Hormone-dependent recruitment of NF-Y to the uteroglobin gene enhancer associated with chromatin remodeling in rabbit endometrial epithelium.

Expression of the rabbit uteroglobin gene is hormonally induced in cells of the endometrial epithelium during the preimplantation phase of pregnancy. Here we show that progesterone activation of the gene is mediated by two clusters of hormone responsive elements located between 2.4 and 2.7 kilobase pairs upstream of the transcriptional start site. Between these two clusters, genomic footprinting studies in the intact endometrial epithelium reveal the hormone-inducible occupancy of several cis-acting elements. One of the protected elements shows sequence homology to the consensus binding site of the transcription factor NF-Y, which binds to the element in gel shift experiments. This uteroglobin Y box is essential for enhancer activity in transient transfection experiments with endometrial and non-endometrial cell lines, in accordance with the ubiquitous expression of NF-Y. To understand why binding of this ubiquitous factor to the uteroglobin Y box in endometrium depends on hormone induction, we examined the chromatin structure of the relevant gene region. In the uninduced state, the enhancer region appears to be organized into positioned nucleosomes. Upon hormone induction, this nucleosomal pattern is lost and the enhancer region becomes hypersensitive to nucleases, suggesting that a hormone-induced change in the local chromatin structure unmasks previously unaccessible binding sites for transcription factors. Our results emphasize the limitations of using transient transfection assays for the functional analysis of cis-acting elements and underline the need for including the native chromatin organization in this kind of studies.

Uteroglobin is a small globular protein originally described as the main protein component in the uterine secretion of rabbits in the preimplantation phase of pregnancy (1, 2) (for a review, see Refs. 3 and 4). Later, the protein was also found in the oviduct (5), the male genital tract (6), and the lung (7). Although its general physiological role remains unclear, in the rabbit uterus the protein might be involved in implantation of the trophoblast and/or in anti-inflammatory reactions (4,8,9). Expression of the uteroglobin gene in cells of the endometrial epithelium is transcriptionally regulated by estrogens and progestins (10 -12). The effect of estrogens is mediated by a regu-latory composite unit located around 250 bp 1 upstream from the transcription start site, which comprises a single estrogen responsive element and an adjacent GC/GT box (13,14). The potential cis-acting elements for progesterone activation of the gene have only been mapped in DNA binding experiments in vitro. Progesterone receptor-binding sites have been assigned to the first intron of the gene (15), as well as to a region between 2.4 and 2.7 kb upstream of the transcription start site, where two clusters of three imperfect binding sites for glucocorticoid and progesterone receptors have been found (16,17). The relevance of these upstream receptor-binding sites was supported by the appearance of adjacent progesterone inducible DNase I-hypersensitive sites in chromatin of isolated nuclei from endometrial epithelium (17).
In order to identify physiologically relevant regulatory elements within the uteroglobin gene upstream region, we performed gene transfer and genomic footprinting studies in native endometrial epithelium following hormonal treatment of female rabbits. We show that the two clusters of progesterone receptor-binding sites function as bona fide hormone responsive elements. They flank several sequence motifs that are protected in vivo following hormone induction. One of them is a Y-box encompassing a reverse CCAAT motif, which is recognized by the ubiquitous transcription factor NF-Y. The NF-Y binding element is an essential part of a short enhancer region which is functional in various transiently transfected cell lines. Despite its ubiquitous expression, in vivo NF-Y binds to the uteroglobin enhancer only in endometrial cells and only after hormonal induction of the gene. This apparent paradox is likely due to the nucleosomal organization of the enhancer region. We show here that the relevant region is packed in positioned nucleosomes in the uninduced state. Hormonal induction leads to a loss of the regular nucleosomal pattern and to the appearance of nuclease-hypersensitive sites reflecting a remodelled chromatin structure. We hypothesize that these changes in chromatin structure may be the mechanism underlying the progesterone induced recruitment of the ubiquitously expressed transcription factor NF-Y, and other factors, to the upstream uteroglobin gene enhancer.

EXPERIMENTAL PROCEDURES
Animals and Treatments-Adult 1/2 to 1-year-old female rabbits (New Zealand White or Chinchilla-Bastard, 3-4 kg) were housed in individual cages under controlled conditions of temperature and light (12 h light-dark) and kept separated from male rabbits to prevent pheromone-triggered ovulation. Pseudopregnancy was induced by two * This work was supported by the Deutsche Forschungsgemeinschaft and Fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Present address: University of California, San Diego, Dept. of Medicine, 9500 Gilman Dr., La Jolla, CA 92093-0648.
