|Home | About | Journals | Submit | Contact Us | Français|
Reconstitution of the glucocorticoid receptor (GR)-regulated mouse mammary tumor virus (MMTV) promoter in Xenopus oocytes was used to monitor the effects of different transcription factor contexts. Three constitutively binding factors, nuclear factor 1 (NF1), octamer transcription factor 1 (Oct1), and the Forkhead box A1 (FoxA1), were shown to act in concert, to direct the chromatin structure, and to enhance the GR response. FoxA1 has a dominant effect in the absence of hormone and induces a cluster of DNase I-hypersensitive sites in the segment comprising bp −400 to +25. This FoxA1-mediated chromatin remodeling does not induce MMTV transcription, as opposed to that of the GR. However, the robust FoxA1-dependent chromatin opening has the following drastic functional consequences on the hormone regulation: (i) GR-DNA binding is facilitated, as revealed by dimethyl sulfate in vivo footprinting, leading to increased hormone-induced transcription, and (ii) the GR antagonist RU486 is converted into a partial agonist in the presence of FoxA1 via ligand-independent GR activation. We conclude that FoxA1 mediates a preset chromatin structure and directs a context-specific response of a nuclear receptor. Furthermore, the alternative nucleosome arrangement induced by GR and FoxA1 implies this to be determined by constitutive binding of transcription factors rather than by the DNA sequence itself.
Eukaryotic chromatin is organized as an array of nucleosomes (32) separated by internucleosomal linker DNA (28). Chromatin participates in the regulation of gene expression by restricting access of transcription factors to DNA (20, 34). In eukaryotes, gene activation generally requires the disruption of DNA-histone interactions to allow factor binding, enhanceosome assembly, and the recruitment of basal transcription factors (8). Pioneer transcription factors, such as the Forkhead box A1 (FoxA1), are able to bind constitutively to their cognate DNA sites with high affinity and thereby render a more accessible chromatin (12). Other factors with lower affinity for chromatinized DNA may achieve enhanced binding by cooperative interactions. An example of the latter is the cooperative binding of the octamer transcription factor 1 (Oct1) and nuclear factor 1 (NF1); they reciprocally facilitate each other's constitutive binding to the mouse mammary tumor virus (MMTV) long terminal repeat (LTR), leading to enhanced hormone-dependent glucocorticoid receptor (GR)-DNA interaction (7).
In Xenopus laevis oocytes, chromatin assembly of injected MMTV LTR DNA leads to the formation of an array of randomly positioned nucleosomes. However, the combined expression of Oct1 and NF1 results in partial nucleosome positioning (7). Addition of GR and glucocorticoid hormone renders a fully activated MMTV LTR, containing six translationally positioned nucleosomes (5). This recapitulates the MMTV LTR nucleosome arrangement originally described by Richard-Foy and Hager in tissue culture cells, and the nucleosomes are referred to as nucleosomes A to F (35). Hormone induction leads to remodeling of the B-nucleosome that harbors a cluster of GR binding sites (49). The B-nucleosome forms a hormone-dependent enhanceosome complex, as defined by its resistance to micrococcal nuclease (MNase) (5). In the absence of Oct1- and NF1-induced prepositioning of nucleosomes, the addition of hormone-liganded GR is still capable of activating the MMTV LTR, albeit at a lower level and at a lower rate of induction; furthermore, the GR-induced enhanceosome becomes smaller and less stable (7). Hence, a defined context of constitutively bound transcription factors determines both a specific nucleosome arrangement and a specific hormone-induced enhanceosome complex at the MMTV LTR. These structural features are part of a phenomenon that we refer to as preset chromatin, since it also includes a faster and more robust hormone response (7).
The FoxA1 transcription factor was shown to participate in the regulation of liver- and other gut-specific genes (15, 30) and to strongly influence GR-mediated gene regulation (11, 36, 39) and GR-DNA binding in vitro (45). FoxA1 is expressed in many other tissues, including prostate and mammary gland, and binds to a large number of genes involved in organogenesis, metabolism, signaling, and the cell cycle (17, 33). The structure of the FoxA1 forkhead domain resembles the globular domain of linker histones (14). The C-terminal domain of the protein was reported to interact with core histones H3 and H4 and to contribute to its ability to mediate an open chromatin in a non-ATP-dependent manner in vitro (12). Recently, we reported on two FoxA1 sites in two separate locations within the MMTV LTR, i.e., in the −225 position and the location comprising positions −51 to −39 (−51/−39) relative to the transcription start site (23). When expressed in oocytes from Xenopus laevis, the FoxA1 was able to bind both to the inactive and to the hormone-activated MMTV LTR. In the inactive state, this was accompanied by a slightly increased basal transcription. FoxA1 binding in the presence of GR resulted in the formation of a hormone-dependent enlarged enhanceosome complex. We also showed that FoxA1 partially inhibited the transcription of hormone-activated MMTV LTR (23). However, recent FoxA1 and GR titration experiments showed that FoxA1 stimulates hormone-induced MMTV transcription at low GR concentrations (see Fig. Fig.6)6) (P.-H. Holmqvist, S. Belikov, and Ö. Wrange, unpublished results).
Here we exploit the high copy number achieved by the DNA-microinjected Xenopus oocytes to study the structural and functional in vivo effects of different transcription factor contexts at the MMTV promoter. This strategy generates chromatin structural information of a higher resolution than what has been achieved previously and shows that FoxA1 also mediates a distinct opening of the chromatin structure in a more distal region of the MMTV LTR. Furthermore, we demonstrate that the different transcription factor contexts result in diverse nucleosome organizations. Importantly, FoxA1 facilitates GR binding to its targets and sensitizes the MMTV promoter to glucocorticoid hormone. Unexpectedly, the presence of FoxA1 alters the MMTV promoter response of the glucocorticoid antagonist RU486 into a partially agonistic one. FoxA1 also supports DNA binding and MMTV transcription mediated by a truncated GR protein lacking the ligand binding domain (LBD). Taken together, this implies that the “pioneer protein” FoxA1 programs the transcriptional response of a nuclear hormone-regulated promoter by mediating a local chromatin remodeling effect.
