Combined Action of PHD and Chromo Domains Directs the Rpd3S HDAC to Transcribed Chromatin

See allHide authors and affiliations

Science  18 May 2007:
Vol. 316, Issue 5827, pp. 1050-1054
DOI: 10.1126/science.1139004


Nucleosomes must be deacetylated behind elongating RNA polymerase II to prevent cryptic initiation of transcription within the coding region. RNA polymerase II signals for deacetylation through the methylation of histone H3 lysine 36 (H3K36), which provides the recruitment signal for the Rpd3S histone deacetylase complex (HDAC). The recognition of methyl H3K36 by Rpd3S requires the chromodomain of its Eaf3 subunit. Paradoxically, Eaf3 is also a subunit of the NuA4 acetyltransferase complex, yet NuA4 does not recognize methyl H3K36 nucleosomes. In Saccharomyces cerevisiae, we found that methyl H3K36 nucleosome recognition by Rpd3S also requires the plant homeobox domain (PHD) of its Rco1 subunit. Thus, the coupled chromo and PHD domains of Rpd3S specify recognition of the methyl H3K36 mark, demonstrating the first combinatorial domain requirement within a protein complex to read a specific histone code.

Histone modifications are important for almost all DNA-related processes, but studies of the role of histone methylation in transcription regulation have become a major focus in the field (1). Methylated lysines K4me, K36me, and K79me are enriched around regions of active transcription (2, 3). These methylation marks do not simply facilitate transcription, because K4me does not affect transcription per se in vitro (4). Moreover, for the majority of the yeast genome, transcription occurs normally in the absence of K4me, K36me, or K79me, whereas methylation appears to be dependent on active transcription (5). These observations imply that methylation acts in maintaining the architecture of transcribed-chromatin templates, rather than directly facilitating transcription. Consistent with this hypothesis, recent studies demonstrate that K36me is recognized by the chromodomain of its Eaf3 subunit (CHDEaf3) within Rpd3S, thereby tethering Rpd3S to the coding region of actively transcribed genes. Once targeted, Rpd3S creates a hypoacetylated state, which in turn suppresses transcription initiated within the body of the gene (68).

Many chromatin-related complexes contain multiple domains that can recognize specific histone marks, but their contributions to the specificity and function of the complexes remain elusive. We discovered a critical role of the plant homeobox domain of the Rco1 subunit (PHDRco1) in Rpd3S targeting. Our data suggest that CHDEaf3 and PHDRco1 contribute combinatorially to the overall affinity and specificity of Rpd3S for its nucleosomal targets, and both domains are essential for Rpd3S-mediated control of global-acetylation levels at transcribed chromatin in vivo.

Previous studies have demonstrated that the CHDEaf3 in Rpd3S preferentially binds to K36me2 histone peptides (6, 7). We wanted to further test Rpd3S binding in a nucleosomal context. To measure the binding of Rpd3S to modified nucleosomes, we developed an assay. Reconstituted recombinant nucleosomes were immobilized on magnetic beads, sequentially modified, and then washed to remove the modifying enzymes. The resulting nucleosomes were released by restriction digestion, as illustrated in Fig. 1A. We examined the following combinations of modifications: mock-modified, acetylated by Spt-Ada-Gcn5-acetyltransferase and NuA4, methylated at H3K36 by the recombinant hSet2 (Fig. 1, B and C), and both acetylated and methylated. We found that Rpd3S bound to methylated nucleosomes with a higher affinity than to either unmodified or acetylated nucleosomes, and acetylation further enhanced the binding of Rpd3S to the methylated nucleosomes (Fig. 1D). However, Rpd3S was unable to bind the 147–base pair nucleosome lacking linker DNA, even when the appropriate modifications were present (Fig. 1E). The recognition of K36me by Rpd3S is specific because H3K79me does not stimulate Rpd3S nucleosomal binding (fig. S3).

Fig. 1.

Rpd3S preferentially binds to nucleosomes methylated at K36. (A) Outline of the experimental strategy, using immobilized templates to make covalently modified nucleosomes. O/N, overnight; CoA, coenzyme A; RT, room temperature; SAM, S-adenosylmethionine. (B and C) hSet2 and rDot1 specifically methylate recombinant nucleosomes at the desired sites. Error bars in (B) indicate SD. HMT, histone methyltransferase; HYPB SET, the SET domain of HYPB; LON, long oligo-nucleosomes. (D) Rpd3S preferentially binds to K36-methylated nucleosomes, and acetylation further enhances its association. -, unmodified; Ac, acetylated; Me, methylated; M/A, methylated and acetylated; TSA, trichostatin A. (E) The linker DNAis required for Rpd3S binding to the hyper-methylated and hyperacetylated nucleosomes.

