The NuA4 histone acetyltransferase complex is required for gene regulation, cell cycle progression, and DNA repair. Dissection of the 13‐subunit complex reveals that the Eaf7 subunit bridges Eaf5 with Eaf3, a H3K36me3‐binding chromodomain protein, and this Eaf5/7/3 trimer is anchored to NuA4 through Eaf5. This trimeric subcomplex represents a functional module, and a large portion exists in a native form outside the NuA4 complex. Gene‐specific and genome‐wide location analyses indicate that Eaf5/7/3 correlates with transcription activity and is enriched over the coding region. In agreement with a role in transcription elongation, the Eaf5/7/3 trimer interacts with phosphorylated RNA polymerase II and helps its progression. Loss of Eaf5/7/3 partially suppresses intragenic cryptic transcription arising in set2 mutants, supporting a role in nucleosome destabilization. On the other hand, loss of the trimer leads to an increase of replication‐independent histone exchange over the coding region of transcribed genes. Taken together, these results lead to a model where Eaf5/7/3 associates with elongating polymerase to promote the disruption of nucleosomes in its path, but also their refolding in its wake.
The multisubunit NuA4 complex is an essential gene regulatory acetyltransferase in yeast. Dissection of its subunits reveals a trimeric submodule independently associating with gene coding regions, which may act as a histone chaperone during RNA polymerase II transcriptional elongation.
Eaf5/7/3 forms a submodule within the NuA4 acetyltransferase complex and also functions as an independent native trimer.
High‐resolution genome‐wide analysis reveals NuA4 binding to promoters and coding regions of actively transcribed genes.
The Eaf5/7/3 trimer independently associates with the coding region of transcriptionally active genes and affects RNA polymerase II processivity.
Eaf5/7/3 interacts with phosphorylated, elongating RNA polymerase II and H3K36me3‐containing nucleosomes.
Eaf5/7/3 cooperates with FACT to promote transcription‐coupled nucleosomes disruption, as well as their refolding behind the polymerase to suppress histone exchange.
Specific post‐translational modifications of histones within chromatin play critical roles in nuclear functions. Evolutionarily conserved acetylation of histone N‐terminal domains was the first covalent modification to be shown to be a key regulator of a nuclear process, that is, transcription. Multiple other histone modifications such as phosphorylation, methylation, and ubiquitination have also since emerged as important players in transcription regulation, defining the chromatin state (Bannister & Kouzarides, 2011; Zentner & Henikoff, 2013). Methylation of histone H3 on lysines 4, 36, and 79 has been linked to the elongation of transcription (Shilatifard, 2006). While some modifications have the potential of directly modulating chromatin structure, they can also be recognized as specific marks by protein modules within regulators (Yun et al, 2011; Musselman et al, 2012). Classical examples of these protein module–histone mark interactions are the bromodomains with specific acetylated lysines and chromodomains (CHDs) with methylated lysines. These interactions are believed to play an important role for retention/accumulation of specific regulatory factors at chromosomal loci. Furthermore, cross talk between different histone modifications has been demonstrated (Suganuma & Workman, 2011).
Nucleosome acetyltransferase of H4 (NuA4) is a large multisubunit complex that acetylates histone H2A, H2A.Z and H4 N‐termini in chromatin (Allard et al, 1999; Boudreault et al, 2003; Babiarz et al, 2006; Keogh et al, 2006). Its catalytic subunit Esa1 is the only essential histone acetyltransferase (HAT) in yeast (Smith et al, 1998; Clarke et al, 1999). Initial purifications of the NuA4 complex showed 13 stable subunits including ATM‐related factor Tra1 (shared with the SAGA complex), enhancer of polycomb homolog Epl1, actin‐related protein Arp4, leukemogenic factor ENL/AF9 homolog Yaf9 and initially 7 Esa1‐associated factors, Eaf1‐7, two of which now bear the names Yng2 (Eaf4) and Swc4 (Eaf2) (Doyon & Cote, 2004; Altaf et al, 2009). NuA4 has been implicated in the regulation of gene transcription through local recruitment by DNA‐bound transcription factors (Utley et al, 1998; Reid et al, 2000; Rohde & Cardenas, 2003; Nourani et al, 2004; Mitchell et al, 2008; Joo et al, 2011) or by histone marks and other regulators (Morillon et al, 2005; Ginsburg et al, 2009; Uprety et al, 2012). But NuA4 function is not only restricted to transcription as it was also shown to be important for the cellular response to DNA damage and to play a direct role for the repair of DNA double‐strand breaks (Bird et al, 2002; Choy & Kron, 2002; Boudreault et al, 2003; Downs et al, 2004; Lin et al, 2008). NuA4 is truly multifunctional as it was also proposed to play a role in chromosome segregation and establishment of chromatin boundaries (Le Masson et al, 2003; Krogan et al, 2004; Marston et al, 2004; Zhang et al, 2004; Babiarz et al, 2006; Zhou et al, 2010; Mitchell et al, 2011) and to regulate cellular life span and autophagy through acetylation of non‐histone substrates (Lin et al, 2009; Lu et al, 2011; Yi et al, 2012).
It is interesting to note that NuA4 subunits contain several protein domains typical of activities interacting with chromatin, for example, PHD finger, SANT, YEATS, actin‐related, and CHDs (Doyon & Cote, 2004) . In fact, NuA4 contains 2 of the 3 yeast proteins that harbor CHDs. In particular, the Eaf3 CHD protein has been shown to regulate the global pattern of histone acetylation in vivo, keeping coding regions of long genes at low levels of acetylation, while promoter regions contain more acetylated histones (Reid et al, 2004). It is also part of the Rpd3S histone deacetylase complex that binds the coding region of genes to stabilize nucleosomes in the wake of the RNA polymerase II (RNAPII) and suppresses intragenic cryptic/spurious transcription (Carrozza et al, 2005; Joshi & Struhl, 2005; Keogh et al, 2005; Li et al, 2007c). This binding involves an interaction between Set2‐dependent H3K36 methylation and Eaf3 chromodomain (Li et al, 2007b; Sun et al, 2008; Xu et al, 2008). Rpd3S, like Set2, follows elongating RNAPII through an interaction with its C‐terminal domain (Drouin et al, 2010; Govind et al, 2010). Competition between Eaf3‐containing HAT (NuA4) and HDAC (Rpd3S) complexes is thought to occur during transcription (Biswas et al, 2008), and NuA4‐dependent acetylation of chromatin has been suggested to help chromatin remodeling during transcription elongation (Carey et al, 2006; Ginsburg et al, 2009).
