A PWWP Domain-Containing Protein Targets the NuA3 Acetyltransferase Complex via Histone H3 Lysine 36 trimethylation to Coordinate Transcriptional Elongation at Coding Regions*

Post-translational modifications of histones, such as acetylation and methylation, are differentially positioned in chromatin with respect to gene organization. For example, although histone H3 is often trimethylated on lysine 4 (H3K4me3) and acetylated on lysine 14 (H3K14ac) at active promoter regions, histone H3 lysine 36 trimethylation (H3K36me3) occurs throughout the open reading frames of transcriptionally active genes. The conserved yeast histone acetyltransferase complex, NuA3, specifically binds H3K4me3 through a plant homeodomain (PHD) finger in the Yng1 subunit, and subsequently catalyzes the acetylation of H3K14 through the histone acetyltransferase domain of Sas3, leading to transcription initiation at a subset of genes. We previously found that Ylr455w (Pdp3), an uncharacterized proline-tryptophan-tryptophan-proline (PWWP) domain-containing protein, copurifies with stable members of NuA3. Here, we employ mass-spectrometric analysis of affinity purified Pdp3, biophysical binding assays, and genetic analyses to classify NuA3 into two functionally distinct forms: NuA3a and NuA3b. Although NuA3a uses the PHD finger of Yng1 to interact with H3K4me3 at the 5′-end of open reading frames, NuA3b contains the unique member, Pdp3, which regulates an interaction between NuA3b and H3K36me3 at the transcribed regions of genes through its PWWP domain. We find that deletion of PDP3 decreases NuA3-directed transcription and results in growth defects when combined with transcription elongation mutants, suggesting NuA3b acts as a positive elongation factor. Finally, we determine that NuA3a, but not NuA3b, is synthetically lethal in combination with a deletion of the histone acetyltransferase GCN5, indicating NuA3b has a specialized role at coding regions that is independent of Gcn5 activity. Collectively, these studies define a new form of the NuA3 complex that associates with H3K36me3 to effect transcriptional elongation. MS data are available via ProteomeXchange with identifier PXD001156.

Post-translational modifications of histones, such as acetylation and methylation, are differentially positioned in chromatin with respect to gene organization. For example, although histone H3 is often trimethylated on lysine 4 (H3K4me3) and acetylated on lysine 14 (H3K14ac) at active promoter regions, histone H3 lysine 36 trimethylation (H3K36me3) occurs throughout the open reading frames of transcriptionally active genes. The conserved yeast histone acetyltransferase complex, NuA3, specifically binds H3K4me3 through a plant homeodomain (PHD) finger in the Yng1 subunit, and subsequently catalyzes the acetylation of H3K14 through the histone acetyltransferase domain of Sas3, leading to transcription initiation at a subset of genes. We previously found that Ylr455w (Pdp3), an uncharacterized proline-tryptophan-tryptophan-proline (PWWP) domain-containing protein, copurifies with stable members of NuA3. Here, we employ mass-spectrometric analysis of affinity purified Pdp3, biophysical binding assays, and genetic analyses to classify NuA3 into two functionally distinct forms: NuA3a and NuA3b. Although NuA3a uses the PHD finger of Yng1 to interact with H3K4me3 at the 5-end of open reading frames, NuA3b contains the unique member, Pdp3, which regulates an interaction between NuA3b and H3K36me3 at the transcribed regions of genes through its PWWP domain. We find that deletion of PDP3 decreases NuA3-directed transcription and results in growth defects when combined with transcription elongation mutants, suggesting NuA3b acts as a positive elongation factor. Finally, we determine that NuA3a, but not NuA3b, is synthetically lethal in combination with a deletion of the histone acetyltransferase GCN5, indicating NuA3b has a specialized role at coding regions that is independent of Gcn5 activity. Collectively, these studies define a new form of the NuA3 complex that associates with H3K36me3 to effect transcriptional elongation. MS  Eukaryotic DNA is wrapped around octamers of the evolutionarily conserved core histone proteins H3, H4, H2A, and H2B, forming nucleosomes-the fundamental unit of chromatin. Chromatin acts as a barrier to the transcriptional machinery, and therefore precise coordination of nucleosome organization is required for the passage of RNA polymerase II (RNAPII) 1 (1)(2)(3). Nucleosome organization is regulated by chromatin remodelers, histone chaperones, and other complexes that enzymatically add, remove, or bind ("write," "erase," or "read," respectively) post-translational modifications (PTMs) on histones (4 -7). Together, these factors ensure that genes remain accessible to transcription factors, activators or coactivators, and RNAPII (5). Critically, these factors also function to restore chromatin structure following the passage of RNAPII during transcription elongation (3,8).
