and 24 h of stimulation, and protein levels remained elevated between 2 and 6 days of treatment. These kinetics are congruent with the delayed cAMP-stimulated induction of the dPRL promoter in ESC which occurs after 18 h and then persists (17).
Our results provide evidence that the composite C/EBP consensus binding sites D-B located at Ϫ310/Ϫ285 are required for the cAMP response of the dPRL promoter. Mutation of five bases in this region in reporter construct dPRL-332/D-Bmut/ luc3 drastically reduced cAMP inducibility compared with wild type dPRL-332/luc3. Involvement of C/EBPs is supported by the finding that the same mutation completely abolished transactivation by VP16/LAP-DBD. The latter observation also indicates that the downstream C/EBP consensus binding sequence designated site F in this report does not interact with LAP-DBD.
Decidualization of ESC is a process that requires tight control in vivo to establish a delicate balance between proliferation and differentiation. We believe that at the molecular level this is achieved by a complex orchestration of both transcriptional activators and repressors. Our understanding of the mechanisms controlling decidualization is still limited, but we know that a sustained elevation of intracellular cAMP, a down-regulation of the regulatory subunit RI␣ of PKA, and a subsequent elevation of catalytic subunit activity are part of the pathway (17). One of the transcriptional repressors induced in the course of decidualization is the inducible cAMP early repressor ICER, a product of the CREM gene (43). In this report we show that members of the C/EBP family are also expressed in ESC, and that positive and negative regulators of transcription (LAP and LIP) are induced by a decidualization stimulus while the activating form C/EBP␦ is present at a constant level. Changing the ratio between C/EBP␤ and C/EBP␦ would be predicted to exert profound effects on dPRL promoter activation via a complex control mechanism at the level of varying patterns of homo-and heterodimerization between LAP, LIP, and C/EBP␦. A regulatory role for shifting isoform ratios has been elucidated in the rat mammary gland: while in pregnancy the LAP/LIP ratio is low, during lactation a more than 100-fold increase in the LAP/LIP ratio is observed. This hormonally regulated balance of C/EBP isoforms controls milk protein gene expression (23).
Induction of the dPRL gene in response to cAMP occurs in a delayed fashion, indicative of the necessity of intermediate steps for transcriptional activation. We speculate that the C/EBP␤ gene may be one of the intermediate targets of the PKA pathway. We were intrigued by the finding that the C/EBP␤ gene itself, in contrast to numerous reports in the literature (26,41,49,60,61), is a slow responder to cAMP in ESC. We do not know at present if this is a feature characteristic of the promoter of the human C/EBP␤ gene, most other data having been raised in rodent systems. We rather tend to believe that the slow response is characteristic of this particular tissue, the endometrial stroma. Under the assumption that first the human C/EBP␤ promoter, as reported for the mouse promoter (59), underlies autoregulation, it would be subject to the same restrictions as outlined above for the dPRL promoter: a certain balance between LAP and LIP, and possibly other family members, may have to be established before positive autoregulation can set in. Provided secondly that the human C/EBP␤ promoter contains functional CREs, as reported for the rat promoter (49), it would not only receive a positive signal from activating CREB and CREM isoforms that are phosphorylated by PKA catalytic subunit, it would also receive an opposing signal from the repressor ICER which is up-regulated by sustained PKA activity in ESC (43). Again, a certain balance between these conflicting inputs would have to be established to allow for transcriptional activation of the C/EBP␤ gene.
Taken together, our study establishes the role of C/EBPs (particularly LAP and C/EBP␦) as important transducers of the cAMP signal to a delayed-response gene. The binding of activator C/EBPs to a composite C/EBP response element at Ϫ310/ Ϫ285 in the proximal dPRL promoter region is instrumental for cAMP responsiveness of the promoter. Mutation of the response element reduces, and expression of the dominantnegative C/EBP isoform LIP abrogates cAMP inducibility. Further studies will be needed to identify potential cofactors required for the tight control of differentiation-specific dPRL gene expression to achieve a deeper knowledge of the process of human endometrial decidualization in particular and of specific physiological functions of individual members of the C/EBP family of transcription factors in general.