Asf1-like structure of the conserved Yaf9 YEATS domain and role in H2A.Z deposition and acetylation

Abstract

Chromatin can be modified by posttranslational modifications of histones, ATP-dependent remodeling, and incorporation of histone variants. The Saccharomyces cerevisiae protein Yaf9 is a subunit of both the essential histone acetyltransferase complex NuA4 and the ATP-dependent chromatin remodeling complex SWR1-C, which deposits histone variant H2A.Z into euchromatin. Yaf9 contains a YEATS domain, found in proteins associated with multiple chromatin-modifying enzymes and transcription complexes across eukaryotes. Here, we established the conservation of YEATS domain function from yeast to human, and determined the structure of this region from Yaf9 by x-ray crystallography to 2.3 Å resolution. The Yaf9 YEATS domain consisted of a β-sandwich characteristic of the Ig fold and contained three distinct conserved structural features. The structure of the Yaf9 YEATS domain was highly similar to that of the histone chaperone Asf1, a similarity that extended to an ability of Yaf9 to bind histones H3 and H4 in vitro. Using structure–function analysis, we found that the YEATS domain was required for Yaf9 function, histone variant H2A.Z chromatin deposition at specific promoters, and H2A.Z acetylation.

Histones are the major protein constituent of chromatin and can be modified in several fundamental ways, including the addition of posttranslational modifications, ATP-dependent chromatin remodeling, and incorporation of histone variants (1). H2A.Z, encoded by the HTZ1 gene in Saccharomyces cerevisiae, is an H2A variant with roles in transcriptional repression and activation, chromosome segregation, DNA replication and repair, and heterochromatin-euchromatin boundary formation (2). H2A.Z is incorporated into nucleosomes by the conserved ATP-dependent chromatin remodeling complex SWR1-C, the first complex discovered to be dedicated to variant histone deposition (3–5). SWR1-C shares 4 subunits with the NuA4 histone H4 acetyltransferase complex, which, among other substrates, also acetylates H2A.Z to restrict spreading of heterochromatin and to prevent chromosome missegregation (6–8). One of the shared subunits, Yaf9, is important for cellular response to spindle stress, proper DNA repair and metabolism, H2A.Z chromatin deposition and acetylation, and histone H4 acetylation at telomere-proximal genes (7, 9–11). Yaf9 contains an evolutionarily conserved YEATS domain found in more than 100 different eukaryotic proteins (12).

In humans, GAS41 is the closest relative of Yaf9 based on sequence identity, domain organization, and its presence in human SRCAP and TIP60 complexes, which respectively correspond to yeast SWR1-C and NuA4 (9, 11). Yaf9 and GAS41 have an N-terminal YEATS domain followed by a conserved region of unknown function called the A-box and a C-terminal coiled-coil domain (11). Interestingly, YEATS domain proteins have several connections to cancer. First, GAS41 is highly amplified in human glioblastomas and astrocytomas (13) and is required for repression of the p53 tumor suppressor pathway during normal cellular proliferation (14). Second, 2 other YEATS domain containing proteins, ENL and AF9, are frequent fusion partners of the mixed-lineage leukemia protein in leukemia (15).

This study established the conservation of YEATS domain function from yeast to human. To better understand the physical organization of the YEATS domain, we determined the structure of this region from Yaf9 to 2.3 Å resolution. Interestingly, the Yaf9 YEATS domain structure was highly similar to the structure of the Asf1 histone chaperone, a congruence that extended to an ability of Yaf9 to bind histones H3 and H4 in vitro. Futhermore, in yeast the YEATS domain was required for Yaf9 function, H2A.Z deposition at specific promoters, and H2A.Z acetylation.

Results

YEATS Domain Function Was Conserved from Yeast to Human.