Recent work has begun to identify functional modules/subcomplexes within the large NuA4 complex, with a central role of Eaf1 as assembling platform (Auger et al, 2008; Mitchell et al, 2008). For example, picNuA4 is formed of Esa1, Epl1, and Yng2 and has been implicated in global non‐targeted acetylation of chromatin (Boudreault et al, 2003; Friis et al, 2009). Tra1 is an interface for interaction with activators and is shared with the SAGA HAT complex (Brown et al, 2001). Arp4 and Act1 are also found in the INO80 chromatin remodeling complex involved in DNA repair, while Swc4, Yaf9, Arp4, and Act1 are also part of the SWR1 complex (SWR1‐C), which is responsible for incorporation of the H2AZ histone variant in euchromatin (Shen et al, 2000; Krogan et al, 2003; Kobor et al, 2004; Mizuguchi et al, 2004; Zhang et al, 2004; Auger et al, 2008). We showed that NuA4 is conserved in higher eukaryotes and corresponds to the human Tip60 complex (Doyon et al, 2004). While homologs of 12 of the 13 yeast NuA4 subunits are present in the human complex, additional human subunits led us to propose that the Tip60 complex is a physical merge of two yeast complexes, NuA4 and SWR1‐C, through a fusion of the Eaf1 and Swr1 proteins into p400 in mammals (Doyon & Cote, 2004; Auger et al, 2008; Altaf et al, 2009). The two separate yeast activities have been shown to have close functional ties, primarily converging on the biology of histone variant H2A.Z (Kobor et al, 2004; Babiarz et al, 2006; Keogh et al, 2006; Millar et al, 2006; Durant & Pugh, 2007; Auger et al, 2008; Kim et al, 2009; Altaf et al, 2010; Zhou et al, 2010).
In order to define specific function(s) of subunits in the NuA4 complex, we pursued genetic and biochemical analysis of a group of poorly characterized Esa1‐associated factors. Here, we demonstrate that Eaf7 associates with Eaf5 and the CHD‐containing protein Eaf3 and that this trimer defines a new functional subcomplex in NuA4. Our data indicate that this trimer also exists independently of NuA4 in the cell and is found on the coding regions of actively transcribed genes. Even though it is functionally linked to the Set2 histone H3K36 methyltransferase, its genome‐wide localization appears independent of Eaf3 and physically linked to elongating RNAPII. We demonstrate that loss of the trimer affects Pol II kinetics, can partially suppress intragenic spurious transcription arising in set2 mutant cells, and leads to increased histone exchange on a transcribed coding region. Altogether, our results identify a new functional trimeric protein complex important for the transcription elongation process that plays a role in nucleosome destabilization and recycling.
Purification of Eaf5, Eaf6, and Eaf7 shows that they are stable subunits of the NuA4 HAT complex
When we first affinity purified the yeast NuA4 HAT complex to homogeneity, we identified 6 uncharacterized yeast proteins, named Esa1‐associated factors 1–6 (Eaf1‐6) (Galarneau et al, 2000; Eisen et al, 2001; Nourani et al, 2001). Neither Eaf5 nor Eaf6 had known domains or obvious phenotypes. No clear homolog of Eaf5 is found in higher eukaryotes, while Eaf6 has a homologous protein in the human Tip60/NuA4 complex (Doyon & Cote, 2004; Doyon et al, 2004), HBO1, and MOZ/MORF HAT complexes (Doyon et al, 2006). Later on, using the tandem affinity purification (TAP) approach, we purified a large amount of NuA4 complex using the subunit Epl1 (Boudreault et al, 2003; Auger et al, 2008; see Fig 1A, lane 1), and a thirteenth subunit, Eaf7, was identified based on high peptide coverage obtained by mass spectrometry, similar to the other subunits (48 peptides, 62% of protein length). This protein, which shows no special features aside from being highly charged, was also identified in NuA4 purifications by other groups (Krogan et al, 2004; Mitchell et al, 2008). Eaf7 was likely initially missed because its highly charged nature (pI 5.0; predicted molecular weight 49 kDa) generates a smeary band that varies from one gel to another and stains poorly (Fig 1A and Supplementary Fig S1B, C and D). The use of Eaf5 instead of Epl1 as a bait in TAP purification not only confirms the stable association of Eaf5 itself with the NuA4 complex but also the protein band corresponding to Eaf7 (Fig 1A, lane 2). Interestingly, Eaf5, Eaf7, and the chromodomain protein Eaf3 are recovered at higher levels than the other NuA4 subunits (compare lane 1 and 2). This is not the case when Eaf6 is TAP‐tagged, which allows the purification of the usual set and intensities of NuA4 components (lane 3). The stable association of Eaf7 with NuA4 was confirmed by directly tagging and purifying the protein (Fig 1A, lane 4). Eaf7 is clearly a stable subunit of NuA4 as it copurifies with the same set of polypeptides (note that the relative intensities are similar to the ones obtained with Eaf5‐TAP). Histone acetyltransferase activity of the Eaf7‐purified complex was also analyzed and is again identical to NuA4 complexes purified through other tags or methods (Supplementary Fig S1A).
Eaf5 anchors Eaf7/Eaf3 to the rest of the NuA4 complex
In order to understand the role of Eaf7 in the structure of NuA4, we purified the complex from cells lacking the EAF7 gene. The Western blot analysis (Fig 1B) and the silver‐stained gel (Supplementary Fig S1B) clearly show that the loss of Eaf7 subunit in NuA4 does not disrupt its assembly since 11 of the original subunits are still present. On the other hand, the data indicate that association of the Eaf3 chromodomain subunit is completely lost in the absence of Eaf7. We also used the Eaf5‐TAP‐expressing strain for NuA4 purification and found similar results as with the Epl1‐TAP strain (Fig 1C and Supplementary Fig S1C). In eaf7Δ cells, NuA4 loses subunits Eaf3 and Eaf7, while purification of Eaf5‐TAP from an eaf3Δ strain shows apparently only loss of Eaf3 (lane 2). No definitive answer is obtained for Eaf7 in the latter case though the smeary signal on gel appears unaffected (Supplementary Fig S1C, lane 2). To pursue Eaf7 interacting partners within NuA4, we used Eaf7‐TAP‐expressing cells (Fig 1D and Supplementary Fig S1D). While Eaf7‐TAP purification yields the full native NuA4 complex, deletion of EAF3 does not disrupt this purification as all NuA4 subunits are recovered, except Eaf3 (lane 2). This indicates that Eaf3 is not required for Eaf7 association to NuA4, while conversely Eaf7 is necessary for Eaf3 (Fig 1B, lane 2). Interestingly, when EAF5 is deleted, all NuA4 subunits are lost in the Eaf7‐TAP purification, except Eaf3 (Fig 1D, lane 3). This indicates that Eaf7 is not associated with NuA4 in the absence of Eaf5 but is still bound to Eaf3. Finally, we purified Eaf7 from a strain that contains a truncated version of NuA4 subunit Epl1. In these cells, the catalytic HAT subcomplex of NuA4, picNuA4, is separated from the rest of the complex, allowing only global non‐targeted acetylation of H4/H2A in chromatin (Boudreault et al, 2003). Eaf7‐TAP purification clearly shows that Eaf7 is still associated with Eaf3 and Eaf5 but not with picNuA4 components (Esa1, Epl1, Yng2) (Fig 1D, lane 4 and Supplementary Fig S1D), which is in clear contrast to purification of Eaf6 in the same background (Fig 1E and Supplementary Fig S1G). Western blot analysis suggests that Eaf3/5/7 are not associated with the rest of NuA4 either, but closer analysis of the silver‐stained gel shows weak specific bands for Eaf1 and Tra1 subunits (Supplementary Fig S1D). This concurs with our previous results in which we showed that Eaf1 was the platform for NuA4 assembly and that purification of Eaf5‐TAP from an eaf1Δ strain yielded only Eaf5/7/3 (Auger et al, 2008). Altogether, our results indicate that Eaf7 forms a bridge between Eaf5 and Eaf3 proteins and Eaf5 anchors the trimer to Eaf1 and the rest of NuA4. In contrast, Eaf6 resides within the piccolo NuA4 subcomplex, consistent with recent results we obtained with the homologous human protein (Avvakumov et al, 2012).
