|Home | About | Journals | Submit | Contact Us | Français|
We showed previously that the strong PHO5 promoter is less dependent on chromatin cofactors than the weaker coregulated PHO8 promoter. In this study we asked if chromatin remodeling at the even stronger PHO84 promoter was correspondingly less cofactor dependent. The repressed PHO84 promoter showed a short hypersensitive region that was flanked upstream and downstream by a positioned nucleosome and contained two transactivator Pho4 sites. Promoter induction generated an extensive hypersensitive and histone-depleted region, yielding two more Pho4 sites accessible. This remodeling was strictly Pho4 dependent, strongly dependent on the remodelers Snf2 and Ino80 and on the histone acetyltransferase Gcn5, and more weakly on the acetyltransferase Rtt109. Importantly, remodeling of each of the two positioned nucleosomes required Snf2 and Ino80 to different degrees. Only remodeling of the upstream nucleosome was strictly dependent on Snf2. Further, remodeling of the upstream nucleosome was more dependent on Ino80 than remodeling of the downstream nucleosome. Both nucleosomes differed in their intrinsic stabilities as predicted in silico and measured in vitro. The causal relationship between the different nucleosome stabilities and the different cofactor requirements was shown by introducing destabilizing mutations in vivo. Therefore, chromatin cofactor requirements were determined by intrinsic nucleosome stabilities rather than correlated to promoter strength.
Nuclear eukaryotic DNA is packaged into nucleosomes, where DNA is wrapped around a protein core consisting of eight histone proteins (48). The nucleosome forms the basic unit of a complex protein-nucleic acid structure termed chromatin. Chromatin structure has a strong influence on the regulation of gene transcription as the accessibility of DNA regions, for example, promoter elements and transactivator binding sites, is restricted and modulated by their incorporation into nucleosomes. Therefore, it has become an important field of research to understand the mechanisms by which transcription activators or repressors and the transcriptional machinery gain access to their binding sites and navigate the chromatin environment (51).
Many yeast nucleosomes are clearly positioned in relation to the DNA sequence (45, 49, 67, 82, 85), and nucleosomes are shown to occlude transactivator binding sites (47, 80). Nonetheless, it has become clear that nucleosomes, despite their intrinsic mostly repressive function, are highly dynamic. Especially in yeast promoter regions, there is a constant turnover of histones (20, 34, 62). The dynamics of chromatin are mediated by an intricate interplay of chromatin-related cofactors. For example, the so-called remodeling complexes, like the SWI/SNF, Ino80, or ISWI complexes, use the energy of ATP to either slide nucleosomes along the DNA, to alter the nucleosome structure to provide more accessible DNA, to exchange histones from the octamer core for variant histones, or even to completely disassemble nucleosomes and evict the histones from the previously nucleosomal region (10, 24, 46, 79). Remodeling complexes work in concert with a great variety of histone-modifying enzymes that add or remove chemical modifications like acetyl, methyl, or phosphate residues (11, 40). Further, free histones that are not part of a nucleosome are highly aggregation prone and are therefore bound by a diverse group of histone chaperones that assist nucleosome assembly and disassembly (56). At present it is not possible to predict which chromatin cofactors are required for chromatin remodeling in a particular case, as no comprehensive rules for cofactor requirements have been established.
The yeast PHO5 promoter is a classical example for the role of chromatin in promoter regulation (74). Upon induction, an array of four positioned nucleosomes at the repressed promoter becomes mostly remodeled, leading to an extended nuclease-hypersensitive site that is largely depleted of histones (3, 14, 58). That way an additional binding site for the specific transactivator Pho4 becomes accessible, which is a critical prerequisite for gene induction (25, 26). The PHO8 promoter is coregulated by the same transactivator as PHO5 and also shows a pronounced chromatin transition upon induction (5) but has much lower promoter strength, i.e., the transcriptional activity in the fully induced state is much lower (52). In the past, we and others studied extensively the mechanisms that lead to promoter chromatin opening at these two promoters. At both promoters the SWI/SNF and Ino80 remodeling complexes, the histone acetyltransferase Gcn5, and the histone chaperone Asf1 are involved in chromatin remodeling (6). However, the degree of cofactor requirement is markedly different. Whereas the PHO8 promoter strictly depends on the ATPase subunit Snf2 of the SWI/SNF complex and on Gcn5 for promoter opening (28), there are redundant pathways for PHO5 promoter chromatin remodeling, and no essential cofactor downstream of the transactivator Pho4 has been identified yet (6). Previously, we suggested that different intrinsic stabilities of promoter nucleosomes could be the reason for the differential cofactor requirement at these two promoters (31). Now, we wondered if it was a general trend that stronger promoters are packaged into less stable nucleosomes and show less dependency on chromatin cofactors.
In order to address this question without further complication by comparing different transactivation mechanisms, we turned to the PHO84 promoter, which is coregulated with the PHO5 and PHO8 promoters but is even stronger than the PHO5 promoter (54). The PHO84 gene encodes a high-affinity phosphate transporter (15), and its mechanism of transcriptional regulation via regulation of Pho4 activity, as it is common to the phosphate-regulated genes, is mostly known (35, 37, 55). A comparative study of the transcriptional induction of the two coregulated PHO5 and PHO84 genes in response to phosphate starvation showed a lower threshold for PHO84 induction. Cells grown in medium with intermediate phosphate concentrations activate transcription of PHO84 but not of PHO5 (71). Even growth in rich yeast extract-peptone-dextrose (YPD) medium, which is mostly repressive for PHO5 induction, leads to significant levels of PHO84 transcription (23, 53). Also, the induction of PHO84 occurs more rapidly than induction of PHO5. However, this is not an intrinsic feature of the PHO84 promoter but a consequence of the lower threshold of induction. Polyphosphate stores in the cell buffer the physiological signaling pathway of phosphate starvation, leading to a gradual increase in signal strength and an earlier response of the PHO84 promoter than of the PHO5 promoter. Mutants that are defective in polyphosphate storage induce PHO5 and PHO84 with similar kinetics (77).
The role of chromatin in the regulation of the PHO84 promoter has not been explicitly studied yet. Nonetheless, there are several reports on effects of chromatin-related cofactors on the activity of PHO84 under repressing or inducing conditions. This argues that also the PHO84 promoter is regulated on the level of chromatin structure and makes it a promising model for the study of promoter chromatin remodeling mechanisms.
Genome-wide expression analyses in rich YPD medium revealed that PHO84 is downregulated in the absence of Gcn5 or Snf2 (44). Shukla et al. (68, 69) demonstrated reduced acetylation of histone H3 and reduced recruitment of TATA binding protein and polymerase II at the PHO84 promoter under such conditions in a gcn5 mutant. The recruitment of Snf2 to the PHO84 promoter in YPD medium has been directly shown, and this recruitment is dependent on Pho4 and vice versa (23). Also, both Snf2 and Ino80 are present at the induced PHO84 promoter, and induced PHO84 mRNA levels are reduced in the absence of these cofactors (36, 72). Further, basal transcription is increased in the absence of the histone methyltransferase Set1 (16). Very recently, during preparation of the manuscript, a comprehensive study of PHO regulon promoters explained very convincingly that the low threshold of PHO84 induction and its high dynamic range are due to the affinities of the five Pho4 binding sites and their positions in relation to positioned nucleosomes at the PHO84 promoter (41). That study also showed that PHO84 promoter nucleosomes become remodeled upon induction, but the role of chromatin cofactors was not addressed.
We have now characterized the chromatin states at the PHO84 promoter under repressing and inducing conditions and present findings of our comprehensive investigation of the role of Pho4 binding sites, i.e., UASp elements, and chromatin cofactors in PHO84 promoter chromatin dynamics. The PHO84 promoter in the repressed state exhibited a short hypersensitive region that was flanked by two positioned nucleosomes and harbored two high-affinity Pho4 binding sites. Upon induction, this chromatin structure was remodeled into an extensive hypersensitive region that was depleted of histones and allowed access to two additional UASp elements. This chromatin transition was strongly dependent on Snf2, Ino80, and Gcn5, weakly dependent on the histone acetyltransferase Rtt109, and even more weakly on the histone chaperone Asf1. Strikingly, remodeling of each of the two nucleosomes flanking the short hypersensitive region in the repressed state showed a markedly different degree of cofactor requirement. Remodeling of one was critically dependent on Snf2, whereas remodeling of the other one was not. In addition, remodeling of the latter was less dependent on Ino80 than remodeling of the former and was even remodeled in the simultaneous absence of both Snf2 and Ino80. Therefore, the strong PHO84 promoter appeared to be a hybrid between the PHO5 and PHO8 promoters with regard to the presence of both a stable, strictly Snf2 dependent nucleosome and a less stable, redundantly remodeled nucleosome at the same promoter. We show that this differential cofactor requirement was caused by different intrinsic stabilities of the two nucleosomes, as manipulation of nucleosome stability resulted in corresponding changes in the degree of remodeling cofactor requirements. We suggest that cofactor requirements for remodeling of promoter nucleosomes are mainly determined by intrinsic stabilities of individual nucleosomes and that promoter strength is not stringently predictive for cofactor requirements.