The primary sequence similarity between human GAS41 and yeast Yaf9 predicted a functional conservation. We tested whether any of the 3 domains (YEATS domain, A-box, and coiled-coil) of GAS41 could substitute for its Yaf9 counterpart in yeast. Genes encoding for hybrid proteins, each carrying a distinct module from either GAS41 or Yaf9 (Fig. 1A), were tested for their ability to complement growth phenotypes caused by loss of YAF9. For complete list of yeast strains, see Table S1. Strains lacking Yaf9 are sensitive to chemical stressors including formamide, hydroxyurea (HU), and benomyl (12). Hybrid proteins carrying the YEATS domain of GAS41 and at least the coiled-coil region of Yaf9 rescued most of the growth defects caused by loss of Yaf9 (Fig. 1B). In contrast, hybrid proteins carrying the GAS41 coiled-coil and the Yaf9 YEATS domain were unable to support growth of strains lacking Yaf9, irrespective of the source of the A-box (Fig. 1B). Expression of the human GAS41-protein construct alone was unable to complement any of the phenotypes caused by loss of YAF9 (Fig. 1B), as was expected from the GAS41-Yaf9 hybrid results. The function of the GAS41-Yaf9 hybrid proteins reflected their incorporation into SWR1-C and NuA4. Functional TAP-GAS41-Yaf9 hybrids and full length TAP-Yaf9 protein all copurified with SWR1-C subunits Swr1, Swc2, and Swc3, as well as NuA4 subunits Eaf1, Epl1, and Tra1 (Fig. 1C). In contrast, much less SWR1-C and NuA4 copurified with hybrids containing the GAS41 coiled-coil or with full-length TAP-GAS41.

Fig. 1.

YEATS domain function was conserved from yeast and human. (A) Schematic representation of Yaf9-GAS41 hybrid proteins constructed with modules originating from Yaf9 in green and GAS41 in yellow. The nomenclature indicates the origin of a given module where “G” refers to GAS41 and “Y” to Yaf9, in order of YEATS domain, A-box, and coiled-coil. For simplicity, the N-terminal TAP-tag present in all constructs was omitted. (B) GAS41 YEATS domain conferred resistance to genotoxic stress in yeast when fused to Yaf9 coiled-coil. Tenfold serial dilutions of yaf9Δ strains carrying the indicated plasmids were plated and incubated on CSM plates with the following concentrations of chemicals: 2% formamide, 30 μg/ml benomyl and 75 mM HU. (C) GAS41 YEATS domain fused to Yaf9 coiled-coil was incorporated into SWR1-C and NuA4. Analytical-scale affinity purifications of TAP-Yaf9-GAS41 hybrids from cells containing affinity-tagged versions of 3 representative SWR1-C and 3 representative NuA4 subunits were performed and immunoblotted for the indicated proteins.

The YEATS Domain of Yaf9 Adopted an Ig Fold.

To better understand the function of the evolutionary conserved Yaf9 YEATS domain, we determined its high-resolution structure using x-ray crystallography. The crystallized fragment (Yaf9 amino acids 8–171 determined by mass spectrometry) contained the YEATS domain and A-box, suggesting that these components of Yaf9 form a stable structural element, with a less-structured linker connecting to the coiled-coil domain.

The Yaf9 YEATS domain crystallized in the space group P6₅22. The structure of the selenomethionine-substituted protein was determined by multiwavelength anomalous dispersion, and refined to a final resolution of 2.3 Å (Table 1). Three copies of the YEATS domain were present in the asymmetric unit. The final model included residues 8–119 and 143–169 for each of the 3 protomers, and was refined to an R_work/R_free of 21.8/25.8% with no residues in disallowed regions of Ramachandran space (Table 1). The 3 protomers were highly similar to each other, displaying an average rmsd of ≈0.25 Å across all atoms.

Table 1.

Data collection, phasing analysis, and refinement statistics

Yaf9(8-171)	Native	Se-Derivative
	λ	Se-λ1	Se-λ2
Wavelength, Å	1.11587	0.9795	1.0199
Space group	P6522	P6522
Cell dimensions, Å	84.725, 288.212	84.582, 286.531	84.636, 286.639
Resolution, Å	50–2.3 (2.42–2.3)	50–2.90 (3.02–2.90)	50–2.90 (3.02–2.9)
Redundancy	6.3 (5.0)	3.8 (3.8)	3.8 (3.8)
Completeness, %	95.8 (92.8)	100 (100)	100 (100)
Rsym,* %	7.5 (36.8)	13.5 (43.1)	12.3 (41.4)
I/σ (I)	16.2 (4.4)	8.7 (2.9)	9.2 (3.0)
	Phasing analysis
Resolution		40-2.9
No. sites		6
Mean figure of merit		0.36
	Refinement statistics
Resolution, Å		20-2.3
Rwork,† %		21.8
Rfree,‡ %		25.8
RMSD bonds, Å		0.009
RMSD angles, °		1.143
No. protein atoms		3671
No. water atoms		200
	Ramachandran % (no res.)
Most favored		89.7
Allowed		10.0
Gen. allowed		0.3