Interestingly, Eaf3 chromodomain protein, but not Eaf5/Eaf7, is also part of the Rpd3S histone deacetylase complex (Carrozza et al, 2005; Keogh et al, 2005). Using Eaf3‐TAP‐expressing cells in parallel with Epl1‐TAP and Sin3‐TAP, we found that a much larger portion of cellular Eaf3 seems associated with the HDAC complex (Supplementary Fig S1E and F, compare lanes 1, 4 and 5). As expected, deletion of EAF5 or EAF7 has no effect on Eaf3 association with Rpd3S (lanes 2 and 3). The presence of Eaf3 in both HAT and HDAC complexes suggests that its chromodomain plays a similar role in the interaction/spreading of each complex on the chromatin fiber through an interaction with Set2‐dependent H3K36me.
Eaf3/5/7 form an independent native complex in yeast cells
Upon closer examination of NuA4 purifications using Eaf3, Eaf5, and Eaf7‐TAP strains in comparison with other tagged subunits, we noticed that each protein of the trimer increased the recovery of the two others (e.g., see Fig 1A). While this could be due to their direct physical association, we suspected that this could also reflect their presence in an additional protein complex outside NuA4. To test this hypothesis, we purified Eaf7 and Eaf5 by tandem affinity as before and loaded the purified material on a gel filtration column (Fig 2A and B). Western blot analysis and silver‐stained gel of the fractions clearly demonstrate that, while the NuA4 complex elutes as the usual 1.3 MDa complex, the majority of Eaf3, Eaf5, and Eaf7 proteins coelute as a small protein complex (~300–400 KDa). A smaller portion of Eaf5/7/3 also coelutes with the large NuA4 complex. Depending on their three‐dimensional conformation, the lower molecular weight fractions could contain uniquely the trimeric complex with no other stably associated proteins, as indicated by mass spectrometry analysis (data not shown). The existence of a native separate Eaf5/7/3 trimer in the cell indicates that these three proteins do represent a functional entity that works within NuA4 but also by itself. Based on our results and previous work, a schematic representation of the different functional modules/subcomplexes of NuA4 is presented in Fig 2C.
Eaf3, eaf5, and eaf7 mutant cells share similar phenotypes
In order to get clues about Eaf3/5/6/7 roles in NuA4 function, we tested deletion mutant strains on different media (Fig 3 and Supplementary Fig S2). All mutant strains have no defects in normal growth conditions in rich media. On the other hand, eaf3, eaf5, and eaf7 cells are sensitive to temperature, while eaf6 cells are not (Fig 3A). Growth defects in the presence of DNA‐damage‐inducing drugs like methyl methanesulfonate (MMS) are a typical phenotype of NuA4 mutants (Boudreault et al, 2003; Downs et al, 2004; Auger et al, 2008), but none of the cells tested here show sensitivity. In contrast, eaf3/5/7 mutant cells are incapable of growth in presence of rapamycin, an inhibitor of the Tor kinase pathway and another typical phenotype of NuA4 mutants, presumably because of its role in ribosomal protein gene expression (Reid et al, 2000; Doyon & Cote, 2004; Auger et al, 2008). Other potential phenotypes analyzed include growth on glycerol, at 16 degrees, in the presence of formamide, caffeine, 6‐azauracil, MPA, or hydroxyurea and defect in telomeric silencing. Significant growth defects are detected for eaf5/7 mutant cells on 3% formamide, but no other media (Fig 3B and data not shown). Phenotypic analysis of eaf3 mutant cells is complicated by the presence of the protein in both NuA4 and Rpd3S, two activities with clearly opposing functions (Biswas et al, 2008). Nevertheless, almost identical phenotypes between eaf5 and eaf7 mutant cells and some similar ones with eaf3 mutants support our findings of these three proteins contacting each other within NuA4 and suggest that the trimer forms a functional subcomplex.
A strong functional interaction is established between NuA4 and SWR1‐C, responsible for incorporation of Htz1 (H2A.Z) in chromatin (reviewed in Altaf et al, 2009; Altaf et al, 2010). Thus, not surprisingly, eaf5 and eaf7 mutant cells show synthetic lethality when combined with swr1 mutation (Supplementary Fig S2A). On the other hand, some phenotypes and genetic interactions support an independent role of the Eaf5/7/3 trimer, outside of NuA4, in agreement with the fractionation study. Large‐scale genetic analysis indicates that eaf5 and eaf7 mutants show synthetic negative interactions with several other mutants from the NuA4 complex, including eaf1, esa1, epl1, arp4, yng2, yaf9, and swc4, suggesting a distinct but related function (Lin et al, 2008). Accordingly, growth sensitivity to formamide appears characteristic of the Eaf3/5/7 trimer since a commonly used esa1 mutant does not show such a phenotype (Fig 3C).
Eaf5 and Eaf7 show functional interactions with the Set2 histone methyltransferase
The function of the Eaf5/7/3 trimer is likely related in part to the presence of Eaf3 chromodomain and its interaction with H3K36me2/3 (Eisen et al, 2001; Li et al, 2007b; Sun et al, 2008; Xu et al, 2008), a histone mark deposited by the Set2 methyltransferase during the transcription elongation process (reviewed in Li et al, 2007a). On the other hand, set2 mutant cells do not show strong sensitivity to formamide, in contrast to eaf5/7 cells and cells lacking the H3K4 methyltransferase Set1, a critical player in the early steps of transcription (Fig 3C). Analysis of set2‐eaf5/7 double‐mutant cells indicates an aggravated sensitivity to caffeine, suggesting functional interaction during gene expression (Fig 3D). A possible role of the trimer could therefore be during transcription elongation, where Set2 associates with RNAPII and methylates H3K36 in its wake (Li et al, 2007a). Sensitivities to nucleotide depletion drugs mycophenolic acid (MPA) and 6‐azauracil (6‐AU) are frequently used as hallmark of transcription elongation deficiency in yeast cells (Reines, 2003). While our eaf5/7 and set2 mutants show no sensitivity on their own, growth defects become evident in the double mutants, most clearly with MPA (Fig 3E and Supplementary Fig S2B). Taken together, these results suggest that the Eaf5/7/3 trimer is functionally linked to histone H3 methylation during transcription elongation but also plays specific distinct roles, partly independent of NuA4 acetyltransferase activity.
