INTRODUCTION

Regulation of transcription in mammalian cells is strongly influenced by the presence of nucleosomes within promoters and by the ensuing higher order chromatin structures. The nucleosomal organization of a promoter is defined by the rotational and translational positioning of nucleosomes on DNA. If a nucleosome occupies a fixed position relative to the underlying sequence of DNA, the nucleosome is said to be translationally positioned (1, 2). Likewise, an invariant orientation of the DNA helix relative to the surface of the histone octamer with respect to a specific sequence of DNA defines it as rotationally positioned. Several distinct steps of the transcription process, as dissected in vitro, can be affected by nucleosomes. One consequence of nucleosomal packaging of DNA can be limited accessibility of the DNA to DNA-binding proteins (for review see Ref. 3). Particular transcription factors, such as TATA binding protein (TBP)¹ (4), NF-1 (5-8), HSF (9), and GAL4 (9), bind with a lower affinity to sites located on a nucleosome core particle compared with sites on DNA alone. Accessibility can be influenced further by the rotational positioning of the DNA, with some alignments completely precluding factor binding (4, 10). Since transcription factors vary in their ability to overcome the limited accessibility of translationally and rotationally positioned nucleosomal DNA, the interplay between transcription factors and the nucleosome positioning at target promoters may define the specific role of each factor in the activation of gene expression (9).

In some cases, transcription factors that are able to bind nucleosomal DNA sites can facilitate the binding of additional factors to the promoter (11-15). Such transcription factors may either alter chromatin structure directly or target chromatin remodeling proteins such as SWI/SNF (16), NURF (17), RSC (18), or CHRAC (19) to particular sites. As a clear demonstration of this phenomenon, binding of glucocorticoid receptor to a nucleosomal site in the mouse mammary tumor virus promoter results in a rapid structural alteration in the nucleosome and the concomitant binding of NF-1 (5, 20, 21), which cannot normally bind its site in the nucleosome core (7).

Although often a general repressive force that must be overcome for transcriptional activation, as described above, nucleosomes have also been implicated in facilitating the transcriptional activation process (14, 22, 23). One example is the vitellogenin B1 promoter, which is transcriptionally potentiated upon assembly into a chromatin structure in vitro (24). Likewise, the tissue-specific, serum albumin enhancer exists in a precisely positioned nucleosomal array only in the liver, in which the enhancer is active (25). In both cases, it was suggested that nucleosomes may assist in the transcriptional activation process by spatially juxtaposing factor binding sites and facilitating protein-protein interactions (see also Refs. 26 and 27). Proteins bound to sites separated by approximately 200 bp can be brought into close proximity when a nucleosome core loops out the intervening DNA (28). The compaction state of the chromatin must also modulate protein-protein interactions, as the relative positions of proteins bound to adjacent nucleosomes are determined both by the specific placement of binding sites upon each nucleosome and by positions of the nucleosomes in a higher order structure (29).

To decipher the intricate complexities of the involvement of chromatin structure in transcriptional regulation, it is of the utmost importance to understand the in vivo nucleosome arrangement at a given promoter. To that end, we have analyzed the chromatin structure of the human pS2 promoter, an excellent model system for the study of transcriptional activation from a chromatin template by an inducible transcription factor. Transcription of the pS2 promoter (30) is activated by estrogen receptor (ER) (31), which binds to a 13-base pair imperfectly palindromic sequence located at 405 to 393, approximately 375 bp from the TATAA box (32).

A role for chromatin in the transcriptional regulation of the pS2 promoter is suggested by the characteristics of ER-dependent transcription from other nucleosomal templates. First, the binding of ER and resulting estrogen-dependent transcription from artificial promoters in yeast (33-35) and from the rat prolactin gene on a bovine papilloma virus episome (36) result in a disruption of chromatin structure at these promoters. Second, ER-dependent transcription in yeast is dependent upon a functioning SWI·SNF complex (37). Finally, in the previously mentioned vitellogenin B1 promoter, reconstituted nucleosomes potentiate activation of transcription by ER (24).

We have mapped at a nucleotide level of resolution the locations of two nucleosomes and a linker region in the human pS2 promoter, in a region encompassing the ERE and the TATAA box from nucleotides 440 to +40. This is the first in vivo detailed characterization of chromatin structure of a natural, estrogen-responsive promoter. Additionally, we have identified particular chromatin features that correlate with the transcriptional potential of the pS2 promoter. Thus, our data provide the basis for understanding the role of chromatin in the transcriptional regulation of the pS2 promoter.

EXPERIMENTAL PROCEDURES

Three different human breast tissue cell lines were used. Normal human mammary epithelial cells (HMEC), specimen name 184 spiral, were graciously provided by Martha Stampfer. These cells were propagated at 1% CO₂ in HEPES-buffered Mammary Epithelium Cell Basal Medium supplemented with Growth Medium SingleQuots (Clonetics). Two human breast adenocarcinoma cell lines, MDA-MB 231 and MCF-7, were propagated in minimum essential medium  modification with 10% fetal calf serum at 5% CO₂.

CMT cells (an African green monkey kidney cell line that stably produces SV40 T antigen) (38) were grown in Dulbecco's modified Eagle's medium with 10% fetal calf serum at 5% CO₂; these cells were used for propagation of SVSpS2 viral stocks.