*R_sym=Σ|l_i− <I_i>|/Σ|l_i|, where I_i is the scaled intensity of the i^thmeasurement and <I_i> is the mean intensity for that reflection.

^†R_work=‖F_o|−|F_c‖/|F_o|, where F_o and F_c are the observed and calculated structure factor amplitudes, respectively.

^‡R_free was calulated as R_work using 5% of randomly selected data not included in refinement.

The YEATS domain folded into an elongated β-sandwich consisting of 8 antiparallel β-strands capped on one end by 2 short α-helices (Fig. 2 A and B). Structural comparison of the YEATS domain against the protein databank using the DALI server showed that it adopted an immunoglobulin (Ig) fold (16). Ig folds are common macromolecular interaction modules, and are found in proteins with a variety of cellular functions (17). The YEATS domain was structurally homologous to the Ig folds of a broad number of factors (over 200 hits with Z scores smoothly varying from 3 to 6), with no evidence of any particular standout among fold homologs.

Fig. 2.

Secondary structure and amino acid conservation of YEATS domain. (A) Sequence alignment of YEATS domains showing amino acid conservation. Secondary structure elements and specific classes of mutations are labeled. (B) Ribbon diagram of the Yaf9 YEATS domain. Structural elements are labeled and conserved sequence motifs highlighted. This and all other molecular graphics figures were generated using PYMOL (29). (C) View of the 3 YEATS domains protomers (green, cyan, and yellow) as related by noncrystallographic symmetry. The extended N-terminal tail of each domain docks into the hydrophobic groove of a neighboring molecule, forming a pseudocontinuous sheet with strand β7 across the domain. (D) Surface view of the Yaf9 YEATS domain (boxed region in C) highlighting the conserved cleft motifs (purple, magenta, and pink), and hydrophobic groove (yellow). The N terminus of an adjacent protomer is shown as cyan sticks.

Structural Features of the YEATS Domain.

Our structure revealed 3 interesting physical features of the YEATS domain. One feature was the presence of a highly conserved cleft located on the end of the Ig fold opposite the 2 capping helices (Fig. 2 B and D). This region was composed of 3 surface-exposed loops emanating from the core β-sheet region. The first was a His-Thr-His (HTH) triad, which preceded strand β2 and was invariant in Yaf9/GAS41 family members but not in Taf14 and Sas5, the 2 other YEATS domain-containing proteins in yeast (12). The other 2 loops included a Leu-His-X-Ser/Thr-Tyr/Phe [LHx(S/T)(Y/F)] pentad that connected strands β3 and β4, and a Gly-Trp-Gly sequence wedged between strands β5 and β6. Both of these segments were essentially invariant across all YEATS proteins and constituted the defining signature sequence motifs of the clade. The presence of this conservation within the external loops of an Ig fold was intriguing, as these regions often form the principal surface used by such proteins to engage client molecules. A second feature consisted of a relatively shallow groove near the N- and C-termini of the YEATS domain (colored yellow in Fig. 2D). Formed in part by the capping helices, the groove was relatively nonpolar and displayed only a modest degree of surface amino acid conservation. The groove was distinguished, however, by the presence of a narrow, but deep, hydrophobic pocket that extruded approximately 7–8 Å down into the core of the β-sheet region (Fig. 2D; see also Fig. 4A). The hydrophobic character of the groove and pocket was conserved among all YEATS domains. Moreover, in the crystal, the groove served to bind the extended N-terminal arm of an adjacent YEATS domain protomer, such that the N-terminal tail formed an extra strand of the β4–3-6–7 sheet, in trans. This interaction was recapitulated between the 3 YEATS domains present in the asymmetric unit to form a trimeric array of protomers with rotational, threefold noncrystallographic symmetry (Fig. 2C). Although the YEATS fragment we crystallized was monomeric in solution (as determined by gel filtration and dynamic light scattering), the surface features of the groove and its ability to associate with the N terminus of a partner molecule suggested that the grove may also constitute a peptide binding surface. The third feature, between the cleft and putative peptide-binding groove, was a region rich in conserved charged residues, particularly basic amino acids. Although the charged patch did not have any notable structural features, such as a deep depression or any particularly marked curvature, it was one of the most electropositive surfaces on the YEATS domain (Fig. 4B).