The construction of pMMTV:M13, harboring the 1.2-kb MMTV LTR fused to the herpes simplex thymidine kinase gene at position +137 of the MMTV LTR, was previously described (5) and is referred to as the MMTV reporter.
Insertions into RN3P vectors of cDNA coding for rat GR (5), mouse FoxA1 (23), human Oct1, and pig NF1-C1 (3), and the production of corresponding mRNA were described previously. The plasmid DNAs were linearized with Asp718 and in vitro transcribed by a mMESSAGE mMACHINE kit (Ambion). The LBD-truncated GR expression vector pSTC3-556 harbors the wild-type N terminus and DNA binding domains of rat GR, and the expression is driven by the cytomegalovirus promoter and was described previously (43).
Xenopus oocyte preparation and injection were carried out as previously described (1). Cytosolic injection of 23 nl of mRNA was done in the equatorial plane between the animal and the vegetable poles, and 5 to 7 h later, a blind nuclear injection of 18.4 nl containing 3 ng of single-stranded DNA, pMMTV:M13, was done in the center of the pigmented animal pole using an oil-driven Nanoliter 2000 injector pump (World Precision Instruments), with the oocytes submerged in OR2 medium at ambient temperature. To control for variation in the hit rate of the intranuclear injection, each analysis involved the use of two independent double samples using 8 to 12 oocytes two to three times for each data point. In a typical experiment, mRNA coding for the indicated proteins was injected with the following amounts: pig NF1-C1, 0.35 to 0.7 ng; human Oct1, 2.3 ng; rat GR, 2.3 ng; mouse FoxA1, 1.4 ng. These amounts were adjusted according to the DNA binding activity of each protein monitored by in vivo dimethyl sulfate (DMS) footprinting (see below).
Ten oocytes for each treatment were homogenized in 180 μl of homogenization buffer (5% glycerol, 15 mM HEPES [pH 7.8], 60 mM KCl, 15 mM NaCl, 5 mM MgCl2, 1 mM CaCl2, 0.05% NP-40, 1 mM dithiothreitol). A total of 60 μl was taken to each of three tubes containing 0.375 U, 0.75 U, and 1.5 U of MNase or 0.85 U, 1.7 U, and 3.3 U of DNase I in 40 μl of homogenization buffer. After incubation at 15°C for 5 min, an equal amount (i.e., 100 μl) of 2× stop solution (20 mM EDTA, 10 mM Tris [pH 8.0], 2% sodium dodecyl sulfate [SDS], 1 mg/ml proteinase K) was added, and samples were incubated overnight at 37°C, followed by two phenol-chloroform extractions and isopropanol precipitation. For indirect end labeling experiments, DNA was digested with EcoRV in the presence of RNase A and purified by phenol-chloroform extraction and isopropanol precipitation. DNA was resolved on agarose gel, blotted, and hybridized with a random-primed-labeled EcoRV-SacI DNA fragment (5). For analysis by primer extension, DNA was treated in 0.2 M NaOH for 30 min at 90°C, followed by neutralization by equal volumes of 0.2 M HCl, purification by phenol extraction, and isopropanol purification (5).
Intact oocytes in OR2 medium were treated with DMS, followed by DNA purification and development of a methylation pattern by primer extension and quantified as described previously (6).
The S1 nuclease protection assay was carried out as described previously (23).
Protein expression was monitored by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) analysis of microdissected oocyte nuclei or soluble whole-cell extracts from mRNA-injected oocytes after incubating 4 to 6 oocytes in 0.1 ml of OR2 medium also containing 0.5 μCi [14C]lysine (specific radioactivity, 309 mCi/mmol; PerkinElmer) for 24 h. Analysis of dried gels was performed with Fuji bio-Imaging analyzer BAS-2500 using the Image Gauge version 3.45 software. An estimation of the relative nuclear content of expressed protein is obtained by phosphorimager-based measurement of the radioactive content in the corresponding protein bands and corrected for their lysine content (data not shown). There is a linear correlation between the amount of injected mRNA and the amounts of expressed protein within the amounts used in this work (4).
Six oocyte nuclei isolated by manual dissection from Xenopus oocytes injected the day before with 17 ng of FoxA1 mRNA or from noninjected control oocytes in binding buffer (1 mM EDTA; 20 mM Tris-HCl, pH 7.8; 10% [vol/vol] glycerol; 1 mm dithiothreitol) were collected in 120 μl of the same buffer also containing 0.1 mg/ml insulin to reduce nonspecific binding. The nuclei were homogenized by pipetting with a 200-μl tip and centrifuged for 5 min at 13,000 rpm and 4°C. The supernatant was divided into aliquots and preincubated on ice with the indicated amounts of an unlabeled single-stranded 20-mer oligonucleotide for 1 h to reduce nonspecific DNA binding activity. Then, 6 μl of nuclear extract, corresponding to one-third of a nucleus, was incubated with ~60 fmol of a 33P-labeled double-stranded DNA (dsDNA) probe harboring the MMTV −374/−319 or −239/−213 sequence for 15 min at 18°C followed by 0°C. Then, 0.3 μl (0.13 μg) of a polyclonal FoxA1 antibody (ab23738; Abcam) or an unrelated antibody (anti-green fluorescent protein [anti-GFP], A6455; Life Technologies) was added to the indicated incubations and left on ice for 30 min, followed by electrophoresis on a 3.6% polyacrylamide gel (acrylamide/bisacrylamide ratio, 37.5:1; Gibco BRL) in 0.5× Tris-acetate-EDTA buffer (pH 7.7) with 0.05% Triton X-100 in the gel and run at ~14 V/cm for 3 h at 4°C, followed by drying and phosphorimager analysis as described above.