Using recent advances in mass spectrometry, we previously found that Ylr455w, an uncharacterized PWWP domaincontaining protein (59 -61), copurifies with stable members of the NuA3 HAT complex (19). We propose that Ylr455w be called Pdp3: PWWP domain protein in NuA3. Moreover, Pdp3 (Q09842) is also the name of a PWWP domain-containing protein in Schizosaccharomyces pombe that interacts with homologous subunits of the NuA3 complex (62).
Here, we find evidence that two functionally distinct forms of NuA3 exist. One form of NuA3 binds H3K4me3, through the PHD finger of Yng1, and acetylates H3K14 at the 5Ј-ends of actively transcribed genes to promote transcription initiation (19,24,26,27). In contrast, an interaction between Pdp3 and NuA3 recruits the complex to chromatin, through the Pdp3 PWWP domain binding to H3K36me3. Deletion of the PDP3 gene decreases NuA3-directed transcription and results in growth defects when combined with transcription elongation mutants, suggesting Pdp3-associated NuA3 functions in the transcription elongation process. Although H3K36me2/3 can act as a repressive mark that protects chromatin integrity during transcription elongation (3,8,32), the work described in this study suggests that H3K36me3 can also act to positively regulate transcription elongation through the recruitment of the NuA3 complex via Pdp3.

EXPERIMENTAL PROCEDURES
S. cerevisiae Strains-All strains are described in supplemental  Table S1.
Escherichia coli and S. cerevisiae Plasmids-All constructs are described in supplemental Table S2.
Peptide Sequences-All peptide sequences are listed in supplemental Table S3.
Primer Sequences-Primers were designed by Primer3 to target the 5Ј-end of our genes of interest. All primer sequences are listed in supplemental Table S4.
Mass Spectrometric Protein Identification-Pdp3-TAP protein complex purification was performed with S. cerevisiae grown to log phase in yeast extract peptone dextrose (YPD), essentially as described, to maintain complex integrity (61,63). Proteins copurifying with Pdp3 were subjected to tandem MS analysis of peptides with a Thermo LTQ-XL mass spectrometer coupled to an Eksigent nanoLC 2D system as previously described (30). Spectral counts and proteins were identified with Mascot (Matrix Science, London, UK; version 2.3.01).
Database Searching-Tandem mass spectra were extracted by Thermo ExtractMSn version 1.0.0.8. Charge state deconvolution and deisotoping were not performed. All MS/MS samples were analyzed using Mascot (Matrix Science, London, UK; version 2.3.01). Mascot was set up to search the SwissProt_57.15 database (selected for S. cerevisiae, 57.15, 6973 entries), assuming the digestion enzyme was nonspecific. Mascot was searched with a fragment ion mass tolerance of 0.60 Da and a parent ion tolerance of 2.0 Da. The iodoacetamide derivative of cysteine was specified in Mascot as a fixed modification. S-carbamoylmethylcysteine cyclization at the N-terminus, oxidation of methionine, pyro-carbamidomethyl formation at the N-terminus, acetylation of asparagine, and the presence of proline at the N-terminus were specified in Mascot as variable modifications.
Criteria for Protein Identification-Scaffold (version Scaffold_4.0.1, Proteome Software Inc., Portland, OR) was used to validate MS/MSbased peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 20.0% probability by the Peptide Prophet algorithm (64). Protein identifications were accepted if they could be established at greater than 95.0% probability and contained at least one identified peptide. Protein probabilities were assigned by the Protein Prophet algorithm (65). Proteins that contained similar peptides, and could not be differentiated based on MS/MS analysis alone, were grouped to satisfy the principles of parsimony. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http:// proteomecentral.proteomexchange.org) via the PRIDE partner repository (66) with the data set identifier PXD001156 (67)(68)(69).