Plasmids-For CAT reporters the uteroglobin upstream sequences were inserted into a unique HindIII site located in front of various deletion constructs of the uteroglobin promoter (Ϫ258 to ϩ14) as described in Ref. 18, but lacking the two copies of the SV40 enhancer. For the construction of the uteroglobin promoter driven luciferase reporter plasmid, the HindIII to XhoI (Ϫ35 to ϩ14) uteroglobin promoter fragment (18) was inserted into the pXP2 plasmid (19) and uteroglobin upstream sequences were inserted into the unique HindIII site. Similarly uteroglobin upstream sequences were inserted into the unique HindIII site of pT81luc (19), to generate the fusion constructs with the heterologous herpes simplex virus thymidine kinase promotor (Ϫ81 to ϩ52). The different upstream DNA fragments were generated by restriction enzyme cleavage of genomic clones (20), followed by a filling reaction with Klenow DNA polymerase or degradation of termini with T4 DNA polymerase, and blunt end ligation of HindIII linkers (Pharmacia, Uppsala, Sweden). The minimal activating fragment (Ϫ2523 to Ϫ2453) and its mutation were obtained by oligonucleotide-directed PCR. An expression plasmid for the rabbit progesterone receptor, pKSV10-rPR, was supplied by E. Milgrom, Paris (21).
Genomic DNase I and Dimethyl Sulfate Footprinting in Vivo-DNase I genomic footprinting was performed as described (22). For dimethyl sulfate genomic footprinting minor variations of a published procedure (23) were used. In brief, endometrial epithelium or liver cells were exposed to 0.2% dimethyl sulfate in Dulbecco's modification of EagleЈs minimal essential medium (DMEM) for 5 min. The reaction was stopped by washing several times with phosphate-buffered saline. Epithelial cells were separated and genomic DNA prepared. Modified DNA was cleaved with piperidine and submitted to ligation-mediated (LM)-PCR as described (22). The following gene-specific oligonucleotides were used. For the upper strand: En1, CTTTGCTTGATTGGCC; En2, CTT-GATGTTCACTAAACAGGCACCTTGG; En3, GCACCTTGGAACGAA-TCAGTGAACAGGCC; for the lower strand: EAI, GTCTTGTTCTCCC-CTCC; EAII, ATGCCTTGTTTTTCACTCTGCAGCC; EAIII, CTCTGC-AGCCTGCCTTGGCAATCATCTC.
EMSA and Methylation Protection Analysis in Vitro-Nuclear extracts for EMSA were prepared according to Andrews and Faller (24). The following double-stranded oligonucleotides were used: a E␣ gene fragment comprising the Y box (Ϫ65 to Ϫ44) (25); the uteroglobin gene enhancer fragment (Ϫ2523 to Ϫ2451) comprising the reverse Y box (Ϫ2495 to Ϫ2499), or the respective point mutant C to A at Ϫ2495 cloned in pBluescript (Stratagene Inc.) by ligation to HindIII linkers; an unrelated control oligonucleotide with the upper strand sequence 5Ј-GAAGATCTGTGGAAAGTCCCACTAGAGC-3Ј. The uteroglobin gene fragments were cut from the plasmid, purified by gel electrophoresis, and 32 P-labeled by Klenow filling-in reaction. Unincorporated nucleotides were removed through gel filtration. The antiserum against NF-Y B, pR␣YB (26), or an unrelated control serum, anti-Sp1 (27), were preincubated with nuclear extracts for 15 min at room temperature. Poly(dI-dC) and calf thymus DNA were added as nonspecific competitor. Binding buffer were as described (28). The binding reaction was incubated at room temperature for 15 min before loading on a 5% nondenaturing polyacrylamide gel. Gels were run in 0.5-fold Tris borate/EDTA buffer at 7 V/cm and room temperature for 120 min, dried, and autoradiographed. Methylation protection in vitro was performed as described previously (14).
Transient Gene Transfections-Ishikawa cells (29) were donated by E. Gurpide (Mount Sinai Hospital, New York). RBE-7 cells were from A. Mukherjee, National Institutes of Health, Bethesda (30). The cells were maintained in Dulbecco's modified EagleЈs medium (Life Technologies, Inc.) supplemented with 10% (Ishikawa) or 4% (RBE-7) fetal calf serum (c.c. pro GmbH, Germany), penicillin (100 IU/ml), and streptomycin (100 g/ml) and cultivated at 37°C (Ishikawa) or 33°C (RBE-7) and 5% CO 2 . Transient transfections were performed by the calcium phosphate DNA co-precipitation method (31). A 90-mm dish received 10 g of reporter plasmid, different amounts of co-expression plasmids, and sheared calf thymus DNA to a total of 20 g of DNA and the precipitate was left on the cells for 10 h. After 2.5 h of chloroquine treatment (0.1 mM) medium was changed and hormones were added directly to the culture medium which was supplemented with 10% fetal calf serum treated with dextran-coated charcoal (32). RBE-7 cells were then shifted to 39°C. Anti-estrogen, ICI 164.384 (10 Ϫ7 M), was eventually applied to noninduced cells to further reduce estrogen-mediated effects on basal reporter expression (33). Cells were harvested 48 h after transfection and extracts prepared. CAT, luciferase, and ␤-galactosidase activity were determined by routine procedures (34).