For a complete list of the Saccharomyces cerevisiae strains used in this study see Table Table1.1. Strain CY338 is a derivative of CY337 where the PHO4 locus was disrupted by transformation with a linear DNA fragment of the PHO4 locus with a URA3 marker gene cassette inserted into the PHO4 open reading frame (ORF). CY339, CY409, and other pho5 derivatives of strains were constructed by transformation with a linear fragment that inserted a URA3 cassette instead of the BamHI-SalI fragment at the PHO5 locus. Yeast strains were grown under repressive conditions (high phosphate [+Pi]) in YPD with 0.1 g/liter adenine plus 1 g/liter KH2PO4, in yeast nitrogen base selection medium supplemented with the required amino acids for plasmid-bearing strains, and in phosphate-free synthetic medium for induction (3, 6).
The plasmids pCB84a and pCB84b are derivatives of pCB/WT (26) in which a LEU2 marker cassette is inserted into the HindIII site and where the PHO5 promoter is exchanged for the PHO84 promoter. In more detail, a PCR product, generated with the primers PHO84(do) (5′-AGATTTAAACATTTGGATTGTATTCGTGG-3′) and either PHO84(up-885) (5′-CAGGATCCAAAGTGTCACGTG-3′) for pCB84a or PHO84(up-479) (5′-CAGGATCCCGTTCCTCTCACTG-3′) for pCB84b and genomic DNA as template, was ligated via BamHI and DraI into the PHO5 promoter. As there are multiple DraI sites in the vector, the vector was opened via BamHI and SalI and the DraI-SalI fragment 5′ of the PHO5 ORF was prepared separately and added to the ligation mixture, resulting in a triple ligation of PCR product, BamHI-SalI-digested vector backbone, and the DraI-SalI fragment. Plasmid pCB84a was used as template for generating the UASp variants UASpCmut, -Dmut, and -Emut by the Megaprimer method (63) with the following primers that introduced the point mutations: PHO84-mutC, 5′-GCCAATTTAATAGTTCATCGATGATCAGTTATTTCCAGCACGTG-3′; PHO84-mutD, 5′-GGACGTGTTATTTCCACATCGATGGGCGGAAATTAGCGAC-3′; PHO84-mutE, 5′-GCTTATTAGCTAGATTAAAACTAGTCGTATTACTCATTAATTAAC-3′. The following primers were used as reverse primers for generating UASpCmut, -Dmut, and -Emut, respectively: PHO84-rev1, 5′-CCACAATAGTAAGTGG-3′; PHO84-rev2, 5′-CTGGTGATCTACGAG-3′. The point mutations introduced a ClaI site each instead of UASpC and UASpD and a SpeI site instead of UASpE. The combined mutations of UASpCEmut and UASpDEmut were generated by inserting the BsgI-MstII fragment from the UASpEmut plasmid into the UASpCmut and UASpDmut plasmids, respectively. The UASpBmut plasmid and plasmid pCB84a-10A were generated using pCB84a as template and the QuikChange kit (Stratagene) with the following mutagenesis primers: pho84-mutBfor, 5′-GAAATGACAGCAATCAGTATTACGGAATTCGGTGCTGTTATAGGCGCCCTATAC-3′, and pho84-mutBrev, 5′-GTATAGGGCGCCTATAACAGCACCGAATTCCGTAATACTGATTGCTGTCATTTC-3′ for pCB84a-Bmut and pho84-A10for, 5′-GTATAGGGCGCCTATAACAGCACCAACGTGCGTAAAAAAAAAAGCTGTCATTTCTTGGCATGTTTTCT-3′, and pho84-A10rev, 5′-AGAAAACATGCCAAGAAATGACAGCTTTTTTTTTTACGCACGTTGGTGCTGTTATAGGCGCCCTATAC-3′, for pCB84a-10A, respectively. The point mutation in pCB84a-Bmut introduced an EcoRI site instead of UASpB. Plasmid pCB84a-19A was generated with the QuikChange kit and pCB84-10A as template and the primers pho84-A19for, 5′-TGCTGCACGTATAGGGCGCCTATAACAGCACCAAAAAAAAAAAAAAAAAAAGCTGTCATTTCTTGGCAT GTTTTC-3′, and pho84-A19rev, 5′-GAAAACATGCCAAGAAATGACAGCTTTTTTTTTTTTTTTTTTTGGTGCTGTTATAGGCGCCCTATACGTGCAGCA-3′. Plasmid pUC19-PHO84 was prepared by ligating a 3.5-kb PCR product, generated with the primers 5′-CCGGAATTCTCGAGTCATGATTTGGAACAGCTCC-3′ and 5′-CGCGGATCCGCAGAGAGATGTGAGGAAAT-3′ and genomic DNA from strain BY4741 as template, via EcoRI and BamHI, into pUC19. Plasmids pUC19-PHO84-10A and -19A were generated from pUC19-PHO84 and from pUC19-PHO84-10A with the primers pho84-A10for/-rev and pho84-A19for/-rev, respectively, and the QuikChange kit. The DNA sequence of the PHO84 promoter region in all plasmids constructed in this study was confirmed by dideoxy sequencing (data not shown). The Pho4 overexpression plasmid pP4-70L corresponds to YEpP4 (75) but carries the LEU2 instead of the URA3 marker.
Acid phosphatase assays were done as described previously (29). The preparation of yeast nuclei (3) and chromatin analysis of nuclei by restriction nucleases and DNase I digestion with indirect end labeling were as described previously (27, 76). Secondary cleavage for DNase I indirect end labeling was done with HindIII for both the chromosomal and the plasmid locus (at bp −1453 and −1239 from the ATG of the PHO84 ORF for chromosomal and plasmid locus, respectively). For secondary cleavage after chromatin digestion with BsrBI, HhaI, MfeI, PacI, AgeI, SpeI, and FokI, we used HindIII for the chromosomal locus and HindIII/SalI for the plasmid locus. The probe for the chromosomal locus is a PCR product corresponding to bases −1428 to −1083 from the ATG of the PHO84 ORF, and the probe for the plasmid locus corresponds to the HindIII-BamHI fragment of pBR322. Due to the presence of multiple HhaI sites in the plasmid probe region, i.e., the HindIII-BamHI fragment of pBR322, BamHI and EcoRV were used for secondary cleavage and a PCR product from −557 to −310 was used as probe in order to monitor HhaI accessibility at the plasmid locus. Due to the frequent occurrence of TaqI sites, AvaII/ClaI were used for the chromosomal and BamHI/SalI for the plasmid locus for secondary cleavage and a PCR product from −736 to −371 was used as a probe for monitoring TaqI accessibilities.
Yeast cultures with a density of 1 × 107 to 2 × 107 cells/ml were treated with 1% formaldehyde for 20 min at room temperature. Cross-linking was quenched by adding glycine to a final concentration of 125 mM. The cells were washed two times with ice-cold 0.9% NaCl, resuspended in HEG150 buffer (150 mM NaCl, 50 mM HEPES, pH 7.6, 10% glycerol, 1% Triton X-100, 1 mM EDTA, 1 mM dithiothreitol [DTT], 1 mM phenylmethylsulfonyl fluoride) and lysed with a French press (three times at 1,100 lb/in2) or by sonication (Bioruptor; Diagenode; three times for 30 s with a 60-s pause, position high, ice water bath). In this last step, chromatin was sheared to an average size of 500-bp fragments. Chromatin immunoprecipitation (ChIP) was performed as described before (73). The anti-histone H3 C-terminal antibody was obtained from Abcam (ab1791-100). Immunoprecipitated DNA was quantitatively measured in triplicates with the ABI Prism 7000 sequence detection system using the following amplicons: TEL-1, 5′-TCCGAACGCTATTCCAGAAAGT-3′; TEL-B, 5′-CCATAATGCCTCCTATATTTAGCCTTT-3′; TEL-probe, 5′-6-carboxyfluorescein [FAM]-TCCAGCCGCTTGTTAACTCTCCGACA-6-carboxytetramethylrhodamine (TAM)-3′; ACT1-A, 5′-TGGATTCCGGTGATGGTGTT-3′; ACT1-B, 5′-TCAAAATGGCGTGAGGTAGAGA-3′; ACT1-probe, 5′-FAM-CTCACGTCGTTCCAATTTACGCTGGTTT-TAM-3′; PHO84 UASpC-A, 5′-GAAAAACACCCGTTCCTCTCACT-3′; PHO84 UASpC-B, 5′-CCCACGTGCTGGAAATAACAC-3′; PHO84 probe, 5′-FAM-CCCGATGCCAATTTAATAGTTCCACGTG-TAM-3′.
Salt gradient dialysis was performed as described previously (42). A typical assembly reaction mixture contained 10 μg supercoiled plasmid DNA (Qiagen preparation), 20 μg bovine serum albumin (A-8022; Sigma), and variable amounts (for example, 6 or 10 μg) of Drosophila melanogaster embryo histone octamers (70) in 100 μl high-salt buffer (10 mM Tris-HCl, pH 7.6, 2 M NaCl, 1 mM EDTA, 1 mM β-mercaptoethanol, 0.05% Igepal CA630 [I-3063; Sigma]) and was dialyzed for 15 h at room temperature while slowly diluting 300 ml of high-salt buffer with 3 liters of low-salt buffer (same as the high-salt buffer, but with 50 mM NaCl) using a peristaltic pump. A final dialysis step versus low-salt buffer ensured a final NaCl concentration of 50 mM.