The hormone regulation of the MMTV promoter was reconstituted in the Xenopus oocyte system either in the presence or in the absence of FoxA1. The oocytes were first injected with mixtures of mRNAs coding for the indicated combinations of proteins, i.e., FoxA1, GR, NF1, and Oct1. This resulted in expression of the corresponding proteins, as confirmed by SDS-PAGE analysis after the incubation of injected oocytes in the presence of [14C]lysine (data not shown). About 5 to 7 h after the mRNA injection, the DNA reporter in single-stranded circular form was injected into the oocyte nucleus. Injection of single-stranded DNA results in second-strand synthesis and a replication-coupled chromatin assembly. After 16 h, the synthetic glucocorticoid hormone triamcinolone acetonide (TA; 10−6 M) was added, and the oocytes were incubated for another 6 to 7 h. The biological activity of the expressed proteins in terms of DNA binding, the effects of chromatin structure, and MMTV transcription was analyzed, as shown below.
Analysis of the DNase I digestion pattern in situ by indirect end labeling from the EcoRV site at position +425 (5, 48) showed no signs of DNase I hypersensitivity within the MMTV LTR in the oocytes injected only with GR mRNA in the absence of hormone (Fig. (Fig.1,1, lane 1). Addition of hormone resulted in the appearance of a strong DNase I-hypersensitive site (DHS) within the −200/−75 segment covering the MMTV glucocorticoid receptor elements (GREs) (5, 35, 49) (Fig. (Fig.1,1, compare lanes 1 and 2 and their corresponding scans). In addition, we observed two hormone-dependent sites moderately sensitive to DNase I located at ~200 bp upstream and ~200 bp downstream of the main DHS, respectively (Fig. (Fig.1,1, compare lane 2 and the corresponding scan with lane 1). These weak hypersensitive sites (HSs) most probably mark the positions of the internucleosomal linkers distal to the positioned nucleosomes flanking the hormone-activated DHS in the GRE region of the MMTV promoter (5).
Coexpression of GR and FoxA1 without hormones gave rise to a strong double HS around position −350. The segment spanning from −200 to +25 was also sensitive to digestion with DNase I, albeit to a lower extent (Fig. (Fig.1,1, compare lanes 1, 2, and 3). Hormone activation caused a robustly increased hypersensitivity over the −250/+25 region, whereas the HS around −350 was diminished compared to that of the nonactivated oocytes (Fig. (Fig.1,1, compare lanes 3 and 4). Recently, we demonstrated that concomitant binding of NF1 and Oct1 to the MMTV promoter enhanced the GR-DNA interaction and resulted in a faster and stronger hormone response (3). Likewise, the expression of NF1 and Oct1 in addition to GR and FoxA1 also resulted in a faster and more robust hormone-induced transcriptional response (see Fig. Fig.6D).6D). In spite of the effect on transcription, the inclusion of NF1 and Oct1 had little effect on the chromatin structure of the MMTV promoter in the presence or absence of hormone, as assayed by the DNase I cleavage and indirect end labeling (Fig. (Fig.1,1, compare lanes 1 and 5, 2 and 6, 3 and 7, and 4 and 8). We conclude that FoxA1 binding mediates a robust constitutive remodeling of the MMTV chromatin structure, seen as a DHS region extending over the region comprising −400 to +25. Hormone activation does not extend the borders of this DHS but changes the relative intensity of DNase I digestion toward the proximal region containing the GREs (Fig. (Fig.1,1, compare lanes 3 and 4).
The DNase I digestion patterns analyzed by indirect end labeling assay (Fig. (Fig.1)1) were also developed by the +65/+92 primer extension. In agreement with our previous results, we observed strong DHSs, a hallmark of FoxA1 binding (13, 39), at positions −225 (Fig. (Fig.2A)2A) and −39/−51 (23). These DHSs were seen both in the presence and in the absence of hormone, but were enhanced by hormone addition (Fig, (Fig,2A,2A, compare lanes 5, 6, 7, and 8 for FoxA at −225). The strong DHS in the −375/−325 segment observed in indirect end labeling experiments (Fig. (Fig.1)1) called for a high-resolution analysis of this area by primer extension using a more proximal primer, i.e., −97/−126. This revealed two FoxA1-induced DHSs at positions −360 and −332 (coding strand) (Fig. (Fig.2A,2A, lanes 5 to 8). These sites were seen both in the presence and in the absence of hormone. However, in contrast to the −225 and −39/−51 sites where hormone addition led to enhanced cutting, the DHSs at −360 and −332 were reduced by hormone activation (Fig. (Fig.2A,2A, compare lanes 5 to 8 and their corresponding scans).
