|Home | About | Journals | Submit | Contact Us | Français|
In S. cerevisiae, histone variant H2A.Z is deposited in euchromatin at the flanks of silent heterochromatin to prevent its ectopic spread. We show that H2A.Z nucleosomes are found at promoter regions of nearly all genes in euchromatin. They generally occur as two positioned nucleosomes that flank a nucleosome-free region (NFR) that contains the transcription start site. Astonishingly, enrichment at 5′ ends is observed not only at actively transcribed genes but also at inactive loci. Mutagenesis of a typical promoter revealed a 22 bp segment of DNA sufficient to program formation of a NFR flanked by two H2A.Z nucleosomes. This segment contains a binding site of the Myb-related protein Reb1 and an adjacent dT:dA tract. Efficient deposition of H2A.Z is further promoted by a specific pattern of histone H3 and H4 tail acetylation and the bromodomain protein Bdf1, a component of the Swr1 remodeling complex that deposits H2A.Z.
The association of eukaryotic DNA with histone octamers to form nucleosomes has profound implications for all aspects of DNA metabolism. Epigenetic control mediated through chromatin is now recognized as a major form of genetic regulation that functions during both normal development and pathogenic processes such as tumorigenesis. Therefore, a critical challenge faced by dividing eukaryotic cells is the faithful maintenance of both active and inactive epigenetic states of specific genomic regions. Three known biochemical mechanisms exist to control the states of chromatin: histone posttranslational modifications (on both the unstructured N-terminal tails and core regions), ATP-dependent chromatin remodeling by Swi2/Snf2 family members, and histone variant substitution. The current goal of the field is to link these mechanisms to epigenetic regulation. Substantial progress has been made in understanding how silent heterochromatin is generated and maintained. Compared to heterochromatin, less is understood about how euchromatin is generated, maintained, and inherited. Indeed, euchromatin has widely been viewed as a default state. More recently, however, several chromatin modifications have been identified that promote the euchromatic state by antagonizing silencing. These include the replacement of histone H2A by H2A.Z (Meneghini et al., 2003) and three histone modifications: acetylation on lysine 16 of the H4 tail (Kimura et al., 2002; Suka et al., 2002) and methylation of lysines 4 and 79 of H3 (Ng et al., 2003a; Santos-Rosa et al., 2004; van Leeuwen et al., 2002). In this paper, we focus on the deposition pattern of H2A.Z in euchromatin and its implications.
In previous work, we demonstrated that in S. cerevisiae, the evolutionarily conserved histone variant H2A.Z functions in euchromatin to antagonize the spread of Sir-dependent silencing. Furthermore, we showed that at the right border of the HMRa silent mating-type cassette, H2A.Z functions in parallel with a well-characterized boundary element (Meneghini et al., 2003). Thus, H2A.Z is a component of euchromatin that functions to antagonize the opposite chromatin state. One key question, therefore, is whether H2A.Z is randomly distributed through euchromatin and if not, how its deposition to specific sites is determined. We and others have also identified a 13 subunit ATP-dependent chromatin remodeling complex, the Swr1 complex, that is required for the deposition of H2A.Z in vivo (Kobor et al., 2004; Krogan et al., 2003b; Mizuguchi et al., 2004). Where the Swr1 complex acts and how its specificity is determined is not known. A subunit of this complex is Bdf1, a protein containing two tandem bromodomains known to bind acetylated histone tails (Ladurner et al., 2003; Matangkasombut and Buratowski, 2003). This suggests recognition of histone acetylation as one potential mechanism for the targeting of H2A.Z deposition to euchromatin.
Early chromatin immunoprecipitation (ChIP) experiments performed by Smith and coworkers suggested a relative enrichment of an epitope-tagged version of H2A.Z at the promoter regions of the highly inducible GAL1–10 and PHO5 genes in yeast (Santisteban et al., 2000). Moreover, these experiments demonstrated enrichment under noninducing conditions for the linked genes, and this enrichment decreased upon gene induction. However, it is difficult to make general conclusions from these studies for three reasons. First, since only four intergenic regions were examined, their correlation with higher H2A.Z levels could have been coincidental. Second, since no intergenic regions lacking a promoter were examined, the correlation with H2A.Z levels could have reflected preferential H2A.Z deposition in intergenic regions rather than in promoters per se. Third, since nucleosome density was not examined in the gene induction experiments, whether H2A.Z was selectively removed upon gene activation relative to H3, for example, was not clear. Thus, the following issues remain unresolved: (1) where is H2A.Z deposited in general?, (2) what is the relationship between H2A.Z deposition to transcription?, and (3) what are the signals that induce its deposition?