Yeast whole-cell extract was prepared as previously described (31) with the following modifications. Commercially available baker's yeast concentrate (Deutsche Hefewerke GmbH, Nürnberg, Germany) was used as starting material for an upscaled version of the preparation. The extraction buffer was modified to 0.2 M HEPES-KOH, pH 7.5, 10 mM MgSO4, 10% glycerol, 1 mM EDTA, 390 mM (NH4)2SO4, 1 mM DTT, and 1× Complete protease inhibitor without EDTA (Roche Applied Science), and the buffer for resuspension after the ammonium sulfate precipitation was 20 mM HEPES-KOH, pH 7.5, 10% glycerol, 80 mM KCl, 1 mM EGTA, 5 mM DTT, and 1× Complete protease inhibitor without EDTA. For the final dialysis the same buffer as for resuspension but with 0.1 mM phenylmethylsulfonyl fluoride and 1 mM sodium metabisulfite instead of the Complete protease inhibitor was used and exchanged to completion.
A 100-μl reconstitution reaction mixture with 1 μg DNA preassembled by salt gradient dialysis was incubated with or without yeast extract (~250 μg protein, judged from Coomassie-stained gel lanes in comparison to standard protein) and with or without a regenerative energy system (3 mM ATP-MgCl2, 30 mM creatine phosphate [Sigma], and 50 ng/μl creatine kinase [Roche Applied Science]) in assembly buffer (20 mM HEPES-KOH, pH 7.5, 10% glycerol, 80 mM KCl, 0.5 mM EGTA, 2.5 mM DTT) for 2 h at 30°C.
Aliquots (25 μl) of a reconstitution reaction mixture were mixed with an equal volume of digestion buffer (20 mM HEPES-KOH, pH 7.5, 12% glycerol, 5.5 mM MgCl2, 5.5 mM CaCl2, 2.5 mM DTT, 80 mM NaCl, 0.1 mg/ml bovine serum albumin) containing DNase I (04716728001; Roche Applied Science) at a concentration in the range of 0.005 to 0.02 U/ml (free DNA), 0.02 to 0.1 U/ml (salt gradient dialysis chromatin), or 2 to 10 U/ml (salt gradient dialysis chromatin with extract) and incubated at room temperature for 5 min. The digestion reactions were stopped by adding 10 μl of Stop buffer (10 mM EDTA, 4% sodium dodecyl sulfate), deproteinated by proteinase K digestion overnight, and ethanol precipitated. SspI (bp −1440 from the ATG of the PHO84 ORF) was used for secondary cleavage instead of HindIII. For direct comparison between in vitro-reconstituted chromatin and in vivo chromatin (see Fig. Fig.7A,7A, below), SspI was used for all loci.
Prior to restriction enzyme digestions, ATP was removed from the reconstitution reaction mixtures to inhibit ATP-dependent remodeling during the restriction digestion by adding 0.1 U of apyrase (M0393L; New England Biolabs) to the reaction mixtures and incubating for 30 min at 37°C. Two-microliter aliquots of an apyrase-treated reconstitution reaction mixture were combined with 30 μl of RE digestion buffer (20 mM HEPES-KOH, pH 7.5, 4.5 mM MgCl2, 2.5 mM DTT, 80 mM NaCl, 0.5 mM EGTA) and treated with two different enzyme concentrations for each restriction enzyme, similar to the in vivo RE digests. The reactions were stopped by adding 7.5 μl Stop buffer, deproteinated by proteinase K digestion overnight, and ethanol precipitated. Secondary cleavage was performed as described above for the chromosome locus.
We characterized the PHO84 promoter chromatin structure under repressing conditions, i.e., in rich or synthetic medium with additional phosphate to ensure full repression, and under inducing conditions, i.e., synthetic phosphate-free medium. By DNase I indirect end-labeling analysis of the repressed state (+Pi) we detected a short hypersensitive (sHS) region (about 150 bp), roughly between the MfeI and ApaI restriction sites, that was flanked by one positioned nucleosome upstream and one downstream (Fig. 1A and B, upstream nucleosome and downstream nucleosome). This sHS region contained two closely positioned high-affinity Pho4 binding sites, UASpC and UASpD, whereas the two low-affinity sites, UASpB and UASpE, were occluded by the positioned upstream and downstream nucleosomes, respectively (Fig. (Fig.1B)1B) (54). In addition, we observed a broad hypersensitive region upstream of the BsrBI restriction site. Upon induction (−Pi), the upstream nucleosome and at least one nucleosome downstream of the sHS region were remodeled, leading to an extended hypersensitive (eHS) region of about 500 bp. Its upstream border was almost fused to the broad hypersensitive region and the downstream border faded into the core promoter region around the TATA box and the transcriptional start site (Fig. 1A and B; see also Fig. Fig.4B,4B, ,5A,5A, and and8A,8A, below). This way UASpB and UASpE became accessible (Fig. (Fig.1B).1B). Sometimes the eHS region appeared to contain a short region of lower DNase I accessibility between the MfeI and ApaI sites (see Fig. Fig.4B4B and and8A),8A), which may reflect Pho4 and recruited factors bound to UASpC and UASpD. In Fig. Fig.1A1A the intensity of the broad hypersensitive region upstream of the BsrBI site appeared to change somewhat upon induction, which was probably attributable to an overall lower degree of digestion. However, in the majority of cases it did not undergo major changes upon induction (see Fig. Fig.4B,4B, ,5A,5A, and and8A,8A, −Pi panels, below; also, data not shown). Therefore we refer to it as a constitutive hypersensitive region (cHS).
The chromatin transition was fully dependent on the transactivator Pho4, as the PHO84 promoter chromatin pattern under inducing conditions in a pho4 deletion strain was virtually the same as the wild-type (wt) pattern of the repressed state (Fig. (Fig.1A).1A). Interestingly, the unchanged nucleosome organization in a pho4 mutant suggested that the nucleosome positioning at the repressed promoter did not depend on binding of Pho4, e.g., to its linker binding sites UASpC and UASpD.
In addition to DNase I indirect end labeling, we mapped the PHO84 promoter chromatin structure of the repressed and the induced state more quantitatively by assaying the accessibility for several restriction enzymes along the promoter region that underwent the chromatin structure transition (Fig. 1C and D). Under +Pi conditions, the accessibilities for the various restriction enzymes were rather different, as would be expected for an organization into nucleosomes and nucleosome-free linker regions. The accessibilities at the HhaI and TaqI sites were the lowest, speaking for their protection by the upstream and downstream nucleosome, respectively. The BsrBI site was fully accessible under both repressing and inducing conditions, which was in agreement with its localization at the downstream start of the cHS region (Fig. (Fig.1A).1A). The MfeI site was substantially but not fully accessible in the repressed state, indicating a location at the very border between the downstream nucleosome and the sHS region (Fig. (Fig.1A).1A). Interestingly, a region of about 100 bp between the downstream nucleosome and the TATA box was only semiprotected in the repressed state, as the accessibilities for PacI, AgeI, and FokI were in the range of 43% (FokI) to 57% (AgeI) (Fig. 1C and D). This argued against a clearly positioned but rather for a less-organized nucleosome or for a chromatin structure with increased plasticity. Alternatively, some other DNA-protecting entity, e.g., an assembly of general transcription factors, could be responsible for this semiprotection.
In the induced state, all restriction enzyme sites tested in the promoter region of more than 500 bp upstream of the TATA box were highly accessible (Fig. 1C and D), confirming the presence of an extended hypersensitive region as observed by DNase I indirect end labeling (Fig. (Fig.1A)1A) and suggesting that the whole region was mostly nucleosome free. Restriction enzyme accessibility assays also confirmed that the transition to this open chromatin state was dependent on Pho4 (Fig. (Fig.1D).1D). For unknown reasons, the accessibilities at the HhaI, PacI, and AgeI sites, but not at the TaqI site, were even decreased under inducing compared to noninducing conditions in the pho4 strain.
In the wt strain, the accessibility of the FokI cleavage site, which overlaps with the TATA box sequence (15), also increased upon induction, but not to the same high level as for the other restriction enzyme sites. In addition, the accessibility of the FokI site in the induced state was quite variable. This altogether may be due to the poor performance of this restriction enzyme on chromatin templates or may indicate the presence of an unstable or partially remodeled nucleosome or of components of the general transcription machinery recruited to the TATA box under inducing conditions.
In summary, the restriction enzyme accessibility data in connection with the DNase I indirect end-labeling analysis led us to map the upstream and downstream nucleosome as shown in Fig. 1B and D. The main guidelines were the location of the ApaI and MfeI sites just at the borders of the nucleosomes toward the sHS region. For the reasons stated above, we have not assigned clear nucleosomal positions to the region between the downstream nucleosome and the TATA box region but suggest a less-organized DNA protective structure there.