Next, we asked if the DHSs at positions −360 and −332 are caused by direct protein binding or by alterations in the DNA/chromatin structure due to secondary structural effects of the binding of FoxA1 at positions −225, −39, and −51. This was addressed by in vivo DMS methylation protection, a sensitive probe for sequence-specific DNA-protein interactions (3, 6, 7, 23). Hormone induction results in distinct protection of the specific guanines in GREs I to IV from DMS methylation but only in the oocytes injected with GR mRNA (3, 7). Likewise, the protection from DMS-induced methylation at the FoxA1 binding sites is dependent on the FoxA1 mRNA injection (Fig. (Fig.2B)2B) (23). In agreement with our DNase I data (Fig. (Fig.2A),2A), we observed a hormone-dependent reduction of the DMS methylation over the corresponding guanine at position −224 of the FoxA1 site after hormone activation, thus reflecting a hormone-dependent increase in FoxA1-DNA binding (Fig. (Fig.2B,2B, right, compare lanes 1, 3, and 4). This was, however, not the case for the −360 and −332 sites. Here, a weak but FoxA1-dependent protection against DMS methylation was instead reduced upon hormone induction. This corroborates the results of DNase I footprinting (Fig. (Fig.2A)2A) and the hormone-dependent reduction in the upstream DHS (Fig. (Fig.1).1). Inspection of the DNA sequences of the MMTV LTR shows that FoxA1 binding sites at −360 and −332, CATGTAC and TGCATTC, respectively, possess only 5 and 4 bp out of 7 bp of the FoxA1 consensus sequence, TRTTKRY (where R is purine, Y is pyrimidine, and K is T or G). However, the −360 site has only one base pair substitution in comparison with the previously described −51 site (Fig. (Fig.2C)2C) (23). We note that the deviation of the −360 and −332 sites from the FoxA1 consensus sequence makes them difficult to identify by computer-based motif analysis. In order to gather further proof that FoxA1 is indeed contacting these degenerate FoxA1 DNA sites, we also performed band shift experiments, using a −374/−319 MMTV LTR sequence as a dsDNA probe and nuclear extracts of oocytes injected or not injected with FoxA1 mRNA (Fig. (Fig.3).3). This revealed two strictly FoxA1-dependent DNA complexes. Importantly, both of these complexes were supershifted by a FoxA1-specific antibody (Fig. (Fig.3,3, lanes 7 and 12) but were unaffected by an unrelated antibody (compare lanes 3, 5, and 7). Interestingly, preincubation of the nuclear extract with more competition caused a redistribution of the two FoxA1-dependent complexes toward the smaller component (Fig. (Fig.3,3, compare lanes 3 and 10). This argues for the binding of either one or two FoxA1 molecules to the probe. This is in agreement with two FoxA1 sites, at −360 and at −332, being localized within the probe. As a positive control, we also conducted band shift analysis with the −225 FoxA1 site using the −239/−213 as dsDNA probe since this harbors a single FoxA1 consensus sequence (cf. Fig. Fig.2C2C and and3).3). The result was a single FoxA1-dependent complex that was again supershifted by the FoxA1-specific antibody (data not shown).
In order to address the possible interdependence of the FoxA1 binding sites, we conducted DNase I indirect end labeling experiments using FoxA1 site mutants in which either the −225 or the −39/−51 site had been mutated (data not shown) (the mutants were described in reference 23). This revealed a FoxA1-dependent DHS at around position −350 in both ΔFox−225 and ΔFox−51/−39 mutants, thus arguing that FoxA1 binding at positions −360 and −332 is independent of FoxA1 binding at positions −225 and −39/−51 (data not shown). We conclude that the MMTV LTR harbors at least five FoxA1 binding sites at positions −39, −51, −225, −332, and −360 and that their locations correlate with the DNase I hypersensitivity induced by FoxA1 expression already in the absence of hormone (Fig. (Fig.11 and and2),2), thus confirming its previously demonstrated capacity to bind constitutively to chromatinized DNA in vitro (12) and in vivo (23), i.e., to act as a pioneer factor. The constitutive FoxA1 effect on the MMTV chromatin structure dominates over the much weaker effects induced by NF1 and Oct1. Likewise, the DHS pattern induced by hormone activation is determined by FoxA1 and GR, while the NF1 and Oct1 effects are moderate, with their main effect being a slightly more distinct pattern of hypersensitivity in the −250/−75 segment (Fig. (Fig.1,1, compare lanes 4 and 8).
We used MNase digestion together with indirect end labeling (Fig. (Fig.4)4) and primer extension (Fig. (Fig.5)5) to address nucleosome organization along the MMTV LTR in different transcription factor contexts. MNase digestion was done at 15°C to minimize the effect of nucleosome sliding (4). Oocytes previously injected with GR mRNA in the absence of hormone showed a random pattern of digestion similar to that of the naked DNA control (Fig. (Fig.4,4, lanes 1 and 5 and their corresponding scans), thus arguing for a random nucleosome distribution. Hormone induction resulted in distinct changes in the digestion pattern typical for the establishment of translationally positioned nucleosomes with distinct protection over the nucleosome-bound segments and hypercutting in the linker regions (Fig. (Fig.4,4, lanes 1 and 2). These findings are in agreement with our previous results showing that hormone activation leads to the establishment of specific translational positioning of nucleosomes, despite the lack of significant positioning in the absence of hormone (5). Expression of FoxA1 together with GR in the absence of hormone gave rise to a different pattern (Fig. (Fig.4,4, compare lanes 1 and 3). Here, distinct HS-to-MNase digestion appeared in the −475/−350 region, in agreement with the DNase I data (Fig. (Fig.1).1). Addition of hormone resulted in a distinct reorganization of the digestion pattern. The borders of the nucleosome structure organized around the −200/−60 GRE region spread out to form an “enlarged” nucleosome (Fig. (Fig.4,4, compare lanes 2 and 4, and Fig. Fig.5;5; see below). Concomitantly, a reduced MNase digestion was observed over the distal HS region along the −375/−300 segment (Fig. (Fig.4,4, compare lanes 3 and 4). This coincides with an extended protection that is compatible with the formation of a phased nucleosome in this region (Fig. (Fig.4,4, lanes 2 and 4).
As described above, we also coexpressed GR together with NF1 and Oct1. This results in a more distinct pattern of translationally positioned nucleosomes after hormone induction and stronger protection over the GRE region (Fig. (Fig.4,4, compare lanes 2 and 7). Coexpression of GR, NF1, and Oct1 together with FoxA1 gives rise to a pattern similar to that seen in the GR- and FoxA1-injected oocytes (Fig. (Fig.4,4, compare lanes 3 and 8 and lanes 4 and 9; Fig. Fig.5).5). However, upon hormone induction, the translationally positioned enlarged nucleosome covering the −225/−25 segment is more distinctly protected in the presence of NF1 and Oct1 (Fig. (Fig.4,4, compare lanes 4 and 9; Fig. Fig.55).