Previous studies have described a prominent role for H2A.Z at heterochromatin-proximal regions to antagonize the spread of silencing; however, we were curious to examine whether H2A.Z might play a broader role in the genome. Such additional roles could be elucidated through knowledge of the deposition profile of H2A.Z across chromosomes. We chose to examine the H2A.Z deposition profile in S. cerevisiae chromosome III because it contains the HMRa and HMLα silent mating-type cassettes and is well characterized with respect to the location of replication origins, cohesion sites, and transcription initiation sites. This analysis was conducted with a strain carrying an allele of H2A.Z with an amino-terminal influenza hemagglutinin epitope tag (HA3-HTZ1) that was integrated at the endogenous locus as the sole genomic copy. This allele is functional in that it can complement the synthetic lethality of htz1Δ with bre1Δ (Hwang et al., 2003). ChIP and quantitative real-time PCR (QPCR) were used to determine H2A.Z enrichment at 300 bp segments whose 5′ ends were spaced at 1000 bp intervals across chromosome III. We observed a highly nonuniform and chromosome-wide distribution of H2A.Z (Figure 1A; see Table S1 in the Supplemental Data available with this article online). Further analysis of our data indicated that the level of H2A.Z deposition at a given ChIP probe region was positively correlated with its proximity to the nearest 5′ end of a gene (Figure 1B). However, we observed no apparent correlation with proximity to 3′ ends of genes that are not near 5′ ends, the transcription rate of the nearest gene, cohesion sites, or origins of replication (M.Z.B., M.D.M., and H.D.M., unpublished data and see below).
We next increased the resolution of our chromosome III analysis to a single intergenic region flanked by two nonconverging ORFs. The intergenic region upstream of SNT1 was chosen because it is significantly smaller relative to the SNT1 coding region. A 4.2 kb continuous region starting from 2 kb upstream of the SNT1 initiation codon to 2.2 kb downstream was assayed for H2A.Z enrichment by ChIP and QPCR using primer sets that tiled the region. This assay revealed a striking intergenic enrichment for H2A.Z with a sharp decline in the coding region of SNT1 and in the upstream gene BPH1 (Figure 1C).
We then identified a larger region of chromosome III (the LEU2-YCL012W interval) containing a mixture of gene orientations: genes whose 5′ ends share an intergenic region (5′-5′); genes whose 5′ ends are adjacent to a 3′ end (5′ only); and genes whose 3′ ends converge (3′-3′). We assayed the H2A.Z deposition profile within this 11 kb region by ChIP and QPCR. This tiling analysis revealed that for every H2A.Z peak of enrichment, there was a corresponding 5′ end (Figure 1D). In most cases, the peak enrichment of H2A.Z was close to or directly upstream of the initiation codon. The only shared 5′ region in this data set (the DCC1-BUD3 intergenic region) had two distinct peaks of H2A.Z enrichment, one corresponding to the 5′ end of each gene. The observed H2A.Z peaks in these promoter regions were not due to increased crosslinkability or nucleosome density of these regions because additional ChIP analysis of histone H3 across the same region revealed slightly lower, not higher, ChIP signals in intergenic regions (Figure S1). Finally, the two regions with converging ORFs (3′-3′) had no observable peak of H2A.Z, supporting the notion that H2A.Z is indeed selectively enriched at 5′ regions of genes.
Our initial analyses of H2A.Z deposition relied on a ChIP protocol that involved shearing DNA to an average size of 500 bp, which meant that QPCR analyses of immunoprecipitated material resolved multiple nucleosomes, thereby obscuring finer details of H2A.Z localization. In addition, the tiling methods we used to assay H2A.Z deposition at an appropriate resolution are not feasible for rapidly examining much larger regions such as whole chromosomes. To overcome these two limitations, we used a ChIP and microarray hybridization protocol to analyze the distribution of endogenous, untagged H2A.Z at the resolution of single nucleosomes; the data were normalized for nucleosome density (see Experimental Procedures). The microarrays tiled the majority of chromosome III and 223 additional regulatory regions at a resolution of 20 bp. These experiments yielded a nucleosome-resolution map of H2A.Z enrichment patterns across nearly half a megabase of the S. cerevisiae genome (Table S2).
Analysis of the data recapitulated our initial conclusions about the specific deposition of H2A.Z at 5′ ends of genes (Figure 1) and extended these conclusions to a larger portion of the yeast genome. Figure 2A shows the microarray data for the regions analyzed by QPCR in Figures 1C and and1D.1D. Consistent with the QPCR data, the region upstream of SNT1 contains H2A.Z, and this H2A.Z signal has now been resolved into a striking distribution pattern in which two consecutive nucleosomes contain H2A.Z. Two H2A.Z nucleosomes are also found upstream of the BPH1 gene, one upstream of the FEN1 gene and two in the RRP43-RBK1 intergenic region that is flanked by the 5′ ends of the respective genes. In contrast, no H2A.Z peak was observed in the FEN1-RRP43 intergenic region that is flanked by the 3′ ends of those two genes. Also shown in Figure 2A is a portion of the NFS1-YCL012C region analyzed in Figure 1D; LEU2 is missing because this gene is deleted in the strain profiled using the microarray method. The QPCR analysis of this region was precisely recapitulated by the microarray data: H2A.Z was found specifically in intergenic regions that contain at least one 5′ end of a gene. Indeed, analysis of the entirety of chromosome III revealed H2A.Z upstream of most euchromatic genes and not at intergenic regions flanked by two converging 3′ ends (Table S2). Genes on chromosome III that lacked detectable H2A.Z in their promoter regions correspond to genes in the HMLα silent cassette, genes near the telomeres of chromosome III, ORFs annotated as “dubious,” and seven apparently bona fide euchromatic genes (HIS4, POL4, ADY2, AGP1, RPS14A, PMP1, and YCR006C). While it is not obvious why these genes lack H2A.Z in their promoter regions, we note that YCR006C contains a tRNA gene in its upstream regions. tRNA genes have been shown to contain boundary activity (Donze et al., 1999) and have been shown to inhibit expression of adjacent Pol II-transcribed genes (Bolton and Boeke, 2003). Since genes lacking H2A.Z in their promoters represent a small minority, we conclude that H2A.Z generally marks the 5′ ends of genes in euchromatin.
