生物谷报道：有关基因转录的过程人们并不是太清楚，核小体中的组氨酸的乙酰化和甲基化的一对相反过程是可能控制核心小体转录的核心因素。但人类对组氨酸的乙酰化和甲基化认识十分少。最新一期Cell上以特色文章刊登了麻省科学家Richard A. Young采用基因芯片技术，对酵母全基因组的组氨酸的乙酰化和甲基化进行全面解读，这一重大发现，使人类更深入地认识基因转录的精细调控的机理带来更深地认识。也为将来进一步进入高等生物的研究打下基础。生物谷将继续关注这一报道。下面是提供的主要摘要，以及全文提供下载：
摘要：
Eukaryotic genomes are packaged into nucleosomes whose position and chemical modification state can profoundly influence regulation of gene expression. We profiled nucleosome modifications across the yeast genome using chromatin immunoprecipitation coupled with DNA microarrays to produce high-resolution genome-wide maps of histone acetylation and methylation. These maps take into account changes in nucleosome occupancy at actively transcribed genes and, in doing so, revise previous assessments of the modifications associated with gene expression. Both acetylation and methylation of histones are associated with transcriptional activity, but the former occurs predominantly at the beginning of genes, whereas the latter can occur throughout transcribed regions. Most notably, specific methylation events are associated with the beginning, middle, and end of actively transcribed genes. These maps provide the foundation for further understanding the roles of chromatin in gene expression and genome maintenance.
Introduction:
The genomes of eukaryotes are packaged into chromatin, the fundamental unit of which is the nucleosome. Nucleosomes consist of approximately 146 base pairs of DNA wrapped around a histone octamer (Luger et al., 1997). The histone components of nucleosomes and additional chromatin proteins can interact to form higher-order chromosomal structures. Nucleosomes are thus critical to the organization and maintenance of genetic material, and their position and modification state can profoundly influence genetic activities such as regulation of gene expression (Kouzarides, 2002; Narlikar et al., 2002).
Several recent studies have explored the relative occupancy and modification state of nucleosomes across the yeast genome by using chromatin immunoprecipitation of histones and DNA-microarray analysis, a technique known as “genome-wide location analysis” or “ChIP-chip” (Bernstein et al., 2002; Bernstein et al., 2004; Kurdistani et al., 2004; Lee et al., 2004; Robyr et al., 2002; Santos-Rosa et al., 2002). The results of these studies, which used DNA microarrays that probed a single site in the intergenic or transcribed portions of the genome, suggested that the intergenic regions of S. cerevisiae are less densely occupied by nucleosomes than transcribed regions (Lee et al., 2004) and that nucleosome density in both regions inversely correlates with transcription rates (Bernstein et al., 2004; Lee et al., 2004). Similar studies have explored the relationship between transcription and genome occupancy by chromatin regulators (Humphrey et al., 2004; Kurdistani et al., 2002; Lieb et al., 2001; Ng et al., 2002; Ng et al., 2003; Robert et al., 2004; Robyr et al., 2002) or transcription and modification of nucleosomes (Kurdistani et al., 2004; Robyr et al., 2002; Roh et al., 2004). Some conclusions from these studies are difficult to reconcile with one another; for example, the histone acetyltransferases Gcn5 and Esa1 appear to be recruited to genes upon transcriptional activation (Robert et al., 2004), but acetylation of the amino acid residues they target in histones does not appear to be strongly associated with transcriptional activity (Kurdistani et al., 2004).
To more accurately define nucleosomal occupancy and modification and their relationships to transcription, we have investigated genome-wide chromatin structure at considerably higher resolution than that afforded by the experimental designs that have been used thus far. Such designs have typically used DNA microarrays with a single feature to capture signals from each intergenic or transcribed region (Bernstein et al., 2004; Kurdistani et al., 2004; Lee et al., 2004). We describe here profiling of protein-DNA occupancy at high resolution and accuracy using improved protocols and DNA microarrays that tile the yeast genome. The results we obtain with this approach provide a more complete picture of nucleosome occupancy and the nature of modifications associated with transcriptional activity, resolve discrepancies in previous reports, and allow us to produce the first high-resolution maps of histone acetylation and methylation in the yeast genome.