This less-organized structure together with the somewhat elevated accessibilities at the HhaI and TaqI sites suggested to us that there may be a low level of Pho4 present at the promoter even under repressive conditions. Under +Pi conditions Pho4 is mostly phosphorylated at multiple sites and mainly located in the cytosol (37), but some Pho4 may still be nuclear. For example, earlier we showed a Pho4 footprint at the repressed PHO8 promoter (52) and sin mutations in histone H4 showed significantly derepressed PHO5 activity in a UASp element-dependent, i.e., presumably Pho4-dependent, manner under otherwise-repressing conditions (81). Such nuclear Pho4 may bind especially to the accessible high-affinity sites UASpC and UASpD in the sHS region. This could lead to some basal recruitment of chromatin remodeling activities and a partially remodeled chromatin structure. We tested this by restriction enzyme analysis of the PHO84 promoter region in a pho4 deletion strain under high-phosphate conditions (Fig. (Fig.1D).1D). However, only the accessibility of the HhaI site was decreased significantly, arguing that there was some basal Pho4-dependent remodeling only of the upstream nucleosome in the repressed state. This may also be noticeable based on the slightly more spread out sHS region in the presence of Pho4 (Fig. (Fig.1A,1A, compare wt +Pi and pho4 −Pi). In contrast, the structure between the downstream nucleosome and the TATA box region was maintained semiopen also in the absence of Pho4.
The generation of an extended hypersensitive region at the induced PHO84 promoter was reminiscent of our previous findings for the PHO5 and PHO8 promoters (3, 5). Such hypersensitivity was found by ourselves and others to reflect not just altered nucleosomal structures but also nucleosome disassembly leading to histone eviction from the promoter regions (1, 14, 38, 58). We checked if histones were lost also from the induced PHO84 promoter. During PHO84 induction kinetics, the histone H3 occupancy was monitored by ChIP using an antibody directed against the C terminus of histone H3. The histone H3 occupancy dropped after 2 hours of induction to about 10% of the level under repressing conditions (Fig. (Fig.2).2). At the same time there was no significant change of the histone H3 occupancy at a telomere control locus. Therefore, chromatin remodeling at the PHO84 promoter eventually led to histone eviction.
A special feature of the PHO84 promoter is the presence of five Pho4 binding sites, UASpA to UASpE, which makes it one of the strongest PHO promoters (54). Ogawa et al. (54) showed previously by using a PPHO84-lacZ reporter construct and deleting an extensive upstream region that UASpA and UASpB were not required for full PHO84 activity. They further showed by site-directed mutagenesis that the low-affinity site UASpE in combination with either of the high-affinity sites UASpC or UASpD was necessary and sufficient for PHO84 regulation. We wished to check if any of these effects on promoter activity actually reflected effects on chromatin remodeling.
We set up an analogous reporter system by constructing plasmid pCB84a, for which the PHO84 promoter was coupled to the PHO5 coding gene. Thereby we avoided possible chromatin structure artifacts due to the close presence of the bacterial lacZ DNA sequence (unpublished observations). The enzymatic activity of the secreted acid phosphatase Pho5 can be measured easily with intact cells and PHO5 transcriptional activity fully correlates with acid phosphatase activity, indicating no significant posttranscriptional regulation of PHO5 expression (8). Importantly, the endogenous copy of PHO5 was always deleted in strains where PHO84 reporter constructs were used.
Using the pCB84a construct we observed phosphate-regulated PHO84 promoter activity with a substantially higher basal and final level of Pho5 acid phosphatase activity than seen with the PHO5 promoter (Fig. (Fig.3A;3A; see also Fig. 9A and B, below). This was expected for the stronger PHO84 promoter.
The PHO84 promoter chromatin structure on the plasmid underwent the same regulated transition as the endogenous chromosomal locus (compare Fig. Fig.3B3B and and1A1A for DNase I mapping; data not shown for restriction enzyme accessibilities). It should be noted that the region far upstream of the PHO84 promoter, which is used for probing in indirect end-labeling techniques, was different between the plasmid and the chromosomal locus, thus allowing for a distinction of both loci within the same cell by differential probing and therefore excellent internal control. Due to the different relative position of the secondary cleavage site at the plasmid and chromosomal locus, the DNase I indirect end-labeling fragments at the plasmid locus were 214 bp smaller, leading to a more stretched out appearance of the plasmid chromatin patterns on the blot. Possible minor changes in nucleosome positions between the chromosomal and the plasmid locus could still be undetected by this low-resolution approach.
Using this reporter plasmid, a set of promoter variants similar to the ones of Ogawa et al. (54) was constructed: a truncated version, plasmid pCB84b, in which effectively the upstream nucleosome and UASpA and UASpB were deleted (ΔΔUASpAB [schematic in Fig. Fig.3A]),3A]), and point mutants for either one of the Pho4 binding sites, UASpC, UASpD, and UASpE, or for two sites together, i.e., UASpCEmut or UASpDEmut. For the truncated promoter the proper positioning of the downstream nucleosome in the repressed state and the generation of the corresponding extended hypersensitive region (truncated eHS type) upon induction were confirmed by DNase I indirect end labeling (data not shown).
Induction of the truncated promoter ΔΔUASpAB as monitored by acid phosphatase activity was very similar to the wt promoter (Fig. (Fig.3A).3A). Mutation of the accessible high-affinity sites, UASpC or UASpD, affected the final promoter activity rather slightly, with the effect of the UASpD mutation being a bit more pronounced (Fig. (Fig.3A).3A). In contrast, the absence of the intranucleosomal low-affinity site, UASpE, had a much stronger effect, reducing the final promoter strength by more than 50%. The combination of mutations in the UASpE and either UASpC or UASpD sites drastically reduced the final promoter activity to about 25% and 15% of the wt activity, respectively. We conclude, in agreement with Ogawa et al. (54), that the contribution of UASpC and UASpD was redundant, whereas UASpE contributed about half the promoter activity by itself. Further, there was some cooperativity between the intranucleosomal UASpE site and the accessible site UASpD and maybe also UASpC, as the effects of the double mutants were larger than the sum of the effects of each single mutant.
Next we examined if the effects on promoter strength were a consequence of inefficient promoter chromatin remodeling or of an effect downstream of chromatin opening. The DNase I indirect end-labeling patterns under inducing conditions of the UASpCmut or UASpDmut promoter variants were the same as for the wt promoter (data not shown), which was in agreement with a rather slight effect of these mutations on promoter activity. The finding that one UASp element in the sHS linker was sufficient for full remodeling of the upstream and downstream nucleosome is similar to the PHO8 but different from the PHO5 promoter, where the linker site UASp1 alone was not sufficient for chromatin remodeling (25). This may be because UASp1 at the PHO5 promoter is a low-affinity binding site, in contrast to the high-affinity linker sites at the PHO84 and PHO8 promoters (5, 7, 54).
Any promoter variant lacking UASpE showed a hypersensitive region under inducing conditions that was less extensive in the downstream direction (eHS*) (Fig. (Fig.3C,3C, schematic). This was especially clear in the DNase I patterns of the induced UASpCEmut and UASpDEmut promoter variants (Fig. (Fig.3C3C and data not shown), in which the extended hypersensitive region (eHS*) extended only up to about the SpeI marker band (−259 bp) (Fig. (Fig.3C),3C), which was introduced with the UASpEmut point mutation and marked therefore the position of UASpE. In contrast, the eHS region of the induced wt promoter pattern reached further downstream beyond the AgeI marker (bp −172) (Fig. 1A and B and and3B).3B). This less-extensive eHS* region was less clearly visible in the DNase I pattern of the UASpEmut variant (Fig. (Fig.3C),3C), but less extensive remodeling downstream of the SpeI site was confirmed also for this variant by a reduced final accessibility of the AgeI site (Fig. (Fig.3C,3C, table). We concluded that UASpE is essentially required for remodeling of the region between the downstream nucleosome and the TATA box.
Previously, we found that remodeling of the chromatin structure at the weak PHO8 promoter was critically dependent on Gcn5 and Snf2 (28). At the stronger PHO5 promoter only the rate of chromatin remodeling was strongly decreased in the absence of Gcn5 or Snf2 (6, 8, 19), but eventually full remodeling was achieved. We wondered if remodeling at the even stronger PHO84 promoter would be mostly or even fully independent of the presence of these cofactors.
First, we examined induction kinetics of the PHO84 promoter in gcn5 cells and found a strong delay in comparison to wt cells, even though the final induction level was unaffected (Fig. (Fig.4A).4A). In agreement with this, the DNase I pattern of the fully induced promoter in the gcn5 mutant was the same as observed in wt cells (Fig. (Fig.4B).4B). Therefore, the Gcn5 activity had no essential role for the final opening of the PHO84 promoter chromatin. This was confirmed further by restriction enzyme analysis of DNA accessibility at the entire promoter region under fully inducing conditions (Fig. 4C and D, −Pi).
In analogy to our earlier findings at the PHO5 promoter (8), we assumed that the kinetic delay on the activity level in the gcn5 mutant (Fig. (Fig.4A)4A) was caused by a delay in the chromatin remodeling step. We quantified chromatin opening for wt and gcn5 cells by restriction enzyme accessibility at 1.5 h after shift to phosphate-free medium and by histone H3 ChIP during an induction time course. To our surprise, we did not catch much of a delay in the increase of restriction enzyme accessibility at this time point of induction. There was only a slight delay compared to wt in opening at the AgeI site, i.e., in the region between the downstream nucleosome and the TATA box (Fig. 4C and D, ,1.51.5 h, −Pi). For comparison, chromatin remodeling at the PHO5 promoter, as probed by ClaI accessibility, was still strongly delayed after 3 hours of induction in a gcn5 strain (8). Nonetheless, we did observe a strong delay in histone eviction kinetics as monitored by histone H3 ChIP (Fig. (Fig.4E).4E). Even after 2 hours of induction, there was six to seven times more histone H3 still present at the promoter in the gcn5 mutant than in the wt cells. Therefore, we observed for the first time a large disparity between restriction enzyme accessibility and histone H3 eviction kinetics during induction of a PHO promoter. We conclude that histone eviction, rather than an initial increase of DNA accessibility, appeared to be the rate-limiting step in PHO84 promoter opening in a gcn5 mutant.