The MNase digestion pattern was also developed by primer extension to achieve higher resolution (Fig. (Fig.5).5). As mentioned above, hormone induction of GR-containing oocytes causes nucleosome positioning in the previously randomly organized MMTV LTR (5, 6) and results in a highly remodeled so-called B-nucleosome (5, 35, 46). At the same time, during hormone induction, the GRE region is organized into an enhanceosome particle, as defined by MNase digestion (Fig. (Fig.4,4, see HSs flanking B-nucleosome in lanes 2, 4, 7, and 9 and in their corresponding scans). The hormone-induced enhanceosome contains DNA of variable length, depending on the protein context. In the presence of GR, the size of the DNA segment is 120 bp, i.e., the −190/−70 region (Fig. (Fig.5,5, lane 2) (3). Addition of NF1 protein increases the size of the DNA in the enhanceosome complex to 130 bp, covering the −190/−60 segment (3). This finding is confirmed in Fig. Fig.55 (compare lanes 2 and 6). The protected DNA segment in both of these complexes is shorter than the 145 to 146 bp that is characteristic for the canonical nucleosomal core particle and is hence dubbed the subnucleosome (5, 7). Expression of FoxA1 in the presence of hormone caused an extension of the borders of the protected segment in both the 3′ end and 5′ end directions, thus converting it into a larger enhanceosome (enlarged) complex covering ~200 bp, i.e., from −225 to −27 (Fig. (Fig.5,5, compare lanes 2 and 4). Coexpression of FoxA1 with GR, NF1, and Oct1 does not change the size of the enlarged enhanceosome particle, a signature for FoxA1 binding, but increases its protection significantly and thus seems to enhance its stability (Fig. (Fig.4,4, compare lanes 4 and 9; Fig. Fig.5,5, compare lanes 4 and 8). NF1 in the context of FoxA1 is still occupying its site and cooperates with FoxA1 for DNA binding, as indicated by a DNase I footprint showing stronger protection over the NF1 and GR sites in the presence of FoxA1 (data not shown). However, FoxA1 and Oct1 have overlapping binding sites on the MMTV promoter, referred to as the −39 and −51 sites. This makes it difficult to estimate the binding level of each protein. The NF1- and Oct1-dependent increase in the protection from nucleases within the FoxA1-induced enlarged enhanceosome and a distinct stimulation of Oct1 on MMTV transcription (data not shown) argue that both FoxA1 and Oct1 contribute to the active MMTV promoter structure at the −39 and −51 regions.
Expression of FoxA1 has definite effects on the nucleosome organization of the MMTV LTR. Strong binding of FoxA1 to its distal sites at −360/−332 results in the formation of a strong hypersensitivity to DNase I (Fig. (Fig.1),1), to MNase (Fig. (Fig.44 and and5),5), and to methidiumpropyl-EDTA-Fe(II) (MPE) footprinting (23). This strong hypersensitivity to these three DNA cutting agents argues for the disruption of the nucleosome structure in this region upon FoxA1 binding in the absence of hormone. Our results imply that this highly remodeled chromatin segment serves as a strong positioning signal for the nucleosomes flanking these FoxA1 binding sites. Hormone induction partially compromises FoxA1 binding to the distal sites at −360/−332 and enhances binding to the proximal sites (see above). This hormone-induced change in FoxA1 binding correlates with a different translational nucleosome architecture; namely, a nucleosome seems to be formed over the distal FoxA1 sites, and we propose that this positioned nucleosome compromises the FoxA1 binding to the −360/−332 sites (Fig. (Fig.11 and and2).2). This nucleosome is less stable than the corresponding C-nucleosome, formed in this region in the presence of GR or GR-NF1-Oct1, as judged from its moderate sensitivity to DNase I (Fig. (Fig.1)1) and MNase (C-nucleosome region) (Fig. (Fig.44 and and5)5) and hypersensitivity to MPE digestion (23). Thus, FoxA1 is a strong chromatin organizer of the inactive MMTV promoter, while GR takes over as the dominant chromatin organizer in the presence of hormone. Taken together, these results show distinct differences in the nucleosome architecture of the MMTV LTR as a function of the different transcription factor contexts, and thus, we conclude that this is what specifies the nucleosomal architecture in the MMTV promoter rather than the DNA sequence.
To evaluate the effect of FoxA1 on GR-DNA binding and on the kinetics of hormone-induced transcription, we performed a time study using oocytes in which GR alone or GR and FoxA1 were expressed. The amounts of GR alone and GR plus FoxA1 were monitored in these oocyte populations (see Materials and Methods), and the concentration of GR protein was shown to be identical in both groups (data not shown). After addition of hormone, oocytes were harvested, and GR binding and hormone-induced transcription were analyzed at the indicated time points (Fig. (Fig.6A).6A). Hormone (TA) was added to oocytes at a saturating, 1,000 nM, or a half-saturating, 10.5 nM, concentration; the latter is based on previous analysis showing that 10 to 11 nM of TA induces a half-maximal hormone-induced translocation of GR to the nucleus and half-maximal transcription (6). Quantification of specific binding of GR by DMS methylation protection assay in vivo disclosed a markedly faster GR binding in the presence of FoxA1 than that achieved in the presence of GR alone (Fig. (Fig.6B,6B, top). GR alone rendered 82 to 83% of DMS methylation at the GREs after 4 h compared to 40% of that in the presence of GR and FoxA1 (Fig. (Fig.6B,6B, top). The presence of FoxA1 thus resulted in a 3.4-fold enhanced GR binding at the latest time point. At earlier time points, the difference was larger (Fig. (Fig.6B,6B, top). Likewise, the MMTV transcription also showed an increased rate of induction and a higher level at all times for GR and FoxA1 compared to those for GR alone (Fig. (Fig.6B,6B, bottom). Both GR-DNA binding and the MMTV transcription displayed the strongest difference at early time points, i.e., at 1 and 2 hours.
When a low nuclear GR concentration is decreased even further by the use of 10.5 nM of hormone, a barely detectable level of DMS protection is seen with GR alone (Fig. (Fig.6C,6C, top), which is translated into low hormone-induced transcription (Fig. (Fig.6C,6C, bottom). The presence of FoxA1, however, has a drastic stimulatory effect both on GR binding and on MMTV transcription. Specifically, at the 4-hour time point, there was 60% DMS methylation in GR and FoxA1-expressing oocytes compared to 98% DMS methylation when expressing GR only, signifying a more than 10-fold increase in the GR binding efficiency and in hormone-induced transcription caused by FoxA1 (Fig. (Fig.6C,6C, top and bottom).