Details of the method used to isolate, cultivate, and transfect endometrial epithelium cells will be published elsewhere. 2 Briefly, the inner luminal surface of the uterus was incubated with collagenase at 37°C, flushed, and cells were collected by centrifugation. The suspension of cells was maintained in DMEM on polystyrene culture dishes. The medium was supplemented with 5% newborn calf serum (Life Technologies, Inc.), penicillin (100 IU/ml), and streptomycin (100 g/ml) and the cells were cultivated at 37°C and 5% CO 2 . The purity of the primary cell preparation was determined by immunocytochemistry employing an antibody against rabbit uteroglobin (35). Transient transfection was performed by the calcium phosphate DNA co-precipitation method (31). After transfection the cells were washed with phosphate-buffered saline and the medium changed for phenol red-free DMEM (Life Technologies, Inc.) supplemented with 5% fetal calf serum treated with charcoal dextran (32). Hormones were added to the medium as ethanolic solution immediately after transfection and incubation was continued for 48 h. R5020 (gift from Roussel UCLAF, Romainville, France) was used as synthetic progestin, an equivalent amount of ethanol as solvent control.
Mapping of Nucleosome and Nuclease Hypersensitive Sites-Digestion of cell nuclei with methidiumpropyl-EDTA-Fe II (MPE), separation of the DNA fragments in agarose gels, blotting, and indirect end labeling (36,37) were performed as described previously (22). Treatment of cell nuclei with DNase I was essentially as described (22). 6 g of genomic DNA was cleaved with NdeI, resolved on a 1.2% agarose gel, and blotted on Biodyne A Nylon membrane (Pall, Dreieich, Germany). Hybridization was carried out with a radioactively labeled (NdeI-StuI)uteroglobin gene fragment (Ϫ1843 to Ϫ2083) according to Church and Gilbert (38).
Quantitation of Radiolabeled DNA-Quantitative evaluation was performed directly from the dried gel using a PhosphorImager and ImageQuant software (Molecular Dynamics Inc., Sunnyvale, CA) or from autoradiographs using a laser scanner.

RESULTS
Tissue-specific and Hormone-dependent Occupancy of an Upstream Region around Ϫ2.5 kb in Vivo-To identify transcription factors bound to the uteroglobin enhancer we performed genomic footprinting analyses with DNase I and dimethyl sulfate in vivo. We focused on the upstream region of the uteroglobin gene between Ϫ2.7 and Ϫ2.4 kb that encompasses two clusters of non-canonical putative progesterone responsive elements (PRE) and was previously shown to exhibit characteristic steroid hormone inducible and tissue-specific DNase I-hypersensitive sites (17) (Fig. 1A).
Cell nuclei from the endometrial epithelium and liver were treated with DNase I and the digested DNA was analyzed by LM-PCR genomic footprinting (22,23,39). Close inspection of the cleavage patterns of liver nuclei and of naked genomic DNA revealed only small differences in the reactivity of individual cleavage sites ( Fig. 2A). Very likely, these differences result from the organization of the DNA sequences in chromatin and not from bound transcription factors (see genomic footprinting with dimethyl sulfate below). The DNase I cleavage pattern found in uteroglobin expressing tissue, i.e. induced endometrial epithelium of pseudopregnant animals ( Fig. 2A, lanes 5 and 6), was different from that of free DNA ( Fig. 2A, lanes 1 and 2) and from the pattern observed in non-expressing tissues, such as liver ( Fig. 2A, lanes 3 and 4) or endometrial epithelium from estrous animals (data not shown). In addition to the region with potential PREs (17), three other regions were strongly protected against DNase I cleavage (schematically indicated by boxes on the right margin in Fig. 2A). The DNase I footprints are flanked by nuclease-hypersensitive sites (indicated by arrows in Fig. 2A) and were absent in uninduced endometrial epithelium and in other nonexpressing tissues, such as liver. In addition, the induced endometrial epithelium exhibited a prominent cluster of DNase I-hypersensitivite sites between Ϫ2455 and Ϫ2450 (indicated by arrowheads in Fig. 2A), a region which also encompasses potential PREs (17).
To specify the DNA contacts of proteins bound to these regions in vivo we performed genomic footprinting with dimethyl sulfate (22,23). In cells which do not express the uteroglobin gene, such as liver cells, no difference in guanine methylation was observed compared with control DNA methylated in vitro, suggesting that there is no factor bound to the major groove of the DNA double helix over the enhancer region. A characteristic pattern of hypermethylated and protected guanine residues was detectable in epithelial cells of the endometrium from induced animals, suggesting binding of several transcription factors to the regions protected against DNase I digestion. Here we will focus on a region in the upper strand exhibiting a prominent pattern of two protected guanine residues, at Ϫ2495 and Ϫ2496, and a characteristic hypermethylation at Ϫ2493 in induced endometrium (Fig. 2B, compare lanes 1 or 2 with lane 4). The corresponding region in the lower strand encompasses the conserved sequence CCAAT of a Y box and did not show differences in the cleavage pattern due to the lack of guanine residues (data not shown). Although the changes in dimethyl sulfate reactivity were particularly prominent in the induced endometrium, a weaker hypermethylation at the guanine residue Ϫ2493 was also detectable in endometrial cells from adult estrous or ovariectomized rabbits (Fig. 2B, compare lanes 1 or 2 with lane 3, and data not shown), suggesting occupancy of this site in a subpopulation of cells known to express the uteroglobin gene under these conditions (11,40).