The Eaf5/7/3 trimer is enriched over the coding region of specific genes
Since NuA4‐dependent acetylation of chromatin has been proposed to facilitate transcription elongation in vitro (Carey et al, 2006) and in vivo by association to coding regions (Ginsburg et al, 2009), it was important to determine where Eaf5/7/3 is localized during transcription of a gene. NuA4/Esa1 had already been shown to be enriched on the promoter of active genes (Robert et al, 2004; Venters et al, 2011), so Epl1‐myc was used in parallel to Eaf5/7‐myc in ChIP‐qPCR experiments on the actively transcribed PMA1 gene (Fig 4A). When the signals are normalized to the promoter/UAS region of the gene, it becomes apparent that Eaf7 and Eaf5 are preferentially located on the body of the gene, unlike Epl1. These results support a partial physical disconnection of the trimer from the rest of NuA4, as our biochemical fractionations indicated (Fig 2). Eaf7 binding on the PMA1 coding region is affected by both EAF5 and EAF3 deletions but still clearly detected (Fig 4B). Similar results were obtained on the PGK1 gene (Fig 4C). Since Eaf7 association is still detected in eaf5 and eaf3 mutants, this indicates that neither H3K36me nor the NuA4 complex is essential for Eaf7 to bind the coding region of these genes. We next determined whether transcription is essential for this binding using the inducible GAL1 promoter either in its natural context or fused to the long FMP27 gene (Mason & Struhl, 2005). As shown in Fig 4D and E, both Epl1‐myc and Eaf7‐myc signals increase during gene activation and Eaf7 shows the strongest stimulation throughout the coding region. These results demonstrate that Eaf7 binding to coding regions is dependent on ongoing transcription.
NuA4 and the Eaf5/7/3 trimer are globally associated with highly transcribed genes with overlapping but distinct profiles over the transcription units
Previous genome‐wide analyses of NuA4 localization have been performed by ChIP‐on‐chip using microarrays carrying 2–4 probes for each gene (Robert et al, 2004; Venters et al, 2011). These studies have suggested an enrichment of NuA4 at the promoter of active genes. To obtain very precise profiles of NuA4 and Eaf5/7/3 throughout the yeast genome, we have performed ChIP‐on‐chip using a RNA‐based labeling protocol and very high resolution tiled arrays (overlapping 25 bp probes covering the entire genome), as was used in other recent studies (Schulze et al, 2009, 2011). Average signal profiles over open reading frames and intergenic regions were obtained from two independent experiments and are presented in Fig 5 and Supplementary Fig S3. Epl1‐myc signals, depicting the NuA4 complex, clearly correlate with the level of gene transcription (Fig 5A and Supplementary Fig S3A). Interestingly, the profile over the transcription unit is different from what has been concluded in previous studies, showing the strongest signal at the promoter/transcription start site but also strong signals throughout the coding region. As a control, anti‐Myc ChIP from an untagged strain does not show such strong signals over the coding region (Supplementary Fig S3B). A large portion of the most highly transcribed genes in yeast cells are the ones encoding for ribosomal proteins, which tend to be relatively short. Thus, profiles depicted through the CHROMATRA visualization tool provide a more complete view since it represents signals on all individual genes and throughout their variable lengths (Supplementary Fig S3A) (Hentrich et al, 2012). This strong association with ribosomal protein genes certainly accounts for the usual high sensitivity of NuA4 mutant cells to rapamycin, a drug targeting the Tor pathway that regulates ribosome biogenesis (Boudreault et al, 2003; Auger et al, 2008). Gene‐by‐gene analysis indicates that most genes bound by Epl1 are also bound by Eaf7, although their distributions along those genes are different (Supplementary Fig S3A; examples in Fig 5E and F, with untagged control in Supplementary Fig S3C and D). Importantly, in comparison, Eaf7‐myc signals, while also linked to transcriptional activity, do not show prevalence toward the promoter/transcription start site, but a relatively smooth distribution reaching a maximum over the coding region (Fig. 5B). In other words, while Epl1 and Eaf7 signals over the coding regions are often similar, Epl1 tends to be significantly stronger over the promoter regions. Deletion of EAF3 does not greatly change Eaf7 profiles, but signals tend to be more evenly distributed between promoter and coding regions, resembling those of Epl1 (Fig 5C; Supplementary Fig S3A). This decrease in Eaf7 preference toward coding regions is likely due to the absence of interaction with H3K36me through Eaf3 over coding regions, consistent with ChIP‐qPCR showing that Eaf3 affects but is not essential for Eaf7 binding (Fig 4). Finally, deletion of EAF5, which completely disconnects Eaf7–Eaf3 from NuA4 (Fig 1), has a clear effect on the Eaf7‐myc signals, specifically affecting gene‐specific and average profiles on the promoter but not the coding regions (Fig 5D, E and F; Supplementary Fig S3A). These results indicate that Eaf7 binding to promoter regions is dependent on its association with NuA4, while it is not the case over the coding regions. We conclude that the remaining Eaf7 signals over the body of genes represent native binding sites of the Eaf5/7/3 trimeric complex. Interestingly, when comparing Eaf7 average binding profile over ribosomal protein (RP) genes versus highly transcribed non‐RP genes, the preferential association to coding regions is even more pronounced (Supplementary Fig S4A–F).
Deletion of EAF7 affects RNA polymerase II dynamics during transcription elongation
Based on the preferred binding of Eaf5/7/3 to the coding region of actively transcribed genes, we performed multiple ChIP‐qPCR analyses of different histone marks in mutant backgrounds to shed light on the specific function of this small complex. Transcription‐linked marks like H3ac, H4ac, H3K4me, and H3K36me do not seem greatly affected in a reproducible manner (data not shown). Of course, histone acetylation/deacetylation during transcription elongation is thought to be very dynamic, and an effect on steady‐state levels could be difficult to detect, more so if directly linked to nucleosome stability. Thus, we turned our attention to the RNA polymerase, using the integrated GAL‐FMP27 system to characterize RNAPII progression in different mutant backgrounds. This system utilizes the transcription repression obtained by shifting cells to glucose media in order to characterize the last wave of RNAPII passing through the coding region, by ChIP analysis. This allows determination of in vivo transcription elongation rates and polymerase processivity (Mason & Struhl, 2005). Using this system, it has been argued that NuA4 and SAGA HAT complexes cooperate to stimulate transcription elongation rate (Ginsburg et al, 2009). Looking at fully activated GAL‐FMP27, we see that RNAPII steady‐state binding over the coding region is not greatly affected in set2, eaf3, eaf5, and eaf7 mutant backgrounds (Fig 6A). On the other hand, set2 and eaf3 seem to slightly increase pol II signal near the transcription start site (+46), while eaf7 decreases it. During the time course of repression in glucose media, Pol II is seen disappearing rapidly at the 5′ of the coding region but does so at later time points over downstream regions (Fig 6B). In the set2 mutant background, Pol II elongation appears generally delayed at the first time point while the defect is clearly seen toward the 3′ end of the gene at later time points (Fig 6C). An interesting pattern of Pol II dynamics can be seen in the eaf7 mutant where signals remain basically unchanged at all loci after 2 min of repression, followed 2 min later by a rapid disappearance throughout the coding region, most clearly at the 3′end of the gene (Fig 6D and Supplementary Fig S5). This result suggests a role of Eaf7 in Pol II processivity more than elongation rate, meaning that, in the mutant background, the polymerase tends to dissociate from the template (Mason & Struhl, 2005). In agreement with an important role during transcription elongation, the eaf7 deletion mutant from the homozygous diploid yeast gene knockout collection was identified in a large‐scale screen for mutants that show sensitivity to high concentration of 6AU and MPA (Riles et al, 2004).