Results and Discussion
High-Resolution Genome-wide ChIP-Chip
To increase the resolution and accuracy of genome-wide location analysis, we designed a DNA microarray that contains over 40,000 probes for the yeast genome and developed hybridization methods that maximized signal-to-noise ratios on this array (see Experimental Procedures). To test whether these modifications improved the resolution and accuracy of genome-wide binding analysis, we explored the genome-wide occupancy of Gcn4, a transcriptional regulator of amino acid-biosynthetic genes with a well-characterized DNA binding specificity (Hope et al., 1985; Oliphant et al., 1989), crystal structure (Ellenberger et al., 1992; O’Shea et al., 1991), and previously identified target genes (Arndt et al., 1986; Natarajan et al., 2001). The binding data for individual target genes are shown in Figure 1A and in Figure S1 in the Supplemental Data available with this article online. For those regions for which there is strong evidence for Gcn4 binding, we found that the peak of Gcn4 binding occurred directly over that binding site.
Figure 1. Nucleosome Occupancy across the Yeast Genome with High-Resolution Genome-wide Location Analysis
(A) Occupancy of the HIS1 promoter by Gcn4. The genomic positions of probe regions are arrayed along the x axis, with the ratio of enrichment of Gcn4 for probes along the y axis. ORFs are depicted as gray rectangles, and arrows indicate the direction of transcription. Red boxes represent sequence matches to the Gcn4 binding specificity within promoter regions.
(B) Composite profile of Gcn4 binding at the set of 84 high-confidence Gcn4 target genes. Promoter and downstream regions were aligned with each other according to the position of a sequence match to the Gcn4 binding specificity. Aligned probes were then assigned to 50 bp segment bins, and an average of the corresponding enrichment ratio was calculated. Standard error of the mean is shown in gray. Genetic elements are depicted as in Figure 1A, except that dashed lines represent sites including both ORFs and intergenic regions.
(C) Nucleosome occupancy at the promoter of CPA1, a gene encoding an amino acid-biosynthetic enzyme. The genomic positions of probe regions are arrayed along the x axis, with the ratio of enrichment of histone H3 (blue) or H4 (red) for probes along the y axis. ORFs are depicted as gray rectangles, and arrows indicate the direction of transcription.
(D) A composite profile of histone occupancy at 5324 genes. The ends of ORFs were defined at fixed points according to the position of translational start and stop sites. The length of the ORF was then subdivided into 40 regions of equal length, and probes were assigned according to their nearest corresponding relative position. Probes in promoter regions were similarly assigned following subdivision into 20 regions. The average histone H3 (blue) or H4 (red) enrichment for each subdivided bin is plotted.
To test the accuracy of the new method, we identified a test set of 84 genes most likely to be targeted by Gcn4 in vivo (Figure 1B, Table S1, Experimental Procedures) and a set of 945 genes least likely to be targeted by Gcn4; the selection criteria for these sets of genes is described in Experimental Procedures. Based on these positive and negative Gcn4 targets, analysis of Gcn4 binding with the new method suggests a false-positive rate of less than 1% and a false-negative rate of ∼25%, corresponding with a total of 210 genes whose promoters are bound within the optimal p value threshold of 6 × 10−6 (Experimental Procedures). These results demonstrate that the new array and protocol modifications provide substantially higher resolution and accuracy than our previous method using self-printed arrays (Harbison et al., 2004; Lee et al., 2002).
Global Nucleosome Occupancy
The improved accuracy and resolution of this ChIP-Chip method was used to investigate nucleosome occupancy and modification throughout the yeast genome. When we examined histone occupancy with antibodies against core histone H3 or histone H4, using genomic DNA as the reference channel, we found a relatively high density of nucleosomes over transcribed regions and a lower density over intergenic regions (Figure 1C and 1D). Figure 1C shows a stereotypical example of histone occupancy at a portion of chromosome XV. Figure 1D presents composite profiles of histone H3 and H4 for 5324 genes aligned according to the location of translation initiation and termination sites. There was an ∼20% reduction in histone occupancy in intergenic sequences relative to genic sequences for the average gene. These results are consistent with previous observations (Lee et al., 2004) and suggest that the majority of yeast genes have higher nucleosome density over transcribed regions relative to intergenic regions.