Second, we examined PHO84 promoter induction kinetics in a snf2 mutant and observed a similar delay as with the gcn5 mutant, again with hardly any effect on the final level of induction (Fig. (Fig.4A).4A). In marked contrast and much to our surprise, this final activity of the snf2 strain corresponded to an only partially open DNase I pattern of the induced PHO84 promoter, both on the chromosomal and the plasmid locus (Fig. (Fig.4B4B and data not shown). The downstream nucleosome was remodeled, but the upstream one was not at all. In addition, we noticed that the spreading of the eHS region was less extensive in the downstream direction than in the wt case (eHS**) (Fig. (Fig.4B,4B, schematic) and confirmed this by a reduced final accessibility of the AgeI and PacI sites (Fig. 4C and D, −Pi). This reduced downstream spreading of the eHS** region was similar to the reduced spreading of the eHS* region in the UASpEmut variant (Fig. (Fig.3C).3C). It was even somewhat more severe, as also the PacI site accessibility was reduced in the eHS** but not in the eHS* region (Fig. (Fig.4C4C and and3C,3C, tables). Even though the eHS** region in the snf2 mutant was less remodeled than the eHS* region in the UASpEmut variant, it was still compatible with full final activity levels (Fig. (Fig.4A).4A). So, we concluded that the lower final activity in the UASpEmut, and even more so in the UASpCEmut and UASpDEmut variants (Fig. (Fig.3A),3A), was less due to compromised chromatin remodeling but mainly due to the reduced number of UASp elements (see also reference 41). As the transition from the semiopen to the fully open state in the region between the downstream nucleosome and the TATA box was compromised in both the snf2 mutant and the UASpEmut variant, we suggest that recruitment of the SWI/SNF complex by UASpE-bound Pho4 was essential for chromatin remodeling in this region.
Restriction enzyme probing of the induced state in the snf2 mutant also confirmed the lack of remodeling of the upstream nucleosome, i.e., persistently low HhaI accessibility, and full remodeling of the downstream nucleosome, i.e., high TaqI accessibility (Fig. 4C and D, −Pi). Altogether, this chromatin pattern constituted a third type of extended hypersensitive region (eHS**) (Fig. (Fig.4B,4B, schematic), where the upstream nucleosome was still present, the downstream nucleosome fully remodeled, and the region between the downstream nucleosome and the TATA box not fully remodeled.
The same partially remodeled DNase I pattern was also observed in the snf2K798A strain, which bears a point mutation in the Snf2 ATPase domain (Fig. (Fig.5A),5A), confirming that the ATPase activity of Snf2 rather than some other feature of the SWI/SNF complex was responsible for the observed effect.
In analogy to the gcn5 mutant, we examined whether the kinetic delay of PHO84 promoter induction in the snf2 mutant (Fig. (Fig.4A)4A) corresponded not only to the aforementioned reduction in the final extent of remodeling but also to a kinetic delay of chromatin opening, for example, at the TaqI site in the downstream nucleosome. After 1.5 h of induction there was not much delay in opening of the TaqI or any other site, based on the 1.5-h values for the snf2 strain compared to wt and normalized to their respective −Pi values (Fig. 4C and D). However, histone eviction kinetics measured by histone H3 ChIP in snf2 cells showed a strong delay (Fig. (Fig.4E).4E). At present we are unsure why the final level of histone occupancy at the induced PHO84 promoter in snf2 cells as measured by histone H3 ChIP was not much higher than for the wt and gcn5 strains. This would be expected due to the continued presence of the upstream nucleosome in the snf2 strain. The resolution of our ChIP analysis (about 500 bp) cannot distinguish between the upstream and the downstream nucleosome, because the amplicon used (Fig. (Fig.2,2, schematic) will score fragments from both nucleosome regions. However, as the upstream nucleosome was not remodeled at all and as the downstream region close to the TATA box was remodeled to a lesser extent than in the wt (see above), we assume that histone H3 ChIP mainly monitored remodeling of the downstream nucleosome. Therefore, the delayed histone eviction in the snf2 mutant argues for a role of Snf2 in remodeling of the downstream nucleosome. Similar to the case of the gcn5 mutant, also here histone eviction seemed to be the rate-limiting step.
As remodeling of the downstream nucleosome was eventually complete but kinetically delayed at the histone eviction step in both the snf2 and gcn5 single mutants, we wondered if the downstream nucleosome may not open up at all in a snf2 gcn5 double mutant. This was not the case, as the DNase I pattern of the fully induced PHO84 promoter in the snf2 gcn5 double mutant was indistinguishable from that found in snf2 cells (Fig. (Fig.5B5B).
Previously, it was shown by us and others that submaximal induction conditions can exacerbate the dependency of PHO5 promoter chromatin remodeling on chromatin cofactors (19, 38). Such submaximal induction conditions may be achieved by using low-phosphate rather than phosphate-free medium (19) or by overexpression of Pho4 in high-phosphate medium (25). We tested under the latter conditions whether the differential requirement of Snf2 for remodeling of the downstream and the upstream nucleosome still persisted at submaximal induction. DNase I indirect end-labeling analysis under these submaximal induction conditions showed the same pattern as under fully inducing conditions, for both the wt as well as the snf2K798A mutant (Fig. (Fig.5C).5C). So, even at such low induction levels the downstream nucleosome could be remodeled without Snf2 activity, demonstrating further the different degree of Snf2 requirement for remodeling of the upstream and downstream nucleosome.
The pho4, snf2, and gcn5 mutants all had a decreased basal level of transcription (Fig. (Fig.4A4A and data not shown) (69). In all these three mutants the semiopen less-organized chromatin structure between the downstream nucleosome and the TATA box was not affected in the repressed state. Therefore, this semiopen structure was not sufficient for sustaining substantial basal transcription under repressing conditions.
Nonetheless, in all three mutants the accessibility of the HhaI site under repressing conditions was reduced in comparison to wt, in snf2 and gcn5 cells even more so than in the pho4 mutant (Fig. 4C and D, +Pi, and 1D, table). The reduced HhaI accessibility might have been responsible for the reduced basal transcription. In the wt, the targeted recruitment of Snf2 and Gcn5 by Pho4 could keep the upstream nucleosome in a partially remodeled state, which would allow partial access to UASpB and lead to even more remodeling of the upstream nucleosome and high basal transcription. To test this, we introduced a point mutation in UASpB and found indeed that the HhaI site accessibility under +Pi conditions (19 ± 2%) was significantly lower than at the wt promoter and similar to that of the wt promoter in the pho4 mutant (17 ± 2%) (Fig. (Fig.1D).1D). However, despite this lower HhaI accessibility there was hardly any effect on the basal level of activity for the UASpBmut construct (data not shown), arguing that UASpB and basal remodeling of the upstream nucleosome were not necessary for the substantial basal transcription. In addition, mutation of the other intranucleosomal site, UASpE, which analogously may have been involved in basal remodeling of the downstream nucleosome, did not affect basal transcription either (Fig. (Fig.3A3A).
As we had already observed a cooperation between Snf2 and Ino80 for chromatin remodeling at the PHO5 and PHO8 promoters (6), and as others have shown a recruitment of both Snf2 and Ino80 to the PHO84 promoter upon induction (23, 36, 72), we investigated the role of Ino80 for PHO84 promoter opening. In particular, there was the possibility that Ino80 would be the alternative remodeler for remodeling of the downstream nucleosome in the absence of Snf2.
The absence of Ino80 by itself did not prevent full remodeling of the PHO84 promoter chromatin structure, i.e., the DNase I pattern of an ino80 mutant under fully inducing conditions corresponded to the eHS type of the wt (Fig. (Fig.6A)6A) and the accessibility of restriction enzymes along the promoter region increased to almost-wt levels (Fig. (Fig.6C).6C). Further, the DNase I pattern of the induced promoter in the snf2 ino80 double mutant was indistinguishable from the pattern of the snf2 single mutant (Fig. (Fig.6B).6B). Together, these results argue that Ino80 was neither essentially required for remodeling under fully inducing conditions in the wt strain nor for remodeling of the downstream nucleosome in the absence of Snf2. Nonetheless, the chromatin opening kinetics in the ino80 strain was strongly delayed over the entire promoter region after 1.5 h of induction as examined by restriction enzyme accessibility (Fig. (Fig.6C).6C). Therefore, Ino80 is clearly involved in the wt chromatin remodeling pathway at the PHO84 promoter.
In contrast to Snf2 and Gcn5, Ino80 was not involved in keeping the upstream nucleosome in a partially remodeled state under repressing conditions (+Pi), as the HhaI accessibility was not affected in the ino80 mutant (Fig. (Fig.6C,6C, table, +Pi). A slight decrease in PacI accessibility may indicate that Ino80 has a minor role in positioning the downstream nucleosome under repressing conditions.