Recent work by Yuan and coworkers (2005) demonstrated that nucleosomes are generally uniformly distributed across yeast promoters and ORFs but nearly all yeast genes contain an ~150 bp nucleosome-free region (NFR) centered ~200 bp upstream of the initiation codon. cDNA hybridization studies demonstrated that these regions contain the initiation site for transcription of their associated genes (Yuan et al., 2005). The genes represented in the H2A.Z microarray data were aligned by the center of their NFRs to generate a cluster hierarchy shown in Figure 2B. Remarkably, for about 2/3 of the genes analyzed, the NFR is flanked by two nucleosomes that contain H2A.Z. The remainder of these genes appears either to have H2A.Z present only at one nucleosome or lack H2A.Z entirely for potential reasons explained above (Table S3). Thus, not only does H2A.Z mark the 5′ ends of genes, but two positioned H2A.Z nucleosomes typically flank the transcription initiation site. These data demonstrate that H2A.Z nucleosomes are placed in a highly stereotyped and organized manner at the 5′ ends of genes.
The striking localization of H2A.Z at most gene promoters suggested that there could be a relationship between H2A.Z and gene transcription. To address this issue, we selected from the H2A.Z ChIP microarray data those genes that contain two H2A.Z nucleosomes flanking a NFR and compared the levels of H2A.Z enrichment at each of the two nucleosomes to two distinct measurements of transcriptional activity for the corresponding gene (Figure 3). We used genome-scale data from either an analysis of initiation rates (Fraser et al., 2004; Figures 3A and and3C)3C) or RNA polymerase II occupancy (Kim et al., 2004; Figures 3B and and3D).3D). Comparison of H2A.Z enrichment at genes to either data set showed no correlation between H2A.Z enrichment and transcriptional activity. In other words, the transcription rate of a gene does not predict the levels of H2A.Z at a given promoter.
To further assess whether H2A.Z requires active transcription for its selective enrichment at gene promoter regions, we examined several promoter regions under conditions known to produce their tight repression. We first chose to examine the sporulation/meiosis-specific genes DIT1, DIT2, HOP1, and SPO22 in a haploid strain grown in rich media. These genes are transcriptionally inactive in haploid cells and in nonmeiotic diploid cells (Chu et al., 1998). Additionally, these four occur in two pairs in which their 5′ ends flank an intergenic promoter region. Strikingly, we observed peaks of H2A.Z enrichment at both of the shared promoter regions (Figures 4A and 4B). These patterns were not explained by underlying nucleosome density since H3 is relatively evenly distributed across these intervals (Figure S2). To attempt to observe de novo deposition of H2A.Z at these loci, we constructed a galactose-inducible HA epitope-tagged allele of HTZ1 with which we could selectively induce or repress the transcription of H2A.Z. As expected, under the repressive glucose condition, virtually no H2A.Z is detectable by ChIP (Figure S3). However, after growth for several generations in galactose, peaks of H2A.Z enrichment were observed at the divergent promoters of both meiotic gene pairs (Figure S3). Thus, H2A.Z can be deposited at inactive genes.
Another region we examined is the highly regulated mating-type specific gene AGA2. In yeast, a-specific genes (asgs) such as AGA2 have been well studied and are known to be active in MATa strains but extremely tightly repressed by the α2-Mcm1 complex in MATα and MATa/α strains (Galitski et al., 1999). We utilized isogenic strains harboring the chromosomal HA3-HTZ1 allele and differing only in the allele present at the mating type locus (MATa or MATα). Using ChIP and QPCR, we observed a peak of H2A.Z signal at AGA2 in MATa and its continued presence in MATα strains (Figure 4C). Although well above those seen in the ORF of the BUD3 gene (Figure 4C), H2A.Z levels were approximately 2-fold lower at the repressed AGA2 locus compared to the active locus even though promoter histone H3 signals were similar in a versus α cells (Figures 4C and S2).
Previous work showed that asg promoters display relative hypoacetylation on the histone H4 tails in MATα strains relative to MATa strains (Deckert and Struhl, 2001). We performed ChIP using antibodies raised against a tetra-acetylated peptide derived from the N-terminal tail of histone H4 (Ac4H4), and confirmed this result—an approximately 2-fold reduction of acetylation was observed in the MATα strains (Figure 4D). Interestingly, both the positioning and relative level of acetylation in a versus α cells closely parallels those of HA3-HTZ1 at AGA2, suggesting potential interplay between acetylation and H2A.Z deposition.
Finally, we identified two genes involved in mating in the microarray data that have been shown not to be expressed under vegetative conditions: FIG2 and PRM1. Previous work has shown that expression of these genes only occurs in response to mating pheromone (Erdman et al., 1998; Heiman and Walter, 2000). Analysis of H2A.Z enrichment at these loci revealed peaks in their promoter regions (Figure S4).
Our analysis revealed no correlation between H2A.Z levels normalized for nucleosome density and transcription rates or RNA polymerase II occupancies, suggesting no general relationship between transcription and H2A.Z levels. As described in the Introduction, previous studies of H2A.Z levels at GAL1 and PHO5 promoters revealed that it decreased upon gene induction, although whether this represented exchange of H2A.Z for H2A or general nucleosome depletion was not determined. In contrast, we observed that while the inactive AGA2 promoter contains H2A.Z, its levels are higher when the gene is active.