For deoxyribonuclease I (DNase I) and micrococcal nuclease (MNase) treatment, HMEC 184, MDA-MB 231, and MCF-7 cells were grown as monolayers. Proliferating MDA-MB 231 and MCF-7 cells were grown in phenol red-minus EX-CELL 320, serum-free media (JRH Biosciences) for 72-90 h prior to nuclease digestion to eliminate any estrogenic activity from the medium. The anti-estrogen ICI 164,384 (ICI Pharmaceuticals) was added at 1 µM for 24-72 h prior to nuclease digestion, while 100 nM -estradiol (Sigma) was added either 1 h or 14-24 h before nuclease digestion. HMEC cells were grown in their original media until nuclease digestion. All cells were prepared and digested as described previously (39). Briefly, cells were permeabilized by treatment with 0.05% lysolecithin (Sigma) and then treated for 2-8 min with 800-1600 units of MNase (Worthington) in 6 ml or for 2-10 min with 20-1000 units of DNase I (Worthington) in 5 ml at either room temperature or 37 °C. Following addition of a stop buffer containing SDS and proteinase K, the cell extract was treated with RNase, extracted multiple times with phenol-chloroform, and precipitated with ethanol.

Control, MCF-7 genomic DNA was isolated in the same manner as the nuclease-treated genomic DNA. This deproteinated DNA was then treated in vitro with either dimethyl sulfate (DMS)/piperidine, DNase I (0.0125-0.05 units in 0.4 ml for 4 min at 37 °C), or MNase (0.05-0.2 units in 0.4 ml for 4 min at 37 °C). The in vitro digestion times and enzyme concentrations were determined experimentally to mimic the extent of total genomic DNA digestion in the in vivo digestion experiments. The extent of DNA digestion was monitored by electrophoresis of ethidium bromide-stained DNA through agarose gels.

LMPCR (40) was performed on DNase I and DMS/piperidine-cleaved DNA (5 µg) essentially as described (41). All reactions were carried out with Vent polymerase (New England Biolabs) (42). For the initial primer extension step, a 5-biotinylated primer was used for extension product capture (43). Following ligation of the linker oligonucleotides, the resulting template was amplified through 18 cycles of PCR. A portion of the PCR product was subjected to seven rounds of end-labeled linear PCR amplification (25) and was visualized on pre-flashed Reflection NEF-495 autoradiography film (DuPont) following electrophoresis through a 6% polyacrylamide, 7 M urea gel (1:15, bisacrylamide/acrylamide).

For LMPCR of MNase-digested DNA, two changes were made in the protocol used for DNase I. LMPCR of DNase I-treated samples examined single-stranded nicks in the DNA, thereby necessitating the initial primer extension reaction to produce the blunt end required for ligation of the unidirectional linker. However, only double-stranded cleavages from MNase-treated samples, which produce the blunt end for ligation without further manipulation, were analyzed. In addition, MNase-treated DNA (5 µg) was phosphorylated with 10 units of T4 polynucleotide kinase (New England Biolabs) supplemented with 0.1 mM ATP for 1 h at 37 °C before the ligation reaction.

Optimal hybridization temperatures were experimentally determined to be the calculated T_m (41) of the primer + 3 °C. The sequences of the specific primers, their hybridization temperatures (in parentheses), and the terminal nucleotide positions of each set (in parentheses) are as follows: NT-T1, 5-biotin CAG GGC GCA GAT CAC CTT GTT CT (60 °C), NT-T2, 5-GGC CAT TGC CTC CTC TCT GCT CCA A (68 °C), and NT-T3, 5-CCT CTC TGC TCC AAA GGC GAC CCC GAG T (72 °C) (+7); NE-T1, 5-biotin GGT CAT CTT GGC TGA GGG ATC TG (60 °C), NE-T2, 5-CCC TCT TTC CCA TGG GAG TCT CCT C (68 °C), and NE-T3, 5-CCC ATG GGA GTC TCC TCC AAC CTG ACC (71 °C) (264); NE-B1, 5-biotin GCT TAG GCC TAG ACG GAA TGG (60 °C), NE-B2, 5-GGA ATG GGC TTC ATG AGC TCC TTC CC (68 °C), and NE-B3, 5-GGG CTT CAT GAG CTC CTT CCC TTC CCC C (72 °C) (419); and NTV-T1, 5-biotin ACA CAG TTG TGT CAA AAG CAA (60 °C), NTV-T2, 5-CAA GTG TAA GCA GCT GCC AAG CTA (63 °C), and NTV-T3, 5-GTA AGC AGC AGC CAA GCT AAT TCC CAG (67 °C) (+7). The same unidirectional double-stranded linker was used with all primer sets, 5-GCG GTG ACC CGG GAG ATC TGA ATT C and 5-GAA TTC AGA TC (41).