As presented above for the case of Snf2, we checked if PHO84 promoter opening became more dependent on Ino80 under submaximal conditions. Strikingly, the DNase I patterns of the snf2K798A and the ino80 mutants at submaximal induction were indistinguishable, i.e., under these conditions the upstream nucleosome became strictly dependent also on Ino80 (Fig. (Fig.6D6D).
As shown above, remodeling of the upstream nucleosome was strictly dependent on Snf2, whereas remodeling of the downstream nucleosome was not (Fig. (Fig.4B4B and and5A).5A). In addition, remodeling of the upstream nucleosome was more dependent on Ino80 than remodeling of the downstream nucleosome (Fig. (Fig.6D).6D). This constitutes a case of differential cofactor requirements for nucleosome remodeling within one and the same promoter.
We found earlier that the differential cofactor requirements for chromatin remodeling at the PHO5 and PHO8 promoters correlated with differential intrinsic stabilities of the positioned nucleosomes (31). These stabilities were measured using our yeast extract chromatin assembly system that is able to generate the proper in vivo nucleosome positioning de novo in vitro (31, 39). In this system, plasmids bearing the yeast locus of interest are assembled by salt gradient dialysis into a chromatin structure with a specific but usually not proper, i.e., not in vivo-like, nucleosome positioning pattern. The in vivo-like pattern is induced in the next step by the addition of yeast whole-cell extract in the presence of energy. A so-far-unidentified energy-dependent activity in the yeast extract apparently constitutes the thermodynamic conditions for in vivo-like nucleosome positioning. In a next step, it is possible to compare the intrinsic stability of properly positioned nucleosomes by titrating the histone concentration. Under conditions of limiting histones (underassembled chromatin) there are fewer nucleosomes deposited onto the DNA than there are nucleosome positions available. Therefore, the multitude of alternative and mostly overlapping nucleosome positions will compete for nucleosome occupancy. Positions that are already occupied in equilibrium in underassembled chromatin are more stable than those that are occupied only in fully assembled chromatin (for a full discussion of this methodology see reference 31). Using this approach, we observed previously that the proper positioning over the PHO5 promoter region could only be generated in fully assembled chromatin, whereas the proper PHO8 promoter pattern was also achieved in underassembled chromatin. Therefore, the intrinsic stability of the PHO8 promoter nucleosomes was higher than the stability of the PHO5 promoter nucleosomes.
With the same methodology we compared the intrinsic stability of the upstream and downstream nucleosome at the PHO84 promoter (Fig. 7A and B). First, we prepared fully assembled salt gradient dialysis chromatin (histone octamer:DNA mass ratio set as 100%) using a plasmid with a 3.5-kb PHO84 insert as template and tested if the yeast extract would generate the in vivo pattern. Much to our surprise, we observed that the DNase I pattern of the salt gradient dialysis chromatin was already very similar to the in vivo pattern (Fig. (Fig.7A,7A, compare SGD and in vivo). This pattern was clearly different from a digest of free DNA and did not change much, as expected (31), with the addition of yeast extract in the absence of energy. This was the first case out of 14 tested yeast loci (C. Wippo and P. Korber, unpublished results) where salt gradient dialysis by itself was already able to generate a very in vivo-like chromatin structure. This suggests that rather strong nucleosome positioning sequence elements in the PHO84 promoter lead to in vivo-like nucleosome positioning already under pure salt gradient dialysis conditions. Nonetheless, incubation with yeast extract and energy did make the pattern more similar to the in vivo pattern, especially regarding the relative band intensities and the upper part of the lane, i.e., the coding region (Fig. (Fig.7A,7A, compare SGD +Yex/ATP with in vivo). Therefore, the PHO84 promoter is one more example where our yeast extract in vitro assembly system constitutes conditions more similar to in vivo conditions for nucleosome positioning than salt gradient dialysis alone.
Second, we repeated the salt gradient dialysis chromatin assembly with limiting histones (histone octamer:DNA mass ratio of 60%) and still obtained a rather in vivo-like pattern (Fig. (Fig.7B).7B). This in vivo-like pattern again did not change upon the addition of yeast extract without energy. However, incubation with yeast extract in the presence of energy, i.e., conditions that should be closer to the in vivo conditions, had a differential effect on the regions upstream and downstream of the sHS region. The upstream nucleosome and the cHS region again became even more like the in vivo pattern, but the sHS region was so much extended further downstream that the position of the downstream nucleosome was compromised. The sHS region was always somewhat sharper in the pure salt gradient dialysis chromatin pattern and became fuzzier upon addition of yeast extract and energy, also with fully assembled chromatin templates (Fig. (Fig.7A;7A; compare widths of brackets). But whereas the more fuzzy sHS region in the fully assembled chromatin (100%) resembled more the in vivo case, it was stretched too far downstream to be compatible with a proper positioning of the downstream nucleosome for the underassembled chromatin templates (60%) (Fig. (Fig.7B;7B; compare widths of brackets). We stress that the more extensive sHS region under underassembled conditions compared to fully assembled conditions (Fig. (Fig.7B)7B) was not due to the use of different degrees of DNase I digestion, as we saw such a difference significantly and repeatedly over a wider range of DNase I digestions (data not shown).
This differential effect on the upstream and downstream nucleosome was confirmed by restriction enzyme accessibility assays. The accessibility of the TaqI site in the downstream nucleosome increased much more (from 15% to 69%) (Fig. (Fig.7B)7B) upon addition of extract and energy to underassembled chromatin than the accessibility of the HhaI site in the upstream nucleosome (from 10% to 36%). The overall lower accessibilities in the fully assembled chromatin compared to the in vivo situation probably reflected here a subpopulation of aggregated, i.e., indigestible, templates in vitro, which may form especially at such high histone concentration. Altogether, these results suggested that the downstream nucleosome was intrinsically less stably positioned in vivo than the upstream nucleosome. This correlated with its more relaxed cofactor requirements.
The finding of higher intrinsic stability of the upstream nucleosome also correlated strikingly with the prediction of the N-score algorithm (84) (Fig. (Fig.7C).7C). The N-score algorithm was trained on in vivo yeast nucleosome positioning data and used to predict the probability for nucleosome occupancy (positive values) or depletion (negative values) rather than exact positions. It showed a positive peak right in the middle of the upstream nucleosome, maybe suggesting an especially stable nucleosome here in vivo. In contrast, the DNA sequence underlying the downstream nucleosome was rather neutral, or even negative at its 3′ end, with regard to the propensity for nucleosome occupancy.
So far, we correlated, in this and our previous study (31), intrinsic nucleosome stability and the cofactor requirement. Next we wished to test directly if stability was causative for requirement. Extended stretches of poly(dA-dT) are known to be unfavorable for nucleosome formation in vivo and in vitro (4, 33, 57). So we replaced a stretch of 10 or 19 consecutive bases with adenine deoxynucleotides (plasmids pCB84a-10A and −19A, respectively) in the middle of the upstream nucleosome region (Fig. (Fig.7C).7C). As expected, such replacements led to increasingly more negative N-scores for the region that was occupied by the upstream nucleosome in the wt promoter (Fig. (Fig.7C7C).
We needed to check if the upstream nucleosome would still form in vivo on these mutated DNA templates. DNase I mapping confirmed the presence of the upstream nucleosome for both variants in the wt and snf2 backgrounds (Fig. (Fig.7D7D and data not shown). Restriction enzyme accessibility assays showed that there was no increase in HhaI site accessibility for the 10A replacement compared to the wt promoter (data not shown) but an increase for the 19A variant (from 25 to 40% in wt and from 15 to 48% in the snf2 background) was observed (Fig. (Fig.7D).7D). This suggested a destabilized upstream nucleosome for the 19A variant already under repressive conditions. There was also a subtle shift in positioning as the sHS region extended more upstream beyond the ApaI marker (compare Fig. Fig.7D7D and and3B).3B). This region of additional hypersensitivity at the 3′ border of the upstream nucleosome correlated with the region of the most negative N-score at about −550 (Fig. (Fig.7C7C).
The reduced stability of the 19A variant was directly assessed in our in vitro chromatin assembly assay (Fig. (Fig.7E).7E). First, the upstream nucleosome formed neither with a limiting (60%) nor with the full (100%) complement of histones during salt gradient dialysis, but the DNase I pattern in this region was similar to that of the free DNA digest. This speaks for the lower nucleosome positioning power of the mutated DNA sequence under these conditions. Second, the addition of yeast extract and energy induced a more in vivo-like chromatin structure in the fully assembled (100%) chromatin template, with accessibilities for the HhaI and TaqI sites that were very similar to the in vivo values (Fig. 7D and E; compare 19A in the wt background [D] and 100% with yeast extract and energy [E]). This confirmed again that the unidentified energy-dependent activity in the yeast extract constitutes conditions for more in vivo-like nucleosome positioning. Third, addition of yeast extract and energy to the underassembled (60%) chromatin templates increased not only the TaqI site accessibility (from 22 to 66%) (Fig. (Fig.7E),7E), similar as seen before for the wt promoter (from 15 to 69%) (Fig. (Fig.7B)7B) but now also the HhaI site accessibility (from 47 to 73%). This argued for a low stability of both the upstream and downstream nucleosome.