To extend these results, we examined H2A.Z and H3 levels at a gene that is highly inducible by mating pheromone, FIG 1 (Erdman et al., 1998). As shown in Figure S5, FIG 1 expression is dependent on mating pheromone—treatment of cells with mating pheromone strongly induced mRNA accumulation over a 1 hr time course as determined by quantitative RT-PCR. ChIP analysis revealed that H2A.Z is depleted during gene induction. However, H3 was also depleted from the FIG 1 promoter during the time course such at the 5, 15, and 30 min time points, the ratio of H2A.Z to H3 was constant (Figure S5). At the 60 min time point, an apparent depletion of H2A.Z relative to H3 was observed; however, it seems likely that the promoter H3 signal at this time point was elevated as an artifact of signal from flanking nucleosomes that were not separated by sonication from the promoter nucleosomes prior to ChIP (Figure S5). The H2A.Z signal would not be subject to this problem since H2A.Z nucleosomes are distributed in a punctate pattern whereas H3-containing nucleosomes are distributed homogenously. Thus, with the possible exception of this late time point, activation of FIG 1 results in nucleosome loss rather than the specific replacement of H2A.Z with H2A.
We performed a reporter-based genome-wide screen of the S. cerevisiae knockout collection to identify genes that antagonize the spread of silencing from the HMRa silent mating type cassette (R.M.R. and H.D.M., unpublished data). This screen identified Eaf1, a nonessential component of the essential NuA4 HAT complex and the bromodomain-containing proteins, Bdf1 and Bdf2. Bdf1 is a component of the Swr1 complex responsible for H2A.Z deposition, and both Bdf1 and Bdf2 bind to acetylated histone tails (Ladurner et al., 2003; Matangkasombut and Buratowski, 2003). To test whether histone acetylation is important for H2A.Z deposition, we generated strains bearing the HA3-HTZ1 allele containing deletions of the genes encoding the H4-specific histone acetyltransferase (HAT) Eaf1 or the H3- and H4-specific HAT Elp3. In addition, we created a strain lacking both HATs. ChIP analysis revealed a dependence upon histone tail acetylation for robust H2A.Z enrichment (Figure 5A). At a majority of loci examined, deletion of EAF1 resulted in a reproducible defect in H2A.Z levels. The defects varied from approximately 1.5- to 3-fold in magnitude. Likewise, deletion of ELP3 also led to a defect at most loci, albeit more quantitatively modest than those of the eaf1Δ mutant. The severity of the defects in the eaf1Δ elp3Δ double mutant is not significantly greater than either of the single deletions (Figure 5A), suggesting Eaf1 and Elp3 may act in the same pathway to mediate H2A.Z deposition.
To further test the role of histone acetylation in H2A.Z deposition, we utilized a series of histone H3 and H4 mutants in which specific target lysine residues have been mutated to arginine which prevents acetylation. We observed a consistent quantitative defect in H2A.Z enrichment values at most of the 10 loci examined (Figures 5B and 5C). In general, the strongest defects were observed in cells harboring the H3-K9R mutant or the H4-K5R,K12R mutant. For the H4-K5R,K12R mutant, we performed ChIP using antibodies against H3 as well and found no differences in nucleosome density at the loci examined in Figure 5C, indicating that the defect in H2A.Z deposition was not caused by general nucleosome loss (Figure S6). Surprisingly, a deletion mutant in the H4 tail displayed a less-severe defect than the H4-K5R,K12R mutant (Figure 5C). The H4-K8R,K16R mutant displayed no defect, indicating that not all mutants in acetylatable tail lysines produce a defect in H2A.Z deposition (Figure 5C).
Having established a role for histone tail acetylation for complete H2A.Z deposition, we hypothesized that acetylation could be acting to recruit targeting of the Swr1 complex via binding of its subunit Bdf1 to acetylated tails. This is an attractive model because in addition to being important for antisilencing, Bdf1 is known to bind preferentially to acetylated forms of histone H4 and is enriched in intergenic regions throughout the genome (Kurdistani et al., 2004). However, ChIP analysis using polyclonal Htz1 antibody raised against the C-terminal tail region showed that a bdf1Δ strain has little or no defect in H2A.Z enrichment at euchromatic loci (Figure 5E). We reasoned that this could be due to compensatory activity by the redundant gene BDF2, which when deleted yielded no detectable defect in H2A.Z enrichment (data not shown). Unfortunately, bdf1Δ bdf2Δ strains are inviable, precluding a test of this hypothesis using null alleles. Therefore, we elected to generate a “knockdown” allele of BDF2 by replacing its 3′UTR region with a MX6 marker cassette. This maneuver has been found to consistently cause destabilization of the cognate mRNA (Schuldiner et al., 2005). We refer to this allele as bdf2-utrΔ, and as is the case for the bdf2Δ, it also has no defect for H2A.Z enrichment (data not shown). As seen by tetrad dissection, the bdf1Δ bdf2-utrΔ double mutants grow more slowly than either single mutant, indicating a defect produced by bdf2-utrΔ (Figure 5D). Examining these strains by ChIP, we found that although bdf1Δ cells showed little or no defects in H2A.Z deposition, the bdf1Δ bdf2-utrΔ displayed a reproducible defect in H2A.Z deposition at a majority of loci examined (Figure 5E), while no defect was observed in the ORF of the control locus PRP8. These experiments clearly demonstrate a dependence on Bdf1 and its redundant homolog Bdf2 for full H2A.Z deposition at the 5′ regions of genes. However, since acetylation of promoter nucleosomes generally correlates with transcription (Liu et al., 2005), the requirement for Bdf1/2 and histone tail acetylation for efficient deposition of H2A.Z does not explain how it can be deposited at inactive genes in euchromatin.