Protein Expression-PDP3 constructs were made with an N-terminal HIS6-pfuMBP (60 -434) -FLAG tag (pET28a derivative vector obtained from the G. Bowman Laboratory, Johns Hopkins University) and/or an N-terminal Thioredoxin-HIS6-S•tag tag (pET32a vector, Millipore). Proteins were exogenously expressed in chemically competent BL21 E. coli (Invitrogen, Carlsbad, CA) after overnight induction with 1 mM IPTG at 18 -20°C in LB medium. Point mutants were made with the QuikChange Site-Directed Mutagenesis Kit (Stratagene, LaJolla, CA) and expressed as described.
Peptide Synthesis for Pull-down Assays-Peptides were synthesized as previously described by the C. D. Allis Laboratory (The Rockefeller University) or the UNC Peptide Synthesis and Arraying Core Facility (70).
Peptide Synthesis for Fluorescence Polarization Assays-Fluorescent peptides were synthesized using standard Fmoc-solid phase peptide chemistry on a Prelude Peptide Synthesizer (Protein Technologies). Peptides were made on a 0.05 mmol scale with four equivalents of amino acids using Rink Amide AM resin (Novabiochem) to generate peptide amides. 5-Carboxyfluorescein (5-FAM) (Chempep) was coupled to the peptides using Lys(ivDde) (Chempep). The ivDde protecting group was orthogonally removed using standard deprotection procedures. The resulting peptides were typically cleaved using TFA/thioanisole/water/triisopropylsilane/phenol (87.5:2.5:2.5:2.5:5 v/v) and purified with a Varian Dynamax Microsorb C18 preparative column (Agilent). Purified peptide was lyophilized and its mass was confirmed with an Applied Biosystems Voyager DE-STR MALDI-TOF mass spectrometer (Invitrogen). Of note, to obtain the H3K79me3 5-FAM-linked peptide, it was necessary to install Fmoc-(FmocHmb)Phe-OH (Novabiochem) at F84.
Fluorescence Polarization Assays-Binding assays were performed as described in (71) with the following modifications. Full-length purified wild-type Pdp3, F18A, and W21A proteins were exchanged into FP buffer (50 mM Tris pH 7.5, 150 mM NaCl, and 5 mM DTT) and concentrated to ϳ260 -430 M using Amicon Centrifugal Filter Units MWCO 30,000 (Millipore). Binding assays were performed in a 60 l volume with 96 well half area black flat bottom nonbinding surface plates (Corning). Protein was serially diluted with FP Buffer in twofold increments and incubated with 120 nM of 5-FAM-labeled histone peptides. Following a 30 min equilibration period, fluorescence was detected at room temperature with an Infinite M1000 plate reader (Tecan, Durham, NC) using a 470 nm excitation filter and 527 Ϯ 20 nm emission filter. Binding curves were analyzed by the total binding equation Y ϭ B max * X/(K d ϩ X) ϩ NS * X ϩ Background, where B max ϭ 1 and nonspecific (NS) and background variables are constrained to be equal between peptides, using Prism 5.0 (GraphPad Inc., San Diego, CA). Error bars represent the S.D. of a representative experiment (n ϭ 2) performed in triplicate.
Yeast Strains and Cell Spotting Assays-S. cerevisiae strains were created using heterologous gene replacement (72). Strains were grown on YPD or synthetic complete (SC) media as indicated. BUR1 delete shuffle strains were grown on SC-Ura plates to maintain the wild-type BUR1 plasmid prior to plating on media containing the drug 5-Fluoroorotic acid (5-FOA) (73). For cell spotting assays, either 0.5 or 2 ODs of cells were fivefold serially diluted, spotted onto the appropriate plates, and incubated at 30°C for 2-3 days as indicated.
Chromatin Association Assay-Strains were grown overnight in YPD to confluence. Each strain was diluted to 0.1 ODs in 50 mLs of YPD and grown to an OD of ϳ0.8 -1. 40 -50 ODs of cells were then isolated, washed with water and SB buffer (1 M Sorbitol, 20 mM Tris.Cl pH 7.4), and frozen at Ϫ80°C until ready for isolation. Cells were resuspended in 1 ml of PSB buffer (20 mM Tris.Cl pH 7.4, 2 mM EDTA, 100 mM NaCl, and 10 mM ␤-ME), followed by addition of 1 ml of SB buffer. Cells were then spheroplasted with Zymolyase (Seikagaku Biobusiness, Rockville, MD) for 30 min at room temperature. Spheroplasts were spun down at 2000 ϫ g and washed twice with LB (0.4 M Sorbitol, 150 mM potassium acetate, 2 mM magnesium acetate, and 20 mM PIPES pH 6.8). TritonX-100 was added to LB (final concentration of 1%). Cells were lysed for 15 min on ice. Chromatin was isolated by spinning down lysates at 5000 ϫ g for 15 min. The supernatant was collected and saved as the "soluble" fraction. The chromatin was washed once more with LB and then resuspended into an equal volume to that of the "soluble" fraction. Volume equivalents were resolved on 15% SDS-polyacrylamide gels and subjected to immunoblot analysis.