A twofold-greater stimulation of MMTV transcription is achieved by coexpressing NF1 and Oct1 together with GR and FoxA1 (Fig. (Fig.6D).6D). We also addressed the functional contribution of NF1 alone or together with Oct1 in a FoxA1-plus-GR context. This revealed a 3-fold, a 4-fold, and a 7.6-fold increase in MMTV transcription when adding FoxA1, NF1, and Oct1 to the GR-containing mRNA mix injected into the four groups of oocytes (data not shown). Hence, each added factor stimulated the transcriptional activity. This finding correlates with the increased enhanceosome protection/stability by the presence of all four factors (Fig. (Fig.1,1, lane 8; Fig. Fig.3,3, lane 9; Fig. Fig.5,5, lane 8) and shows that factor-induced changes in chromatin structure correlate with effects on transcription.
Collectively, our data argue for the FoxA1-remodeled chromatin to provide a faster and stronger GR-GRE binding reaction that in turn results in the enhanced hormone-induced transcription, with NF1 and Oct1 enhancing this effect even further. We postulate that the latter effect is due to a cooperative positive interaction between FoxA1, NF1, and Oct1. The level of stimulation by each factor is concentration dependent, and hence, the functional importance of each factor may be modulated within a wide spectrum.
RU486 is a well-known antagonist of glucocorticoids, progestins, and to some extend, androgens (9). Binding of RU486 to the C-terminal LBD of GR blocks its recruitment of coactivators, as revealed by a loss of chromatin remodeling (6). The robust FoxA1-induced chromatin remodeling of the MMTV promoter posed the question of whether FoxA1 could modulate the RU486-mediated effect of GR. Oocytes expressing GR alone or GR and FoxA1 were divided into six groups, as defined by the added hormone agonist/antagonist (Fig. (Fig.7A,7A, lanes 1 to 6 for GR alone and lanes 7 to 12 for GR plus FoxA1). After a 6-hour incubation with or without a ligand(s), the oocytes were assayed for MMTV transcription and for specific protein-DNA binding (Fig. (Fig.7).7). The effect of the different transcription factor contexts was monitored both with RU486 alone (Fig. (Fig.7,7, lanes 6 and 12) and with RU486 together with the agonist corticosterone (lanes 4, 5, 10, and 11). The strategy was used in order to address both the intrinsic effect of the antagonist as such and its ability to compete with, and thereby block, the effect of an agonist. For this experiment, we found it more appropriate to use corticosterone, a physiological glucocorticoid agonist, since it has a 10-fold-lower affinity for GR than the synthetic agonist TA (38).
The addition of 15 nM corticosterone leads to induction of the MMTV transcription (Fig. 7A and B). As expected, a higher dose of hormone, 100 nM corticosterone, results in increased transcription (Fig. 7A and B, compare lanes/bars 2 and 3 for GR-injected oocytes to lanes/bars 8 and 9 for GR- and FoxA1-injected oocytes). In good agreement with our previous results, FoxA1 strongly enhances the hormone-induced transcription by 4.8-fold and 3-fold at 15 nM and 100 nM of corticosterone, respectively, thus also confirming that the stimulatory effect of FoxA1 is more prominent at lower concentrations of hormone-activated GR (Fig. 7A and B, compare lanes/bars 2 and 8 and lanes/bars 3 and 9). Addition of the agonist corticosterone results in GR binding and, thus, in a decreased accessibility for DMS methylation (Fig. 7C and D, compare bars 1, 2, 3, 7, 8, and 9). As expected, a higher concentration of corticosterone correlates with enhanced GR binding.
The presence of FoxA1 has a robust, positive effect on GR-DNA binding. In GR-injected oocytes, 94% and 87% of DMS methylation is observed with 15 nM and 100 nM corticosterone, respectively, while in oocytes injected with GR plus FoxA1, the percentages of DMS methylation are 61% and 50%, respectively (Fig. 7C and D, compare bars 2 and 3 to bars 8 and 9). The level of DMS methylation of GR-specific guanines in the absence of hormone was set to 100%. Addition of the hormone antagonist RU486 either alone or together with corticosterone to GR-containing oocytes has no detectable effect on GR binding (Fig. 7C and D, compare bars 1 and 4 to 6) and transcription, thus confirming its capacity to act as an antagonist in this context (Fig. 7A and B, compare lanes/bars 1 and 4 to 6). However, in the GR-plus-FoxA1 context, addition of RU486 alone gives rise to a significant level of the MMTV transcription, equal to 22% of that detected in the GR-injected oocytes treated with 100 nM of corticosterone (set to 100%) (Fig. 7A and B, compare lanes/bars 7 and 12 to lanes/bars 3 and 12). Furthermore, in the context of GR plus FoxA1, addition of RU486 together with 15 nM corticosterone results in the transcription equaling 28%, and addition of RU486 together with 100 nM corticosterone equaling 84% of transcription generated by 100 nM of corticosterone alone in the absence of FoxA1 (Fig. 7A and B, compare lanes/bars 10 and 11 to 5). This demonstrates that in the GR-plus-FoxA context, RU486 fails to block corticosterone-induced transcription (Fig. 7A and B, compare lanes/bars 2 and 3 to 10 and 11) and can even induce a significant level of transcription on its own. In line with these results, oocytes injected with GR and FoxA1 and treated with RU486 demonstrate significantly enhanced GR-DNA binding (Fig. 7C and D, compare bars 7 and 12, bars 4 and 5, and bars 10 and 11). Thus, we conclude that the presence of FoxA1 converts the RU486-activated GR effect from an antagonistic to a partially agonistic response.
FoxA1 binding to the −51 DNA site was also evaluated by in vivo DMS footprinting (Fig. (Fig.7E).7E). In concordance with our previous data (23), we detect constitutive FoxA1 binding to the MMTV LTR in the absence of hormone (Fig. 7C and E, compare bars 1 and 7). Addition of the hormone agonist results in stronger binding of FoxA1 (Fig. (Fig.7E,7E, compare bars 7, 8, and 9). Interestingly, interaction of FoxA1 with its DNA binding sites is also enhanced by treatment of RU486 either alone or together with corticosterone (Fig. 7C and E, compare bars 7 and 10 to 12). This shows that FoxA1 and GR liganded with either corticosterone or RU486 reciprocally facilitate each other's binding to the MMTV LTR.