The SV40/pS2 recombinant plasmid, pSVSpS2, was constructed by cloning the human pS2 promoter (1100 to +10) (44) linked to a partial rabbit -globin cDNA (+213 to +362, +664 to +1553) in the place of the T antigen coding region, SV40 fragment BclI-(2770) to EcoNI-(5027). The parental recombinant SV40 plasmid consisted of SV40 776 DNA in a derivative of pBR322 (45), in which an XhoI site was inserted into one of the HpaI sites of SV40 DNA-(2666). The pS2--globin fragment was inserted into this XhoI site in an antisense direction with respect to the SV40 early promoter. The deletion in the rabbit -globin cDNA from 362 to 664 (a BspMI site to a Bpu1102I site, joined by a linker adding the sequence 5 CTGTTGGC) was constructed to ultimately maintain the appropriate viral genome packaging size.

The recombinant plasmid pSVSpS2 was partially digested with limiting amounts of EcoRI to remove pBR322 DNA. EcoRI-digested DNA of the appropriate size was circularized by ligation and transfected by limiting dilution into CMT cells to initiate viral production. Viral DNA from individual wells of transfected cells was analyzed to ensure the desired genomic structure.

Viral DNA was prepared essentially as was the genomic DNA, described above (39). Host CMT cells were permeabilized and treated with DNase I or MNase. However, the CMT cells were infected with SVSpS2 36 h prior to the permeabilization and nuclease digestion. Subsequently, viral DNA, along with some smaller fragments of host genomic DNA, was isolated by the Hirt extraction procedure (46). DNA in the supernatant was precipitated with ethanol and used as the template for the LMPCR procedure. Due to the mixture of viral and genomic DNA, the amount of DNA for LMPCR analysis was determined experimentally, as a percentage of the total DNA isolated from a fixed number of cells.

Nucleosome core reconstitution was performed by the serial dilution method (47). Radiolabeled DNA was amplified by PCR from 460 to +56 of the pS2 promoter contained in the virus SVSpS2, using primers NTV-T1 and NE-B1. The DNA (0.5 to 1 µg) was incubated with 5 µg of purified HeLa cell histone octamers (48) to form nucleosomes. Purified DNA or reconstituted nucleosomes (30 ng) were digested with 1-500 units of exonuclease III (New England Biolabs) as suggested by the manufacturer, for 2-8 min at 23 °C. The resulting products were precipitated and electrophoresed through a 6% polyacrylamide, 7 M urea gel prior to autoradiography.

RESULTS

To understand the role of chromatin in the transcriptional regulation of the estrogen-inducible pS2 promoter, we examined the nucleosome structure of the pS2 promoter in human breast cells derived from both normal and carcinomatous breast tissue. The cells differed both in the levels of ER and in the inducibility of pS2 mRNA. The normal HMEC, 184 EP cells (49), do not contain detectable levels of ER or of pS2 mRNA, even upon incubation with estradiol (50). The two tumor cell lines that we examined, MDA-MB 231 and MCF-7, were derived from human breast adenocarcinomas. Similarly to HMEC cells, MDA-MB 231 cells do not express detectable levels of either ER protein (51) or pS2 mRNA (data not shown). In contrast, the mammary epithelial-like MCF-7 cells express high levels of ER (52). In these cells, the pS2 gene is highly inducible by estradiol (30). Under the conditions described under "Experimental Procedures," we observe an average of 10-fold induction of pS2 mRNA by 100 nM estradiol, as opposed to ethanol carrier (data not shown). Maximal induction ratios require the exclusion of estrogenic activity from the non-inducing cell culture media. However, even under non-inducing conditions, basal levels of pS2 mRNA are evident in MCF-7 cells as compared with the non-expressing MDA-MB 231 and CMT cells (data not shown).

The chromatin structure of the endogenous, single copy pS2 promoter was defined by treating human breast cells with nucleases sensitive to nucleosome positioning and resolving the specific cleavage sites at the nucleotide level by LMPCR. A positioned nucleosome at the TATA box in the pS2 promoter is likely to influence transcriptional regulation from this promoter, due to the differential binding affinity of TBP for a naked versus nucleosomal template (53) plus the additional influence on TBP binding of the orientation of the TATA sequence relative to the surface of the histone octamer (4). We first established the rotational phasing of nucleosomes covering the pS2 promoter by digesting genomic DNA with DNase I, which reveals the 10-base pair periodicity of the double helix when DNA is positioned on the surface of a histone octamer through a 10-bp ladder of alternating nuclease-accessible and nuclease-protected sites (54). Therefore, a triple-nested primer set adjacent to the TATA box in the pS2 promoter and reading into the upstream sequences of the promoter was initially designed for LMPCR. This primer set is designated NT-T (+7): (+7) indicates that nucleotide +7 of the pS2 promoter corresponds to the 3 nucleotide of the end-labeled primer, NT stands for the nucleosome at the TATA box and -T stands for cleavages in the top strand of the promoter DNA. With NT-T (+7), DNase I treatment revealed a 10-bp ladder in the genomic pS2 promoter in all three cell types, between nucleotide positions 28 and 150 (arrows, Fig. 1, lanes 4-11 and 13-16). Bases hypersensitive to or protected from DNase I cleavage in the chromosomal context were established by comparison with cleavage products from genomic, deproteinized DNA digested with DNase I in vitro (Fig. 1, lanes 2, 3, 17, and 18).