Fig. 4.

Mutations in conserved surface residues in the Yaf9 YEATS domain affected protein function. (A) Yaf9 _(8–171) surface representation showing YEATS residue conservation as mapped by Consurf (dark green = invariant, light green = conserved, white = variable)(31), based on an alignment of more than 30 Yaf9 and GAS41 orthologs (SAS5 and Taf14 were excluded). The peptide (cyan sticks)/hydrophobic groove interaction seen between Yaf9 protomers is highlighted. (B) Surface stereorepresentation of Yaf9 _(8–171) showing charge distribution (red = negative, blue = positive, contoured at +/− 5k_BT by APBS)(32). The view is rotated approximately 60° to that shown in A. (C) Surface of Yaf9 _(8–171) showing the where the 3 mutant classes map with respect to the structure and each other. The orientation is the same as in (A). (D) Cells with yaf9 alleles of Classes A and B were sensitive to genotoxic stressors. (E) Yaf9 YEATS domain was involved in similar process with Asf1. The alleles carrying mutations in the charged surface or conserved cleft were unable to rescue growth defects of the yaf9Δasf1Δ double mutant. Note that the yaf9-4, yaf9-23 and yaf9-28 alleles caused approximately a tenfold greater growth retardation of the asf1Δ yaf9Δ strain than empty vector alone in cells grown at 37 °C.

Yaf9 Shared Structural and Biochemical Properties with the Asf1 Histone Chaperone.

In the course of analyzing the Yaf9 YEATS domain, we noted a marked similarity to the histone chaperone Asf1 (Fig. 3 A and B). Like Yaf9, the core region of Asf1 is also predicated on an Ig fold. Moreover, of the several different topological groups of Ig fold categorized to date, both proteins belong to the same “switched” Ig subclass (Fig. S1). Although the surface features of the 2 proteins are relatively distinct—for example, Asf1 lacks the deep conserved cleft formed by apical loops of the Yaf9 Ig fold—there were some intriguing similarities beyond the level of fold topology. In particular, Asf1 binds short peptide segments using the edges of its β-sheets, such as with the Hir1 B-box of CAF1 or the C-terminal tail of histone H4 (18–20). In our structure, the Yaf9 YEATS domain used one of its β-sheets and a hydrophobic groove to associate with the N-terminal peptide of an adjoining protomer in a similar fashion (Fig. 2C).

Fig. 3.

Yaf9 YEATS domain structure was similar to histone chaperone Asf1. (A) Topology diagram of the Yaf9 YEATS domain and Asf1 switched-type Ig folds. Conserved β-strands are shown as purple arrows. A swap of the last strand has occurred between the 2 proteins, pairing the “h” strand of Yaf9 and Asf1 with either the “a-b-e” sheet (gray) or “g-f-c-d” sheet (green). The corresponding β-strands (30) are shown. (B) Structural comparison of the Yaf9 YEATS domain and Asf1 core. The ribbon diagram is rainbow colored from the N terminus (blue) to the C terminus (red) of each protein. (C) YAF9 and ASF1 interacted genetically. Cells lacking both ASF1 and YAF9 were hypersensitive to high temperature (37 °C) and HU. (D) YAF9 and HTZ1 interacted genetically with H3K56R. Cells expressing H3K56R in combination with either yaf9Δ or htz1Δ were hypersensitive to HU. Cells expressing H3K56R htz1- K3,8,10,14R showed no obvious synthetic interaction. (E) Yaf9 bound to histones H3 and H4 in vitro. Calf thymus histones bound to the indicated amounts of purified GST or GST-Yaf9 from yeast as judged by immunoblotting with α-H3, α-H4, and α-H2B histone antibodies.