NuA4 and Eaf5/7/3 interact with elongating RNA polymerase II
Based on the previous results and since it has been argued that both Rpd3S and NuA4 can interact with elongating RNA Pol II (Ginsburg et al, 2009; Drouin et al, 2010; Govind et al, 2010), we tested whether the Eaf5/7/3 trimer could also interact with RNAPII. We compared our Eaf7 genome‐wide data with the signals previously obtained following the same protocol with the Rpb3 subunit of RNAPII (Schulze et al, 2011), and we observed high level of overlap between the signals throughout the transcription units (Supplementary Fig S6). Epl1, Eaf5, and Eaf7 were then used as TAP‐tagged forms to investigate interaction with Pol II by Western blot on TEV eluates from IgG magnetic beads (Fig 6E). A clear Pol II signal is detected in all wild‐type fractions but not in the untagged control. Using eaf1 and eaf5 mutant backgrounds that allow full separation of the trimer from the rest of NuA4, we find that NuA4 requires the Eaf5/7/3 trimer to bind RNAPII (Fig 6E, compare lanes 2 and 3). On the other hand, the presence of Eaf1/NuA4 enhances the association of Pol II with the trimer (lanes 4 and 5, 7 and 8). Interestingly, the loss of Eaf7 does not seem to affect the interaction (lanes 4 and 6). These data indicate that Eaf5 plays a critical role in NuA4‐RNAPII interaction but also that other important points of contact might exist. Since eaf5 and eaf7 mutants show strong negative genetic interactions with genes encoding for Ctk1/3 (Mitchell et al, 2008; Wilmes et al, 2008; Bandyopadhyay et al, 2010), the main kinase responsible for phosphorylation of Pol II C‐terminal domain on serine‐2 of the repeats and the subsequent association of the Set2 methyltransferase (Xiao et al, 2003), we determined which isoform of phosphorylated elongating RNAPII associates with NuA4 and Eaf5/7/3. Interestingly, serine‐5 phosphorylation by Kin28, which is associated with the early steps of elongation, is detected in both Epl1 and Eaf5 fractions, but not in Eaf7, and EAF7 deletion does not affect interaction with Eaf5 (Fig 6F, lanes 2, 4, 5 and 6). In contrast, serine‐2 phosphorylation, linked to the downstream elongating polymerase, is clearly detected in the Eaf7 fraction (lane 6). Importantly, while deletion of EAF7 has no effect on RNAPII Ser5‐P association with Eaf5, it has a clear detrimental effect on association of RNAPII Ser2‐P (compare lanes 4 and 5). Altogether, these results argue that the Eaf5/7/3 associates with elongating polymerases and that Eaf7 favors interaction with the CTD Ser2‐P isoform located further downstream in the coding regions of active genes, an interaction reminiscent of Set2 H3K36 methyltransferase and Rpd3S (Xiao et al, 2003; Drouin et al, 2010).
We can also detect signals for H3K36me3 in the TEV eluates, an interaction that clearly depends on the Eaf5/7/3 trimer, presumably through Eaf3 chromodomain (Fig. 6F). Thus, we postulate that a dual physical interaction is important over the coding region, one with the phosphorylated CTD of elongating polymerase and the other one with nucleosomes modified by the associated Set2 methyltransferase (similar to what is proposed for Rpd3S)(Drouin et al, 2010; Govind et al, 2010). To test this hypothesis, we performed ChIP‐qPCR to analyze Eaf7 binding to coding regions in the absence of Ctk1, the main kinase for CTD Ser2. As shown in Fig. 6G, loss of Ctk1 clearly affects binding of Eaf7 to the body of several transcribed genes. Interestingly, the combined losses of Ctk1 and Eaf3 (interaction with H3K36me) cripple binding of Eaf7 to near background level.
The Eaf5/7/3 trimer plays a role in nucleosome destabilization during transcription elongation
Even with independent functions, the physical association of the Eaf5/7/3 trimer with NuA4 suggests a role in chromatin dynamics during transcription elongation. It has already been shown that NuA4 competes with Rpd3S in vivo (Biswas et al, 2008). Rpd3S is known to associate with elongating polymerase, which allows its binding to nucleosomes methylated on H3K36 by Set2 and their deacetylation (Carrozza et al, 2005; Joshi & Struhl, 2005; Keogh et al, 2005; Drouin et al, 2010; Govind et al, 2010). This action is thought to stabilize nucleosomes in the wake of elongating polymerases, preventing spurious transcription initiation (Carrozza et al, 2005; Li et al, 2007c). To investigate the role of NuA4 and Eaf5/7/3 in this pathway, we used a system designed by the Winston lab in which a HIS3 reporter gene is integrated within the coding region of FLO8, next to a known cryptic TATA box that becomes active upon disruption of nucleosome maintenance during transcription elongation (Fig 7A)(Nourani et al, 2006; Cheung et al, 2008). In this system, deletion of SET2 leads to cryptic initiation of HIS3 transcription and growth in the absence of histidine in the media (Fig 7B). Deletion of EAF3 also leads to growth in the absence of histidine because of its role in Rpd3S deacetylase activity, but to a somewhat lower level. In contrast, deletion of EAF5 and EAF7, or mutation of Esa1 acetyltransferase subunit of NuA4 does not allow expression of HIS3. On the other hand, these same mutants do partially suppress HIS3‐dependent growth of the SET2 deletion. In these conditions, even EAF3 deletion decreases growth of the set2 mutant. To confirm these effects in the natural context and at the transcription level, we analyzed genes that were previously shown to produce intragenic cryptic transcripts in a set2 mutant background (Venkatesh et al, 2012). Using reverse transcription‐qPCR with primer sets corresponding to the 5′ end and the 3′end of these genes, we can clearly detect in set2 mutant cells appearance of shorter transcripts that use cryptic initiation sites within the coding region (Fig 7C). Deletion of EAF7 has no apparent effect on the expression of these genes but again partially suppresses the cryptic transcripts that arise in set2 mutants. Altogether, these data indicate that NuA4 HAT activity and the Eaf5/7/3 trimer function contrary to the Set2‐Rpd3S pathway, favoring nucleosome destabilization during the elongation process. Loss of this function partially suppresses the transcriptional defect caused by the loss of nucleosome methylation/deacetylation in the absence of Set2.