One hypothesis for how H2A.Z is deposited at inactive as well as active genes is that there exist specific DNA elements in promoters that program its deposition. Although there is no precedent for a DNA element that specifically induces variant histone deposition, we decided to pursue this model by systematically mutagenizing a typical promoter that contains two positioned H2A.Z nucleosomes (Figure S7. For this analysis, we chose to analyze the SNT1 promoter region described in Figure 1 because it was well separated from nearby promoters by the relatively large BPH1 and SNT1 ORFs.
To localize sequences required for H2A.Z deposition, we divided the BPH1-SNT1 intergenic region into 75 bp segments and then precisely replaced each segment in the chromosome with a 75 bp fragment of the bacterial cloning vector pBluescript (Figure S7). Mutants in either of two adjacent intervals (termed 5 and 6 in Figure S7) resulted in a modest 2-fold reduction in H2A.Z enrichment (Figure S7). However, a mutant that replaced both intervals resulted in a dramatic defect in H2A.Z enrichment (Figure S7. Interestingly, these two intervals roughly correspond to the nucleosome-free region between the two H2A.Z nucleosomes that lie upstream of the SNT1 gene. RT-QPCR analysis of SNT1 expression revealed only a 2-fold drop in SNT1 mRNA levels (P.D.H. and H.D.M., unpublished data). These data suggested the presence of partially redundant signals for H2A.Z deposition in intervals 5 and 6.
To further define these signals, we constructed 14 additional substitution mutants in the NFR of the SNT1 promoter (Figure 6A). For these mutants, we replaced varying segments within intervals 5 and 6 with identical-sized segments from the ORF of BUD3, which lacks H2A.Z deposition (Figure 1). We examined H2A.Z deposition using primers that span the two flanking positioned H2A.Z nucleosomes. Of the 14 mutants tested, only two, mu1 and mu3, abolished H2A.Z enrichment (Figure 6A). The sequences replaced in mu3 represent a subset of those in mu1, defining the minimal segment that must be mutated to produce a complete loss of H2A.Z deposition in the SNT1 promoter. Substitution of smaller segments resulted in increased H2A.Z enrichment. For example, mu4 has the identical 5′ endpoint as mu3 but substitutes 10 fewer bp on the 3′ end and displays robust H2A.Z enrichment (Figure 6A). These 10 bp are therefore critical for H2A.Z deposition in the context of mu3. Likewise, mu10 substitutes 20 bp fewer than mu3 on the 5′ end and shows increased H2A.Z enrichment (Figure 6A), indicating that there are sequences that promote H2A.Z deposition in the 20 bp that distinguish mu3 from mu10. We note that for mutants that display an intermediate level of H2A.Z deposition, our analysis does not distinguish between a decrease in H2A.Z deposition versus a shift in the position of the H2A.Z nucleosomes. Nonetheless, our identification of mutants that eliminate H2A.Z deposition in this region suggest that the segments identified play a role in deposition per se. Taken together, these data suggest the presence of two redundant signals for H2A.Z deposition, one that includes the 10 bp that distinguishes mu3 from mu4 and another that includes the 20 bp that distinguishes mu3 from mu10.
Our analysis of sequences necessary for H2A.Z deposition at the promoter of SNT1 identified two discrete regions. We next tested whether these regions also sufficient to promote H2A.Z deposition at a novel site. To date, we have not succeeded in identifying a fragment containing the 20 bp 5′ region that is sufficient to promote H2A.Z deposition. Therefore, we focus below on a signal that contains the 3′ 10 bp segment hypothesized above to contain a H2A.Z deposition signal.
A magnified view of this 10 bp sequence and flanking sequences is shown in Figure 6B. Two notable features of this sequence are a binding site for the general regulatory factor Reb1 and an adjacent tract of seven dT:dA base pairs. Both sequence elements are disrupted in mu3 compared to mu4. Moreover, previous studies had shown that a similar arrangement of sequences in the yeast PFY1 promoter was important for the formation of a NFR in that promoter (Angermayr et al., 2003). Therefore, we tested whether a DNA segment containing this region could generate an NFR flanked by H2A.Z nucleosomes when placed elsewhere in the genome.
We inserted the 22 bp segment containing the Reb1 site and (dT:dA)7 tract at an arbitrarily chosen site in the middle of an inactive gene, PRM1 (Figure 7A). PRM1 was selected because it had been shown previously to only be expressed in cells exposed to mating pheromone, and we sought to avoid the potentially complicating effects of transcription on H2A.Z deposition (Heiman and Walter, 2000). Examination of H2A.Z deposition using probes flanking the insertion site revealed robust H2A.Z enrichment in the strain containing the insertion (Figure 7B). Replacement of the three G residues in the Reb1 consensus site abolished the effect of the insertion as did a deletion of the (dT:dA)7 tract.