We were surprised to find that differential enrichment of intergenic and genic regions also occurred in control experiments lacking antibody (compare Figure 2A and Figure 1D). Results similar to those in Figure 2A were obtained in control experiments when ChIPs were performed with antibodies directed against nonhistone proteins (data not shown). Others have noted that different relative levels of intergenic and genic DNA are recovered using various extraction strategies (Nagy et al., 2003), but control data of this type have not yet been used to normalize the results of histone ChIP studies (Bernstein et al., 2004; Bernstein et al., 2005; Kurdistani et al., 2004; Lee et al., 2004). When these control experiments were used to normalize the histone H3 data, we found that there were not substantial differences in the relative levels of intergenic versus genic DNA at the average gene (Figure 2B). Nonetheless, approximately 40% of yeast promoters do have lower levels of histones than their downstream transcribed regions, even after the normalization by control experiments (Figure S2), and we show below that these are associated with transcribed genes.
Figure 2. Comparisons of Histone Profiles
(A) A composite profile of enrichment in a control experiment. The profile is created as in Figure 1D, except that enrichment is measured from a mock immunoprecipitation, in which no antibody has been included (Experimental Procedures).
(B) A composite profile of histone occupancy normalized to a control. The profile is created as in Figure 1D, except that enrichment from H3 immunoprecipitation is normalized to enrichment from mock immunoprecipitations.
(C) A composite profile of histone occupancy according to transcriptional activity. All genes for which data were available (Holstege et al., 1998) were divided into five classes according to their transcriptional rate. Composite data were computed for H3 enrichment as in Figure 1D.
(D) A composite profile of normalized histone occupancy according to transcriptional activity. The composite profile is created as in Figure 2C, except that enrichment from H3 immunoprecipitation is normalized to enrichment from mock immunoprecipitations.
To examine the relationship between gene expression and nucleosome occupancy, we assigned genes into five different classes depending on their transcriptional rate (Holstege et al., 1998) and created a composite histone H3 profile for each class (Figure 2C and 2D). The composite histone profile in Figure 2C was generated by using whole genomic DNA in the reference channel, and that in Figure 2D was generated by normalizing to a no-antibody control ChIP. The results in both profiles confirm that nucleosome occupancy at both promoter and transcribed regions inversely correlates with gene activity, in agreement with previous gene-specific and genome-wide studies (Bernstein et al., 2004; Boeger et al., 2003; Lee et al., 2004; Reinke et al., 2003). The results shown in Figure 2D also suggest that nucleosome occupancy is reduced maximally at the promoters of active genes. In contrast, the promoters of transcriptionally inactive genes are as densely populated with nucleosomes as genic regions.
If gene activation leads to reduced nucleosome occupancy, then dynamic activation of specific genes should cause reduced histone levels at these newly transcribed genes. To test this notion, we performed ChIP-Chip with histone antibodies on cells before and after exposure to oxidative stress (Causton et al., 2001). At genes known to be activated by oxidative stress (e.g., HSP30 and HSP82), nucleosome occupancy dropped substantially (Figure S3). These results confirm that gene activation leads to reduced nucleosome density in both promoter and transcribed regions, with the greatest effect occurring at the promoter.
Histone Acetylation
The histone acetylases Gcn5 and Esa1 are generally recruited to the promoter regions of active genes (Robert et al., 2004), and thus we would expect that the amino acid residues that are substrates of these HATs would be preferentially acetylated at active genes. A recent genome-wide study, however, reported little correlation between transcriptional activity and acetylation of the histone H3 and H4 amino acid residues targeted by Gcn5 and Esa1 (Kurdistani et al., 2004). To understand the source of these discrepancies, we used the new methods to investigate selected histone modifications genome-wide.
