|Home | About | Journals | Submit | Contact Us | Français|
The chromatin architecture of eukaryotic gene promoters is generally characterized by a nucleosome-free region (NFR) flanked by at least one H2A.Z variant nucleosome. Computational predictions of nucleosome positions based on thermodynamic properties of DNA-histone interactions have met with limited success. Here we show that the action of the essential RSC remodeling complex in S. cerevisiae helps explain the discrepancy between theory and experiment. In RSC-depleted cells, NFRs shrink such that the average positions of flanking nucleosomes move toward predicted sites. Nucleosome positioning at distinct subsets of promoters additionally requires the essential Myb family proteins Abf1 and Reb1, whose binding sites are enriched in NFRs. In contrast, H2A.Z deposition is dispensable for nucleosome positioning. By regulating H2A.Z deposition using a steroid-inducible protein splicing strategy, we show that NFR establishment is necessary for H2A.Z deposition. These studies suggest an ordered pathway for the assembly of promoter chromatin architecture.
Since the identification of a nucleosome-free region (NFR) in SV40 minichromosomes nearly 30 years ago (Jakobovits et al., 1980; Saragosti et al., 1980), the mechanisms underlying the positioning of nucleosomes have been an area of active study. Recent genome-scale surveys of nucleosome positions in a variety of eukaryotic organisms have revealed a stereotypical promoter chromatin architecture characterized by a nucleosome-free region (NFR) flanked by at least one nucleosome enriched for the histone H2A variant H2A.Z (Albert et al., 2007; Lee et al., 2007; Mavrich et al., 2008a; Mavrich et al., 2008b; Ozsolak et al., 2007; Raisner et al., 2005; Yuan et al., 2005). As a class, NFR-adjacent nucleosomes are the most precisely positioned in the genome, with neighboring nucleosomes displaying less precision in their locations as their distance from NFRs increases. By acting as anchor points, the tight positioning of NFR-flanking nucleosomes may be a dominant mechanism by which nucleosomes are positioned genome-wide (Mavrich et al., 2008a). In S. cerevisiae, NFR-flanking nucleosomes often occlude the transcription start site (TSS) such that the TSS is on average half a helical turn inside the +1 nucleosome and exhibits a rotational phasing which tends to place sites for transcription factors on the accessible surface of nucleosomal DNA (Albert et al., 2007). Significantly, recent detailed analysis of the PHO regulon in S. cerevisiae has shown that chromatin remodeling during phosphate starvation exposes a class of binding sites for the Pho4 activator that are initially masked by nucleosomes and that this plays a key role in shaping the input-output functions of promoters (Lam et al., 2008). Defects in the positioning of some promoter nucleosomes seen in isw2 mutant cells correlates with the accumulation of cryptic antisense transcripts, leading to the proposal that positioning of nucleosomes also prevents erroneous transcription initiation events (Whitehouse et al., 2007). While the fractional occupancy of H2A.Z in NFR-flanking nucleosomes in yeast is not correlated with transcription rates (Raisner et al., 2005), loss of promoter nucleosomes including those containing H2A.Z occurs in response to transcriptional activation (Bernstein et al., 2004; Schones et al., 2008; Shivaswamy et al., 2008; Zanton and Pugh, 2006). It has been reported that H2A.Z nucleosomes are less stable in vitro, and this property has been hypothesized to aid in their removal in vivo (Zhang et al., 2005). In Drosophila and humans, NFRs flanked by nucleosomes enriched in H2A.Z are also a common feature of promoters (Barski et al., 2007; Mavrich et al., 2008b). In flies, the H2A.Z nucleosomes at promoters tend to occur downstream of the NFR, whereas in humans there appear to be H2A.Z nucleosomes both upstream and downstream of NFRs. Interestingly, both NFR formation and H2A.Z deposition seem to correlate with productive transcription in these organisms. These species-specific differences suggest additional complexity in metazoans. H2A.Z nucleosomes are also relatively enriched at flanking non-promoter NFRs that characterize enhancers and insulators in human T-cells (Barski et al., 2007). Taken together, nucleosome-free regions, whether or not associated with gene promoters, tend to be associated with H2A.Z. A conserved function of H2A.Z demonstrated in both S. cerevisiae and Arabidopsis thaliana is to act in euchromatin to antagonize gene silencing (Meneghini et al., 2003; Zilberman et al., 2008).
Despite the high conservation across eukaryotic evolution of these basic aspects of promoter chromatin architecture identified by descriptive genomic studies, the mechanisms by which NFRs flanked by H2A.Z nucleosomes form remain poorly understood. There exists evidence that octamer positioning genome-wide is mediated by a genomic nucleosome positioning code in which intrinsic DNA-octamer affinities, predicted computationally based on dinucleotide periodicity patterns and/or other sequence patterns, are a significant determinant of location, particularly at NFR-flanking nucleosomes (Ioshikhes et al., 2006; Lee et al., 2007; Segal et al., 2006). For example, one study (Segal et al., 2006) reported that 50% of nucleosome positions in S. cerevisiae chromosome III can be accurately predicted computationally. However, there are differences of opinion in the literature regarding how well computational methods predict actual positions determined experimentally compared to so-called random guess predictions (Peckham et al., 2007; Segal, 2008; Yuan and Liu, 2008). A recent study (Yuan and Liu, 2008) compared a number of methods and found that for S. cerevisiae datasets even improved methods required an error of ~70bp (nearly half a nucleosome) to obtain a prediction sensitivity of 80% and required a similar error to yield an specificity of 80%. These errors stand in contrast to the observed precision of nucleosome positioning in vivo relative to TSSs and transcription factor binding sites as described above.