To determine whether an NFR was induced by the insertion, we performed nucleosome-scanning analysis (Sekinger et al., 2005) to determine the positions of nucleosomes containing H3 and H2A.Z in the parental strain and the strain containing the insertion. Cross-linked mononucleosomes were immunoprecipitated with antibodies to either H3 or H2A.Z and then analyzed by QPCR analysis using primer pairs that amplified 100 bp segments every 20 bp across the PRM1 ORF. As shown in Figure 7C, five nucleosomes containing histone H3 were found in the PRM1 ORF in the parental strain. The arrow in Figure 7C indicates the site of insertion, which was in the center of the +4 nucleosome. The 22 bp insert had two effects on the nucleosome pattern (Figure 7D). First, it caused a delocalization of the nucleosome pattern in the first part of the PRM1 ORF. Second, it resulted in a formation of an NFR. This can be deduced by examining the peak-to-peak distance of nucleosomes flanking the insertion site, which is 320 bp in the strain containing the insertion versus 180 bp between the center points of the +3 and +4 nucleosomes in the parental strain. Assuming that 147 bp of DNA is wrapped by the yeast histone octamer, one calculates that the insertion caused an expansion of the linker region between these two nucleosomes from approximately 33 bp to 173 bp.
We next determined the positions of H2A.Z nucleosomes in the parental and insertion strains. As shown in Figure 7E, little H2A.Z enrichment was observed in ORF of the PRM1 gene in the parental strain, as expected. Strikingly, insertion of the 22 bp segment from the SNT1 gene resulted in the appearance of two positioned variant H2A.Z nucleosomes (Figure 7F). Moreover, the peaks were separated by 320 bp, confirming the formation of an NFR in the insertion strain (Figure 7F). Finally, we examined histone H4 acetylation in the PRM1 ORF in the strain containing the insertion and found no difference compared to wild-type (P.D.H. and H.D.M., unpublished data), indicating that this DNA signal functions in a distinct pathway from acetylation and Bdf1.
Our results show that nucleosomes containing the conserved histone variant H2A.Z occur in euchromatin in a highly organized rather than a random pattern. In particular, the experiments decisively demonstrate that H2A.Z is selectively present at the vast majority of gene promoter regions. Most commonly, it occurs as two positioned nucleosomes that flank a NFR that includes the transcription initiation site. The most striking finding is that H2A.Z enrichment is uncorrelated with transcription rates and is observed at promoters of genes that are not detectably transcribed. The implications of this observation are potentially far reaching, as it indicates that cells can identify the 5′ ends of genes in the absence of ongoing transcription. We describe two mechanisms that begin to provide insight into how this remarkable pattern of histone variant deposition occurs. Analysis of the SNT1 promoter resulted in the identification of a 22 bp bipartite DNA element sufficient to promote H2A.Z deposition when placed in a novel context. This signal contains two necessary elements that are generally conserved in yeast promoters: a binding site for the Myb-related general regulatory factor Reb1 and an (dT:dA)7 tract. In addition, we demonstrated that H2A.Z deposition is linked to histone acetylation and Bdf1, a double bromodomain protein that binds acetylated histone tails.
Our results provide the first single nucleosome-resolution global picture of the deposition pattern of a conserved histone variant. Alignment of the microarray data based on the identified NFR of yeast promoters that includes the transcription initiation site (Yuan et al., 2005) revealed that most euchromatic genes contain two positioned H2A.Z nucleosomes which flank the NFR. Our analysis to date cannot distinguish whether each of these nucleosomes contains two copies of H2A.Z or one copy of H2A.Z and one copy of H2A. However, it has been suggested based on structural analysis that heteromeric H2A.Z/H2A nucleosomes may be unable to form due to steric clash (Suto et al., 2000). Because one of the two H2A.Z nucleosomes is typically downstream of the initiation site of transcription and one is not, it is unlikely that passage of RNA polymerase alone plays a role in either depositing or removing H2A.Z nucleosomes in general. Indeed, a small group of genes contains only the downstream H2A.Z nucleosome (Figure 2B). It is not yet obvious why these genes differ in their deposition pattern. Consistent with our previous data that indicated the exclusion of H2A.Z nucleosomes from the HMRa silent mating-type cassette, the microarray analysis (which was performed in a mating type a strain) reveals an exclusion of H2A.Z from the HMLα silent cassette and from subtelomeric regions (see Tables S2 and S3).
Most strikingly, we find that the levels of deposition of H2A.Z in promoters are clearly not correlated with either the transcription rate or RNA polymerase II occupancy of the linked coding sequences (Figure 3). This is in contrast to modifications such as trimethylation of lysine 4 of histone H3 in yeast, which does correlate with transcription rate and typically occurs on the first nucleosomes downstream of the transcription initiation site (Bernstein et al., 2002; Krogan et al., 2003a; Ng et al., 2003b). Indeed, our analysis of genes that are not transcribed and/or tightly repressed demonstrated enrichment of H2A.Z in their promoters. These include two meiotic gene pairs examined in haploid cells in rich media, the a-specific gene AGA2 assayed in α cells, and two genes only expressed in pheromone-treated cells, FIG 2 and PRM1, that were assessed in the absence of pheromone. Although we cannot rule out the possibility that H2A.Z deposition occurs at these genes in response to rare transcription events that produce mRNAs that fail to detectably accumulate, a simpler interpretation of our data is that cells have a transcription-independent mechanism to specify H2A.Z deposition at the 5′ ends of genes.