Histone H3 lysine 9 acetylation (H3K9ac) and histone H3 lysine 14 acetylation (H3K14ac) are among the modifications catalyzed by Gcn5 (Kuo et al., 1996; Utley et al., 1998; Zhang et al., 1998). We used ChIP-Chip to measure the levels of H3K9ac relative to the levels of core histone H3 genome-wide. The results show that acetylation of histone H3 at lysine 9 peaks at the predicted transcriptional start sites of active genes (Figure 3A) and that this modification correlates with transcription rates genome-wide (Figure 3B). We also found that acetylation of histone H3 at lysine 14 peaks over the start sites of active genes (Figure 3C) and correlates with transcription rates genome-wide (Figure 3D). We conclude that there is a positive association between Gcn5, the modifications known to be catalyzed by Gcn5, and transcriptional activity (Figure 3 and Figure S4A).
Figure 3. Nucleosome Acetylation Generally Correlates with Transcriptional Activity
(A) Acetylation of H3K9 at a locus on chromosome XII. Enrichment is depicted as in Figure 1. The number beneath each gene represents the transcriptional frequency of the corresponding ORF (Holstege et al., 1998) in mRNA/hr.
(B) Composite profile of acetylation of H3K9 across the average gene. Composite profiles of acetylation according to transcriptional-frequency class are shown as in Figure 2.
(C) Acetylation of H3K14 at a locus on chromosome XII. Enrichment is depicted as in Figure 1. The number beneath each gene represents the transcriptional frequency of the corresponding ORF (Holstege et al., 1998) in mRNA/hr.
(D) Composite profile of acetylation of H3K14 across the average gene. Composite profiles of acetylation according to transcriptional-frequency class are shown as in Figure 2.
(E) Hyperacetylation of H4 at a locus on chromosome XII. Enrichment is depicted as in Figure 1. The number beneath each gene represents the transcriptional frequency of the corresponding ORF (Holstege et al., 1998) in mRNA/hr.
(F) Composite profile of hyperacetylation of H4 across the average gene. Composite profiles of acetylation according to transcriptional-frequency class are shown as in Figure 2.
Four lysine residues of histone H4 are acetylated by Esa1, an acetyltransferase associated with the NuA4 complex (Allard et al., 1999; Clarke et al., 1999; Vogelauer et al., 2000). We measured the levels of hyperacetylated histone H4 relative to core histone genome-wide using ChIP-Chip with an antibody that recognizes histone H4 acetylated at lysines 5, 8, 12, and 16 (H4K5ac8ac12ac16ac). The results showed that H4 hyperacetylation peaks over the start sites of active genes (Figure 3E) and correlates with transcription rates (Figure 3F), although the association is not as strong as that observed for H3K9ac and H3K14ac. Our analysis cannot exclude the possibility that acetylation of individual lysine residues in the N-terminal tail of histone H4 might correlate differently with transcriptional activity. Nonetheless, our data reveal a positive, albeit modest, correlation between Esa1 occupancy, the modifications known to be catalyzed by this enzyme, and transcriptional activity (Figure 3 and Figure S4B).
To ascertain whether dynamic gene activation leads to the expected increase in histone acetylation at sites catalyzed by Gcn5 and Esa1, we performed ChIP-Chip with the relevant histone antibody on cells before and after exposure to oxidative stress. The results confirm that gene activation leads to increased histone acetylation at sites catalyzed by Gcn5 and Esa1 in the promoter and transcribed regions of activated genes (Figure S5).
In general, we find that histones with the acetylated residues studied here are enriched predominantly at promoter regions and transcriptional start sites of active genes and that enrichment drops substantially across the ORFs (Figure 3 and Figure S5). This is consistent with the model that transcriptional activators generally recruit Gcn5 and Esa1 to promoters of genes upon their activation (Robert et al., 2004) and with the idea that the two HATs acetylate local nucleosomes when recruited to these genes. Our conclusion that there is a strong correlation between transcriptional activity and acetylation of the histone H3 and H4 amino acid residues targeted by Gcn5 and Esa1 is in contrast to that of Kurdistani et al., 2004. This discrepancy is most likely due to differences in the material used in the control channel in the ChIP-Chip procedure. The experiments described here compare ChIP with a histone-modification antibody to a control ChIP with a core histone antibody. The experiments reported in Kurdistani et al., 2004 used whole genomic DNA in the reference channel. We found we could replicate the results in Kurdistani et al., 2004 if we used whole genomic DNA as a reference in ChIP-Chip experiments (Figure S6), but for reasons described above, this method of normalization is inappropriate.