We previously demonstrated that the GR-RU486 complex is translocated to the nucleus and is specifically, albeit weakly, bound to the MMTV GRE but is unable to induce chromatin remodeling and transcription (6). The GR protein is known to harbor a weakly activating N-terminal domain that is independent of the ligand-dependent activation function that is located at its C terminus. This was revealed by pioneering experiments where deletion mutants of the GR LBD were shown to be constitutively translocated to the cell nucleus and to activate transcription from a transiently transfected MMTV reporter gene (19, 22). The FoxA1-dependent conversion of RU486 into a partial agonist (Fig. (Fig.7)7) might thus be explained by the requirement of a chromatin remodeling activity, here mediated by FoxA1 binding (Fig. (Fig.1),1), to assist the weakly activating GR N-terminal domain. Here we tested this possibility by expression of a truncated GR protein lacking the LBD, here dubbed GR-Δ, in oocytes in the presence or absence of FoxA1 (Fig. (Fig.8A).8A). Analysis of the protein content in nuclei that were manually isolated from the injected oocytes demonstrated that GR-Δ is indeed translocated to the nucleus (Fig. (Fig.8B)8B) and that both GR-Δ and FoxA1 are expressed in the relevant pools of injected oocytes (Fig. (Fig.8B,8B, right). The MMTV transcription is dependent entirely on the concomitant presence of GR-Δ and FoxA1 (Fig. (Fig.8C).8C). This implies that the weak transcription activation capacity of GR-Δ requires the FoxA1 chromatin remodeling activity to elicit a transcriptional response and that the same mechanism may explain the FoxA1-mediated effect on the GR-RU486 complex, where the activating function of the C-terminal LBD has been incapacitated by the bound antagonist.
Here we utilize the Xenopus oocyte system to correlate the effect of transcription factor context to the structural and functional effects on a hormone-dependent promoter. Unique features of this experimental system, the high copy number of homogenously responding templates and the high capacity for protein translation, were exploited to retrieve novel chromatin structural information of FoxA1-nuclear hormone receptor interactions. To the best of our knowledge, this is the first in vivo demonstration of constitutive binding of FoxA1 alone leading to a more open chromatin structure. We refer to this process as chromatin presetting, since it results in a faster and stronger GR-mediated hormone response. Unexpectedly, and shown for the first time, FoxA1 binding and remodeling of chromatin structure switch the hormone antagonistic effect of RU486 at the MMTV promoter to a partially agonistic response. The mechanism of this effect is probably a FoxA1-mediated assistance of GR binding and transcription activation via other domains of the GR protein than its LBD. Furthermore, our data imply that the MMTV LTR chromatin structure, rather than what have been claimed previously, is not directed by its DNA structure but rather is a function of the DNA binding protein context.
FoxA1 binds to several sites within the inactive MMTV LTR and, upon binding it, strongly remodels chromatin structure of the segment covering the region comprising −400 to +25. This is demonstrated by DNase I (Fig. (Fig.1)-based1)-based and MNase (Fig. (Fig.44 and and5)-based5)-based experiments and is in agreement with previously used MPE-based experiments (23). The FoxA1 binding has profound effects on nucleosome organization. The robust binding of FoxA1 to its distal sites at −360/−332 in the inactive MMTV promoter results in the formation of a constitutive HS (Fig. (Fig.1,1, ,4,4, and and5)5) (23). This strong hypersensitivity demonstrates that FoxA1 binding induces disruption of the nucleosome structure. We propose that this highly remodeled chromatin segment serves as a positioning signal for nucleosomes flanking the −360/−332 FoxA1 binding site. During hormone induction in the presence of FoxA1, the GRE-containing nucleosome forming an enhanceosome complex is extended upstream and downstream, and the flanking C-nucleosome becomes protected (Fig. (Fig.44 and and5).5). We propose that this positioned C-nucleosome causes the reduction in FoxA1 binding at the distal sites. Conversely, the FoxA1 binding to the proximal sites is enhanced (Fig. (Fig.11 and 2A and B). These changes in FoxA1 binding efficacy may thus be caused by a GR-induced change of the nucleosome architecture, resulting in a strongly positioned C-nucleosome and a remodeled B-nucleosome now turned into a GR- and FoxA1-containing enhanceosome, i.e., an enlarged nucleosome (Fig. (Fig.5).5). In the presence of NF1 and Oct1, the hormone-dependent enlarged 200-bp-long FoxA1-containing enhanceosome, −225/−25, becomes more strongly protected against MNase (Fig. (Fig.4).4). At the same time, the neighboring upstream nucleosome tends to occupy two alternative positions (Fig. (Fig.4,4, compare lanes 7 and 9). We conclude that the different transcription factor contexts specify different nucleosomal organizations over the MMTV promoter, which have drastic effects on the transcriptional response (Fig. (Fig.66 to to8).8). The overlapping Oct1 and FoxA1 binding sites in the −51 and −39 region are intriguing. The maintenance of the FoxA1-dependent enlarged nucleosome (Fig. (Fig.5)5) and the NF1- and Oct1-dependent enhancement in nuclease protection (Fig. (Fig.11 and and4)4) together with the stepwise increase in transcriptional enhancement by each added factor (Fig. (Fig.6D6D and data not shown) indicate that both FoxA1 and Oct1 may bind either concomitantly or sequentially in a dynamic equilibrium. Incidentally, overlapping FoxA1 and Oct1 binding sites also flanking an NF1 site have been described in the 6-phosphofructo-2-kinase gene (31) and were highly enriched in regulatory regions of androgen-responsive genes in prostate cancer cells (26). The commonality of this densely packed regulatory module also found in cellular genes indicates an important role of this troika of factors.