The connection between NFR formation and H2A.Z deposition is likewise not well-defined. One report suggested that H2A.Z deposition plays a role in nucleosome positioning, while another proposed that H2A.Z deposition has no role (Guillemette et al., 2005; Li et al., 2005). H2A.Z nucleosomes have been reported to be poor in vitro substrates of chromatin remodeling enzymes compared to their H2A counterparts (Li et al., 2005). Thus, whether NFR formation is required for H2A.Z deposition or vice versa is unknown.
In the absence of a consensus view of how promoter chromatin architecture is specified with precision, we sought to clarify the underlying mechanisms. In previous work, we identified a segment of the SNT1 promoter required for normal levels of H2A.Z deposition. Remarkably, insertion of a short segment of this region into the middle of a transcriptionally quiescent PRM1 gene resulted in the formation of an NFR flanked by two nucleosomes containing H2A.Z (Raisner et al., 2005). This sequence contained a putative binding site for the Myb family transcription factor Reb1 and an adjacent T tract. Below we describe further studies of this synthetic NFR as well as chromosome-wide studies of the roles of several essential factors in nucleosome positioning and H2A.Z deposition.
We considered three models by which the DNA signal containing the Reb1 binding motif (henceforth called Reb1:dT7) might program promoter chromatin structure (Fig. 1A, Fig. 1B). In Model I, the DNA signal first programs NFR formation, and then the NFR acts as a signal to induce H2A.Z deposition into the flanking nucleosomes. Model II proposes the reverse process of NFR formation such that a DNA signal first induces H2A.Z deposition, and H2A.Z then acts as a signal for NFR formation. Lastly, NFR formation and H2A.Z deposition could occur in an independent, uncoupled fashion (Model III). To distinguish these models, we first sought to define trans-acting factors that mediate NFR formation.
The mechanism by which Reb1:dT7 induces formation of an NFR flanked by two nucleosomes carrying H2A.Z likely involves the recruitment of Reb1, although this was not tested directly in our previous study. We also hypothesized that Reb1 might recruit a chromatin remodeling enzyme to produce an NFR. A systematic study of protein interactions revealed that Reb1 physically associates with Rsc2, Rsc3 and Npl6, which are subunits of the essential chromatin remodeling complex RSC (Cairns et al., 1996; Cairns et al., 1999; Gavin et al., 2002). Since the Reb1-related factor, Abf1, has also been implicated in the formation of a nuclease-sensitive site (De Winde et al., 1993), we pursued its functional role.
As Reb1, Abf1, and the catalytic subunit of RSC (Sth1) are all essential proteins, we generated a series of conditional alleles Specifically, we used the temperature-sensitive degron system to engineer yeast strains in which we could control degradation of these proteins via the N-end rule pathway (Dohmen and Varshavsky, 2005). This pathway operates through the recognition of destabilizing N-terminal amino acids by the nonessential E3 ubiquitin ligase Ubr1. A N-terminal arginine is the strongest signal for degradation by Ubr1, and traditional degron alleles encode proteins capped by an arginine followed by a temperature-sensitive murine dihydrofolate reductase (DHFRts) peptide fused to the target protein. As translation initiates at an ATG codon, an ATG-initiated segment encoding a ubiquitin moiety is placed before the arginine codon. Once synthesized, this segment is cleaved in cells by ubiquitin C-terminal proteases to expose the N-terminal arginine. Such degron systems also place the modified allele of interest under the control of a regulatable promoter, which traditionally has been the copper-inducible promoter pCUP1.
We constructed pCUP1UBI4DHFRtsc-mycSTH1 and pCUP1UBI4DHFRtsc-mycREB1 alleles in strains that had UBR1 under the control of the pGAL1 promoter. We were unable to achieve substantial degradation of these proteins or growth arrest under degron-inducing conditions, although such an allele for STH1 has been reported (Parnell et al., 2008). We achieved more complete degradation of Reb1 and Sth1 under degron-inducing conditions with a different construct (pMET3UBI4DHFRts3xHA) that utilized the methionine-repressed pMET3 promoter (Fig. 1C). Yeast strains carrying Reb1-degron or Sth1-degron alleles were inviable under degron-inducing conditions (Fig. 1D). While an Abf1 DHFRts degron has been reported in the W303 strain background (Reed et al., 1999), we were unable to construct a viable pMET3UBI4DHFRts3xHAABF1 degron allele in our S288C background despite several attempts and strategies. We, however, successfully created a pMET3UBI4abf1(M1R) allele in a strain carrying a pGAL1UBR1 allele. As described above, the ubiquitin moiety of the translated protein is cleaved off soon after translation, leaving an Abf1 protein capped with a destabilizing N-terminal arginine. Under inducing conditions, this Abf1-degron yielded growth arrest and displayed Abf1 depletion (Fig. 1C, 1D). We also constructed an Abf1 Reb1 “double degron” strain in order to simultaneously deplete both factors from cells (Fig. 1C, 1D). For the studies described below, we determined nucleosome positions by hybridizing mononucleosomal versus genomic DNA-derived probe with in-house printed custom tiling arrays that span S.cerevisiae chromosome III at 20bp resolution (Yuan et al., 2005). The arrays also included oligonucleotides that tiled sequences corresponding to the PRM1 gene to allow us to observe nucleosome positions programmed by Reb1:dT7 (see Experimental Procedures for details). Unless otherwise specified, all experiments represent averages of four independent, biological replicates.