Transcription Factor NF-Y Binds to the Y Box of the Uteroglobin Gene-To identify potential nuclear proteins that bind to the Y box in the uteroglobin gene enhancer, we performed in vitro DNA binding experiments employing nuclear extracts from induced endometrial epithelium. In EMSA we observed a retarded complex that was specifically competed by an excess of the corresponding unlabeled oligonucleotide (Fig. 3A, lanes 4  and 5) but not by an unrelated control oligonucleotide (Fig. 3A,  lanes 2 and 3; B, lanes 4 -6). We also performed competition assays with a series of consensus oligonucleotides for different transcription factors (data not shown). Efficient competition was only observed with a Y-box oligonucleotide comprising the binding site for the heteromeric transcription factor NF-Y (25) (Fig. 3B, lanes 1-3). We tested the relevance of this element for complex formation by introducing a point mutation into the potential NF-Y-binding site of the uteroglobin enhancer. Substitution of the protected guanine residue at position Ϫ2495 by a thymine (Fig. 1B) considerably diminished the ability of the mutated oligonucleotide to compete for NF-Y binding in EMSA (Fig. 3A, compare lanes 6 and 7 to lanes 4 and 5). Thus the point mutation interferes efficiently with the DNA binding of the NF-Y complex in accordance with the importance of this guanine for NF-Y binding observed in binding site selection experiments (25).
The identity of the specifically shifted protein was further confirmed using a specific antiserum directed against the subunit B of the heteromeric NF-Y complex (26). Addition of the antiserum, but not of a control serum, led to a characteristic supershift of the retarded complex (Fig. 3B, lanes 7-9). This demonstrates that transcription factor NF-Y from induced endometrial epithelium specifically binds to the CCAAT box in the uteroglobin gene upstream region in vitro.
To verify that the transcription factor bound in vitro is identical to the protein that caused the footprint observed in vivo, we carried out a methylation protection analysis in vitro. We incubated a binding reaction containing the radioactively labeled uteroglobin gene fragment and nuclear extract from induced endometrial epithelium with dimethyl sulfate and sepa-  6 and 7). B, identification by EMSA of a Y box binding activity in nuclear extracts from induced endometrial epithelium. Nuclear extracts were incubated with a radioactive oligonucleotide encompassing the putative uteroglobin Y box. Competition assays were performed with increasing amounts of bona fide Y box oligonucleotide (lanes 1-3; Ref. 25 Fig. 2B). The lower strand is not shown as no changes in dimethyl sulfate reactivity were observed upon binding of nuclear proteins. rated bound and free oligonucleotides by EMSA. A guaninespecific cleavage reaction revealed changes in the methylation pattern of the factor-bound oligonucleotide (Fig. 3C). A doublet of protected guanine residues adjacent to a single hypermethylated guanine residue was detected in the upper strand, whereas no differences were detectable over the corresponding region in the lower strand (data not shown). The observed changes in reactivity of distinct guanine residues toward dimethyl sulfate in vitro coincided precisely with the methylation pattern obtained in vivo, suggesting that the protein bound in vivo is the nuclear transcription factor NF-Y (note that the orientation of the sequence in Fig. 3C is opposite to that shown in Fig. 2B).
The Uteroglobin Y Box Functions as a Ubiquitously Active Enhancer Element-We analyzed the potential function of the 5Ј-flanking region of the uteroglobin gene by measuring its ability to transcriptionally activate reporter genes in transient transfection experiments in primary epithelial cells from rabbit endometrium. A 3.9-kb fragment of 5Ј-flanking sequences fused to a luciferase reporter gene exhibited only moderate activity (Fig. 4, row 1), in accordance with the low expression level of the gene in the uninduced endometrial epithelium (41). Analysis of a series of 5Ј-and 3Ј-deletions (Fig. 4, rows 3-8) linked to the virtually inactive uteroglobin core promoter (row 2) delineated a core enhancer region that spans from Ϫ2523 to Ϫ2451 and encompasses the reverse Y box (Fig. 4, row 9). The 4-fold increase in transactivation observed after deletion of the sequences between Ϫ2288 and Ϫ2451 (compare rows 4 and 7) could be due to a distance effect or/and to the removal of negative regulatory elements. The point mutation at Ϫ2495 in the Y box, which was shown to abolish NF-Y binding (see Fig.  3A), consistently reduced the activity of the corresponding reporter gene construct (Fig. 4, row 10), indicating that NF-Y binds to the transiently transfected Y box of the uteroglobin enhancer and contributes to the transcriptional activation mediated by the enhancer.