[View Larger Version of this Image (41K GIF file)]

The edge of the 10-bp periodicity of DNase I accessibility was a pair of adjacent cleavage sites at nucleotides 27 and 28, located within the TATA box (Fig. 1, lanes 4-11 versus lanes 2 and 3, designated by the bottom arrow adjacent to lane 11). Overall, the ladder of DNase I cleavages spans 123 bp in this region of the pS2 promoter, covering the TATA sequence. This 10-bp repeating pattern from 27 to 150 was revealed in MCF-7, MDA-MB 231, and 184 cells (arrows, Fig. 1). However, the MCF-7 genomic pS2 promoter DNA had a slightly altered accessibility to DNase I within this region compared with the other cells (Fig. 1, lanes 7-10 and 15 and 16 versus lanes 4-6, 11, 13, and 14, labeled CS₁). The 10-bp repeat of hypersensitivities to DNase I relative to naked DNA was still obvious (Fig. 1, arrows and dashes within CS₁ bracket), yet the contrast between this repeat and the intervening sequences was softened by cleavages within the protected areas of the 10-bp ladder in the MCF-7 cells, which were not accessible in the other cells examined. An especially distinctive, enhanced sensitivity to DNase I cleavage was apparent at nucleotides 41 and 42 (bottom of the bracketed CS₁ area). This altered accessibility was highly reproducible among repeated genomic DNA digestions and LMPCR amplifications.

The translational positioning of the nucleosome(s) encompassing this rotationally phased nucleosomal DNA was probed with MNase, which preferentially cuts DNA in the linker regions between nucleosomes (55). When LMPCR analysis is limited to double-stranded MNase cleavages, MNase-hypersensitive sites figure prominently at the edges of nucleosomes (25). MNase digestion of the genomic pS2 promoter in cells revealed several double-stranded cleavages, not observed upon digestion of deproteinized DNA (Fig. 2), bracketing the rotationally phased DNA (28 to 150). The downstream edge of the nucleosome covering the TATA box (nucleosome T) was demonstrated by a pronounced set of bands located at nucleotides 22, 23, and 24, within the 3 edge of the TATA box (Fig. 2A, lanes 2-11 and 15-19, as compared with lanes 1 and 12-14; lower starred bracket). This 3 boundary is located 4 bp downstream from the beginning of the DNase I ladder (Fig. 2C, position of star (lane 9) compared with arrows (lanes 4-7)). Hypersensitive sites around 165 defined the 5 boundary of nucleosome T (upper starred bracket, Fig. 2A; expanded in Fig. 2B), which is located the length of a nucleosome, 142 bp, upstream of the 3 boundary. The DNA outside the 5 edge of nucleosome T, linker TE, was defined by a series of novel and hypersensitive MNase double-stranded cleavages in the region of the promoter between nucleotide positions 165 and 241 (bracket, Fig. 2A, lanes 2-11; expanded in Fig. 2B). Comparison of the MNase digestion patterns in the different cells (MCF-7, MDA-MB 231, and 184) demonstrated that nucleosome T is translationally positioned in all cases (Fig. 2A, lanes 2-11). Of note, estrogen-induced, maximal transcriptional activation of the promoter (Fig. 2A, lanes 3, 5, and 19) did not displace nucleosome T, even though the TATA box is contained within the extreme 3 edge of the nucleosome. The presence of nucleosome T was confirmed at both 1 h after estradiol addition (Fig. 2, lane 19) and between 14 and 23 h after induction (Fig. 2, lanes 3, 5, 15, and 16). Transcriptional activity of pS2 is maximized at 1 h after estrogen treatment and is maintained at this level for at least 24 h, as determined by a nuclear run-on assay (56).

[View Larger Version of this Image (46K GIF file)]

The chromatin structure of DNA encompassing the ERE of the pS2 promoter is likely to determine the accessibility of this binding site to ER, which stimulates pS2 transcription. To investigate the nucleosomal positioning of the genomic DNA surrounding the ERE at 400, two triple-nested primer sets were designed to probe both the top (NE-T (264)) and bottom (NE-B (419)) strands of this region (NE stands for nucleosome encompassing the ERE). First, chromosomal DNA was digested with DNase I to determine the rotational phasing of nucleosomal DNA in the region. The location of the NE-B (419) primer set was technically restrained by the repetitive Alu sequence DNA 85 bp upstream of the ERE (52), restricting the extent of upstream DNA that could be examined. Nonetheless, a 10-bp ladder of DNase I cleavages was visualized from 398, within the ERE, to 278 (arrows, Fig. 3A, lanes 4-14). From the top strand, where the upstream end of the 10-bp ladder could be more distinctly resolved, rotational phasing clearly extended upstream to nucleotide 411, beyond the ERE (note in particular the upper arrow, comparing lanes 11-14 to lane 15 in Fig. 3B). Thus, rotational phasing of nucleosomal DNA was detected between positions 278 and 410, in all three types of breast cells (Fig. 3, A and B) and independent of activation of transcription by addition of estradiol in the MCF-7 cells (Fig. 3A, lanes 7, 9, and 12; Fig. 3B, lanes 3, 5, and 13). Due to the inclusion of the ERE in this rotationally phased nucleosomal DNA, the nucleosome(s) involved was named nucleosome E. 