As expected from our previous study (Raisner, et al., 2005), nucleosome position analysis of the PRM1 ORF with or without a Reb1:dT7 insertion revealed that the insertion produces an NFR (Fig. 2A). NFR formation was unaffected in strains carrying the Reb1-degron or Sth1-degron alleles under conditions in which the degron system was inactive (Fig. 2B). We mapped nucleosome positions in strains carrying the Reb1-degron or the Sth1-degron five hours after activating the degron system, and these positions were compared to nucleosome positioning data from a control strain isogenic to the degron strains except that it lacked the Reb1 or Sth1 degron allele. The latter control strains were subjected to the identical culture growth protocol as the experimental strains. Depletion of Reb1 resulted in loss of the NFR programmed by Reb1:dT7 inserted into the PRM1 ORF, consistent with a direct role for Reb1 (Fig. 2B). Likewise, inactivation of the RSC complex through depletion of Sth1 resulted in complete loss of the NFR programmed by the Reb1:dT7 sequence (Fig. 2B), supporting the hypothesis that Reb1 functions by recruiting RSC.
Model II above proposes that NFR formation requires the prior deposition of H2A.Z into chromatin, which is mediated by the Swr1 chromatin remodeling complex. We tested whether H2A.Z or Swr1 is required for NFR formation induced by the Reb1:dT7 signal. We constructed htz1Δ and swr1Δ strains and mapped the nucleosome positions in these strains. NFR formation meditated by insertion of Reb1:dT7 at PRM1 did not require H2A.Z or Swr1 (Fig. 2C).
To test whether NFR formation at PRM1 might be a consequence of transcription, we examined whether NFR formation induced by Reb1:dT7 was associated with the production of transcripts. Transcript levels were examined using RT-PCR analysis of extracted total RNA. In a control strain in which the PRM1 gene was induced with mating pheromone for 1hr, a specific signal was obtained (Fig. S10A). We examined strains containing the insert and the Sth1 degron under both degron-inducing and permissive conditions as well as a strain lacking the insert. In these three cases, multiple peaks rather than a single peak were obtained in the QPCR melting curves indicating that cross-reacting cDNAs, instead of products specific to the PRM1 locus, were being detected. Moreover, no quantitative differences were observed (Fig. S10A). Taken together, these data suggest that NFR formation was not associated with transcription that could be detected by these methods.
We next examined the chromosome-wide requirement for Reb1 in NFR formation. We mapped and compared nucleosome positions in the Reb1-degron strain and an isogenic control strain that lacked the Reb1-degron under conditions described above. Fig. 3A shows a gene-by-gene “difference map” of the positioning data in which genes were aligned to each other based on the position +1 nucleosome downstream of the NFR in control strains and then the mutant signal subtracted from the control signal (yellow indicates more nucleosomal DNA signal in the mutant than in wild-type cells). The data were organized by K-means clustering. Two clusters are evident, one in which positioning was affected (Cluster I affecting 12% of assayed promoters), and another where no effect was evident (Cluster II). Line traces of the average signals of the control and degron strains for these two clusters are shown in Fig. 3B. Inspection of Cluster I indicates that the two NFR-flanking nucleosomes move inward towards the center of the NFR, and this movement propagates further such that other flanking nucleosomes also shift their positions (Fig 3B) Figure 3C shows that the changes in the size of the trough signal representing the NFR between degron and control strains were dependent on the induction of the degron.
A previous study of transcription factor association at promoters assigned likelihood scores for a given transcription factor binding to a given promoter (Harbison et al., 2004). We compared these scores for Reb1 to the highest fold-change probe in the NFRs of promoters we assayed that were assigned scores. As shown in Fig. 3D, there was a significant correlation (p=7.81×10-7 based on a hypergeometric test using a likelihood score cutoff of p<0.05). Consistent with this finding, promoters that contained at least one copy of the most conserved Reb1 binding site (TTACCCG, (Liaw and Brandl, 1994)) tended to experience changes in NFR structure (Fig. 3E). Cluster I was enriched for the most conserved Reb1 consensus site; indeed, 14 out of the 18 promoters in Cluster I contained this motif (p= 5.07×10-13, Fig. 3F). Relaxing the consensus to reflect the poorer conservation of the first two residues of the consensus still yielded highly significant enrichments (Fig. 3F).
We next performed the same analysis on Abf1 degron strain. Difference map analysis and clustering (Fig. 4A) show that 9.3% of promoters were affected, and these promoters were distinct from promoters affected by Reb1 depletion (see below). As with the Reb1 degron, the affected cluster displays a smaller NFR and movement of flanking nucleosomes towards the NFR (Fig. 4B), and these changes were dependent on induction of the degron (Fig. 4C). Likewise, affected NFRs were enriched for Abf1 binding (Fig. 4D) and for an Abf1 consensus site (Fig. 4E, F). The latter correlations are weaker than for the Reb1 site, perhaps because of the higher degeneracy of the Abf1 consensus site (Beinoraviciute-Kellner et al., 2005).
We next examined the effects of Sth1 depletion on nucleosome positioning using strains carrying the Sth1-degron as described above. Strikingly, our analysis showed that Sth1 depletion affected a majority (55%) of promoters (see Cluster I in Fig. 5A). The affected cluster displayed shrinking of the NFR and movement of flanking nucleosomes, whereas little change in nucleosome position was apparent for members of Cluster II (Fig. 5B). As with the Reb1 and Abf1 degron strains, growth under degron-inducing conditions was required to observe these differences (Fig. 5C). Fig. S10B shows a superposition of a histogram of the locations of mapped TSSs (Nagalakshmi, et al., 2008) and the positioning data. Consistent with previous studies, TSSs tend to lie just inside the downstream nucleosome and the movement observed in Sth1-depeleted cells moves these sites further into the nucleosome core (Fig. S10B). This may explain why RSC depletion has been reported to cause cessation of transcription by RNA polymerase II (Parnell et al., 2008).