The uteroglobin core enhancer element was active in combination with the uteroglobin promoter or the heterologous herpes simplex virus TK promoter (Table I). It activated to similar extent in cells of endometrial origin, including primary cells from rabbit endometrial epithelium, the rabbit endometrial cell line RBE-7 (30), the human endometrial adenocarcinoma-de-rived Ishikawa cell line (29), and in a cell line of unrelated origin, the African green monkey kidney cell line CV-1 (Table  I). Furthermore, the uteroglobin core enhancer functioned in both orientations and activated transcription downstream as well as upstream of the reporter promoter (data not shown). We conclude that in transient transfections the minimal region containing the NF-Y-binding site behaves as an enhancer in combination with different promoters and irrespective of the cellular context. The

FIG. 4. Functional identification and delineation of a constitutive enhancer fragment in the uteroglobin gene upstream region.
Delineation of a core enhancer region by deletion mapping and transient transfection in primary cells of the rabbit endometrial epithelium. Transfections were performed by the calcium phosphate DNA co-precipitation method. Symbols for factor-binding sites in the region around Ϫ2.5 kb are as in Fig. 1A. The relative luciferase activity generated by each construct, normalized for co-transfected ␤-galactosidase, is expressed in arbitrary units. An unspecific background of luciferase activity as determined by transfection of the promoterless reporter plasmid pXP2 was subtracted. The Y box mutant (row 10) contained the G to T exchange shown in Fig. 1B. The data correspond to a representative experiment performed several times (see also neously or after transfection of an appropriate expression vector. A representative result obtained with the human endometrial cell line Ishikawa (29) is shown in Fig. 5. In this particular experiment a fragment of the uteroglobin promoter extending up to position Ϫ258 was used to drive transcription of the CAT reporter. The promoter itself was not stimulated by addition of the synthetic progestin R5020 (Fig. 5, row 1). A DNA fragment encompassing the two clusters of receptor-binding sites as defined in vitro (17) conferred a 4-fold induction to the uteroglobin promoter in the presence of R5020 (Fig. 5, row 2), in agreement with the level of progesterone activation of the endogenous uteroglobin gene, as measured by nuclear run off analysis (10,11). We also tested various subfragments of the relevant region although no systematic analysis was performed. Each cluster of receptor-binding sites was functional independently (Fig. 5, rows 3-6), and the enhancer activity in the absence of hormone depended on the presence of adjacent transcription factor-binding sites, including the Y box (compare rows 2 and 5 with rows 3, 4, and 6). As a control, a small fragment just downstream of the PREs did not affect basal expression nor conferred homone induction (row 7). The minimal enhancer (Ϫ2523 to Ϫ2451) did not respond to hormone treatment either, although it included one of the three receptor-binding sites found in the proximal cluster (row 8). These results confirm the functionality of the two clusters of receptorbinding sites previously identified in vitro, and are compatible with the notion that the hormone independent activity of the enhancer is determined by the core enhancer sequence. Since the extent of hormonal induction is not increased in the presence of the Y box, the results do not provide evidence for a functional synergism between hormone receptors and NF-Y in transient transfection experiments. The question remains of why the uteroglobin Y box is occupied in vivo only after hormonal stimulation.
The Regular Nucleosomal Organization of the Upstream Enhancer Region Changes upon Hormone Induction-One explanation for the lack of binding of the ubiquitous transcription factor NF-Y to the uteroglobin Y box in the uninduced endo-metrial epithelium could be the organization of the DNA in chromatin. In this scenario, hormonal induction would remodel the chromatin structure and unmask a previously unaccessible site (42). To test this hypothesis we examined the chromatin organization of the uteroglobin gene enhancer region in nuclei from different tissues, by digestion with MPE, a reagent that preferentially cleaves nucleosomally organized DNA in the internucleosomal linker (43). In liver (data not shown) and in the non-induced endometrial epithelium we observed a pattern of regularly spaced preferential cleavage sites with a characteristic repeat length of 180 to 200 bp (Fig. 6A, lanes 2 and 3), indicating that the uteroglobin enhancer is covered by regularly spaced nucleosomes. The amount of of DNA applied to each lane and the extent of digestion were the same for uninduced and induced endometrium, as judged by ethidium bromide staining (data not shown). Since the signal corresponding to the uteroglobin enhancer is weaker after induction (Fig. 6A,  compare lanes 1-3 and 5-7), we assume that this chromatin region becomes more sensitive to MPE cleavage upon induction. The samples of hormone-induced endometrial cells display a classical nucleosomal ladder in ethidium bromidestained gels (Fig. 6B), but this pattern is not detectable after hybridization with a probe specific for the uteroglobin enhancer region (Fig. 6A, lanes 6 and 7), confirming that the chromatin organization of the enhancer region is altered after induction.