To test whether the structural similarity between Yaf9 and Asf1 was reflected in a genetic interaction, we used an assay for which double mutant haploid segregants were obtained from a diploid strain with only one copy of ASF1 and YAF9. The asf1Δ yaf9Δ double mutants were viable but grew dramatically more slowly than either single mutant, particularly at a higher temperature or on media with low levels of HU (Fig. 3C). Consistent with ASF1 being required for acetylation of H3K56 (21), the yaf9Δ H3K56R double mutant had a synthetic growth defect (Fig. 3D). Moreover, this interaction was likely due to Yaf9's role in H2A.Z deposition but not H2A.Z acetylation, as the htz1Δ H3K56R double mutant also exhibited a synergistic growth defect while the htz1K3,8,10,14R H3K56R double mutant did not (Fig. 3D).

Given the diversity of folds involved in chromatin remodeling, it was unexpected to find that the Yaf9 YEATS domain was a structural homolog of Asf1, a chaperone for histones H3-H4. Consistent with the structural similarity between the Yaf9 YEATS domain and Asf1, in vitro protein-interaction assays showed that GST-Yaf9 bound to histones H3 and H4 (Fig. 3E). These associations were not likely to reflect nonspecific binding to basic charged proteins as Yaf9 did not bind H2B.

Conserved Residues on the YEATS Domain Surface Were Important for Yaf9 Function.

To better understand the role of the YEATS domain in vivo, we constructed 3 classes of mutant Yaf9 proteins. Each mutant carried multiple amino acid substitutions of conserved residues on the surface of one of the 3 structural features. Class A mutants were in the charged surface region of the protein (yaf9-2, yaf9-4, and yaf9-24). Class B mutants were in the conserved cleft (yaf9-1, yaf9-3, yaf9-34, along with yaf9-23 and various dissections of it: yaf9-26, yaf9-27, yaf9-28, yaf9- 29, yaf9-30, yaf9-31, yaf9-32 and yaf9-33). Class C mutants were in the groove that associated with the N-terminal segment of Yaf9 with potential to be a peptide binding groove (yaf9-17, yaf9-18, yaf9-19, yaf9-20, yaf9-21, and yaf9-22). The locations of the altered residues in the proteins encoded by yaf9 alleles are shown in Figs. 2A and 4C, with an exact description in Table S2.

Yaf9 YEATS domain mutants were tested for their ability to complement the sensitivity of yaf9Δ strains to genotoxic agents formamide, HU and benomyl. Drug-sensitivity phenotypes were conferred by mutations in 2 of the 3 conserved areas to varying degrees, with Class A allele yaf9-4 and Class B alleles yaf9-23, yaf9-27, yaf9-28 having growth defects similar to those of yaf9Δ (Fig. 4D). The yaf9-1 strain had growth defects similar to those of yaf9Δ on formamide and HU, but grew similar to wild type on benomyl, while yaf9-3 and yaf9-34 strains had growth defects only on formamide and grew comparable to wild type on HU and benomyl. These results hinted at regional specialization of function, with yaf9-1 being defective in processes leading to formamide and HU sensitivity, whereas the other Class B alleles, yaf9-3 and yaf9-34, were defective in a process leading only to formamide sensitivity. Strains with mutations in the putative peptide-binding pocket (Class C) had no discernable phenotypes (Fig. 4D), suggesting this feature played no discernable role in the Yaf9-dependent functions tested here. In addition to the 20 alleles shown in Fig. 4D and Fig. S3A, 13 alleles with mutations in other conserved residues were tested without revealing noticeable phenotypes (Table S3).