When RSC, Abf1, or Reb1 were depleted, NFRs shrank but were not eliminated. We hypothesized that intrinsic positioning sequences might explain the positions of nucleosomes under these conditions. Therefore, we compared the positions we observed in Cluster I of Sth1-depleted cells with those predicted by Pugh and colleagues (Ioshikhes et al., 2006) based on AA/TT dinucleotide periodicity enrichment. As shown in Figure 5E, the average nucleosome position of the +1 and -1 nuclesomes of Cluster I relax towards positions specified by the NPS signature. For the largely unaffected cluster (Cluster II), a discrepancy between the NPS-predicted and observed positions is still apparent for the +1 nucleosome, whereas the -1 nucleosome is poorly aligned in this despite the sharp NPS prediction peak (Fig. 5E).
To test whether Sth1 depletion results in changes in gene expression, we performed expression profiling of the Sth1 degron strain against a control strain under degron-inducing conditions. As we expected global changes in gene expression, we incorporated external spiked-in RNA controls into our normalization procedure (see Experimental Procedures). We then asked whether there was an enrichment for genes whose expression was reduced in Sth1-depleted cells in Cluster I vs. Cluster II. As shown in Figure 5D, Cluster I is indeed highly enriched for genes whose expression requires Sth1, indicating that that the changes in positioning correlate with changes in expression. Given this result, we tested whether loss of transcription might be responsible for changes in positioning. We mapped nucleosome positions in a temperature-sensitive RNA polymerase II strain (rpb1-1) which ceases transcription within minutes upon shift into restrictive conditions (Nonet et al., 1987). The rpb1-1 mutant was grown at either 25°C or shifted for 1 hour to 37°C. The average nucleosome positions in these conditions were determined for Clusters I and II of the Sth1 degron difference map. There were were no detectable differences in nucleosome positions in either cluster (Fig. 5F, Fig. S10C). Hence, the changes in NFR structure observed upon Sth1 depletion appear to be due to the action of RSC rather than from cessation of transcription per se.
To identify promoters that are redundantly controlled by Abf1 and Reb1, we examined a double degron strain that carried the Abf1-degron and Reb1-degron alleles (Fig.1C, 1D). The resulting difference map was clustered together with difference maps for the Abf1 single degron strain, the Reb1 single degron strain, and the Sth1 degron strain (Fig. 6A; Fig. 6B shows that NFR changes in the double degron are dependent on degron induction). K-means clustering revealed that the promoter NFRs affected by loss of either Abf1 or Reb1 were reproducibly affected in the double degron strain, but there were no other promoter NFRs that were significantly affected in the double degron strain. It is also evident that most NFRs affected by loss of Reb1 also required Sth1 for proper positioning of nucleosomes (Fig. 6A), consistent with the data obtained with the synthetic NFR described above (Fig. 2B). The NFRs affected by loss of Abf1 appeared to have a somewhat lesser degree of dependence for Sth1 for nucleosome positioning (Fig. 6A). Likewise, there clearly are many NFRs that require RSC, but not Abf1 or Reb1, for proper nucleosome positioning (Fig. 6A), suggesting the existence of additional RSC recruitment mechanisms. Analysis of average nucleosome positions for each cluster indicates that the the changes observed (Fig 6A) are due to shifts in nucleosome positions (Fig. S14).
As with the Sth1 degron strain, we examined the Abf1-Reb1-double-degron strain for changes in transcript levels using whole genome microarrays and spiked-in external controls. As shown in Figure 6C, we found a significant correlation between decreases in NFR size and decreases in transcript accumulation.
To complete our analysis of positioning, we used cells lacking H2A.Z (htz1Δ) or lacking the ATPase subunit of its deposition complex (swr1Δ) to determine whether H2A.Z exchange was required for nucleosome positioning chromosome-wide (Fig. 7). Based on the results with the synthetic NFR (Fig. 2C), we expected to see no differences in positioning. Indeed, as shown by line traces of the average positions of aligned promoter nucleosomes, H2A.Z deposition resulted in no detectable changes. Of course, we cannot rule out the possibility that there could be changes too subtle to observe using our 20bp resolution tiling arrays.
Nonetheless, these data argue against Model II (Fig. 1B), which proposed that H2A.Z deposition is essential for NFR formation. We then considered the two remaining models: (1) the deposition of H2A.Z at a promoter requires the presence of an NFR at that promoter (Model II), or (2) H2A.Z deposition occurs independently of NFR formation (Model III). In principle, these models could be distinguished through development of a system in which NFR loss is induced under conditions where H2A.Z is not deposited into chromatin, but then H2A.Z is induced and its deposition examined. The Sth1 degron provides a tool to trigger abrogation of the synthetic NFR programmed by Reb1:dT7 and a shrinkage of bona fide promoter NFRs. However, since global transcription is shut off in RSC-depleted cells, we sought a posttranslational method to control H2A.Z deposition.
We utilized an engineered M. tuberculosis RecA intein whose intrinsic protein splicing is controlled by the human estrogen receptor ligand binding domain (Buskirk et al., 2004). This construct was used previously to interrupt several coding sequences in yeast, and its splicing was shown to be activated in vivo using the estrogen agonist 4-hydroxytamoxifen (4-HT). The chemistry of splicing requires cysteine cleavage sites and leaves a single cysteine residue at the splice junction.