Histone Methylation
Methylation of histones in S. cerevisiae is carried out by three known histone methyltransferases, which are capable of covalently modifying specific lysine residues in histone H3 with up to three methyl groups (Peterson et al., 2004). Since characterization of specific methylation marks has been studied primarily at the level of individual genes (Bannister et al., 2005; Bernstein et al., 2002; Krogan et al., 2003; Ng et al., 2003; Xiao et al., 2003), we sought to systematically profile mono-, di-, and trimethylated residues at K4, K36, and K79 of histone H3 in nucleosomes associated with genomic DNA.
The histone methyltransferase Set1 has previously been shown to be recruited to the 5′ end of actively transcribed genes where it is responsible for histone H3K4 methylation (Bernstein et al., 2002; Briggs et al., 2001; Krogan et al., 2003; Ng et al., 2003; Santos-Rosa et al., 2002). We measured histone H3K4 trimethylation (H3K4me3) using ChIP-Chip and found that the results confirm previous studies (Santos-Rosa et al., 2002) and provide a higher-resolution picture of H3K4 trimethylation across the yeast genome (Figure 4A and 4B). Peaks of histone H3K4 trimethylation occurred at the beginning of actively transcribed genes, and there was a positive correlation between this modification and transcription rates (Figure 4A and 4B).
Figure 4. Nucleosome Methylation Generally Correlates with Transcriptional Activity
(A) Trimethylation of H3K4 at a locus on chromosome XII. Enrichment is depicted as in Figure 1. The number beneath each gene represents the transcriptional frequency of the corresponding ORF (Holstege et al., 1998) in mRNA/hr.
(B) Composite profile of trimethylation of H3K4 across the average gene. Composite profiles of methylation according to transcriptional-frequency class are shown as in Figure 2.
(C) Trimethylation of H3K36 at a locus on chromosome XII. Enrichment is depicted as in Figure 1. The number beneath each gene represents the transcriptional frequency of the corresponding ORF (Holstege et al., 1998) in mRNA/hr.
(D) Composite profile of trimethylation of H3K36 across the average gene. Composite profiles of methylation according to transcriptional-frequency class are shown as in Figure 2.
(E) Trimethylation of H3K79 at a locus on chromosome XII. Enrichment is depicted as in Figure 1. The number beneath each gene represents the transcriptional frequency of the corresponding ORF (Holstege et al., 1998) in mRNA/hr.
(F) Composite profile of trimethylation of H3K79 across the average gene. Composite profiles of methylation according to transcriptional-frequency class are shown as in Figure 2.
We also investigated the profiles of mono- and dimethylated histone H3K4-containing nucleosomes and found that they exhibit a pattern distinct from that observed for trimethylated histone H3K4 (Figure S7). While trimethylated H3K4 peaks at the beginning of the transcribed portions of genes, dimethylated H3K4 (H3K4me2) is most enriched in the middle of genes, and monomethylated H3K4 (H3K4me) is found predominantly at the end of genes.
We measured genome-wide the relative levels of H3K36 trimethylation, which is catalyzed by Set2, a factor associated with the later stages of transcriptional elongation (Strahl et al., 2002). In contrast to the pattern observed with H3K4 trimethylated histones, we found that trimethylated H3K36 (H3K36me3) was enriched throughout the coding region, peaking near the 3′ ends of transcription units (Figure 4C and 4D). H3K36 trimethylation also correlated with transcriptional activity. These results are consistent with the model that Set2 is recruited by the transcription elongation apparatus and that it methylates local nucleosomes during active transcription.
The Dot1 histone methyltransferase modifies histone H3 lysine 79 (H3K79), which occurs within the core domain of histone H3 (Feng et al., 2002; Ng et al., 2002; Ng et al., 2003). Methylation of this residue is estimated to occur in ∼90% of all histones and is associated with telomeric silencing control in yeast (Ng et al., 2003; van Leeuwen et al., 2002), but global ChIP profiles of dimethylated H3K79 in Drosophila (Schubeler et al., 2004) have linked this modification to active transcription. We investigated the genomic profile of H3K79 trimethylation in yeast (H3K79me3) and found that histones with this modification are enriched within the transcribed regions of genes (Figure 4E and 4F). Most genes appeared to have nucleosomes modified at H3K79; there was little correlation between the relative levels of H3K79 trimethylation at genes and transcriptional activity (Figure 4F).