Although H2A.Z can be deposited at inactive genes, our data suggests that transcription can modulate H2A.Z levels in at least two ways. First, at AGA2, we observed higher H2A.Z levels when the gene was active than when it was inactive. Second, at FIG 1, we observed that activation resulted in concomitant depletion of H2A.Z and H3, consistent with the removal of variant octamers. Since the relative levels of H2A.Z and transcription are uncorrelated when considering large numbers of genes (Figure 3), it seems likely that transcription modulates the relative amounts of H2A.Z variant nucleosomes differently at different genes. Further work will be needed to define the relationships between transcription and H2A.Z promoter marking. Nonetheless, our results demonstrate that for cells to identify the 5′ ends of genes and deposit H2A.Z, genes need not be transcribed.
Our genetic experiments led us to investigate the potential connection between histone tail acetylation and H2A.Z deposition. ChIP analyses demonstrated that for various defects in histone tail acetylation, whether produced by mutation of acetylated lysines or deletion of genes encoding histone acetyltransferases, there is a moderate decrease in H2A.Z at most sites assayed. The quantitative rather than qualitative defect in H2A.Z deposition in these mutant backgrounds may reflect either a partial dependence on histone tail acetylation for deposition or that histone acetylation was only partially eliminated in our experiments. Distinguishing between these two possibilities is not trivial since the H3 and H4 N-terminal tails are together essential for viability (Ling et al., 1996). Moreover, cells lacking the catalytic subunit of the NuA4 HAT and cells lacking both the Gcn5 and Sas3 HATs are inviable (Clarke et al., 1999; Howe et al., 2001). We also note that the in vivo deposition assays used here do not measure the rate of H2A.Z deposition. Therefore, the modest defects observed in acetylation mutants at steady state may reflect a more profound defect in the rate of deposition, especially if one considers that as few as one exchange event at a nucleosome per cell cycle might be sufficient to produce wild-type levels of H2A.Z.
We find that the bromodomain proteins Bdf1 and Bdf2 act redundantly to promote H2A.Z deposition. Bdf1 is a subunit of both the Swr1 complex that deposits H2A.Z in vivo and is also associated with TFIID. Because Bdf1 contains two bromodomains and selectively binds acetylated versions of histone H4, we suggest that Bdf1 recognition of acetylated histone tails promotes recruitment of the Swr1 complex and deposition of H2A.Z. In vitro studies of the purified Swr1 complex and acetylated nucleosomal substrates will be required to confirm this model. It is notable that the H4-K8R, K16R mutation did not affect H2A.Z deposition: recent work has shown that deacetylation of H4-K16 is actually necessary for the association of Bdf1 with chromatin in vivo (Kurdistani et al., 2004). Consistent with these observations, recent studies of histone acetylation patterns at the mononucleosome level demonstrated that the two nucleosomes flanking the NFR have a unique acetylation pattern (Liu et al., 2005). In particular, these nucleosomes are both highly deacetylated on H4-K8 and 16, and this deacetylation domain occurs independently of transcription level, thereby precisely paralleling the H2A.Z localization pattern presented here. Moreover, the nucleosome downstream of the NFR is acetylated on H3-K9,14 and H4-K5,12. It is unlikely to be coincidental that lysine-to-arginine mutation of the residues that are deacetylated on the NFR-flanking nucleosomes does not affect H2A.Z deposition, while mutation of acetylated residues inhibits H2A.Z deposition (Table 1). Together with the data showing that Bdf1 binding to chromatin is inhibited by H4-K16 acetylation, these results are consistent with a direct role for Bdf1 in recognizing the acetylation patterns of the NFR-flanking nucleosomes to promote H2A.Z deposition. However, since acetylation of the nucleosome downstream of the NFR correlates with transcription rates (Liu et al., 2005), efficient deposition of H2A.Z at highly deacetylated inactive promoters must involve mechanisms that would not in principle depend on ongoing transcription.
We have defined one such mechanism, namely the existence of DNA signals that program H2A.Z deposition. Our analysis of the SNT1 promoter revealed two segments of DNA that appear to function redundantly since mutations in two segments with the NFR were necessary to eliminate H2A.Z deposition. We showed that the 3′ signal, which contains a site for the Myb-related general regulatory factor Reb1 and an adjacent (dT:dA)7 tract, was sufficient to induce the formation of an NFR and the replacement of H2A with H2A.Z in the two flanking nucleosomes when placed into the middle of the coding sequence of inactive PRM1 gene. Both the Reb1 site and (dT:dA)7 motif were found to be necessary for H2A.Z deposition.
Reb1 was originally identified as an abundant nuclear protein involved in rDNA transcriptional termination but was subsequently shown to associate with a large number of yeast promoter regions (Ju et al., 1990). Recent studies of the conservation of the Reb1 DNA binding motif have shown that it is the single most conserved motif found in yeast promoters and is even more conserved across species than the TATA box (Elemento and Tavazoie, 2005). Several studies have shown that tethering of Reb1 or related Myb-family general regulatory factors (Rap1, Abf1, or Tbf1) to DNA can prevent the spread of silent chromatin, but the mechanism remains unknown (Fourel et al., 2002; Yu et al., 2003). Given our results, it could be that this property of these factors involves the induction of a NFR and/or the deposition of H2A.Z nucleosomes. Consistent with this possibility, there is a near match to the Abf1 binding consensus in the region of the SNT1 NFR that contains the 5′ signal for H2A.Z deposition (P.D.H. and H.D.M., unpublished data).