We initially targeted the Swr1 and Swc2 subunits of the Swr1 complex by inserting the intein construct before the codons for several native cysteine residues in the corresponding genes. These alleles abrogated H2A.Z deposition in vivo, but addition of 4-HT did not restore H2A.Z deposition (unpublished observations), suggesting that the placement of the intein was incompatible with protein stability and/or intein splicing. We next attempted engineering a spliceable HTZ1 allele under the assumption that the smaller size of H2A.Z relative to the intein construct would make the protein context less likely to interfere in proper structural formation of the intein. H2A.Z, however, lacks cysteine residues, so such a spliced allele would by necessity contain a cysteine point mutation. Four H2A.Z residues (Ala46, Thr68, Thr88 and Asp100) that did not confer a significant growth defect in high precision measurements when mutated (S. Braun, D. Breslow, J. Weissman, H. D. M., unpublished observations) were replaced with the intein construct. These alleles initially replaced wild-type HTZ1 at its native chromosomal locus, but none displayed detectable spliced product in the presence of 4-HT (unpublished observations). We therefore placed these alleles under the control of a pGAL1 promoter on a high-copy 2 micron plasmid vector. The allele in which Ala46 was replaced with the intein construct yielded a protein that was spliced in vivo when cultures were treated with 4-HT; Ala46 is a residue in the core histone fold domain (Fig. 8A, 8B). As splicing produced somewhat higher levels of H2A.Z than in that found in wild-type cells (Fig. 8B), we placed the construct on a low-copy CEN-ARS plasmid. Regulated splicing of the H2A.Z intein was also observed (Fig. S13A), and this construct was used in further experiments. We refer to this pGAL1htz1(A46intein) allele on a CEN-ARS plasmid as the H2A.Z intein construct.
We examined the deposition of H2A.Z whose synthesis was directed by this construct using chromatin immunoprecipitation (ChIP). Although splicing in the H2A.Z intein is regulated, a small amount of H2A.Z deposition was observed in the absence of 4-HT, presumably due to low levels of background splicing (Fig. S13C); however, the H2A.Z enrichment signal steadily increases over time in response to 4-HT treatment (Fig S13D), indicating stimulation of H2A.Z deposition by the activation of splicing.
We introduced the H2A.Z intein into htz1Δ strains that carried pGAL1UBR1 and either the Sth1-degron or a wild-type Sth1. Simultaneous activation of the degron system and synthesis of unspliced H2A.Z were accomplished by transferring cells to 37°C media containing galactose. After 5 hours of growth, intein splicing was initiated by the addition of 4-HT, and cells were collected after 3 hours of further incubation at 37°C to allow for H2A.Z deposition. Chromatin immunoprecipitation for H2A.Z and histone H3 were carried out, and quantitative PCR was used to measure H2A.Z enrichment relative to H3 enrichment. The H2A.Z/H3 enrichment values were normalized to an amplicon in the middle of the large BUD3 ORF where there is little detectable H2A.Z (Raisner et al., 2005). Nucleosome positions were also mapped for the Sth1-degron H2A.Z intein strain prior to degradation of Sth1, after 5 hours of Sth1 depletion, and 3 hours after addition of 4-HT. We determined that unspliced H2A.Z was being produced and was spliceable, and we found that Sth1 degradation still occurred in the presence of the H2A.Z intein construct and during 4-HT treatment (Fig S13A, B).
We sought to examine the deposition profiles of H2A.Z at NFRs whose structure is unaffected upon Sth1 depletion and at NFRs that undergo significant changes upon Sth1 depletion. The two NFRs located within an intergenic region containing the DCC1 and BUD3 promoters do not appear to require Sth1 for their organization (Fig. 8C, top panel). The H2A.Z deposition profiles across the DCC1-BUD3 intergenic region in both the Sth1-degron and control strains were similar (Fig. 8C bottom) and indicated that H2A.Z deposition could still occur under these conditions. We next examined how loss of the NFR programmed by insertion of Reb1:dT7 into PRM1 affected the recruitment of H2A.Z. This NFR essentially collapses upon Sth1 depletion in the H2A.Z intein strain carrying the Sth1-degron (Fig. 8D, top panel). In the strain that did not have the Sth1-degron and therefore maintained the NFR programmed by Reb1: dT7 inserted into PRM1, H2A.Z deposition occurred in the middle of PRM1, with its peak deposition at the Reb1: dT7 insertion site (Fig. 8D, bottom left panel). In contrast, upon Sth1 depletion, there was no H2A.Z deposition focus about the Reb1: dT7 insertion site (Fig. 8D, bottom right panel). The apparently undirected, background H2A.Z deposition in the PRM1 ORF is similar to that observed in cells lacking the Reb1:dT7 insertion (Raisner et al., 2005), and similar global patterns of untargeted H2A.Z deposition have been seen in genome-wide studies (Albert et al., 2007). Thus, the focused peak of H2A.Z deposition induced by the Reb1:dT7 DNA signal appears to require the Sth1-dependent formation of an NFR directed by the signal.
RSC depletion did not produce complete collapse of NFRs on endogenous promoters, and, as described above, this may be due to intrinsic positioning signals. Nonetheless, we examined the H2A.Z deposition profile at the promoters of YCR016W and YCR023C, both of which experience significant nucleosome encroachment into their NFRs upon Sth1 depletion (Fig. 8E and 8F, top panels). We observed H2A.Z deposition at these promoters under conditions in which their NFRs were unaffected as well as under conditions where their NFRs were affected (Fig. 8E and 8F, bottom panels). However, the H2A.Z deposition profile at affected NFRs differed in that there was a significant decrease in H2A.Z enrichment in the vicinity of the +1 nucleosome relative to the NFR (see amplicon “D” for Fig. 8E and amplicon “C” for Fig. 8F). Whether the otherwise fairly robust H2A.Z deposition seen at these two promoters under conditions of intein induction is explained by the presence of a residual NFR driven by NPSs or by NFR-independent mechanisms that stimulate H2A.Z deposition such as histone acetylation and its subsequent recognition by Bdf1 (Raisner et al., 2005) is not clear. The latter model is difficult to test since cells lacking H2A.Z and Bdf1 are inviable (Raisner et al., 2005).
Based on the results of a number of genome-scale studies, it has become increasingly clear in organisms as diverse as yeast and humans that gene regulatory regions display stereotypical patterns of nucleosome positioning and identity. Although there are species-specific differences, promoters are generally characterized by an NFR flanked by at least one H2A.Z nucleosome. Despite the power of these descriptive genome-wide studies as well as work that indicates that these characteristics of promoters play key roles in gene regulation (see Introduction for references), they leave open the question of how these structures are programmed.