Finally, both variants showed remodeling of the upstream nucleosome upon induction in a snf2 strain. The extent of remodeling as judged by DNase I indirect end labeling was substantial for both variants in comparison to the internal control of the wt promoter at the chromosome locus (Fig. (Fig.8A)8A) and to the plasmid locus (data not shown). HhaI site accessibility assays confirmed a partial remodeling for the 10A variant and almost full remodeling for the 19A replacement variant (Fig. (Fig.8B).8B). Altogether, these results argue strongly that the intrinsic stability of the upstream nucleosome in the wt promoter caused its strict Snf2 requirement for remodeling.
In addition to the mechanistically interesting relationship between intrinsic stability and Snf2 dependency of remodeling of the upstream nucleosome, we asked further if the critical Snf2 dependency of remodeling the upstream nucleosome was the main cause for the Snf2 effect on overall PHO84 promoter induction kinetics (Fig. (Fig.4A).4A). If so, the kinetic delay in a snf2 background should be reduced if the upstream nucleosome is destabilized (19A variant, plasmid pCB84a-19A) or absent (ΔΔUASpAB variant, plasmid pCB84b). We followed induction kinetics for both variants in the wt and snf2 backgrounds by acid phosphatase assay and compared them to the kinetics of the wt promoter in both backgrounds (Fig. 8C and D). For both variants the delay of induction in the snf2 mutant compared to the wt background was somewhat diminished, more so in the case of the truncated promoter and only very slightly in the case of the mutated promoter. This was more apparent after normalization of the phosphatase activity in the snf2 strains to the respective activity in the wt background at the same time points (Fig. (Fig.8D).8D). Nonetheless, as the delay in the snf2 strains was still substantial in both cases, we reasoned that there was a significant Snf2 dependency of other parts of the PHO84 promoter besides the upstream nucleosome. For example, we showed specifically that the kinetics of remodeling the downstream nucleosome was dependent on Snf2, as histone eviction of the wt promoter was delayed in the snf2 mutant (Fig. (Fig.4E)4E) (see above).
Since the HhaI accessibility of the PHO84 promoter variant in pCB84a-19A was considerably increased under repressive conditions in a snf2 strain (Fig. (Fig.7D)7D) but did not result in a higher basal level of transcription (data not shown), it seemed again (see above) that Snf2 had an effect on basal transcription that was not necessarily linked to basal remodeling of the upstream nucleosome.
We and others found that the histone chaperone Asf1 is involved in the induction of the coregulated PHO5 and PHO8 promoters (1, 38). Recently, several groups reported the critical requirement of Asf1 for the activity of the histone acetyltransferase Rtt109, which acetylates histone H3 at lysine 56 (18, 21, 30, 64, 78). This finding raised the question of whether an involvement of Asf1 reflects its role solely as histone chaperone or rather a role of Rtt109. We checked this for induction of the PHO5 promoter and observed that the delay in induction was virtually the same in the asf1 and rtt109 mutants and that there was no further delay in an asf1 rtt109 double mutant (Fig. (Fig.9A).9A). This argued strongly that Asf1 and Rtt109 function together in the same pathway during PHO5 induction. We also noted that for both the asf1 mutant as well as the rtt109 mutant the basal PHO5 activity levels were slightly but significantly elevated.
In contrast, induction of PHO84 was significantly delayed only in the rtt109 but hardly at all in the asf1 mutant (Fig. (Fig.9B).9B). The induction delay in the rtt109 mutant was due to a delay on the level of chromatin remodeling as monitored by restriction enzyme accessibility and histone ChIP assays (Fig. (Fig.9C,9C, D, and E). However, the effects were much less severe than those in the snf2, gcn5, or ino80 mutants (compare to Fig. Fig.4E4E and and6C),6C), especially as they were rather limited to an early time of induction (45 min). There was hardly any effect on the level of restriction enzyme accessibilities for the asf1 mutant, and only at 45 min of induction was there a slight delay in histone eviction. This may constitute a weaker pendant to the effects in the gcn5 and snf2 strains, i.e., histone eviction being the rate-limiting step.
There was no differential Rtt109 requirement of the upstream and downstream nucleosome discernible, as the kinetics of restriction enzyme site accessibility were similarly delayed for the HhaI and the TaqI sites in the rtt109 mutant (Fig. (Fig.9C).9C). We also checked the effects of the asf1 and rtt109 deletions on induction of the truncated pCB84b construct and got similar results as with the full-length pCB84a plasmid (Fig. (Fig.9F),9F), speaking for a role of Rtt109 in remodeling of the downstream nucleosome but not excluding a role for remodeling of the upstream nucleosome as well.
The effects of the asf1 and rtt109 deletions on PHO5 and PHO84 induction showed some dependency on the strain background. In the BY4741 background, the rtt109 mutant showed a weaker delay for PHO5 induction than the asf1 mutant (data not shown). In the W303 background, the rtt109 mutant had a similar effect on PHO84 induction as in the BY4741 background, but here also the asf1 mutant had an appreciable effect, similar to that of the rtt109 mutant (data not shown).
It was shown that Rtt109 exists in a complex with another histone chaperone, Vps75 (78); however, the absence of Vps75 caused hardly any effect on PHO5 and PHO84 induction (data not shown).
In this study we present a characterization of PHO84 promoter regulation on the level of chromatin structure. The PHO84 promoter in its repressed state harbored an sHS region flanked by two positioned nucleosomes (upstream and downstream nucleosome) and a semiopen and less-organized chromatin structure close to the TATA box. This chromatin organization became extensively remodeled upon induction, leading to an extended hypersensitive region of about 500 bp and the eviction of histones.
At the outset of our study no data on the nucleosomal structure of the PHO84 promoter were available. However, during recent years several groups have undertaken genome-wide nucleosome positioning studies in yeast (45, 49, 67, 82, 85). Very recently, during the preparation of the manuscript, Lam et al. (41) mapped the promoter chromatin structures of PHO regulon genes by tiled PCR amplicons with mononucleosomal DNA as template. Their analysis of the PHO84 nucleosome organization agrees remarkably well with our mapping (Fig. 10A). Even the less-organized structure between the downstream nucleosome and the TATA box region was reflected by a reduced PCR product peak in this region. They also found the same extensive nucleosome-free region in the induced state.
In contrast to this congruence of two locus-specific nucleosome mapping studies using different methods, there is less agreement with the genome-wide approaches. The experiments of Lee et al. (45) did not reveal any nucleosomes in the extended PHO84 promoter region, Whitehouse et al. (82) mapped nucleosomes right in the cHS and sHS regions, and Mavrich et al. (49) correctly assigned the position of the upstream nucleosome and of the cHS and sHS regions but not of the downstream nucleosome. Both our own mapping and that of Lam et al. (41) employed medium with added phosphate to ensure complete repression, whereas the mentioned genome-wide studies used YPD medium, which can lead to a significant level of PHO84 transcription (23, 53). These differences in growth conditions could explain at least the lack of nucleosome detection.
We did the analogous comparison of nucleosome positioning data for the PHO5 and PHO8 promoter regions and found significant disparities as well, especially for the PHO8 promoter (Fig. 10B and C). These differences can be less well explained by differences in growth conditions, as both PHO5 and PHO8 are largely repressed in YPD medium (3, 5, 53). So, it seems that genome-wide nucleosome positioning data, even though they are extremely valuable for detecting genome-wide trends of nucleosomal organization, may need to be verified by locus-specific mapping techniques.
The need for experimental verification is also very important with regard to the prediction of nucleosome positions by DNA sequence-based algorithms. For example, the algorithms of Segal et al. (65) and Ioshikhes et al. (32) predicted the downstream nucleosome and the extended linker at the sHS region rather well (Fig. 10A). However, the upstream nucleosome was not met and the cHS region was missed. As mentioned above, the N-score algorithm of Yuan and Liu (84) accurately predicts low nucleosome occupancy for the cHS region and a peak of high nucleosome occupancy just at the center of the upstream nucleosome. This prediction agrees well with our data that showed a higher intrinsic stability of the upstream nucleosome than for the downstream nucleosome.
The PHO84 promoter appears like a hybrid between the PHO5 and PHO8 promoters with regard to the cofactor dependency for chromatin remodeling upon induction. On one hand, it has a similar degree of cofactor dependency as the PHO5 promoter, because the remodeling of the downstream nucleosome and the overall promoter activation were not essentially dependent on Snf2, Ino80, Gcn5, and Rtt109. It was not even abolished in the snf2 ino80 or snf2 gcn5 double mutants. Nonetheless, all these factors have a more or less important role in remodeling kinetics of the downstream nucleosome. Steger et al. (72) also reported a defect in PHO84 mRNA induction in snf6 (subunit of the SWI/SNF complex) and arp8 (subunit of the Ino80 complex) strains, which corresponds nicely to the promoter-opening delays reported here for the snf2 and ino80 mutants. On the other hand, remodeling of the upstream nucleosome was reminiscent of the PHO8 promoter, as it was strictly dependent on Snf2. In addition, it became critically dependent on Ino80 under submaximal induction conditions, while the downstream nucleosome was still remodeled, i.e., remodeling of the upstream nucleosome appeared to be more dependent on Ino80 than remodeling of the downstream nucleosome.