The amount of Yaf9 mutant protein was similar to the level in wild-type strains in most mutants, including the yaf9-1 and yaf9-3 strains (Fig. S2A), further supporting the view that the different phenotypes resulted from qualitative rather than quantitative differences in Yaf9 function. Overexpression experiments of alleles encoding Yaf9 protein with normal or reduced levels showed that the phenotypes of these mutants were not solely due to reduced protein level and that many of the yaf9 alleles behaved as hypomorphs (Fig. S2B).

The enhanced growth defect of the yaf9Δ asf1Δ double mutant was contributed by the YEATS domain, as a subset of yaf9 alleles were unable to complement the loss of YAF9 in an asf1Δ strain and failed to restore growth on HU and higher temperature (Fig. 4E). Therefore, as the structural similarity implied, it was indeed the YEATS domain, and in particular the charged surface region and cleft of Yaf9, whose function was linked to Asf1. Interestingly, the subset of yaf9 alleles with lower protein levels (yaf9-4, yaf9-23, and yaf9-28) consistently caused approximately a tenfold greater growth retardation of the asf1Δ yaf9Δ strain than empty vector alone in strains grown at 37 °C (Fig. 4E). Thus, although the absence of ASF1 magnified the growth defect of yaf9Δ strains, it magnified the growth defect of the yaf9-4, yaf9-23, and yaf9-28 mutants even further.

Yaf9 YEATS Domain Functioned in Both SWR1-C and NuA4.

To examine the role of the Yaf9 YEATS domain in histone variant H2A.Z biology, we assayed the activity of yaf9 mutants biochemically. Bulk chromatin fractionation assays showed that, as expected for cells lacking a component of the SWR1-C, yaf9Δ strains had reduced H2A.Z levels in the chromatin pellet and increased levels in the nonchromatin supernatant fraction as compared to wild type (Fig. 5A). In contrast, the level of H2A in chromatin and the level of Pgk1 in supernatant were unchanged. Strains with Yaf9 mutations located in the charged surface area (Class A allele yaf9-4) or in the conserved cleft (Class B alleles yaf9-1, yaf9-23, yaf9-27, yaf9-28, and yaf9-34) had less H2A.Z in the chromatin pellet, comparable to the yaf9Δ strain (Fig. 5A and Fig. S4). Therefore, residues in the conserved charged surface (Class A) and cleft (Class B) of the Yaf9 YEATS domain were important for H2A.Z deposition. To further explore the conclusions derived from these crude chromatin association assays, the requirement of the Yaf9 YEATS domain for H2A.Z deposition at specific promoters was determined by ChIP-on-Chip. H2A.Z occupancy was compared between YAF9 and the yaf9-1 and yaf9-3 mutants, which were chosen based on their differences in growth phenotypes (Fig. 4D). Consistent with previous studies, H2A.Z was present at 2,928 promoters in wild-type strains, using our enrichment criteria (Fig. 5 B and C) (2). In contrast, H2A.Z-ChIP efficiency in both yaf9Δ and yaf9-1 mutants was equal to a nonantibody control resulting in background peaks (Fig. 5C), as reported previously for other SWR1-C subunits (22). Interestingly, in the yaf9-3 mutant, H2A.Z was specifically lost at about one-third of promoters but was present at almost wild-type levels at the other two-thirds (Fig. 5 B and C). Further supporting the functional link between Yaf9 and Asf1, promoters reported to contain H3K56ac (23) preferentially lost H2A.Z in the yaf9-3 mutant (P value < 10⁻⁸).

Fig. 5.

Yaf9 YEATS domain was required for H2A.Z chromatin deposition and acetylation. (A) Yaf9 YEATS domain mutant strains had decreased H2A.Z-FLAG chromatin deposition. W, whole cell extract; S, supernatant; C, chromatin pellet. The relative amount of H2A.Z-FLAG in each fraction was determined by immunoblotting in the different strains. Antibodies against histone H3 and Pgk1 were used as loading controls for chromatin pellet and supernatant, respectively. (B) ChIP-on-chip profiles of H2A.Z in YAF9 and yaf9-3 strains. Sample genomic positions for chromosomes 4 and 8 were plotted along the x axis against the relative occupancy of H2A.Z on the y axis. ORFs are indicated as light gray rectangles above the x axis for Watson genes and below the x axis for Crick genes. ARS are indicated as dark gray rectangles. Regions considered enriched above a certain threshold are shown as colored bars on the x axis. (C) Mutation in Yaf9 YEATS domain resulted in loss of H2A.Z at specific promoters. Shown is the number of H2A.Z enriched promoters determined by ChIP-on-Chip in YAF9 and yaf9-3 strains. Both yaf9Δ and yaf9-1 strains had low immunoprecipitation efficiency resulting in no detection of H2A.Z enriched regions above background level. (D) Mutations in the Yaf9 YEATS domain resulted in decreased H2A.Z K14ac in chromatin. Chromatin extracts from the bulk chromatin association assays shown in A were immunoblotted with an antibody against H2A.Z K14ac. Antibodies against Flag and H4 were used as loading controls.