These results suggest that NuA4 and Eaf5/7/3 could help dissociate nucleosomes in front of the advancing polymerase before their Set2‐stimulated reassembly and stabilization in its wake (Li et al, 2007a). In this model, histone chaperones such as Spt6 and FACT play a major role allowing transfer of histones from the front to the back of the polymerase (Avvakumov et al, 2011). Supporting the implication of the Eaf5/7/3 trimer in this process, we detect a physical interaction between Eaf7 and Spt16, a subunit of the FACT histone chaperone (Fig 6F). Furthermore, a negative genetic interaction is clearly detected between mutants in the FACT subunits (Spt16 and Pob3) and eaf5/eaf7 mutants (Fig 7D). Such a role of Eaf5/7/3 in favoring nucleosome destabilization in front of the polymerase should lead to more stable nucleosomes in the mutant cells. Relative nucleosome stability can be more efficiently studied by measuring the dynamics of replication‐independent exchange of histones at specific loci, as previously described (Rufiange et al, 2007). In this type of analysis, newly incorporated histones can be measured in cells blocked in G1 using the H3K56ac mark compared to total histone H3 signal in ChIP experiments. The values obtained at different loci reflect replication‐independent histone exchange that occurs at these loci, which is higher at promoter regions and in the body of highly transcribed genes (Rufiange et al, 2007). In contrast to what we expected, we measure a significant increase of H3K56ac/H3 signal at PMA1 in eaf7 mutant cells, suggesting a higher level of histone exchange and less stable nucleosomes (Fig 7E). Similar results are obtained at other transcribed loci (Supplementary Fig S7A). It has been shown that higher incorporation of new histones can occur when old histones are not properly recycled from in front to behind the elongating polymerase. For example, this happens in cells defective for the FACT histone chaperone when H3‐H4 redeposition is compromised, leading to increased incorporation of new histones (Jamai et al, 2009). Consistent with this concept, we obtain an even higher level of H3K56ac/H3 signal in spt16 mutant cells (Fig. 7E). To determine whether augmented incorporation of new histones is linked to transcription, we compared signals at a non‐transcribed control locus (Supplementary Fig S7B). In contrast to spt16, the eaf7 mutant has a marginal effect on incorporation at the non‐transcribed locus, suggesting after correction that the Eaf5/7/3 trimer and FACT have similar effect on the redeposition of histones evicted by elongating RNA polymerase (Supplementary Fig S7C). No further increase of new histone deposition is detected in eaf7 spt16 double‐mutant cells (Supplementary Fig S7D), and similar results are obtained with the inducible epitope‐tagged H3 system to measure replication‐independent new histone incorporation (Rufiange et al, 2007; data not shown).
It was recently reported that loss of Set2‐mediated H3K36 methylation also leads to replication‐independent increased incorporation of new histones at transcribed genes (Venkatesh et al, 2012). This is thought to occur through Rpd3S/deacetylation‐mediated nucleosome stabilization, preventing the incorporation of new histones carrying the H3K56ac mark for de novo assembly (Smolle et al, 2013). We also detect increased incorporation of new histones in set2 mutant cells, slightly lower than what is seen with the eaf7 mutant, while the double‐mutant cells showed a more even incorporation throughout the PMA1 gene (Fig 7F). Altogether, these results implicate the Eaf7/5/3 trimer in nucleosome dynamics during transcription elongation, favoring RNA Pol II progression through associated nucleosome disruption and transfer for reassembly in its wake.
In this study, we have analyzed poorly characterized subunits of the NuA4 histone acetyltransferase complex. We found that, while Eaf6 resides in the catalytic core of NuA4 (picNuA4), Eaf5 and Eaf7 are responsible for the association of the chromodomain protein Eaf3 to NuA4. Our biochemical data indicate that the Eaf5/7/3 subcomplex is not required for NuA4 structural integrity and is attached through an interaction between Eaf5 and Eaf1 (see Fig 2C). We also found that the Eaf5/7/3 trimer exists as a native complex outside of NuA4. We showed that eaf5/7 mutant cells share similar phenotypes, in which some are shared with other components of NuA4 while others are specific to the trimer, supporting an independent function. Our work presents gene‐specific localization data and high‐resolution genome‐wide mapping of NuA4 and Eaf5/7/3‐binding sites. While both overlap greatly in their target genes and their correlation to transcriptional activity, they present different profiles over the transcription units. In comparison to previous lower resolution studies (Robert et al, 2004; Venters et al, 2011), NuA4 shows a high level of binding throughout the coding regions of these highly transcribed genes, with a polarity of higher enrichment toward the transcription start site and promoter region (Fig 5A). In contrast, Eaf7 shows a polarity of higher enrichment toward the middle of the coding region (Fig 5B). This difference in profiles is exacerbated when the analysis is done in an eaf5 mutant strain, which disrupts any possible interaction between Eaf7 and NuA4 (Fig 5D–F, Supplementary Fig S3A). Surprisingly, while loss of Eaf5 specifically depletes Eaf7 from promoter regions without greatly affecting coding regions, loss of Eaf3 and its H3K36me‐binding chromodomain does not prevent Eaf7 from binding to the body of the gene (Fig 5C, E–F; Supplementary Fig S3A). This clearly demonstrates that Eaf7 requires NuA4 to be bound to the promoter but neither NuA4 nor Eaf3‐H3K36me for its association with the body of active genes. We present data supporting an additional mechanism favoring binding to transcribed regions, related to what has been suggested for the Eaf3‐containing Rpd3S complex (Drouin et al, 2010; Govind et al, 2010). While elongating RNA polymerase II physically interacts with the NuA4 complex (confirming previous findings; Ginsburg et al, 2009) and the Eaf5/7/3 trimer, loss of Eaf7 specifically depletes the Pol II isoform that is phosphorylated on the serine‐2 of its C‐terminal domain repeats (Fig. 6F). Importantly, loss of the CTD Ser2 kinase, Ctk1, decreases binding of Eaf7 to coding regions (Fig 6G). Loss of both Ctk1 and Eaf3 cripples Eaf7 binding to near background levels, indicating that RNAPII CTD Ser2‐P and nucleosomes bearing H3K36me2/3 form a dual interaction surface for the trimer on the body of transcribed genes, at a distance from the transcription start site.
In trying to decipher the precise function of the Eaf5/7/3, we accumulated data supporting a role in nucleosome dynamics during transcription elongation, favoring destabilization in front of the polymerase, but also stabilizing nucleosomes in its wake along with the Set2‐H3K36me‐Rpd3S pathway (see Fig 8). This model is supported by a decrease in elongation processivity of Pol II and partial suppression of spurious transcription appearing in set2 mutant cells, as well as an increase in replication‐independent histone exchange in eaf7 mutant cells. One could easily envision the Eaf5/7/3 trimer traveling with Pol II CTD phosphorylated on Ser2, leading to interaction with H3K36me‐containing nucleosomes through the Eaf3 chromodomain. Destabilization of nucleosomes in front of the polymerase would function in part through tethering of NuA4 acetyltransferase activity. Stabilization of the nucleosomes behind the polymerase, blocking histone exchange/incorporation of new histones, would function through cooperation with histone chaperones to recycle the disrupted nucleosomes (Fig. 8). Supporting this model, we can detect physical interaction between the Eaf5/7/3 trimer and the FACT histone chaperone in cellular extracts (Fig 6F).