Relative Transcript Levels-Total RNA was prepared from TAPtagged YNG1, PDP3, yng1⌬, and pdp3⌬ S. cerevisiae strains via Trizol (Invitrogen) and digested with Turbo DNase (Invitrogen). cDNA was synthesized with the Superscript III First Strand Synthesis System (Invitrogen). Differences in transcript levels of control genes and NuA3 target genes were measured by qPCR using Power SYBR Green PCR Master Mix (Invitrogen) and a Real Time PCR system (Applied Biosystems, Foster City, CA, v2.1). The relative transcript levels from mutant strains as compared with wildtype were calculated using the relative standard curve method. Error bars represent the S.E. of a representative experiment (n ϭ 4) done in triplicate. Statistical significance was determined by an unpaired two-tailed t test.
To confirm that Pdp3 is a member of the NuA3 complex, we tested the ability of full-length recombinant Pdp3 to pull down NuA3 proteins from cellular extracts (Fig. 1B). HIS6-FLAGtagged Pdp3 was purified from E. coli and incubated with lysates from S. cerevisiae endogenously expressing TAPtagged NuA3 subunits. Immunoprecipitated samples were resolved by SDS-PAGE and visualized with an antibody against the TAP tag. HIS6-FLAG-Pdp3 interacts with Yng1, Nto1, and Sas3 (Fig. 1B), supporting the MS/MS results (Fig.  1A, supplemental Table S5). Neither Taf14 nor Eaf6 were detected in recombinant Pdp3 pull-downs, which may indicate that these proteins do not interact with Pdp3 (Fig. 1B, data not shown); however, we cannot exclude the possibility that the TAP tag disrupts native interactions. Also, because Taf14 is a member of multiple chromatin-associated complexes (76), the quantity of soluble Taf14 available to be pulled down in this assay may be insufficient to detect.
Pdp3 Specifically Engages H3K36me3 Through a Conserved PWWP Domain-H3K36me3 localizes to the body of actively transcribed genes, and is associated with transcription elongation in yeast and humans (13,34,35,37,40). BRPF1, a human homolog of yeast Nto1 (23), engages H3K36me3 through a PWWP domain (supplemental Fig. S1) (55,56) that is absent in Nto1. Interestingly, structural modeling predicts that Pdp3 contains a PWWP domain with aromatic residues critical for methyl-lysine binding, as well as beta strands and alpha helices conserved in other PWWPdomain containing proteins that bind methyl-lysine (Figs. 2A,  Fig. S1. B, Schematic representation of the Pdp3 protein. The predicted (red outline) and functional (tan hexagon) PWWP domains are annotated. Aromatic cage residues F18, W21, and F48 are highlighted in yellow. C, Peptide pull-down assays were performed with full-length HIS6-FLAG-Pdp3 and biotinylated histone peptides. Purified protein (input) and immunoprecipitated samples (IPs) were resolved by SDS-PAGE. Binding was monitored by western blotting. D, Peptide pull-down assays were performed with full-length HIS6-FLAG-Pdp3, mutants F18A, W21A, and F48A, truncations PWWP  , PWWP  , PWWP  , PWWP  , and C-term (74 -304) , and biotinylated histone peptides. Purified proteins (inputs) and immunoprecipitated samples (me0 and 36me3) were resolved by SDS-PAGE. Binding was monitored by western blotting. E, Fluorescence polarization assays were used to measure binding affinities of full-length S⅐tag-Pdp3 and mutants F18A and W21A to the indicated 5-FAM-labeled histone peptides. Error bars represent the S.D. of a representative experiment (n ϭ 2) performed in triplicate. Fig. S1). Because of the structural conservation with the BRPF1 PWWP domain, and the homology between the human and yeast HAT complexes, we reasoned that Pdp3 might bind H3K36me3. To test for this possibility, we interrogated the ability of full-length recombinant Pdp3 to bind to trimethylated histone peptides (Fig. 2C). Biotinylated peptides were immobilized on streptavidin resin and incubated with HIS6-FLAG-Pdp3. Immunoprecipitated samples were resolved by SDS-PAGE and visualized with antibodies against FLAG and streptavidin. Pdp3 preferentially engages H3K36me3 compared with all other trimethylated lysine residues tested (Fig. 2C).