[View Larger Version of this Image (57K GIF file)]

To examine the translational positioning of nucleosome E, double-stranded MNase cleavage sites were mapped, initially with primer set NE-B (419) (Fig. 4). A downstream nucleosomal border was established by an intense, novel cleavage site at 270 in the chromosomal DNA in all the breast cells (lower star, Fig. 4A, lanes 4-13), 8 bp closer to the initiation site than the edge of the 10-bp ladder of nucleosome E (Fig. 3A). (Note that the nearby site digested by MNase in naked DNA, at 265, is not accessible to double-stranded MNase cleavage in the nucleosomal DNA.) A double-stranded MNase cleavage site at 420, 150 bp from the pronounced cleavage at 270 (Fig. 4A), defined the upstream edge of nucleosome E, as analyzed by LMPCR using NE-T (264) (Fig. 5, lowest band within starred bracket, lanes 3-6). Thus, the length of DNA separating these borders is in agreement with the expected size of a nucleosome.

[View Larger Version of this Image (40K GIF file)]

[View Larger Version of this Image (57K GIF file)]

As was observed from analysis with NT-T (+7) (Fig. 2, A and B), the downstream linker, linker TE, contained additional hypersensitive sites for double-stranded cleavage by MNase. A particularly distinctive hypersensitive site was observed at 240 (upper star, Fig. 4A, lanes 4-13). Linker TE also contained a striking, cell type-specific MNase cleavage site (CS₂) at 228, which was considerably more pronounced in MCF-7 cells than in either MDA-MB 231 or 184 cells (Fig. 4A, lanes 4-7, as compared with lanes 8-13; see below for further discussion). The upstream linker DNA (L-E) revealed novel MNase double-stranded cleavages at 426 and 440 and a hypersensitive site at 434 (starred bracket, Fig. 5, lanes 3-6). The MNase and DNase I digestion patterns, combined, distinctly define preferred positions of both nucleosomes E and T. These nucleosomes are separated by an approximately 100-bp linker region (165 to 270).

The data presented above establish the presence of two rotationally phased and translationally positioned nucleosomes on the genomic pS2 promoter in three types of breast cells. However, although the positioning of the nucleosomes was ubiquitous, two prominent transcription-associated alterations in the nuclease digestion patterns were observed. First, an altered pattern of DNase I cleavages (CS₁), adjacent to and 5 of the TATA box, was apparent in MCF-7 cells (Fig. 1). The CS₁ alterations include reduced protection from DNase I at positions between the accessible sites in the 10-bp ladder. As a result, the DNase I cleavage ladder in the MCF-7 cells was more subtly defined, and the pattern more closely resembled digestion of naked DNA (note cleavage sites around 34, 42, 54, and 75; lanes 7-10, and 15-16, Fig. 1). This suggests a partially disrupted or looser set of DNA-histone interactions around the TATA box. Second, a distinct, cell type-specific MNase-hypersensitive site located at nucleotide 228 (CS₂, Fig. 4, A and B) occurred only in the pS2 chromatin structure in MCF-7 cells. This site is positioned in the linker region between nucleosomes E and T.

Of the three types of breast cells, only MCF-7 cells contain high levels of ER, and only in these cells is the pS2 promoter estrogen-inducible. Therefore, we investigated the role of ER in establishing the alterations in chromatin structure. MCF-7 cells were treated either with the anti-estrogen ICI 164,384 or with -estradiol. ICI 164,384 prevents ER-mediated transcriptional induction in two ways: 1) by significantly lowering the cellular levels of ER (57) and 2) by keeping ER from its usual nuclear location (58). Also, although not performed in this study, incubation of MCF-7 cells with ICI 164,384 substantially inhibited estradiol-induced reporter gene activity (59). In our experiments, removal of ER from the pS2 promoter by treatment of cells with ICI 164,384 led to a 1.7-fold reduction in the basal level of pS2 mRNA expression (data not shown). However, the unique characteristics of the MCF-7 cell chromatin (CS₁ and CS₂) remained unaltered, regardless of the ability of the ER to specifically bind the chromatin template (Fig. 1, lanes 13-18 and Fig. 4B).

We also compared the pS2 chromatin structure in MCF-7 cells in the uninduced and estrogen-induced states. To this end, cells were treated with either ICI 164,384 or estradiol prior to digestion with DNase I or MNase. As seen in Fig. 1 (lanes 15 and 16), Fig. 2A (lanes 15-16, and 18-19), Fig. 3A (lanes 12-14), and Fig. 3B (lanes 2-5), both the rotational and translational positioning of nucleosomes E and T remained constant. Thus, the pS2 nucleosomal organization in the MCF-7 cells was maintained even upon a 16-fold induction in pS2 mRNA (comparing levels of pS2 mRNA in the ICI 164,384 and estradiol-treated cells).