Questions remain about other possible functions of the Eaf5/7/3 trimer. Consistent with the striking structural conservation of NuA4 from yeast to human, the Eaf7 homolog in higher eukaryotes was found associated with the human Tip60 complex (Cai et al, 2003). In this report, it was apparent that the homologous protein, called MRGBP, was interacting directly with MRG15, the human homolog of Eaf3. Gel filtration and a large‐scale proteomic study also indicate that the MRGBP‐MRG15 complex exists independently of the Tip60 complex (Kirkwood et al, 2013). Results in our lab (K. Jacquet and J. Côté; in preparation) as well as recent published work (Gowher et al, 2012) also demonstrate the independent function of MRGBP‐MRG15. Interestingly, methylation of H3K36 at exon–intron junctions plays a role in co‐transcriptional mRNA splicing, and MRG15 was shown to regulate alternative splicing (reviewed in Luco et al, 2011). Furthermore, the MRGBP‐MRG15 dimer has now been linked to this process in relation to peaks of elongating polymerases carrying Ser2‐P on their C‐terminal domain (Gowher et al, 2012). Since Eaf5 is the only NuA4 subunit having no homolog in human cells, these strong analogies suggest that the MRGBP‐MRG15 dimer is the functional homolog of the yeast Eaf5/7/3 trimer. Keeping this in mind, it is striking to note that most of the intron‐containing genes in the yeast genome are in fact highly transcribed, encoding for ribosomal proteins (RP) and bound by NuA4 and Eaf7. We could not detect significant defect of the mRNA splicing process using a reporter plasmid (Teem & Rosbash, 1983) in eaf5/7 mutant cells (data not shown). Definitive answers about a possible role of the Eaf5/7/3 trimer in mRNA splicing will require additional work in more physiological conditions. It is possible that a role in RNA processing is linked to the strong genetic interaction detected with the SWR1 complex (Supplementary Fig S2A) (Krogan et al, 2004). SWR1‐C and its bromodomain‐containing subunit Bdf1 have been recently linked to transcription elongation and identified as important regulators of mRNA splicing (Santisteban et al, 2011; Albulescu et al, 2012).
Interestingly, secondary mass spectrometry hits in Eaf7‐TAP purifications included proteins associated with the nuclear exosome (Dis3, Srp1, Kap95; Synowsky et al, 2009; data not shown). Furthermore, eaf5 and eaf7 mutants show negative genetic interactions with several components of the nuclear exosome and the THO‐TREX mRNA export machinery (Wilmes et al, 2008). We can detect a reproducible small increase of transcription read‐through in eaf5/7 mutants using a reporter plasmid (Carroll et al, 2004), suggesting some defect in exosome activity/3′ end formation (data not shown). While these results require additional investigation in more physiological systems, they suggest that the Eaf5/7/3 trimer could indeed play a role in co‐transcriptional RNA processing and maturation.
Altogether, the results presented in this study identified an exciting new functional module in the NuA4 histone acetyltransferase, as well as an independent trimeric complex. An important role during transcription elongation is attributed to the trimer, in part through tethering of NuA4 activity, but also independently. This NuA4‐independent role of the Eaf5/7/3 trimer forces us to reexamine several large‐scale genetic studies (e.g., SGA/E‐MAP) that used solely eaf5 and eaf7 interactions as measurement of NuA4 function. The Eaf5/7/3 trimer is an atypical transcription elongation factor as it not only functions in nucleosome disruption to allow better RNAPII processivity, but also facilitates nucleosome recycling/stability behind the polymerase, suppressing histone exchange. It will be most interesting to extend our analysis, dissecting the precise mechanisms by which the Eaf5/7/3 trimer affects both nucleosome disruption in front of the polymerase and refolding in its wake. This will also help to determine additional roles and functional conservation with human MRGBP‐MRG15 in regulating mRNA splicing/processing.
Materials and Methods
Yeast strains, plasmids and growth assays
All the strain genotypes used in this study are based on the S288c background (LPY3431; Clarke et al, 1999 or BY4741(Resgen)) and are listed in Supplementary Table S1. PCR‐based integrations were used to make deletion strains, and epitope‐tagging with KanMX, HisMX, and NatR (plasmid p4339) selection cassettes followed standard protocols (Longtine et al, 1998; Tong et al, 2001). Plasmid pHPL32URA was constructed as follows: EAF5 was amplified from yeast genomic DNA and cloned with BamHI and SmaI in pFEMXE (described in Boudreault et al, 2003) to give pF32L. EAF5 promoter was also amplified from genomic DNA and cloned with HindIII and NcoI in pF32L to give pHPL32. BglII inserts containing URA3 from pFL38 and LEU2 from pHPL32 were swapped to give pHPL32Ura. Plasmid pHPL32ORFEaf7URA was constructed by swapping EAF5 by EAF7 in the pHPL32Ura plasmid using BamHI and PvuII. pSet2‐FlagURA was a generous gift from B. Strahl. Deletion mutants were introduced in strains carrying pGAL‐FMP27, pGAL‐FLO8‐HIS3, and pGAL‐Flag‐H3/Myc‐H3 described in Rufiange et al (2007) and Cheung et al (2008). The double deletion strains were generated by the plasmid shuffle technique where the first deletion is covered by the same gene on a URA3 plasmid followed by the second gene deletion. Yeast strains were then grown overnight in SC at 30°C, diluted to an OD600 of 0.25, and grown for 3 h in SC. 10‐fold serial dilutions of cultures were spotted onto SC or SC supplemented with 0.1% 5′ FOA to evict the URA3 containing plasmid. For the yeast spot sensitivity assay, yeast strains were grown overnight at 30°C in YPD, diluted to an OD600 of 0.5, and grown for 3 h in YPD. Then 10‐fold serial dilutions of cultures were spotted onto different media and grown at the indicated temperatures. YPD plates containing drugs such as methyl methanesulfonate (MMS; 0.015%), rapamycin (25 nM), 6‐AzaUracil (6AU; 200 μg/ml), mycophenolic acid (MPA; 100 μg/ml), formamide (3%), and caffeine (6 mM) were used to test for resistance compared to YPD control plates.
Purification of protein complexes
TAP purification was performed as initially reported by Seraphin and colleagues (Puig et al, 2001) with modifications described in Auger et al (2008). Purified fractions were assayed for HAT activity and by Western blot as described (Allard et al, 1999), and specific protein bands were excised from gels and identified by tandem mass spectrometry at the CHUL proteomic platform. When indicated, calmodulin or TEV elutions were further fractionated by gel filtration on a Superose 6 column as described (Altaf et al, 2010). For co‐IP experiments, TEV elutions were obtained using magnetic beads (Dynal) on which IgG were manually pre‐bound, as described (Mitchell et al, 2008). Antibodies for NuA4 subunits used in Western blot analysis have been described ((Auger et al, 2008) and references therein), while anti‐CTD Ser5‐P and Ser2‐P were obtained from Millipore (3E8, 310), anti‐total RNA pol II from Covance (8WG16), anti‐H3K36me3 from Abcam (ab9050), and anti‐Spt16 was a gift from Tim Formosa.