The Pdp3 PWWP Domain is Necessary and Sufficient for Binding H3K36me3-We next wanted to determine whether the predicted PWWP domain of Pdp3 was sufficient for H3K36me3 binding. Uniprot defined the Pdp3 PWWP domain as the amino acid residues spanning 7-68 ( Fig. 2B) (78). However, we found that truncated Pdp3 (PWWP 1-74 ) is unable to bind H3K36me3 peptide (Fig. 2D), suggesting residues beyond the predicted PWWP domain are necessary for Pdp3 function. From the crystal structure of BRPF1 (56), we observed that two C-terminal alpha-helices support the aromatic cage and may be critical for engagement of H3K36me3 (supplemental Fig. S1). These alpha-helices are conserved in the structural model of Pdp3 (supplemental Fig. S1) (79), yet extend beyond the predicted PWWP domain (Figs. 2A, 2B). We created Pdp3 constructs to include increasing segments of the (modeled) C-terminal alpha-helices and tested the ability of H3K36me3 peptide to pull down PWWP 1-110 , PWWP 2-150 , and PWWP 1-219 . As shown in Fig. 2D, PWWP 2-150 restores binding to H3K36me3 and thus represents the functional PWWP domain of Pdp3 (Fig. 2B). These data suggest Pdp3, and likely other PWWP domain proteins, require extended alpha-helical regions for aromatic cage stability and function. Biophysical Characterization of the Interaction between Pdp3 and H3K36me3-To biophysically quantitate the specificity of the interaction between Pdp3 and H3K36me3, we performed fluorescence polarization assays using full-length recombinant S⅐tag-Pdp3 and 5-FAM-labeled histone peptides (Fig. 2E, Table  I). As expected from our pull-down assays, Pdp3 favors binding to the H3K36me3 peptide over other known targets of PWWP domain proteins, such as H3K79me3 and H4K20me3, with a K d of 69.51 Ϯ 3.7 M (Fig. 2E, Table I) (54,80). Notably, this value is one of the lowest reported dissociation constants for PWWP domain proteins that bind histones (supplemental Table S7) (54,56,57,80,81). We also generated Pdp3 mutants predicted to disrupt the aromatic cage and hence interaction with H3K36me3 (F18A and W21A). These mutants reduced H3K36me3 peptide binding ϳ10-fold compared with wild-type Pdp3 (Fig. 2E, Table I). Pdp3 also showed weak binding to H3K36me2 peptide with a K d of ϳ414 M (Fig. 2E, Table I). Importantly, the ϳsixfold increase in specificity of Pdp3 for H3K36me3, over H3K36me2, suggests Pdp3 has a distinct function from the Rpd3S HDAC complex, which preferentially engages H3K36me2 (18,41,42,45,82).
Pdp3 Requires H3K36me3 to Bind Chromatin-To further understand the biological role of Pdp3, we next determined whether Pdp3 binds H3K36me3 in vivo. We deleted the H3K36me3 methyltransferase, SET2, in the PDP3-TAP background and used a chromatin association assay to separate soluble proteins from those that bind to chromatin. For added controls, we also deleted the H3K4me3 methyltransferase, SET1, and the NuA3 H3K4me3 binding protein, YNG1. Strikingly, in the absence of Set2, and thus H3K36me3, we observe an almost complete ablation of Pdp3 binding to chromatin (Fig. 3A), indicating H3K36me3 is critical for Pdp3 localization in vivo. In contrast, Pdp3 remains bound to chromatin in the absence of Set1 (H3K4me3) and Yng1 (Fig. 3A). The observation that Pdp3 engages chromatin independently of both H3K4me3 and Yng1 suggests that Pdp3 targets NuA3 to active coding regions via an interaction with H3K36me3 (13,14,50). Interestingly, in the absence of Yng1, we note a slight decrease in the amount of chromatin-bound Pdp3 (Fig.  3A), as well as lower levels of cellular Pdp3 (supplemental Fig.  S2A). Together, these data suggest Pdp3 is moderately un- stable in yng1⌬ cells, perhaps because of an inability to be incorporated into NuA3. We next deleted SET1, SET2, and PDP3 in the YNG1-TAP background. In contrast to Pdp3, Yng1 remains bound to chromatin in the absence of either Set1 or Set2 (Fig. 3B). Furthermore, Yng1 levels are not significantly altered in pdp3⌬ cells (supplemental Fig. S2B). These results are consistent with past studies (26,27,83), and further suggest that while Yng1 can engage chromatin through both methyl-dependent and -independent means, the interaction between Pdp3 and chromatin requires H3K36me3.