We tested the uteroglobin gene enhancer region for the presence of translationally positioned nucleosomes by indirect endlabeling analysis of the MPE-digested samples (36,37). In the uninduced endometrial epithelium we found a diffuse but regular pattern of cleavage sites over the uteroglobin enhancer indicative of preferentially positioned nucleosomes (Fig. 7A, left  panel, lanes 1-3). The unsharp nucleosomal pattern could reflect a certain heterogeneity of chromatin structure, with nucleosomes positioned only in a subpopulation of cells. A densitometric scan of the autoradiogram confirmed the periodical spacing of the cleaved fragments with an internucleosomal repeat length of about 200 bp (Fig. 7A, right panel, scan of lane  3). Upon hormone induction the MPE cleavage pattern over the  (21) using the calcium phosphate DNA coprecipitation method. The indicated fragments of the uteroglobin gene enhancer region were inserted upstream of its truncated promoter (Ϫ258 to ϩ14). A RSV-lacZ reporter was co-transfected for normalization of the data. Cells were treated with 10 nM R5020 (black bars) or with the ethanol solvent (shadowed bars). The numbers indicate the normalized CAT activity and are the result of a representative experiment performed several times.
upstream enhancer changed dramatically in the endometrial epithelium, where a new less regular distribution of cleavage sites appeared, including several hypersensitive sites (Fig. 7A,  lanes 4 -7). Mapping the regions of MPE cleavage by Phospho-rImager analysis showed three types of changes. First, the cleavage sites found prior to induction became broader and more diffuse. Second, new hypersensitive cleavage sites appeared over the enhancer region between Ϫ2580 and Ϫ2180. Third, two additional clusters of MPE-hypersensitive regions appeared after induction: one made of a doublet of strong sites at Ϫ3.7 kb and a second one at Ϫ5.1 kb. These findings are compatible with a hormone-dependent long range rearrangement of the chromatin organization over the uteroglobin gene upstream region.
In addition to the MPE cleavage experiments we carried out DNase I digestion of nuclei from the different tissues of hormone-induced animals, and compared the chemical and enzymatic cleavage patterns. In agreement with previous results (17), we observed several DNase I-hypersensitive sites which were specific for the hormonally induced endometrial epithelium (Fig. 7B, compare lanes 6 and 5). In particular, we observed three tissue-specific DNase I-hypersensitive regions located around Ϫ2.5, Ϫ3.7, and Ϫ5.1 kb. The first two cover a broad zone and comprise distinct cleavage sites. Interestingly, the DNase I-hypersensitive zones map to similar positions as the MPE-hypersensitive sites (Fig. 7B, compare lanes 3 and 4  to lane 6). These results show that hormone induction in the endometrial epithelium is followed by a broad change in the regular nucleosomal organization of DNA in the 5Ј-flanking region of the uteroglobin gene leading to better accessibility of the double helix to chemical cleavage reagents and nucleases. DISCUSSION The Experimental in Situ Approach-To understand transcriptional regulation of the uteroglobin gene we have applied in vivo DNase I and dimethyl sulfate footprinting to living cells within the intact rabbit uterus. Earlier approaches using in vitro DNA binding assays and transient transfection of immortalized cells have been largely unsuccessful, because cell lines established from uteroglobin-positive epithelia loose significant expression of the uteroglobin gene in culture (Ref. 44 and references therein). In the few available cell lines that maintain expression of the gene (30,45), the levels of uteroglobin mRNA are very low and significant hormonal induction cannot be reproduced even when expression vectors for the corresponding hormone receptors were stably transfected. 2 This lack of transcriptional competence in cell lines leaves primary cells or native tissue in situ as the only options to study the regulation of uteroglobin gene expression in vivo. Preliminary experiments showed that epithelial cells from the endometrium of hormonally induced rabbits expressed uteroglobin and uteroglobin mRNA at high levels (Refs. 46 -48). 2 Thus, these cells provide an appropriate experimental system to study the protein-DNA interactions taking place in the potentially relevant regions of the uteroglobin gene. In this study we have focused on the induction of the uteroglobin gene by progestins, mediated by the upstream enhancer region encompassing two clusters of receptor-binding sites (17).
Hormone-dependent Occupancy of a Y Box in the Uteroglobin Gene Enhancer-The results summarized here demonstrate that a DNA fragment between 2.7 and 2.3 kb upstream of the start of transcription of the uteroglobin gene, which encompasses two clusters of binding sites for glucocorticoid and progesterone receptors (16,17), can function as a progesteroneresponsive enhancer in transient transfections. Moreover, deletion analysis identified a short core enhancer located between the two clusters of PREs, which is able to activate heterologous promoters in various cell lines independently of the tissue origin and of hormone treatment. Thus, the corre-FIG. 6. Analysis of the nucleosomal organization in the uteroglobin upstream region by MPE-Fe II cleavage. Nuclei of endometrial epithelium from uninduced (End unind.) or hCG-induced (End. (hCG))animals were treated with MPE-Fe II for increasing times. A, the spacing of nucleosomes over the uteroglobin upstream enhancer region was analyzed by electrophoresis of the MPE-Fe II cleavage products from 6 g of genomic DNA through a 1.2% native agarose gel, followed by blotting onto Nylon membrane and hybridizing with a probe which spans the enhancer region (Ϫ2606 to Ϫ2288). Lanes 1-3 correspond to uninduced; lanes 4 -7, to induced endometrial epithelium. The arrowheads on the left margin point to the cleavage maxima, and the numbers indicate the size of the fragments in base pairs. As size markers we used genomic DNA restricted with NdeI (C) or with PstI (M), and a radioactively labeled 100-bp ladder (M100). Numbers indicate the size of the fragments in base pairs. B, the degree of cleavage and the loading of the lanes was estimated by ethidium bromide staining of an aliquot of the MPE-Fe II cleavage products after separation by electrophoresis on a 1.2% native agarose gel. Lanes 1-4 correspond to lanes 4 -7 of A. The position of the mono-, di,tri-, and tetranucleosome bands are indicated on the right margin (1N, 2N, 3N, and 4N, respectively).