Whole-genome studies have attempted to predict nucleosome positioning in the eukaryotic cell based on the DNA sequence (24, 41). However, the predictive capacity of existing algorithms is modest (42). Attempts to reconstruct nucleosome positioning in vitro to reproduce nucleosome organization observed in situ have often failed (5, 21), indicating that translational nucleosome positioning is not determined by the DNA sequence alone and that DNA-binding factors are required. Furthermore, genome-wide studies indicate that specific DNA binding of transcription factors plays an important role in the nucleosome organization of gene regulatory DNA both in yeast (2) and in mammals (40). Our finding that the translational nucleosome organization of the MMTV LTR is directed by the DNA binding factor context may thus be a common feature for gene regulatory DNA segments in metazoans.
The stimulating effect of FoxA1 on GR-GRE binding and on MMTV transcription is enhanced as hormone concentration is reduced (Fig. (Fig.6)6) to the physiological level (37). What is the mechanism of this phenomenon? GR has been classified as a “pioneer factor,” i.e., a protein that is able to bind to its DNA site in inactive chromatin and to initiate chromatin remodeling by itself (47). However, recent studies performed with endogenous GR-regulated genes in tissue culture cells demonstrated a preponderance for GR binding to preexisting DHSs, representing open/remodeled segments of chromatin (27). Importantly, these authors also demonstrated that cell-type-specific distribution of GRE-containing DHSs correlates with cell-type-specific hormone response. In line with these results, genome-wide studies of GREs demonstrated strong conservation of GRE-flanking sequences, emphasizing the composite nature of GREs (44). It appears that the presence of a GR binding site is essential, but not sufficient, for GR binding to occur. Binding of GR apparently needs a partner(s) that contributes with binding energy and/or provides a more open chromatin environment. This implies that chromatin presetting by tissue specifically expressed combinations of constitutively DNA binding transcription/enhancer factors represents a major part of the mechanisms involved in the establishment of the tissue-specific responses exerted by nuclear receptors.
We demonstrate that FoxA1 binding to the inactive MMTV LTR creates an approximately 400-bp area of strongly remodeled DNA segment, i.e., the size of two nucleosomes (Fig. (Fig.11 and and3).3). Thus, FoxA1 binding results in a more open chromatin structure. We propose that the mechanism of FoxA1-stimulated GR binding and transcription depends on its ability to mediate a more open chromatin structure. It remains to be investigated if FoxA1 recruits a chromatin remodeling activity or if the mere DNA binding is sufficient to invoke the remodeling, as suggested previously by in vitro studies (12). An alternative mechanism based on cooperative binding of FoxA1 and GR may also be relevant and might be tested by in vivo cross-linking.
Several previous reports have demonstrated the involvement of FoxA1 in gene regulation by nuclear receptors. FoxA1 binding was shown to overlap with GR binding sites in the tyrosine aminotransferase gene (36), and careful selective mutagenesis of this composite response element demonstrated its positive effects on GR-induced transcription in transient transfections (39). Furthermore, FoxA1 was shown to cooperate with GR binding and transcription in the phosphoenolpyruvate carboxykinase gene (25, 45), and the presence of FoxA1 sites has been found to correlate strongly with estrogen receptor (ER) binding sites in breast cancer cells (10, 33) and the androgen receptor (AR) in prostate cancer cells (26, 33). RNA interference-based reduction of FoxA1 interfered with nuclear receptor binding and gene regulation (16). This suggests that FoxA1 and/or other members of the forkhead transcription factor family may act as licensing factors by opening certain domains within the genome for nuclear receptor binding. Our results presented here demonstrate how this may be achieved.
RU486 in GR-containing oocytes behaves as a pure antagonist (6). Addition of RU486 to previously hormone-activated MMTV LTR results in the cessation of transcription and loss of chromatin remodeling and nucleosome positioning, although RU486 promotes nuclear uptake of GR and a weak GR-GRE binding, seen by DMS in vivo footprinting in oocytes containing a high concentration of GR (6). Here we show that the addition of RU486 together with corticosterone to FoxA1 and GR-containing oocytes fails to block transcription and that RU486 alone acts as a partial receptor agonist in a FoxA1 context (Fig. (Fig.7).7). We also show that FoxA1 binding mediates strong chromatin remodeling around the GREs (Fig. (Fig.1)1) and that the GR-RU486 complex binds specifically, albeit weakly, to a GRE template that has been poised by the constitutive binding of FoxA1 (Fig. (Fig.7D).7D). As mentioned above, GR harbors two activating domains, a ligand-dependent one in the C terminus and a constitutive activating domain in the N terminus (18, 19). It seems that RU486 binding blocks the activating capacity of the C-terminal domain so that it can no longer recruit any chromatin remodeling activity. However, the weakly activating N-terminal domain of GR is able to mediate constitutive activation but only if supported by FoxA1 (Fig. (Fig.8).8). Apparently, the GR N-terminal activating domain is insufficient to cope with the chromatin structure at the GRE. We propose that the FoxA1-mediated chromatin remodeling is required to support the function of this truncated GR protein.
GR, progesterone receptor, ER, and AR belong to the ligand-activated nuclear receptor family (29). We speculate that the FoxA1-dependent conversion of the glucocorticoid hormone antagonist into a partial agonist in the presence of FoxA1 can operate also in transcriptional regulation by AR and ER. If correct, this may have implications for the mechanism of tumor progression in the course of antihormone therapy, for testing of new receptor antagonists, and for the selection of new targets for therapeutic interventions.
We are grateful to Per-Henrik Holmqvist for providing the FoxA1 mRNA required for this work and to Sandro Rusconi for kindly providing the LBD-truncated GR expression vector pSTC3-556.
We thank the Swedish Cancer Foundation (grant 08 0535) and the Swedish Research Council—Medicine (grant K2008-66X) for financial support. C.Å. is sponsored by a postdoc fellowship from the Swedish Research Council—Medicine.
Published ahead of print on 17 August 2009.