Pdp3 is Required for NuA3-regulated Transcription-Mutation of the Yng1 PHD finger results in genome-wide mislocalization of NuA3 and decreased transcription of NuA3-regulated loci (19). To further assess the function of Pdp3 in vivo, we performed RT-qPCR with wild-type, pdp3⌬, and yng1⌬ cells, and calculated the relative transcript levels of NuA3regulated loci (Fig. 4A). As expected, yng1⌬ cells showed a slight but significant decrease in the relative transcript levels of NuA3 target genes, indicating a positive role for Yng1 in NuA3 activity at these genes (Fig. 4A) (19,84,85). Interestingly, pdp3⌬ cells also showed a reduced level of NuA3 target gene transcript (Fig. 4A). These data indicate that NuA3 bind-ing to both H3K4me3, via Yng1, and H3K36me3, via Pdp3, is required for proper NuA3-directed transcription.
PDP3 and NuA3 have Genetic Links to SET2 and Promote Transcription Elongation-Because Pdp3 binds H3K36me3 and is required for proper transcription in vivo, we predicted that Pdp3 would show genetic interactions with genes in the Set2/H3K36me pathway, and might influence transcription elongation. One genetic assay that has been used to measure the influence of Set2/H3K36 methylation on transcription A, RT-qPCR analysis of WT, pdp3⌬, and yng1⌬ cells to determine the relative transcript levels of NuA3-target genes and nontarget genes (SSA1 and SSB1). Transcript levels were normalized to Actin expression. Error bars represent the S.E. of a representative experiment (n ϭ 4) performed in triplicate. Asterisks indicate statistical significance as determined by an unpaired two-tailed t test. * ϭ p Ͻ .05, ** ϭ p Ͻ .01, *** ϭ p Ͻ .001. B, 2 ODs of the indicated yeast strains were 5-fold serially diluted on SC-Ura (left) or SC-Ura ϩ 5-FOA (right) plates and grown at 30°C for 2 or 3 days, respectively. C, 0.5 ODs of the indicated yeast strains were 5-fold serially diluted onto SC-Ura (left) or SC-Ura ϩ 150 g/ml 6-AU (right) plates and grown at 30°C for 2 days. elongation is the BUR1 bypass assay. Bur1, a cyclin-dependent kinase, acts as a positive regulator of transcription through the phosphorylation of several components of elongating RNAPII, including the CTD of RNAPII and the Spt5 C-terminal repeat domain (86 -88). Under normal conditions, deletion of BUR1 is lethal; however, deletion of SET2 or RCO1, a unique member of Rpd3S, bypasses this lethality (34,44). To explore whether Pdp3 and/or Yng1 contribute to transcription elongation, we examined deletions of SET2, RCO1, PDP3, and YNG1 in the bur1⌬ background. Although the bur1⌬set2⌬ and bur1⌬rco1⌬ strains bypass the lethality of BUR1 deletion as expected, neither bur1⌬pdp3⌬ nor bur1⌬yng1⌬ cells are viable (Fig. 4B). This suggests that either the bypass phenotype is too weak to observe, or Pdp3 and Yng1 function as positive elongation factors, and thus would not display a positive growth phenotype in this assay. To distinguish between these two possibilities, we created triple mutant strains with PDP3 or YNG1 deleted in the bur1⌬rco1⌬ background. Because this background shows a bypass phenotype, we could now observe positive or negative growth changes resulting from the absence of Pdp3 or Yng1. Significantly, the triple mutant strains both show a decrease in growth as compared with the bur1⌬rco1⌬ background (Fig. 4B). These data suggest that Pdp3 and Yng1 positively regulate transcription elongation, and are consistent with the idea that NuA3 functions both at promoters and at gene bodies.