Therefore, the PHO84 promoter presents an example of a differential cofactor requirement for histone eviction from two neighboring nucleosomes at the same promoter. This differential cofactor requirement poses even more poignantly the question that was raised earlier after the observation of the differential cofactor requirements at the PHO5 and PHO8 promoters: what makes remodeling of one nucleosome strictly dependent on a certain cofactor, for example, Snf2, while remodeling of another nucleosome is not dependent on this cofactor? In order to answer this question, the two neighboring nucleosomes at the PHO84 promoter constitute a system that is very well internally controlled for the influence of any external factors, like cofactor recruitment, higher-order structure, or nuclear location.
One possible answer to the above question could relate to the presence of a functionally important intranucleosomal activator binding site in nucleosomes that show less cofactor dependency, like the UASpE site in the downstream nucleosome at the PHO84 promoter or the UASp2 site in the −2 nucleosome at the PHO5 promoter (26). However, we tested the UASpEmut, UASpCEmut, and UASpDEmut PHO84 promoter variants in the snf2 background under inducing conditions and saw the same sHS**-type region as for the wt PHO84 promoter in snf2 cells (unpublished results). Therefore, the presence of the intranucleosomal UASpE element did not influence the differential cofactor dependency for remodeling of the upstream and downstream nucleosome.
As an alternative explanation, Dhasarathy and Kladde (19) showed that the stringency of cofactor requirements for chromatin remodeling at the PHO5 promoter was dependent on the amount of Pho4 recruited to the promoter. We found this relationship also at the PHO84 promoter, as the upstream nucleosome became critically dependent on Ino80 if less Pho4 was recruited, i.e., under submaximal inducing conditions. However, this effect is unlikely to explain the promoter-internal difference in cofactor requirements at the PHO84 promoter under the same induction conditions. Here both the upstream and downstream nucleosome should be exposed simultaneously to the same degree of Pho4 recruitment, unless, for example, the higher-order structure makes a difference for the two nucleosomes. But this seems unlikely, as the differential Snf2 dependencies of both nucleosomes were equally observed at the plasmid and the chromosomal locus (Fig. (Fig.4B4B and our unpublished data), which probably differ in higher-order structures.
In this study we provide strong evidence for a hypothesis that we raised previously (31) as an answer to the above question: different intrinsic stabilities of positioned nucleosomes cause different cofactor requirements for their remodeling. We showed previously, using our yeast extract in vitro chromatin assembly system, that the nucleosomes at the PHO8 promoter were intrinsically more stable than those at the PHO5 promoter, thus providing a correlation of nucleosome stability and dependency on cofactors. By the same methodology we measured now a similar, although more subtle, trend while comparing the stabilities of the upstream and downstream nucleosome at the PHO84 promoter. The former was more stably positioned than the latter. This correlated well with the prediction by the N-score algorithm for the PHO84 promoter. Analogously, most of the PHO8 promoter region had a positive prediction for nucleosome occupancy and most of the PHO5 promoter region showed either mildly or strongly negative nucleosome propensity and the only positive peak was located in a linker region in vivo (Fig. 10B and C). This is in agreement with our earlier notion that the nucleosomes at the repressed PHO5 promoter adopt positions in a “loaded spring-like state” (31, 39). Altogether, it appears that nucleosomes that are positioned over DNA regions with more positive N-scores are more strictly dependent on chromatin cofactors for remodeling, and nucleosomes over less favorable DNA sequences according to the N-score can be remodeled by multiple redundant pathways.
We tested this directly for the case of the PHO84 promoter by introducing stretches of homopolymeric poly(dA) at the position of the upstream nucleosome. This progressively lowered the N-score for this region. Indeed, we confirmed in the in vitro assay that the upstream nucleosome was destabilized and observed in vivo that a progressively lower stability of the upstream nucleosome allowed progressively more remodeling of this nucleosome in the absence of Snf2. Importantly, our in vitro assay was an independent measure of nucleosome stability; therefore, we needed not to invoke Snf2 dependency itself as an indicator of stability. A similar approach was undertaken at the RNR3 promoter, where insertion of one or even two 34A stretches close to the TATA box prevented the formation of a positioned nucleosome and relieved the Snf2 dependency of RNR3 induction (86).
We conclude that promoter strength is not necessarily correlated with the degree of cofactor requirement for chromatin remodeling but rather that intrinsic properties of individual promoter nucleosomes determine the cofactor dependency for their remodeling.
We and others showed previously for the PHO5 and PHO8 promoters that chromatin remodeling led to the eviction of histones from the promoter region (1, 14, 38). Genome-wide studies confirmed that histone-depleted regions are a common property of promoters of active genes (13, 43). As discussed earlier (14, 24, 58, 59), there is a significant mechanistic difference if remodeling of nucleosomes leads to increased DNA accessibility while histones are still present or as histones are evicted. Importantly, this distinction cannot be made by techniques based on nuclease digestion, as DNA accessibility and therefore nuclease digestibility changes in both cases. Therefore, it is not necessarily to be expected that chromatin remodeling kinetics as followed by nucleases, e.g., restriction enzyme accessibility, and by histone ChIP will coincide. Even though such kinetic measurements were congruent so far for remodeling at the PHO5 and PHO8 promoters (6, 8), we now observed slower kinetics of histone eviction compared to restriction enzyme accessibility kinetics during induction of the PHO84 promoter in the gcn5 mutant and also specifically for remodeling of the downstream nucleosome in the snf2 mutant. This may argue for an initial phase of nucleosome remodeling leading to altered nucleosomal states that allow more restriction enzyme accessibility but still retain histones associated with DNA. This initial phase may precede the actual, rate-limiting histone eviction phase. For the gcn5 mutation this interpretation is concordant with reports on the stimulatory effect of histone acetylation on histone eviction (17).
The mechanism of histone eviction raises the question of the histone acceptor. We and others suggested in the past that histone chaperones may be the most promising candidates as histone acceptors and showed a role for Asf1 in increasing the rate of opening of the PHO5 and PHO8 promoters (1, 38). However, the recognition of Asf1 as an essential cofactor for the activity of the histone H3 lysine 56-specific histone acetyltransferase Rtt109 (18, 21, 22, 30, 64, 78) raised the alternative possibility that Asf1 functions through the H3 K56ac modification rather than solely as histone acceptor. Indeed, the PHO5 induction kinetics was equally delayed in asf1 and rtt109 strains, and the asf1 rtt109 double mutant showed no synthetic effect. Very recently, just before submitting the manuscript, equivalent result were published by Williams et al. (83). So, Asf1 appears to function in histone eviction at the PHO5 promoter mainly through H3 K56ac, and it is currently unclear if it also serves directly as a histone acceptor.
Surprisingly, in the BY4741 strain background Asf1 seemed to be hardly involved at all in PHO84 induction despite the considerable role for Rtt109. This suggested that Rtt109 may have other targets than H3 K56. This is not unlikely, as Rtt109 exists in a complex with another histone chaperone, Vps75, that seems to be less important for acetylation of H3 K56 in vivo (12, 30, 78). The absence of Vps75 caused only a slight effect on PHO5 induction, much weaker than that observed in the absence of Asf1, and had no significant effect on PHO84 induction (unpublished data). Therefore, Rtt109 could function in PHO84 induction through a so-far-unidentified target that may be acetylated by Rtt109 independently of both Asf1 and Vps75.
All mutants used in this study (besides rtt109) were controlled for causing direct effects on the coregulated PHO5 and PHO8 promoters rather than causing side effects on PHO regulon induction (6, 8, 38). In addition, we observed decreased chromatin remodeling in the snf2K798A and the ino80 mutants under steady-state conditions (overexpression of PHO4 in +Pi medium), under which effects on growth rate should not matter, which otherwise is a concern for effects on PHO induction (6, 38). Other groups have shown a direct role for Snf2, Ino80, and Gcn5 at the PHO84 promoter in ChIP assays (23, 36, 68, 69, 72).
Even though the stable upstream nucleosome poses a very interesting case for the mechanistic study of nucleosome remodeling, it seems to have a rather minor role in the overall regulation of the PHO84 promoter. Given its higher stability and occlusion of the UASpB site, it might play a repressive or fine-tuning role for PHO84 regulation. However, its absence in the pCB84b construct only had a very slight effect on the promoter induction kinetics and on their Snf2 dependence. Further, full final promoter activity was achieved in the snf2 mutant without remodeling of the upstream nucleosome. Finally, the destabilization of the upstream nucleosome in the 19A variant did relieve the Snf2 dependency for remodeling of the upstream nucleosome but had no effect on the basal level of transcription and only mild effects on the promoter induction kinetics. On the other hand, full PHO84 promoter activity was always concomitant with complete remodeling of the downstream nucleosome and every delay in induction kinetics went together with a delay in its remodeling. As its intranucleosomal UASpE site was especially important for PHO84 induction, it seems that controlling the accessibility to UASpE through remodeling of the downstream nucleosome is key to regulating PHO84 promoter induction.
We are grateful to P. D. Kaufman, C. L. Peterson, X. Shen, A. Verreault, and F. Winston for the gifts of yeast strains and to A. Schmid and to V. Fajdetic for technical assistance.
This work was supported by the German Research Community (Transregio 05), the 6th Framework Programme of the European Union (Network of Excellence The Epigenome), and the Ministry of Education, Science, and Technology of the Republic of Croatia, grant 058-0580477-0247 (to S.B.).
This paper is dedicated by P.K. to his wife Marion Rouette, who became indirectly involved enough during the preparation of the manuscript to earn her an honorary coauthorship.
Published ahead of print on 23 March 2009.