Global Map of Histone Marks
We recently mapped the locations of conserved transcription-factor binding sites throughout the yeast genome (Harbison et al., 2004). We used the results described here to generate a complementary genome-wide map of nucleosome occupancy and histone modifications that includes results for eight sets of histone modifications (H3K9ac, H3K14ac, H4K5ac8ac12ac16ac, H3K4me1, H3K4me2, H3K4me3, H3K36me3, and H3K79me3). A portion of this map is shown in Figure 5, and a browsable form of the complete yeast-genome chromatin map is available at the authors’ website (http://web.wi.mit.edu/young/nucleosome).
Figure 5. A High-Resolution Genome-wide Map of Nucleosome States
A map for a region on chromosome IV is depicted as in Figure 1. Conserved binding sites for transcriptional regulators (Harbison et al., 2004) are depicted as colored boxes. Numbers beneath genes represent transcriptional activity (mRNA/hr). Enrichment values from acetylated H3K14, trimethylated H3K4, and histone H3 are depicted in red, green, and black, respectively.
Concluding Remarks
It is well established that nucleosomes play fundamentally important roles in the organization and maintenance of the genome. Nucleosome modifications have been shown to be associated with transcriptional regulation at well-studied genes, and models have emerged that connect regulation of gene expression to histone modification by specific chromatin regulators (Cosma et al., 1999; Gregory et al., 1999; Kuo et al., 1998; Reinke et al., 2003). We have carried out a systematic genome-wide analysis of nucleosome acetylation and methylation at sufficient resolution to determine whether models that connect regulation of gene expression to histone modification (Deckert et al., 2001; Reid et al., 2000; Reinke et al., 2001; Suka et al., 2002) apply to gene regulation throughout the yeast genome.
The results described here, taken together with recent discoveries, are consistent with the following general model connecting gene expression to histone modification. Transcriptional activation by DNA binding regulators generally involves recruitment of Gcn5 and Esa1 to promoters, where these HATs acetylate specific residues on histones H3 and H4 at local nucleosomes (Figure 3) (Bhaumik et al., 2001; Cosma et al., 1999; Larschan et al., 2001; Reid et al., 2000). We were able to find few exceptions to this general rule, where only one or the other HAT acetylates its target residues at the promoters of actively transcribed genes (data not shown). Active transcription is characteristically accompanied by histone H3K4 trimethylation by Set1 at the beginning of genes (Figure 4B) and by H3K4 dimethylation and monomethylation at nucleosomes positioned further downstream in the transcription unit (Figure S7) (Krogan et al., 2003; Ng et al., 2003). As the transcription apparatus proceeds down the transcription unit, increasing levels of histone H3K36 trimethylation are observed at most active genes, catalyzed by Set2 (Figure 4D) (Krogan et al., 2003; Strahl et al., 2002). Histone H3K79me3, which is catalyzed by Dot1 (Feng et al., 2002; Ng et al., 2003; van Leeuwen et al., 2002), is enriched within genes, but, unlike the other modifications studied here, this enrichment is not clearly associated with active transcription (Figure 4F). Correlations between transcriptional activity and histone occupancy or modification at intergenic and transcribed regions are summarized in Table S2.
The genome-wide maps of histone occupancy and modification described here should provide investigators with information useful for further exploring the histone code and its implications for gene regulation and chromosome organization and maintenance. We expect that the approaches used here to map histone occupancy and modification in yeast can also be used to gain insights into the linkage between gene expression and histone modification across the genome in higher eukaryotes.
小知识：
Acetylation level, DNA accessibility, transcription
Hypoacetylation (Blue)
Strong internucleosomal interactions: histone tails constrain wrapping of DNA on nucleosome surface
Hyperacetylation (Yellow)
Weak internucleosomal interactions: histone tails do not constrain DNA, which is accessible to transcription factors