To further connect NuA3 activity to the transcription elongation pathway, we asked whether the absence of NuA3 complex members resulted in sensitivity or resistance to the transcription elongation inhibitor 6-azauracil (6-AU). Strains containing a deletion of SET2, or downstream effectors, show a 6-AU resistance phenotype (34,44). As expected, strains lacking Set2, Rco1, and Eaf3, a member of the Rpd3S HDAC and NuA4 HAT complexes, display resistance to 6-AU as compared with wild-type (Fig. 4C). Of all tested NuA3 members, only pdp3⌬ cells show resistance to 6-AU, a result consistent with the idea that Pdp3 functions in the Set2/ H3K36 methylation pathway (Fig. 4C).
PDP3 is not Synthetically Lethal with GCN5-It has previously been determined that NuA3 complex members, Sas3 and Yng1, display synthetic lethality with the HAT, Gcn5 (89). These data indicate Gcn5 and Sas3 likely collaborate to promote gene activation and/or transcription elongation, as Gcn5 is found in both the promoters and transcribed regions of genes (90,91). Given the physical interaction and functional overlap between NuA3 and Pdp3, we wanted to determine if Pdp3 is also synthetically lethal with Gcn5. We created GCN5 shuffle strains that allowed us to delete YNG1 or PDP3 in a gcn5⌬ background. Upon shuffling out the wild-type GCN5 plasmid, only gcn5⌬yng1⌬ cells display synthetic lethality (supplemental Fig. S3). Surprisingly, gcn5⌬pdp3⌬ cells show no deleterious phenotype (supplemental Fig. S3), indicating PDP3 does not genetically interact with GCN5. These results suggest that Pdp3-containg NuA3 participates in an alternate transcriptional pathway that is distinct from the role of Gcn5 at promoters and in gene bodies. DISCUSSION In this study, we characterize a unique form of the NuA3 HAT complex that contains the PWWP domain protein, Pdp3. Using mass spectrometric, biochemical, and genetic approaches, our collective findings suggest that NuA3 exists in two functionally distinct forms: NuA3a and NuA3b (Fig. 5). This nomenclature has previously been used to distinguish different variations of related protein complexes. For example, although the Isw1a and Isw1b ATP-dependent chromatin remodeling complexes both have the same catalytic protein, Isw1, Isw1b contains a unique PWWP domain protein, Ioc4, required for targeting remodeling activity to H3K36me3 enriched nucleosomes (51,52). For NuA3 complexes, we hypothesize that promoter-associated NuA3a contains the proteins Eaf6, Nto1, Sas3, Taf14, and Yng1, and specifically associates with H3K4me3 using the PHD finger of Yng1 ( Fig.  5A) (19, 24 -27). NuA3a subsequently acetylates H3K14 through the HAT domain of Sas3, initiating transcription at a subset of genes (Fig. 5A) (19,24,25). In contrast, NuA3b, via the H3K36me3-binding PWWP domain protein Pdp3, links to H3K36 methylation-associated transcription elongation in gene bodies (Fig. 5B). Accordingly, our previous mass spectrometry (i-DIRT) data support the model that NuA3b is compositionally distinct from NuA3a (19). However, we cannot exclude the alternate possibility that Pdp3 is a member of both NuA3a and NuA3b. To this end, human H3K36me3binding BRPF proteins resemble a fusion of yeast Nto1 and Pdp3, as if these separate NuA3 proteins are physically linked within the human MOZ/MORF complexes (92). This brings up the intriguing possibility that the PWWP domain of Pdp3 (and potentially BRPFs) has a regulated capacity to bind H3K36me3, becoming active or inactive in a context dependent manner.