Total RNA was purified by hot phenol extraction from 10 ml of cells grown to an OD600 of 0.6 to 1.0. 5 μg of RNA was treated with the Turbo DNase kit (Ambion), and 2 μg of DNA‐free RNA was synthesized into cDNA using qScript cDNA supermix (Quanta Biosciences) according to the manufacturer's instructions. cDNA was then diluted and quantified by real‐time qPCR (Syber Green, Roche) on the LC480 LightCycler (Roche) using primers near the 5′ and 3′ end of the gene (sequences available upon request). Relative amounts were normalized to ACT1 5′. All values are expressed relative to the wild‐type value, set at 1, and error bars denote standard deviations of a biological quadruplicate (four independent cultures, two distinct experiments).
Chromatin Immunoprecipitation (ChIPs)
ChIP experiments were essentially performed as previously described (Utley et al, 2005). Briefly, cell culture of 200 ml of appropriate media grown at 0.5–1 OD600 was cross‐linked for 20 min with 1% formaldehyde. For the replication‐independent histone exchange assays, cells were blocked in G1 with alpha factor as described (Rufiange et al, 2007). After cross‐link, the reaction was stopped by adding glycine at a final concentration of 200 mM. Cell lysate was obtained by bead beating cells with glass beads in FA lysis buffer (50 mM HEPES pH 7.5, 140 mM NaCl, 1 mM EDTA pH8, 1% Triton, 0.1% Na deoxycholate) to reach 70–80% of cells disrupted. Chromatin was sheared by sonication (Bioruptor) to 200–500 bp fragments. Immunoprecipitations were performed on 100 μg of chromatin with 2 μl of anti‐Myc, 2 μl of anti‐Pol II (8WG16), 0.3 μl of anti‐H3K56ac (Upstate, 07‐677), and 0.5 μg of anti‐H3 (Abcam, ab1791) antibodies for at least 12 h at 4°C. ChIPs were then incubated with protein A or G sepharose matrix (GE Healthcare) for 5hrs at 4°C or with Dynabeads coated with protein G (Invitrogen) for the Myc‐tagged proteins. After washes with 140 mM NaCl FA Lysis Buffer, 500 mM NaCl FA Lysis Buffer, Wash Buffer #2 (10 mM Tris pH8, 250 mM LiCl, 0.5% NP40, 0.5% Na deoxycholate, 1 mM EDTA), and TE, the immunoprecipitated chromatin was eluted from the beads with elution buffer (50 mM Tris, 10 mM EDTA, 1% SDS) for 15 min at 65°C and reverse cross‐linked for 12 h at the same temperature. ChIPs were treated 15 min with RNase A and proteinase K, and DNA was purified by standard phenol‐chloroform extraction and precipitated. DNA was then quantified by real‐time qPCR (Syber Green, Roche) on the LC480 LightCycler (Roche). All ChIPs were performed in at least two independent experiments, and data presented in figures represent the average between the biological replicates with the corresponding standard errors. A large non‐coding region of chromosome V (positions 9716–9863) was used as control locus (Nourani et al, 2004). All primer sequences are available upon request.
ChIP‐on‐chip and data analysis
ChIP‐on‐chip experiments were performed as described previously (Schulze et al, 2009) with minor modifications. In brief, yeast cells (500 ml) were grown in a rich medium to an OD600 of 0.5–0.6 and were cross‐linked with 1% formaldehyde for 60 min before chromatin was extracted. The chromatin was sonicated (Bioruptor, Diagenode; Sparta, NJ: 15 cycles, 30 s on/off, high setting) to yield an average DNA fragment of 500 bp. Anti‐Myc (5 μl, Sigma) antibodies were coupled to 60 μl of protein A magnetic beads (Invitrogen). After reversal of the cross‐linking and DNA purification, the immunoprecipitated and input DNA was amplified to about 6 μg aRNA using T7 RNA polymerase in two rounds. Samples were labeled with biotin, and the immunoprecipitated and input samples were hybridized to Affymetrix 1.0R S. cerevisiae microarrays, which are comprised of over 3.2 million probes covering the complete genome. Probes (25‐mer) are tiled at an average of 5 bp resolution, creating an overlap of approximately 20 bp between adjacent probes.
Data analysis was performed as described previously (Schulze et al, 2009) and was submitted to the ArrayExpress database under accession number E‐MTAB‐1780. Annotations for ORFs, ARSs, and centromeres were derived from the SGD database. An ORF was termed enriched if at least 50% of all probes had a MAT score above a threshold of 1.5. Promoters were defined as enriched if all probes 300 bp upstream of the transcriptional start site were above the MAT score cutoff. Promoters that overlap with ORFs of other genes were not considered. Given the transient nature of NuA4 subunits binding to chromatin, we determined a high confidence data set intersecting the two replicates after quantile normalization and performed subsequent analysis with the normalized data. A list of all transcription start and end sites for 4,868 transcripts was kindly provided by Harm van Bakel (Timothy Hughes' lab). The CHROMATRA graphs were generated as described previously (Hentrich et al, 2012). Enrichment scores of each Myc‐tagged protein were calculated across nucleosome‐sized intervals of 150 bp and color‐coded for all known 4,868 transcripts, which were aligned according to their transcription start sites (TSS) and sorted by their length or by length as well as transcriptional frequency (Holstege et al, 1998).
DR, MC, AW, A‐LS, NL, JS, VC, JM‐S, and SP performed experiments. DR, MC, AW, A‐LS, NL, JS, AN, MSK, and JC designed experiments and interpreted the results. JC wrote the manuscript.
Conflict of interest
The authors declare that they have no conflict of interest.
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure S5
Supplementary Figure S6
Supplementary Figure S7
Supplementary Table S1
We are grateful to Anne Rufiange, Dominique Cronier, and Nathalie Bouchard for experiments performed at different stages of this work. We are indebted to Rhea Utley for the construction of some strains used in this study and critical reading of the manuscript, and John Yates for original mass spec analysis that allowed identification of Eaf7 in NuA4. We also wish to thank David Stillman, Swami Venkatesh, Bing Li, Jerry Workman, Andréanne Auger, François Robert, and Rhea Utley for stimulating discussions. We thank John Lucchesi, Patrick Grant, David Stillman, Song Tan, Tim Formosa, and Brian Strahl for anti‐Eaf3, anti‐Tra1, anti‐Arp4/anti‐Sin3, anti‐Yng2, anti‐Spt16, and pSET2/URA3, respectively, and David Stillman for strains used during this study. This work was supported by grants from the Canadian Institutes of Health Research (CIHR) to A.N. (MOP‐81245), M.S.K. (MOP‐119383), and J.C. (MOP‐14308). M.S.K is a Senior Fellow of the Canadian Institute for Advanced Research. A.N. and J.C. hold Canada Research Chairs.
FundingCanadian Institutes of Health Research MOP‐81245 MOP‐119383 MOP‐14308
- © 2014 The Authors