Acetylation of H2A.Z on K14 is catalyzed, at least in part, by NuA4 following H2A.Z deposition into chromatin (6–8). Yaf9 is required for H2A.Z acetylation in vivo; however, it is not known whether this dependency solely reflects Yaf9's contribution to SWR1-C, Yaf9's contribution to NuA4 function, or both (7). To determine the role of the Yaf9 YEATS domain in H2A.Z acetylation, we used the chromatin-associated H2A.Z fraction that remained in each yaf9 mutant (Fig. 5A) and measured acetylation at K14. This approach allowed for the evaluation of NuA4-dependent activities of Yaf9 mutants uncoupled from their H2A.Z deposition effects. Strains carrying yaf9 alleles in the charged surface region (yaf9-4) or in the conserved cleft (yaf9-1, yaf9-3, yaf9-23, yaf9-27, yaf9-28, and yaf9-34) exhibited reduced H2A.Z K14 acetylation, similar to that of yaf9Δ strains (Fig. 5D).

Discussion

YEATS domain proteins are found in many protein complexes involved in chromatin biology and are linked to cancers in humans. Despite these intriguing connections, little is known about the specific functions of YEATS domains, or their interaction partners and structure. Here, we demonstrate that there was a functional conservation between the YEATS domain of yeast Yaf9 and human GAS41. The extent of incorporation of the YEATS domain hybrid proteins into SWR1-C and NuA4 closely paralleled their ability to function in place of Yaf9. Furthermore, these results confirmed earlier findings that the coiled-coil of Yaf9 is required for interactions with NuA4 (11), and extended these findings to define the coiled-coil region to also be important for Yaf9 to associate with SWR1-C. Likewise, the GAS41 coiled-coil is required for interaction with the human TIP60 and SRCAP complexes (14). The ability of the human GAS41 YEATS domain to complement in yeast showed that YEATS domain function was evolutionary conserved and implied that its structure was very similar.

The first structure of a YEATS domain reported here revealed that this domain in Yaf9 consisted of an Ig fold with 3 distinct conserved surface features, of which at least 2 were important for function. The strong defects in H2A.Z chromatin deposition caused by mutations in Yaf9 established that the YEATS domain was required for the proper activity of SWR1-C. Promoter-specific measurements of H2A.Z occupancy in 2 yaf9 mutants suggested that different amino acid changes in the YEATS domain had distinct effects on H2A.Z deposition. The phenotypically moderate yaf9-3 mutant surprisingly lost H2A.Z at a particular set of promoters, while retaining normal levels at the remaining promoters. This contrasted with the more severe yaf9-1 mutant, which completely lost H2A.Z at all promoters, similar to the yaf9Δ. Furthermore, the YEATS domain was important for H2A.Z K14ac by NuA4 in chromatin, which most likely occurs after H2A.Z has been deposited by SWR1-C (6, 7). A common function of Yaf9 as part of the module of subunits shared between SWR1-C and NuA4 (Yaf9, Swc4, Arp4, and Act1) was consistent with recent data suggesting that this module has similar roles in both complexes (24).