Our data clearly place Pdp3 and NuA3b within the Set2-dependent transcriptional elongation pathway. Pdp3 binds to H3K36me3, a Set2-catalyzed histone PTM found almost exclusively within transcriptionally active gene bodies (13, 33-35, 37, 40); and deletion of the PDP3 gene results in growth defects when combined with transcription elongation mutants. The interaction between Pdp3 and Rpb4, as determined from our mass spectrometry studies, further supports a role for NuA3b in transcriptional elongation, given the links between Rpb4 and actively transcribing RNAPII in S. pombe (93). Although our data suggest that NuA3b positively regulates transcription elongation, the exact function of NuA3b remains unclear. Because NuA3-directed transcription is decreased in pdp3⌬ cells, NuA3b could participate in acetylation-dependent nucleosome eviction within the ORF, similar to the proposed function of Gcn5 at coding regions (90,94,95). Although both Gcn5 and Sas3 favor acetylation of H3K14 (19, 24, 25, 96 -99), unlike NuA3a, NuA3b is not synthetically lethal with deletion of GCN5, suggesting Sas3, within the context of NuA3b, has a distinct function from Gcn5 (perhaps acetylating other factors in the transcription elongation complex or an alternate lysine(s) in histones) (Fig. 5B). This result, and the finding that HBO1 (a related human MYST family HAT) can switch between H4 and H3 acetylation depending on its association with JADE1/2/3 or BRPF1, supports the idea that NuA3b may target non-H3K14 substrates (100). It is also possible that Gcn5-mediated H3 acetylation has a distinct function from NuA3b acetylation at H3K14 in ORFs. Future work is needed to determine the precise mechanism of NuA3b function in the transcription elongation process.
A recurring question in chromatin biology is how the addition or removal of methylation groups on specific lysine residues can alter functional properties of the associated chromatin. In yeast, Set2 catalyzes all three forms of H3K36 methylation (33). To date, H3K36me1 is suggested to function in DNA replication (101), and although H3K36me2 and me3 are both linked to transcriptional elongation, H3K36me2 is essential for Rpd3S recruitment (18,(41)(42)(43)(44)(45)(46)(47), whereas H3K36me3 is implicated in nucleosomal positioning via Isw1b and repression of trans-histone exchange (50 -52). Interestingly, unlike H3K36me2, the establishment of H3K36me3 specifically requires Set2 association with the CTD of RNAPII, and H3K36me3 is positively correlated with the rate of transcription (13). Consistent with H3K36me3 links to highly transcribed genes, we speculate that H3K36me3, via association with NuA3b, may have an alternate function on NuA3-regulated genes that is distinct from former pathways determined for H3K36 methylation. The finding that the HDAC Rpd3S is linked to H3K36me2, whereas the HAT complex NuA3b is linked to CTD-dependent H3K36me3, suggests that differential methylation of H3K36 may act as a "chromatin switch" to regulate overall levels of transcription. For example, on lowly transcribed genes, the H3K36me2 state may predominate and recruit Rpd3S to maintain a more repressed chromatin environment suitable for low-level transcription. Conversely, on highly transcribed genes, H3K36me3 may predominate and recruit NuA3b to facilitate nucleosome disruption for RNAPII elongation (on genes regulated by NuA3). Alternatively, both H3K36me2 and me3 may be coordinating HDAC and HAT activities together, within the same coding region, to properly modulate the progression of RNAPII during transcriptional elongation. Although speculative, these intriguing models await testing.
Finally, NuA3 joins a growing list of HAT complexes that can localize to both promoters and gene bodies by modulating complex components (90,94,(102)(103)(104). Whether NuA3b engages a dual H3K4me/H3K36me signature, through the combinatorial action of the PHD finger of Yng1 and the PWWP FIG. 5. The NuA3 HAT complex has two functionally distinct forms that participate in transcription. A, Model of the NuA3a HAT complex. Yng1 binds to H3K4me3 through its PHD finger, thereby recruiting NuA3a to the promoter regions of actively transcribed genes. Sas3 then acetylates H3K14, leading to transcription initiation at a subset of genes (19, 24 -27). B, Model of the NuA3b HAT complex. NuA3b contains a unique member, Pdp3. Pdp3 binds to H3K36me3 through its PWWP domain, thereby recruiting NuA3b to the coding regions of actively transcribed genes. Although the function of NuA3b is not fully defined, we speculate that Sas3 may acetylate histones or nonhistone proteins to facilitate transcription elongation. domain of Pdp3, to fine-tune transcription initiation and elongation pathways, remains an important question for future studies. It is worth noting that mutations of NuA3 human homologs (23,56) are associated with oral squamous cell carcinoma (105) and acute myeloid leukemia (106). Given that HATs and many other chromatin-bound protein complexes are gaining favor as pharmaceutical targets (107)(108)(109)(110), future studies into the PWWP domain of Pdp3 and NuA3b may elucidate how chromatin effectors contribute to human disease, thereby opening up new targets for drug discovery.