Two lines of studies have come to distinct conclusions regarding NFR formation mechanisms. One group of studies has suggested that the direct effects of sequence on DNA-octamer affinity programs NFR formation (see Introduction for references). In contrast, our previous work defined a short signal from the SNT1 gene containing a putative site for a DNA binding protein, Reb1, that is sufficient to program a NFR flanked by H2A.Z nucleosomes when placed into the middle of a positioned nucleosome in an inactive gene (Raisner et al., 2005). Others have also implicated Reb1 and Abf1 in the formation of nucleosome gaps within the specific promoter regions (Angermayr et al., 2003; De Winde et al., 1993). The work described here helps reconcile these two lines of research and provides insight into the relationship between NFRs and H2A.Z deposition. Our principal conclusions are as follows:
A striking result presented here is that at a majority of promoters, the normal positioning of NFR-flanking nucleosomes requires the essential multisubunit chromatin modeling complex RSC. Such a central role for RSC in generating promoter chromatin architecture is consistent with several of its properties: 1) RSC, unlike most chromatin remodeling enzymes in yeast, is essential for viability (Cairns et al., 1996; Cairns et al., 1999), 2) RSC slides nucleosomes in vitro (Lorch et al., 2001), and 3) RSC is required globally for RNA polymerase II transcription (Parnell et al., 2008). Our studies are also consistent with a recent lower-resolution study that concluded that RSC affected histone density at a number of promoters (Parnell et al., 2008). A recent study indicated changes in the positioning nucleosomes at ~12% of promoters in cells lacking the Isw2 chromatin remodeling complex (Whitehouse et al., 2007). The primary function of Isw2 appears to be in transcriptional repression and in suppressing antisense transcription (Whitehouse et al., 2007). Interestingly, in contrast to RSC, Isw2 appears to move nucleosomes in vivo toward the NFR, raising the possibility that it antagonizes the action of RSC at some promoters. The potential for dynamic involvement of multiple ATPases at promoters further underscores the active nature of mechanisms that position nucleosomes in vivo.
The finding in this study and in the previous study that the final resting positions of nucleosomes are strongly influenced by ATP-dependent chromatin remodeling mechanisms argues that that the intrinsic affinity of the octamer for underlying DNA sequences is not determinative for the final positioned state. However, our observation that depletion of Sth1 causes nucleosome positions to relax on average closer to those predicted by an NPS signature strongly suggests that sequence properties play a role in a stepwise mechanism for NFR formation. That is, NPS-mediated positioning exposes binding sites for factors such as Reb1 and Abf1, which in turn induce the action of RSC to move nucleosomes to their steady-state average positions in wild-type cells. Such a model is also consistent with in vitro and in vivo observations that suggest that the Isw2 remodeling enzyme moves nucleosomes into energetically unfavorable sites (Whitehouse and Tsukiyama, 2006). We speculate that, compared to a purely “hard-wired” system, this more dynamic, ATP-dependent mechanism may facilitate binding of DNA binding proteins to nucleosomal sites and transcription initiation. It is important to note that NPS predictions vary in their accuracy considerably at the level of individual genes, suggesting they likely do not predict with full accuracy the intrinsic thermodynamics of octamer-DNA interactions. A histogram of predictions (Ioshikhes et al., 2006) reveals that NPS-predicted positions for individual genes deviate significantly from experimental positions even in the Sth1 degron strain (Fig S15). Nonetheless, the close correspondence of the average profiles supports the two-step model proposed above.
Using a signal for NFR formation/H2A.Z deposition we identified previously, we demonstrated a role for the Reb1 protein and RSC for NFR formation programmed by this isolated signal. Given the previously reported biochemical interactions between Reb1 and subunits of RSC, the simplest interpretation is that recruitment of RSC by Reb1 generates the NFR. Our examination of the generality of this mechanism across chromosome III suggests that a subset of promoters, enriched for Reb1 binding sites, use this mechanism in a nonredundant fashion. Abf1, another essential Myb family member, operates at a distinct subset of promoters. These observations are consistent with studies that show that Reb1 and Abf1 sites are highly enriched in NFRs compared to the binding sites for nearly all other studied DNA binding proteins (Lee et al., 2007). The remaining promoters presumably target RSC and other remodeling mechanisms through other means. In this regard, it is interesting to note that four subunits of RSC contain potential DNA binding domains. Using standard ChIP protocols as well as ones using additional crosslinking agents, we have been unable to detect either wild-type Sth1 or an induced catalytically-dead version of Sth1 at the Reb1:dT7 signal inserted into PRM1, suggesting transient binding of RSC to this site (unpublished data). Likewise, only a fraction of intergenic regions display RSC binding in published ChIP-chip experiments (Ng et al., 2002), despite the global requirement for RSC in pol II transcription (Parnell et al., 2008). We suggest that at many sites of action the off-rate of the RSC complex in vivo may be too high to allow detection by ChIP.
We find no evidence that nucleosome positioning in general requires H2A.Z deposition. While a previous report suggested that H2A.Z controls nucleosome positioning in vivo, this conclusion was largely based on a single 20bp shift observed in the position of a nucleosome in the GAL1 promoter in htz1Δ cells (Guillemette et al., 2005). Another study examined nucleosome positioning in htz1Δ cells at four other loci (SUC2, COQ3, POS5, and COQ1), which are all highly enriched for H2A.Z and saw no differences in positioning (Li et al., 2005). Our results are generally in line with the latter study. However, we note that the technology used in our study, while cost-effective and allowing for multiple experimental replicates, does not have the ability to detect shifts of less than 20bp. Thus, we cannot rule out the possibility that our studies would have missed a more subtle role for H2A.Z deposition in nucleosome positioning.