The second motif that we found to be important for H2A.Z deposition is a tract of dT:dA base pairs which have been noted to be common in yeast promoters, particularly in NFRs (Yuan et al., 2005). Studies of global nucleosome density have also shown that the abundance of motifs containing dT:dA tracts correlate with nucleosome depletion from promoters (Bernstein et al., 2004; Lee et al., 2004). These studies concluded that promoters show transcription-independent reductions in nucleosome density compared coding sequences, but this conclusion has been questioned on technical grounds (Pokholok et al., 2005). Our study is relevant to this issue as it shows the functional importance of an element containing a dT:dA tract flanked by a site for Reb1 in the formation of NFR. Our data may also be relevant to the recent proposal that dT:dA tracts promote the formation of NFRs in vivo because of their intrinsic nucleosome excluding properties in vitro (Sekinger et al., 2005). Although further work is necessary to understand how it functions, it seems unlikely that a sequence as short as 22 bp could act to program the formation of an ~170 bp NFR purely because of its intrinsic properties.
Although both Reb1 sites and dT:dA tracts are common features of yeast promoters, we do not yet know whether this is the sole type of DNA element that programs H2A.Z deposition at promoters. As mentioned above, other Myb-related factors might also be expected to play a role. A previous study identified a Reb1 site and an adjacent dT:dA tract in the NFR in the promoter of the yeast PFY1 gene (Angermayr et al., 2003). This work showed that mutation of the Reb1 site eliminated the NFR; the role of the adjacent dT:dA tract was not assessed. Thus, it may be that Reb1 is generally important for the formation of NFRs in promoters. This raises the question of whether Reb1 promotes H2A.Z deposition and NFR formation through independent or coupled mechanisms. Our preliminary studies show that deletion of HTZ1 or SWR1 does not prevent the formation of the NFR in the strain containing the 22 bp insertion into PRM1 (P.D.H. and H.D.M., unpublished data). Thus, the 22 bp element either promotes NFR formation and H2A.Z independently (e.g., via recruitment of different factors) or the formation of the NFR itself induces H2A.Z deposition. Regardless of the specific mechanisms involved, our studies indicate that DNA- and histone-based mechanisms allow cells to mark the 5′ ends of genes and preserve their euchromatic state.
Strains used in these studies are described in Table S5.
Chromosomal mutations were created as described (Storici et al., 2003). Heterologous sequences used are described in Table S7. Different sequences were used in Figure S7 and Figure 6 to ensure that the results were not dependent on the particular sequence used to replace SNT1 sequences.
Cultures were grown at 30°C. Strains bearing an HA3 epitope-tagged allele of HTZ1 driven by the GAL1 promoter at the endogenous HTZ1 locus were grown to saturation in YPAD, then diluted to an A600 of 0.1, and outgrown in YEP containing 2% glucose to an A600 of 0.6. Fifty milliliters of the cultures were crosslinked and harvested. The remaining cells were washed twice in water and added to YEP containing 2% galactose and 2% raffinose to an approximate A600 of 0.001 and grown for 2 days. These cultures were then back diluted to fresh YEP containing 2% galactose and 2% raffinose to an A600 OD of 0.1 and grown to an A600 of 0.6, crosslinked, and harvested. Three absorbance units were harvested from each and analyzed by immunoblotting with antibodies against H2A.Z.
A wild-type MATa strain was grown in YPAD at 30°C overnight, diluted to an A600 of 0.1, and grown to an A600 of 0.6. Thirty OD units were crosslinked and harvested for ChIP, and three OD units were harvested for total RNA isolation and RT-QPCR analysis of transcript levels using gene-specific primers for FIG 1 and ACT1. The remaining culture was split four ways and α factor was added to a concentration of 10 μM to each and grown to the appropriate time point (5 min, 15 min, 30 min, or 1 hr), whereupon 30 and 3 OD of cells respectively were harvested as described above for ChIP and RT-QPCR analysis.
Mononucleosomes were prepared as described (Liu et al., 2005).
ChIP procedures were as in Meneghini et al. (2003) except for microarray and nucleosome scanning experiments, which were performed as described by Liu et al. (2005) and Sekinger et al. (2005), respectively.
The yeast strain used was BY4741. Hybridization and analysis was performed as described (Liu et al., 2005).
We are grateful to S. Dent for histone point mutants. This work was supported by grants from the NIH-NIGMS (H.D.M., S.L.S., O.J.R.), the Packard Foundation (H.D.M.), the Bauer Center (S.L.S., O.J.R.), and the Burroughs-Wellcome Fund (M.D.M.). We thank W. Marshall for critical reading of the manuscript, R. Wu for pointing out the Reb1 site, and S. Johnson and J. DeRisi for support and advice. Author contributions are as follows: R.M.R. performed the experiments shown in Figures 1C, 1D, ,4,4, ,5,5, S1, S2, S3, S5, and S6. P.D.H. performed the experiments shown in Figures 6, ,7,7, and S7. M.Z.B. and M.D.M. performed the experiments shown in Figures 1A and 1B. C.L.L. and O.J.R. performed the experiments shown in Figures 2, ,3,3, and S4. R.M.R., P.D.H., O.J.R., and H.D.M. wrote the paper.
Microarray data has been deposited in the NIH GEO database (accession number GSE3411).
Supplemental Data include seven figures and seven tables and can be found with this article online at http://www.cell.com/cgi/content/full/123/2/233/DC1/.