From a conceptual viewpoint, the results from the mutant analysis aimed at determining the functional relevance of the 3 conserved YEATS domain features were counterintuitive. A priori, one might assume that a protein such as Yaf9, which must function in 2 different complexes, would be even more constrained than a protein that acts on its own or in only one complex and hence might be more vulnerable to mutation. However, of the 33 mutant alleles that we constructed, 26 had no discernable phenotype despite the 40 mutated amino acids among them. Indeed, no phenotype was observed from single amino acid substitutions even though these were conserved residues. This was an unexpectedly low frequency of phenotypes, especially considering the functional conservation of the YEATS domain from yeast Yaf9 to human GAS41 reported here. Of the 7 mutants with phenotypes, 4 exhibited reduced protein levels (yaf9-4, yaf9-23, yaf9-27, and yaf9-28), and 3 (yaf9-1, as well as yaf9-3 and yaf9-34; differing by the I94M substitution) had normal protein levels. These results have practical significance for forward genetic studies suggesting that even for nonessential subunits of protein complexes like SWR1-C, discovering their function from mutant screens could be unexpectedly difficult.

Given the lack of amino acid sequence similarity, it was unexpected to discover that the Yaf9 YEATS domain was a structural homolog of the histone chaperone Asf1. Similar to Asf1, Yaf9 bound to histones H3-H4 in vitro. The enhanced growth defect of the yaf9Δ asf1Δ mutant indicated an involvement of Yaf9 and Asf1 in a similar process. More precisely, we determined that loss of Asf1-dependent H3K56 acetylation caused enhanced growth defects in the absence of either Yaf9 or H2A.Z but had no effect in the absence of H2A.Z acetylation. This suggested that it was primarilyYaf9's SWR1-C dependent role in H2A.Z deposition, and not its role in NuA4-mediated H2A.Z acetylation, that was affected by loss of Asf1. The lack of interaction with the N-terminal lysines of H2A.Z contrasts with the previously reported requirement for Asf1 and H3K56 acetylation in cells with mutations in the N-terminal lysines of H3 or H4 (25).

The similar process that Asf1 and Yaf9 ordinarily facilitate might be reflected in our finding that 3 yaf9 alleles with lower protein levels had greater growth retardation than the yaf9Δasf1Δ strain with empty vector. It would appear that the 3 mutant proteins interfered with whatever process Asf1 and Yaf9 converge on. A possible explanation may involve nucleosome assembly and disassembly by Asf1, presumably through its ability to bind to H3-H4 dimer intermediates (25, 26). SWR1-C performs a conceptually similar function, disassembling H2A-H2B dimers from nucleosomes and reassembling them with H2A.Z-H2B dimers (5). Perhaps in the absence of Asf1, SWR1-C promotes a more extensive dismantling and reassembly of nucleosomes in a Yaf9-dependent way. If so, the paradoxical phenotype of yaf9-4, yaf9-23, and yaf9-28 being more defective than yaf9Δ in cells lacking Asf1 might reflect their competence in only the disassembly reaction, leaving chromatin in a more compromised state than if Yaf9 were completely absent.

We note that histone binding might be a common feature of YEATS domains since human ENL, through its YEATS domain, specifically binds to histones H3 and H1, without having affinity for H4, H2A, and H2B (27). However, our data suggested that additional relevant targets must exist to facilitate YEATS domain-dependent activities, at least in the case of Yaf9, because mutations in the YEATS domain of Yaf9 that affected H2A.Z chromatin deposition still bound to histones H3 and H4 (Fig. S5). This eliminates a simple model by which diminished interaction of Yaf9 with H3 or H4 results in loss of Yaf9-dependent H2A.Z chromatin deposition and raises the need for more detailed binding studies.

In summary, this study established the structure of the YEATS domain, its evolutionary conservation, a close structural and functional relationship with Asf1, and a differential requirement of the Yaf9 YEATS domain for H2A.Z deposition at specific genes.

Methods

The details of structural, biochemical, and genetic experiments are described in SI Materials and Methods. In short, recombinant His-Yaf9 (amino acids 8–171) was purified from Escherichia coli using nickel affinity and size exclusion chromatography. Model building, structure refinement, superposition, and structure alignment were done using publicly available software described in SI Materials and Methods. Yeast growth assays, analytical-scale affinity purifications, ChIP-on-Chip using Affymetrix tiling arrays, and data analysis were performed as described previously (3, 28).