sponding trans-acting factors must be widely expressed and operate independently of progesterone receptor. Sequence analysis revealed a reverse Y box with the consensus CCAAT (26), between Ϫ2499 and Ϫ2495 within the core enhancer region. A single base mutation of this site, which eliminates binding of NF-Y in vitro, also abolished the enhancer activity, suggesting that NF-Y, or a related factor, participates in uteroglobin enhancer function.
DNase I genomic footprinting over the uteroglobin upstream enhancer in the hormonally stimulated animals reveals three protected regions, one of them covering the putative Y box. The identity of the factor protecting the Y box was further explored in dimethyl sulfate footprinting experiments with intact endometrial cells. We found a pattern of protected and hypermethylated guanine residues corresponding to that previously reported for binding of NF-Y (49) and indistinguishable from the pattern observed in in vitro dimethyl sulfate footprinting experiments with the NF-Y protein complex present in nuclear extracts. The strong DNase I-hypersensitive sites flanking the Y box in genomic footprinting experiments with induced endometrium could reflect the known deformation of the DNA helix accompanying NF-Y binding (49). The binding of NF-Y in vivo strictly followed hormone induction, despite the constitutive expression of NF-Y in various tissues and under different hormonal conditions. Although we have not formally excluded that hormone treatments selectively increase the NF-Y levels in endometrial cells, our gel shift experiments with nuclear extracts from different tissues support ubiquitous and constitutive expression of NF-Y. 2 These results seem to contradict results from transient gene transfection experiments in which the NF-Y-binding site of the uteroglobin enhancer mediates constitutive transactivation independent of the hormone receptor-binding sites. This indicates that NF-Y binds efficiently to the Y box in the transiently transfected uteroglobin enhancer sequences, whereas it does not access this site in the endogenous uteroglobin gene. A similar behavior has been described for binding of NF1 to the MMTV promoter. Whereas NF1 is unable to bind to the nucleosomally organized MMTV promoter (50 -53), the NF1-binding site of transiently transfected MMTV promoters was found to be constitutively occupied by NF1, probably reflecting the poor chromatin organization of transfected DNA (54). These observations suggest that while transient transfection experiments are adequate for identifying potential cis-regulatory elements, their usefulness in the functional characterization of such elements is questionable. Moreover, our findings underline the significance of analyzing endogenous genes in their natural chromatin context. hancer in uninduced endometrial cells. We have previously detected the appearance of DNase I-hypersensitive sites over the uteroglobin enhancer region in the endometrium of progesterone-treated rabbits, which indicates a change in chromatin structure accompanying hormone induction (17). A detailed analysis of the MPE cleavage pattern over the enhancer region prior to hormone induction reveals a regular nucleosomal organization with a significant population of genes exhibiting at least five preferentially positioned nucleosomes. Within the range of resolution of our analysis (Ϯ20 bp), the results locate the core enhancer sequence, and in particular the Y box, in the linker DNA flanked by two nucleosomes, each containing one of the two PRE clusters. Following hormonal stimulation the regular nucleosomal organization is no longer recognizable over the whole region of the enhancer and several clusters of MPE and DNase I-hypersensitive sites appear. One of these nuclease-sensitive clusters covers the core enhancer region and the PREs. Here, in addition to a higher nuclease sensitivity and broadening of the cleavage over the linker DNA, we observe cleavage by MPE over the central region of the two positioned nucleosomes, suggesting a remodeling of their structure. These findings are compatible with the notion that progesterone receptor binding initiates a process of chromatin remodeling which makes the uteroglobin enhancer more accessible for nucleases as well as for transcription factor, and could explain the hormone dependent recruitment of NF-Y. The tissue specificity could in part result from the high concentration of progesterone receptor in epithelial cells of the endometrium.

Recruitment of NF-Y Is
Further upstream in the uteroglobin 5Ј-flanking region, two additional regions of nuclease hypersensitivity are detected, at Ϫ3.7 and Ϫ5.1 kb, whose functional significance is unknown. The region at Ϫ3.7 coincides with a DNase I-hypersensitive region found in the endometrium of rabbits treated with estrogen and progesterone, and contains a putative PRE, able to bind glucocorticoid and progesterone receptor in vitro (17). This region could, therefore, contribute to the hormonal induction of the gene. The sequence of the Ϫ5.1-kb nuclease-hypersensitive region is unknown and its function has not been investigated. It is possible that, as in the rat tyrosine aminotransferase gene (55), several widely spaced enhancer elements exist in the uteroglobin gene. As the gene is differentially regulated in various tissues (3,4), the different enhancers could selectively participate in the physiological regulation of the uteroglobin gene in different epithelial cells.