Because our examination of the pS2 chromatin structure involved nuclease digests of populations of cells, the data represent a composite of the pattern of nuclease-digested DNA from many individual cells. This may reflect either a uniform pattern of digestion products or a combination of different patterns from different cells. This issue is particularly significant when comparing chromatin structure of highly active versus uninduced promoters. If only a small percentage of the cells actually contain a hormone-induced pS2 promoter, any induction-related chromatin alterations may be impossible to detect above the signal from the uninduced population. To determine the percentage of MCF-7 cells in which the pS2 gene is up-regulated in our protocols, pS2 protein was detected by immunofluorescence with anti-pS2 protein antibodies. The non-expressing MDA-MB 231 cells and CMT cells provided negative controls; they showed no cytoplasmic staining with the anti-pS2 antibody, although a background nuclear staining was apparent (data not shown). Less than 1% of the non-induced MCF-7 cells exhibited cytoplasmic staining, whereas in contrast, cytoplasmic staining of the pS2 protein was present in 78% of the total estradiol-induced MCF-7 cells (data not shown). All pS2-positive, MCF-7 cells exhibited a punctate, perinuclear staining, and the more intensely fluorescing cells also displayed diffuse, cytoplasmic staining. Given the high percentage of cells responding to hormone by induction of pS2 protein synthesis, nuclease digestion patterns from the estradiol-treated MCF-7 cells should represent the transcriptionally induced chromatin state of the pS2 promoter.

The DNA Sequence of the pS2 Promoter from 1100 to +10 Is Sufficient for Specific Positioning of Nucleosomes T and E

To determine whether limited promoter sequences were sufficient to dictate the positioning of nucleosomes on the pS2 promoter, we cloned the human pS2 promoter from 1100 to +10 upstream of a -globin cDNA reporter gene, into an SV40 viral vector. It is well documented that the SV40 genome is replicated as minichromosomes consisting of host cell chromatin components (for reviews see Refs. 60 and 61). Therefore, this system enabled us to test whether the region of the pS2 promoter from 1100 to +10 was sufficient to establish and maintain positioned nucleosomes T and E.

Indeed, the patterns of cleavages by both DNase I and MNase on the viral, episomal chromatin retained a high degree of fidelity to the patterns observed from the genomic, human breast cell chromatin (see Fig. 6 for representative data). A complete set of analyses was performed with NE-T (264), NE-B (419), and a viral specific primer set, NTV-T (+7), consisting primarily of sequences complementary to the -globin cDNA. LMPCR of DNase I digests revealed rotational phasing over the same pS2 DNA sequences as in MCF-7, MDA-MB 231, and 184 cells, with accessible sites in the identical sequences in both nucleosome T (arrows, Fig. 6A, and data not shown) and nucleosome E (arrows, Fig. 6B, and data not shown). The MNase cleavage patterns bracketing nucleosomes T and E were also identical to the previously defined cleavage sites in the human breast cells (lower starred arrow, Fig. 6C, and data not shown). Likewise, the linker regions displayed identical double-stranded cleavage sites (upper starred arrow, Fig. 6C, and data not shown).

[View Larger Version of this Image (28K GIF file)]

With regard to the transcription-associated chromatin differences, viral minichromosomes as propagated in the monkey kidney cells contained similar chromatin features to the genomic pS2 promoter from non-inducible, non-expressing human breast cells (184 and MDA-MB 231 cells). In particular, the MNase band at position 228 (CS₂), in the linker region between nucleosomes E and T, was not hypersensitive in the pS2 promoter of the minichromosomes (Fig. 6C). With regard to the DNase I active chromatin pattern (CS₁), the viral pS2 promoter also more closely mimicked the genomic promoter of the non-inducible cells with no enhanced DNase I cleavages at nucleotides 41 and 42 (Fig. 6A). These results strengthen the correlative link between the alterations of the pS2 chromatin structure and the transcriptional potential of the pS2 promoter.

To investigate whether the Alu element 85 bp upstream of nucleosome E was involved in positioning this nucleosome (62, 63), we reconstituted a DNA fragment, encompassing sequences between 460 to +56 of the viral pS2 promoter, into nucleosomes and determined the upstream edge of nucleosome E. The reconstituted nucleosomes prevented exonuclease III digestion beyond the nucleosomal boundary previously documented in the genomic pS2 promoter context (Fig. 7, lanes 7-9). Therefore, nucleosome E is positioned translationally even in the absence of the upstream Alu element.

Reconstitution of nucleosomes establishes the 5 boundary of nucleosome E in the absence of the upstream Alu element.

[View Larger Version of this Image (37K GIF file)]

DISCUSSION

It is well established that nucleosome positioning can exert a tremendous regulatory force on the transcriptional response of a promoter to activating signals. However, the detailed chromatin organization at only a few inducible, higher eukaryotic promoters has been determined. We have mapped the nucleosomal organization of the genomic human pS2 promoter at nucleotide level resolution, identifying the placement of two positioned nucleosomes within the 400 bp upstream of the initiation site. In addition, chromatin structural alterations were evident that correlated with the transcriptional competency of the pS2 promoter in MCF-7 cells.

Two rotationally and translationally positioned nucleosomes (nucleosome E and nucleosome T) were detected between 440 and +5 of the pS2 promoter, encompassing the ERE and TATA box, respectively. Double-stranded MNase-hypersensitive cleavage sites, which figure predominantly at the edges of nucleosomes, bracketed nucleosome length spans of rotationally phased DNA, suggesting a single translational position for both nucleosomes E and T. Overlapping families of nucleosomes have been described on other promoters, including the mouse mammary tumor virus (64, 65) and Hansenula polymorpha methanol oxidase promoters (66). In the latter case, multiple overlapping translational positions of nucleosomes lead to a single in vivo rotational setting. The presence of significant intranucleosomal MNase cleavages in the pS2 promoter, due to the high site preferences of MNase cleavage (22, 25, 39, 67, 68) within these promoter sequences, makes it difficult to rule out the presence of families of nucleosomes. Nonetheless, our data firmly demonstrate the existence of one, predominant pair of translationally positioned nucleosomes.