CHDEaf3 plays a pivotal role in recruiting Rpd3S to chromatin, both in vitro and in vivo (68). We constructed a mutant Rpd3S complex in which CHDEaf3 was deleted (eaf3Δchd). The wild-type (WT) and mutant complexes were purified by tandem affinity purification (TAP)–tagged Rco1. Based on silver-stained gels, the deletion of CHDEaf3 did not affect the integrity of Rpd3S (Fig. 2A). This deletion did, however, substantially reduce the affinity of Rpd3S for nucleosomes (Fig. 2B), as only weak binding of the mutant Rpd3S was detected at concentrations 12 to 24 times that of the WT complex. More importantly, the mutant complex no longer discriminated between methylated and unmethylated nucleosomes. These data indicated that CHDEaf3 not only contributes to the overall affinity of Rpd3S for nucleosomes, but also determines the specificity of Rpd3S binding to K36-methylated nucleosomes.

Fig. 2.

CHDEaf3is important for Rpd3S nucleosome-binding specificity. (A) Deletion of CHDEaf3 does not affect the integrity of the Rpd3S complex. The WT (left) and eaf3Δchd (right) forms of the Rpd3S complex were purified through the Rco1-TAP tag and visualized by silver staining. CBP, calmodulin binding protein. (B) Deletion of the chromodomain reduces overall affinity of Rpd3S for nucleosomes and abolishes its specificity toward K36-methylated nucleosomes.

Eaf3 is also a component of the NuA4 histone acetyltransferase complex. We wished to find out if CHDEaf3 was important in the binding of NuA4 to nucleosomes, in a manner similar to its role in Rpd3S. When purified, NuA4 (Epl1-TAP) and Rpd3S (Rco1-TAP) were directly compared by electrophoretic mobility shift assay (EMSA); NuA4 did not associate with nucleosomes under conditions where Rpd3S bound (Fig. 3A). These data suggest that the Eaf3 subunit, in the context of NuA4, does not support stable binding to nucleosomal substrates.

Fig. 3.

The PHD domain of Rco1 is required for normal function of Rpd3S. (A) Eaf3 plays different roles in NuA4 and Rpd3S. (B) PHDRco1 and CHDEaf3 are essential for regulating global acetylation at ORFs. WT yeast and strains bearing eaf3Δchd (YBL619), rco1Δphd (YBL632), or set2Δ mutations were subjected to ChIP-chip analysis on a high-resolution tiling microarray platform (Agilent Technologies, Santa Clara, CA) with the use of AcH4. The enrichment values were first calculated using the log 2 ratio of immunoprecipitated (IP) versus input. Subsequently, the log 2 ratio for the enrichment of acetylation of H4 (AcH4) in the mutant versus the enrichment of AcH4 in the wild type were pipelined into a modified average-gene analysis (3). The averages of the entire genome are plotted against the relative position of the 5′ intergenic region (IGR), ORF, and 3′ IGR in an average gene. The line in the graph was created by using the R function lowess to smooth the average values. (C) PHDRco1 is required for suppression of spurious transcription. Northern blot analysis of yeast strains grown exponentially was performed using probes against STE11. FL, full length. (D) PHDRco1 is required for Rpd3 nucleosome binding. Silver staining of the mutant Rpd3S complex (rco1Δphd) is shown (left). The amount of complex used here was normalized, based on silver staining and Western blot with an antibody to CBP (fig. S7).

We considered the possibility that Eaf3 might work cooperatively with another subunit of Rpd3S to enhance its specificity for K36me. PHDRco1 represents a potential chromatin-binding domain within the Rpd3S complex. The association of Rco1 with Rpd3S is mutually dependent on Eaf3, making both proteins essential for Rpd3S function (6). To evaluate the role of PHDRco1 in Rpd3S function, we tested whether the mutant bearing a deletion of the PHD domain alone (rco1Δphd) caused a phenotype similar to that of deletion of the entire Rpd3S complex. We employed two different approaches to address this question.

Earlier studies demonstrated that defects in the Set2-Rpd3S pathway led to hyper-acetylation of coding regions at selected yeast genes (68). To determine whether this observation applied genomewide, we performed chromatin immunoprecipitation coupled with microarrays (ChIP-chip) experiments on a high-density tiling microarray platform with the use of an antibody against acetylated histone H4 (AcH4). The enrichment of AcH4 in set2Δ and the wild type was directly compared, and the resulting log 2 ratios were subjected to a modified average-gene analysis (3). This analysis allowed us to evaluate the average distribution profile of acetylation changes genomewide. The resulting profile revealed a peak within the 3′ portion of coding regions across the entire genome (Fig. 3B). Remarkably, using a similar ChIP-chip approach, we found that removal of either eaf3Δchd or rco1Δphd resulted in global-acetylation changes similar to those seen in set2Δ (Fig. 3B). Therefore, CHDEaf3 and PHDRco1 were essential for Rpd3S to regulate global acetylation at open reading frames (ORFs). Second, defects in the Set2-Rpd3S pathway result in the generation of aberrant internally initiated transcripts at the STE11 locus (6). Thus, we performed Northern blot analysis with probes against the 5′ and 3′ portion of STE11. Deletion of PHDRco1 resulted in the appearance of spurious transcripts, similar to those seen in set2Δ (Fig. 3C). Collectively, these results suggest that PHDRco1 is required for Rpd3S function in vivo.