To explore the relationship between NFR formation and H2A.Z deposition we implemented a steroid-regulated protein splicing strategy to induce H2A.Z deposition under conditions in which NFR structure was abrograted by depletion of Sth1 Our data show that deposition of H2A.Z about the NFR programmed by insertion of Reb1:dT7 into PRM1 required the prior action of Sth1, which presumably acts to induce formation of the NFR. This defect in deposition was not due to a general defect in H2A.Z deposition in RSC-depleted cells as normal deposition occurred at the BUD3-DCC1 intergenic region and significant albeit reduced H2A.Z deposition occurred at the promoters of two genes whose NFRs shrank in response to RSC depletion. Our results predict that in vitro studies of the exchange activity of the purified Swr1 deposition complex may show a dependence on adjacent nonnucleosomal DNA. Such a property would not be without precedence as the ACF complex has been shown to have nucleosome-sliding catalytic activity that is stimulated in vitro by flanking DNA (Yang et al., 2006). These observations may explain the general linkage observed in yeast, plants and metazoans between NFRs of various sizes and enhanced deposition of H2A.Z in flanking nucleosomes.
The strains used in this study are described in Table S1. Yeast transformants were generated by conventional lithium acetate and polyethylene glycerol procedures with selectable or counter-selectable transforming DNA. Insertions at the PRM1 ORF were obtained by a two-step process in which a construct containing I-SceI and its restriction site was first inserted and subsequently replaced with a desired sequence (Storici, et al. 2003).
H2A.Z-specific polyclonal antisera were generated against a peptide specific for the C-terminus of S. cerevisiae H2A.Z (custom-generated). The HA epitope tag in the degron alleles was detected using monoclonal antibody HA.11 (Covance). Abf1 was detected using polyclonal antibodies directed towards the Abf1 C-terminus (yC-20, Santa Cruz). Histone H3-specific polyclonal antibodies were directed towards the C-terminus of human histone H3 (ab1791, abcam).
Microarray data can be obtained from NCBI GEO at series accession GSE13446.
For each strain, total RNA from four independently grown cultures was prepared using a TRIZOL procedure and spiked with RNA from the Agilent Dual-color RNA Spike-in Kit. Aminoallyl-dUTP-labeled probe was generated by reverse transcription, and hybridizations were carried out using 4×44k Agilent microarrays that cover 6256 S. cerevisiae features, each of which are replicated 7 times on the array (Agilent design ID 015072). Dye swaps were incorporated such that for each experiment, there were 2 arrays of one dye configuration, and vice-versa. Data normalization was performed using a composite loess procedure that used 1:1 DCP probes for the spike-in loess curve (Yang, et al. 2002). Expression ratios for each gene per array then were derived by calculating the mean of up to 7 technical replicates, while discarding any replicates that were not within 2 standard deviations.
The essential genes ABF1, REB1, and STH1 were studied by regulated degradation of their encoded protein via degron alleles. Each degron allele was under the control of the pMET3 promoter, which is repressed by methionine. The REB1 and STH1 degron alleles had an arginine-capped N-terminal fusion of DHFRts and a triple-HA tag, while the ABF1 degron allele was an abf1(M1R) allele. UBR1, the N-end rule pathway E3 ubiquitin ligase, was placed under the control of a pGAL1 promoter.
To study phenotypes arising from loss of Abf1, Reb1 or Sth1, degron cultures were grown at 30°C to mid-log phase in synthetic complete media lacking methionine and cysteine with 2% raffinose and 0.1% dextrose as carbon sources. Activation of the degron was achieved by first adding galactose to a final concentration of 2% for 30 min, followed by centrifugation at room temperature to collect the cells. These cells were next grown at 37°C in rich media prewarmed at 37°C and supplemented with 2% galactose (YPAG).
Cultures in mid-log phase (which ranged from an OD600 of 0.7-0.9) were crosslinked with 1% formaldehyde for 15 min at the same temperature used for growth, followed by a quenching step for 5 min at room temperature with 0.125M glycine. Cells were washed twice with cold ddH2O prior to storage.
Approximately 20 OD600 units of cells were spheroplasted with 0.25 mg Zymolyase 100-T (Seikagaku) in 2 ml Buffer Z (1M sorbitol, 50mM Tris-Cl pH 7.4, 10mM β-mercaptoethanol) at 30°C with shaking. The spheroplasting time ranged from 30 min to 75 min, depending on the strain and media conditions used for growth. The ideal spheroplasting time was one that yielded appropriately digested chromatin (~90% mononucleosomal-sized DNA) after 20 min of micrococcal nuclease (MNase) treatment. Spheroplasts were collected by centrifugation at 4°C and resuspended in 500μl MNase digestion buffer (0.075% NP-40, 50mM NaCl, 10mM Tris-Cl pH 7.4, 5mM MgCl2, 1mM CaCl2). Chromatin was digested with 3 units of MNase (Worthington) for 20 min at 37°C. Digestions were quenched with 50mM EDTA, and spheroplasts were lysed with 0.1% SDS and centrifuged to transfer the supernatant away from insoluble material. The supernatant containing solublized chromatin was incubated at 65°C overnight with 0.4 mg/ml proteinase K to deproteinize DNA and reverse methylene crosslinks. DNA was recovered by two extractions with phenol and one extraction with chloroform, followed by ethanol precipitation and resuspension in Tris-EDTA (TE) pH 8.0 supplemented with 10 μg/ml RNase A. After a 30 min treatment at 37°C, the DNA was ready for probe generation by linear amplification.