The situation in the uteroglobin enhancer is reminiscent of that found in the MMTV promoter, but there are significant differences. In the MMTV promoter, as in the uteroglobin enhancer, hormone receptor-dependent chromatin remodeling is required for binding of a ubiquitous transcription factor, namely NF1 (22,42). It is known that steroid hormone receptors can bind to positioned nucleosomes provided the rotational orientation of the major groove over the hormone-responsive elements is appropriate (50,56). Although the orientation of the receptor-binding sites in the uteroglobin enhancer is not known, we assume that at least part of the sites are properly oriented and accessible, as this is a prerequisite for receptor binding and initiation of the hormone-dependent chromatin remodeling. There is biochemical and genetic evidence suggesting that receptor binding to the MMTV promoter does not lead to removal or disruption of the positioned nucleosome but rather to a more subtle and localized remodeling of a single nucleosome that enables NF1 to bind (22,57). In contrast, the changes induced by hormonal stimulation within the uteroglobin gene enhancer are more extensive, affecting at least five nucleosomes and leading to the loss of regular nucleosomal spacing over the enhancer region. These changes may result from the presence of two clusters of receptor-binding sites and may be required for binding of NF-Y and other factors to the linker DNA. NF-Y appears to be unable to bind at the linker between two positioned nucleosomes within a regular nucleosomal array, likely because it binds as an heterotrimer that bends DNA rather dramatically upon binding (49). Therefore, changes in the higher order structure of chromatin, determined by changes in the linker histones or in the interactions between nucleosomes, may be required for facilitating access to the NF-Y-binding site (58). Similar changes affecting a large nucleosomal region are induced by binding of an heterodimer of thyroid hormone receptor and retinoic acid X receptor to the Xenopus TR␤A gene (59).
How steroid hormones contribute to nucleosome remodeling is not known, but several possibilities can be envisaged. Remodeling could be mediated by co-activators, such as CBP/ p300, which are known to interact with nuclear receptors (60,61) and have been shown to exhibit histone acetylating activity (62,63). CBP could be recruited directly by the nuclear receptors or via other co-activators, such as SRC-1/ACTR (64,65), which interact on the one side with hormone receptors and on the other with CBP (66) and are themselves histone acetyltransferases (65,67). CBP/p300 itself can recruit an additional histone acetyltransferase, P/CAF (68), and, thus, contribute to further acetylation of core histones or other transcription factors (69). In that respect, it is interesting to note that the MMTV and the HIV-1 promoters can be activated by a moderate increase in histone acetylation, which also leads to remodeling of the positioned nucleosome (70,71). There are also indications for a participation in hormonal induction of the ATP-dependent SWI⅐SNF complex (72,73), which has the capacity to remodel nucleosomes (74). A direct interaction between the hormone receptors and members of the SWI⅐SNF complex has been reported (72,75,76), suggesting that binding of the receptors to chromatin could target the SWI⅐SNF complex to hormone responsive promoters and initiate the remodeling process.
The biochemical nature of the hormone-induced chromatin remodeling is not known, but in vitro data suggest that a dissociation of H2A/H2B dimers from the remodeled nucleosome could facilitate transcription factor access to nucleosomal sequences (77). Since NF-Y B and NF-Y C are homologous to histone H2B and H2A, respectively (78,79), and form a heterodimer in the heterotrimeric NF-Y (80), it is possible that these two subunits of NF-Y replace one histone H2A/H2B dimer and generate a chromatin structure that enables the NF-Y heterotrimer to interact with the Y box (81). Thus, NF-Y could assume the role of an architectural factor, as reported for HNF3␣ in organizing the chromatin structure of the rat albumin enhancer element in the liver (82).
As for the interaction between the enhancer and the promoter region of the uteroglobin gene, it has been shown that a fraction of NF-Y B and NF-Y C interact with TBP (83), and therefore could provide a link to the general transcriptional machinery on the promoter. Another interesting possibility is suggested by the recent finding that activated progesterone receptor interacts with Sp1, directly or indirectly via CBP (84). We have previously shown that estrogen induction of the uteroglobin gene depends on a functional interaction between estrogen receptor and Sp1 bound to adjacent sites in the promoter region (14), and a ligand-dependent interaction of the estrogen receptor with CBP has been reported. One could envision a bridging role of Sp1 and the coregulator CBP (85) between DNA bound progesterone and estrogen receptors which will establish a chromatin loop between the enhancer region and the promoter (86). This model will be consistent with the ob-servation that both the enhancer and the promoter regions become hypersensitive to DNase I during physiological induction of the uteroglobin gene in rabbit endometrium (17).