To further dissect the molecular function of PHDRco1 in Rpd3S, we purified the rco1Δphd mutant Rpd3S complex. Although the deletion of PHDRco1 does not disrupt the integrity of the complex, based on silver-stained SDS–polyacrylamide gel electrophoresis (PAGE) (Fig. 3D, left), it completely abrogates the binding of Rpd3S to nucleosomes (Fig. 3D). This result indicates that one essential role for PHDRco1 is to enhance the overall affinity of Rpd3S for nucleosomes, whereas CHDEaf3 provides specificity for H3K36-methylated nucleosomes.

The inhibitor-of-growth (ING) family of proteins contains PHD domains (PHDING) that preferentially bind to H3K4me3 (9). We found that PHDRco1 is structurally different from PHDING, and it does not recognize methylated H3K4me (fig. S10). Because functional Rpd3S requires both PHDRco1 and CHDEaf3, we hypothesized that Rpd3S nucleosome-binding affinity might be dictated by the combined activities of these two domains. NuA4 contained CHDEaf3 but paired with a structurally different PHDYng2. This combination might not be suitable for persistent chromatin binding. To test this theory, we performed an experiment in which PHD domains of Yng2 and Rco1 were swapped (Fig. 4).

Fig. 4.

Specific domain combinations determine the nucleosomal binding of chromatin modifying complexes. (A and B) Silver staining of WT and indicated mutant complexes (left). EMSA (right). Red arrowheads in (B) indicate mutant Yng2. (C) Cross-linked extracts from the wild type and Yng2-rco1PHD (YBL677) were immunoprecipitated with rabbit–immunoglobulin G (Epl1-TAP). Purified DNA was quantified using real-time polymerase chain reaction with the indicated primer sets. Error bars indicate SD. (D) RNA from ΔchdΔphd (YBL694) and Yng2rco1PHD (YBL677) were included in the Northern blot analysis using STE11 probes.

First, PHDRco1 was replaced by PHDYng2 at the genomic locus, and this mutant Rpd3S was purified though a TAP-tagged Rco1. Although the chimeric Rco1 was stably incorporated into Rpd3S (Fig. 4A, left), the mutant complex was not able to bind nucleosomes (Fig. 4A, right), similar to the rco1Δphd mutant (Fig. 3D). This result suggests that the combination of PHDYng2 and CHDEaf3 does not direct nucleosome binding, even in the context of Rpd3S.

Next, we purified mutant NuA4 complexes in which PHDYng2 was replaced by PHDRco1 at the genomic locus. The fusion Yng2 was stably incorporated into NuA4 and did not change the stability of the complex (Fig. 4B, left). The Yng2-Rco1PHD protein increased the overall affinity of the NuA4 complex for nucleosomes (Fig. 4B, right). This increase in binding was considered notable, because the unbound fraction of nucleosomes was clearly reduced after the addition of Yng2-Rco1PHD–containing NuA4. Using a ChIP assay, we further demonstrated that this altered form of NuA4 is redirected to the coding region of STE11 (Fig. 4C). More importantly, this mistargeting of the mutant NuA4 onto the ORF results in the appearance of cryptic transcripts, even in the presence of WT Rpd3S (Fig. 4D, lanes 6 and 12). These results strongly support the notion that combinatory actions of CHDEaf3 and PHDRco1 determine the targeting of the chromatin-related complexes to H3K36me-enriched regions

Since the histone-code hypothesis was proposed (10), many domains within chromatin-related complexes have been discovered that recognize specific histone modifications. Complexes with different (sometimes even opposite) functions contain identical domains within shared subunits (5). Given this apparent overlap in domain usage, cells must have a system to ensure that the correct complex is recruited to specifically modified nucleosomes. Our study suggests that the combination of multiple domains within a chromatin-related complex is important for interpretation of the histone code. In Rpd3S, PHDRco1 does not bind to naked DNA (data not shown in the figures) nor to the histone K4me mark (fig. S10), yet it is required for robust nucleosome binding. As for CHDEaf3, it is essential for recognition of the K36-methylation mark (Fig. 2B) (68). Moreover, CHDEaf3 contributes to the overall affinity of the complex for nucleosomes because a higher concentration of the mutant eaf3Δchd complex is required to achieve basal binding. Thus, the binding of PHDRco1 to nucleosomes may anchor Rpd3S in a configuration that allows CHDEaf3 to recognize K36me. Without PHDRco1, CHDEaf3 might not be positioned properly or have enough affinity, thereby failing to support nucleosome binding (Fig. 3D). The deletion of both domains affects the integrity of Rpd3S and results in a stronger spurious-transcription phenotype (Fig. 4D and fig. S9). Furthermore, the substitution of PHDYng2 with PHDRco1 can increase the affinity of NuA4 for nucleosomes both in vitro and in vivo (Fig. 4). This result strongly suggests that the specific combination of CHDEaf3 and PHDRco1 in the context of either Rpd3S or NuA4 directs robust nucleosome binding. Therefore, our study presents a mechanism for substrate recognition by a chromatin-related complex. Future studies into how combinations of recognition domains affect the complexity of these enzymes under different physiological conditions are of great interest.

Supporting Online Material

Materials and Methods

Figs. S1 to S11

Tables S1 and S2


References and Notes

Stay Connected to Science

Navigate This Article