Genomic DNA was prepared by purification with a Qiagen 100 column after treating spheroplasts with RNase A and proteinase K. Purified DNA was digested with MNase at room temperature to obtain DNA that ranged in size from 100-300bp. This digested DNA was phenol-extracted twice, chloroform-extracted once, and then ethanol-precipitated and resuspended in TE pH 8.0. Genomic reference probe was then prepared by linear amplification in the same manner as mononucleosomal-sized probe.
Probes for use in microarray experiments were prepared by linear amplification (Liu et al., 2003) but instead of preparing aminoallyl-cDNA probe, aminoallyl-RNA probe was prepared. In brief, DNA obtained by MNase digestion was first treated with calf intestinal phosphatase (New England Biolabs (NEB)) and then thymidine-tailed using terminal dideoxytransferase (NEB). A T7 promoter was adapted to these T-tailed DNA via second strand synthesis with Klenow exo-polymerase (NEB). RNA was next generated using a MegaScript T7 RNA polymerase kit (Ambion) with a 2.3:1 ratio of aminoallyl-UTP to UTP.
Nucleosome positions were mapped using a 20bp resolution tiling microarray with the majority of the probes being identical to a previously described microarray version (Yuan et al., 2005). We designed the remaining probes, which included coverage of the PRM1 ORF. This tiling microarray was printed using a custom arrayer and consisted of 32 print blocks with a total of 16,429 bona fide probes. Unless otherwise specified, nucleosomes were mapped in four independently grown cultures for each strain by hybridizing 5μg of mononucleosomal-sized probe and 5μg of reference genomic DNA probe. Hybridizations were conducted at 65°C for at least 12 hr prior to scanning. Dye swaps were incorporated into the experiments in a balanced manner such that with four mononucleosome:reference replicates, two were labeled as Cy3:Cy5 and the remaining two Cy5:Cy3.
Raw microarray data was processed and analyzed using custom-written software implemented with Python. Algorithms for statistical analysis were provided by the R statistics software package. A mononucleosome:reference ratio was calculated for each feature by first subtracting the feature median background intensity from the feature foreground intensity and then taking the log base 2 transformed ratio. The feature ratios in each print block were normalized for intensity-dependent bias using a LOESS regression algorithm with a smooth value of 0.4 (LOESS function provided by the R statistics package).
Analysis of nucleosome enrichment data was done using custom-written software implemented with Python that integrated statistical algorithms from the R statistics package. Nucleosome enrichment values for each experiment were determined by taking the mean of the experiment replicates and applying a centered moving average with a window of 5 features (100 bp). This moving average was strictly implemented such that no missing values were permitted within the window, and each feature in the window must be offset by 20 bp (the array resolution) from its immediate neighbor.
While results from two experiments can be compared side-by-side, generation of a difference map that represents differences between experiments is useful to highlight how the datasets differ from one another. As nucleosome enrichment data are in log2 space, difference maps were generated by subtracting probe data for a control experiment from corresponding probe data for the second experiment. For example, to generate the difference map shown in Fig. 3A, probe data from a control experiment that lacked the Reb1 degron were subtracted from corresponding probe data from an experiment that contained the Reb1 degron. The strains used for both experiments were grown under the same conditions.
Nucleosome positioning data shown in the Figures were generated by the Matplotlib plotting engine module for Python using a coordinate system with probe data representing nucleosome positions overlaid on the positions of features in the genome.
Comparison of nucleosome positions among a set of ORFs is easily achievable by generating a two-dimensional stack in which each row represents data associated with one ORF and each column represents probe data that is relative to a defined reference point in each ORF. These two-dimensional stacks were generated using custom-written Python software for graphical visualization with Java TreeView.
For the specific purposes of this work, all two-dimensional stacks were generated for ORFs that were associated with probe data. As the resolution of the tiling microarray platform is 20bp, bins of 20bp were defined relative to the translational start site of each ORF, and available probe data 1000bp upstream and downstream of each ORF were assigned to the appropriate bin by the coordinate representing the last nucleotide of the probe. This process generated a two-dimensional stack of data arranged by ORF, but centered at the translational start site. This is not an ideal arrangement for analyzing nucleosome-free regions as the distance between translational start sites and NFRs can vary among ORFs. Therefore we developed a method of analyzing nucleosome positions about NFRs in a two-dimensional stack of data in which data for each ORF was aligned at the nucleosome downstream of the NFR.
Alignments of extracted ORF data at the +1 nucleosome downstream of the NFR were calculated only for experimental data that served as controls. In this work, alignments were calculated using data from a strain isogenic to the degron strains, except that they lacked a degron allele, from a wild-type strain with the Reb1:dT7 sequence inserted into the PRM1 gene, and from a rpb1-1 strain at its permissive temperature. Nucleosome positions for the degron control strain were aligned using data under both degron-inactive and degron-active growth conditions.
Once these alignment maps were generated, experimental data were overlaid onto these maps to determine how nucleosome positioning is affected. For example, nucleosome positioning data from a strain carrying the Sth1-degron grown under degron-inducing conditions were overlaid on the alignment map generated from the control strain grown under the same conditions. Difference maps that highlighted the changes in nucleosome positions between strains were overlaid on alignment maps by first subtracting control probe data from corresponding experimental probe data, and then overlaying the differences onto alignment maps generated from nucleosome positions in the control data.
We thank past and present members of our laboratory for discussions and advice over the course of this research. We thank Leslie Chu and Joachim Li for the pMET3 degron construct. We are especially grateful to Allen Buskirk and David Liu for providing the 4-hydroxytamoxifen-regulated intein construct. This work was supported by a National Institutes of Health Kirschstein predoctoral fellowship (1F31GM083619) to P.D.H. and NIH grant 5R01GM071801 to H.D.M. H.D.M. is a scholar of the Leukemia and Lymphoma Society.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.