The region upstream of the ERE in the pS2 promoter contains a single Alu element. Alu elements have been shown in other contexts to direct translational and rotational positioning of nucleosomes (62, 63). If the pS2 Alu element positioned a nucleosome in a similar manner, its downstream edge would be approximately 16 bp upstream of the ERE. No translationally positioned nucleosome was evident at this position (Fig. 5). In addition, the boundary of nucleosome E closest to the Alu element was appropriately established upon nucleosome reconstitution in the absence of the Alu sequences (Fig. 7). Therefore, although Alu repeats can influence patterns of nucleosome formation over neighboring regions (69), these effects seem to be mitigated by the sequence of the adjacent region in the pS2 promoter.

The positioned nucleosome T in the pS2 promoter contains the TATA box at its 3 edge. The effects on the binding of TBP/TFIIA to TATA sequences positioned at various locations within nucleosomal DNA was previously examined with reconstituted nucleosomes (70). The preferential translational position at which TBP/TFIIA bound the TATA box was indeed at the edge of the nucleosome, and binding was facilitated if contacts made by the amino-terminal tails of the core histones with nucleosomal DNA were eliminated. Various rotational positionings of the TATA box at the edge of the nucleosome did not preclude TBP/TFIIA binding. Thus, TBP is predicted to have access to the TATA box in the genomic pS2 promoter, with binding potentially being facilitated by acetylation of core histone tails. The lack of nucleosome displacement even upon maximal transcriptional activation of the pS2 promoter in MCF-7 cells may reflect this TBP binding site location, which alleviates the need for a dramatic chromatin disruption or rearrangement. Furthermore, in this configuration the transcription start site is located in a linker region between nucleosomes.

The position of the ERE in the pS2 promoter also places it at the edge of a nucleosome, within the 5 edge of nucleosome E. Binding of ER in the major groove of the rotationally positioned nucleosomal pS2 DNA apparently does not change the access of DNase I to the minor groove more than marginally (Fig. 3, A and B), in agreement with results from previous studies of ER (35) and glucocorticoid receptor (16). Furthermore, on the pS2 promoter, the affinity of ER for the ERE is much weaker than its affinity for a consensus ERE² making detection more difficult. Nonetheless, because DNA binding factors bind more readily to DNA at the edge or between reconstituted nucleosomes, rather than near the nucleosome dyad (1, 2), including ER in the context of a vitellogenin promoter (26), we predict that the translational position of the pS2 ERE should be accessible to ER.

By analogy with glucocorticoid receptor, an additional factor in nucleosomal DNA accessibility to steroid receptors is the rotational positioning of the response element (10). The rotational setting of the genomic pS2 ERE, comprised of imperfect palindromic recognition sequences flanking a 3-bp spacer, is seen from data presented in Fig. 3A. The major groove of the perfect pentanucleotide (405 to 401) faces away from the surface of the histone, leaving it accessible to ER. Intriguingly, whereas the rotational positioning of nucleosome E is distinct right to the nucleosomal edge in 184 cells (Fig. 3A, lanes 10 and 11), the DNA in MCF-7 cells becomes nuclease-accessible within a helical turn of DNA inside the ERE, suggesting that the ERE-containing DNA in MCF-7 cells is more loosely associated with the histone octamer. Since the position of nucleosome E is fixed in cells in which the pS2 promoter is either inactive or responsive to hormonal induction, this ERE must be accessible to ER in the chromatin context.

Our in vivo, fine mapping data reveal that the chromatin structure on the human pS2 promoter juxtaposes the ERE and TATA box within 8 to 14 nm of each other (data not shown). This assessment is based on several models of the 30-nm chromatin fiber (71-74), with the ERE positioned at the upstream edge of nucleosome E and the TATA box at the downstream edge of the adjacent nucleosome T. It has previously been shown that nucleosomes, by changing the three-dimensional positioning of transcription factors relative to one another, can facilitate protein-protein interactions among transcription factors and between upstream factors and the basal machinery (27, 28) to promote transcriptional induction (26). Given that the dimensions of a small transcription factor, TBP, are 3 × 4 × 6 nm (75), the proximity of the TATA box to the pS2 ERE in chromatin could readily facilitate protein-protein interactions. Interestingly, according to one model of the chromatin fiber, the zigzag model (71), either the ERE or the TATA box would be on the outside of the chromatin fiber with the other site on the inside. With the ERE on the outside, binding of ER might facilitate subsequent binding of TBP by coordinating an opening of the chromatin structure around the TATA box, previously inaccessible within the chromatin fiber. Although the modelled distances between the ERE and TATA box were calculated for a compacted fiber, the ERE and TATA box would remain juxtaposed even upon relaxation of such a compact, ordered structure to a more accessible structure, as is characteristic of active chromatin (Fig. 8). CS₂, the pronounced site of MNase cleavage between nucleosomes T and E, specific in our experiments to MCF-7 cells, may reflect such a higher order structure transition from an inactive to active state. These models suggest that nucleosomal organization facilitates protein-protein interactions and thereby potentiates transcriptional induction by ER on the pS2 promoter.

[View Larger Version of this Image (23K GIF file)]

As described above, nucleosomes E and T of the pS2 promoter persisted even upon estradiol induction. Only more subtle structural changes correlated with the transcriptional competency of the pS2 gene, as signified by the two chromatin-specific differences in nuclease sensitivities (CS₁ and CS₂). However, because there are persistent levels of basal transcription from the uninduced pS2 promoter in MCF-7 cells, it is not possible to definitively differentiate between the cause and effect of transcription on this chromatin configuration. Nonetheless, because active ER was not required to maintain the altered structure, as shown by treatment of MCF-7 cells with a pure anti-estrogen and because varied levels of transcription did not alter the extent of this disruption, we favor the interpretation that the altered chromatin structure permits basal transcription. There is precedent for this type of persistent, transformed chromatin structure. Nucleosome B of another steroid-inducible promoter, the mouse mammary tumor virus long terminal repeat, is rearranged upon hormonal induction but is neither removed nor shifted (14). Alteration of nucleosome B remained even after treatment with the antagonist, RU486, which was sufficient to block the progesterone induction of transcription (76).

The unique cleavage area CS₁ reflects increased accessibility to DNase I in the region of the pS2 promoter immediately adjacent to the TATA box in MCF-7 cells (Fig. 1). Altered DNase I accessibility near the TATA box also occurs in other estrogen-regulated promoters, including the rat prolactin promoter in mammalian cells (36) and an artificial, multiple ERE-containing promoter construct in yeast (33-35). However, in these cases the details of the disruption were not clear, as the analysis was performed at a grosser level. Our LMPCR data indicate a loosening of histone-DNA contacts near the TATA box while still maintaining a translationally positioned nucleosome.

Given that active ER is not necessary to maintain CS₁ cleavages in the pS2 promoter, this structural alteration may be part of a memory effect that permits subsequent rapid activation of pS2 transcription in response to hormone. Similarly, the major chicken vitellogenin gene contains two DNase I-hypersensitive sites that are induced when the gene is transcriptionally activated by estradiol. Once established, these DNase I sites are stable in that they persist at roughly the same intensity for weeks after hormone treatment and can be propagated to daughter cells in the absence of inducer (77). Since MCF-7 cells require estradiol for effective propagation, the pS2 promoter has been transcriptionally induced during the normal course of cell passage. Therefore, the loosening of nucleosome contacts near the pS2 TATA box may have been established by prior binding of ER and transcriptional activation, in a manner similar to hypersensitivities on the chicken vitellogenin promoter.

The pS2 promoter in the viral context lacked certain accessibility to nucleases at the CS₁ site, and the CS₂ cleavage was absent, similar to genomic DNA in the non-inducible, human breast cells (MDA-MB 231 and 184). Therefore, the characteristic CS₁ and CS₂ sites must be dependent on an as yet unidentified cellular factor or event, lending further support to the hypothesis that either ER, transcriptional activation of the pS2 promoter, or both are necessary to establish but not maintain the altered chromatin structure in the MCF-7 cells.

The major structural determinants of the nucleosomal organization of the pS2 promoter are intrinsic to the DNA sequence. This was evident upon analyzing the nucleosomal structure over the viral pS2 minichromosomal promoter propagated in monkey kidney cells that lack ER and do not express pS2 mRNA. Two positioned nucleosomes T and E were accurately assembled on the episomal pS2 promoter. This intrinsic positioning of nucleosomes on the pS2 promoter is unlike the serum albumin enhancer, which exists in an array of precisely positioned nucleosomes only in cells in which the enhancer is active (25). In addition, the recombinant virus not only reconstructs these two positioned nucleosomes on the pS2 promoter but also establishes the fine chromatin structure found in the genomic promoter of the uninduced breast cells. Therefore, this virus is a valuable tool for exploring not only the effects of chromatin on transcription but also the causes of the nucleosomal alterations.

In summary, the finding that the pS2 promoter is complexed to form two strongly positioned nucleosomes that are not displaced by transcriptional activation suggests that the required transcription factors interact directly with the chromatin template. In fact, this chromatin template may facilitate transcriptional activation by mediating protein-protein interactions through spatial arrangements of DNA-binding sites and by regulating binding of TBP-containing complexes. Nonetheless, the structural alterations in the inducible MCF-7 cells suggest subtle changes in the chromatin environment, without nucleosome displacement, that may be necessary for the process of transcriptional activation. The pS2 promoter may therefore provide an important model for understanding one of the complex mechanisms by which transcription factors interact with chromatin structure to regulate gene expression.

We thank Martha Stampfer for the gift of normal human mammary epithelial cells; Pierre Chambon for the pS2 promoter-containing plasmid; Myles Brown for ICI 164,384 and MDA-MB 231 cells; Ruth Sager for MCF-7 cells; and Konstantin Ebralidse for histone octamers. The critical suggestions of Robert Kingston, Myles Brown, Fred Winston, Han-Fei Ding, and Konstantin Ebralidse are also greatly appreciated.