Cryo-EM structures reveal the acetylation process of piccolo NuA4

Significance

Histone acetylation is one of the most important epigenetic modifications that can induce structural changes in chromatin and regulate gene transcription. NuA4 is the only essential acetyltransferase in yeast that can catalyze the acetylation of histones H2A, H2A.Z, and H4. However, the process by which pNuA4 acetylates its multiple substrates within a nucleosome remains unclear. Here, we solved a series of cryo-EM structures of pNuA4 with various H2A.Z-containing nucleosomes, revealing different acetylation states. Based on the solved cryo-EM structures, we performed a series of single-molecule Förster resonance energy transfer (smFRET) assays. The results demonstrated that pNuA4 adopts a dynamic process to search for its substrates and prefers to access H4 with substantially higher probability than H2A.Z.

Abstract

NuA4 is the only essential acetyltransferase in yeast that can catalyze the acetylation of the histones H2A, H2A.Z, and H4, thereby affecting gene transcription. However, the acetylation process of NuA4, such as how NuA4 acetylates H4 and H2A.Z differently, remains largely elusive. Here, using cryoelectron microscopy (cryo-EM) single particle analysis, we present seven cryo-EM structures of piccolo NuA4 (pNuA4) in complex with wild-type H2A.Z or H2A.Z-mutant-containing nucleosomes in the absence or presence of acetyl coenzyme A (Ac-CoA). We revealed that, in the absence of Ac-CoA, pNuA4 adopts multiple conformations to search for its substrates. After adding Ac-CoA, the single-molecule Förster resonance energy transfer (smFRET) and cryo-EM data indicated that pNuA4 prefers to bind H4 and undergoes a dynamic conformational change to complete the acetylation. We also obtained previously unseen structures in states associated with the acetylation of H2A.Z. These cryo-EM structures and smFRET results suggest a complex acetylation process on H4 and H2A.Z by pNuA4. The results provide a comprehensive picture of the mechanism by which pNuA4 acetylates its substrates within an H2A.Z-containing nucleosome.

The high-resolution crystal structure of a single nucleosome explains in detail how histone octamers are assembled and how the 147 base-pair DNA surrounds the histone octamers (1). Through posttranslational modifications such as acetylation, phosphorylation, and methylation, the N terminus of core histones can trigger specific biological functions (2, 3). Acetylation is one of the earliest studied posttranslational modifications. Histone acetylation is mainly found on enhancers, promoters, and gene bodies (4). It neutralizes the positive charge of histone N-terminal lysine residues and reduces their affinity for DNA, allowing the DNA to be exposed and recruiting other proteins to initiate transcription (5–7).

H2A.Z, the only histone H2A variant in yeast, is encoded by HTZ1 and is highly conserved through evolution (8, 9). It plays important roles in transcription, DNA repair, and chromosome stability through acetylation (10–13). The absence of H2A.Z can cause condition-specific growth phenotypes, as well as sensitivity to chemicals like hydroxyurea, caffeine, formamide, and benomyl in yeast (14–16). The genomic localization of most H2A.Z has been reported at the transcription start site of genes, which typically contain a nucleosome-depleted region where the transcription preinitiation complex can assemble (17, 18). The deletion of the gene encoding H2A.Z, HTZ1, results in the spread of Sir proteins from heterochromatin to euchromatin, thus reducing the transcription of genes (8, 19).

NuA4 is the main histone acetyltransferase (HAT) responsible for the acetylation of H2A, H2A.Z, and H4. The complex is required for gene regulation, DNA repair, and contains 13 subunits, with a total molecular weight of approximately 1.0 MDa (20–23). The subunit Eaf1 serves as a platform that coordinates the assembly of four different functional modules: piccolo NuA4, Tra1, Eaf3/5/7, and the Arp4/Act1/Swc4/Yaf9 (22, 24). The piccolo NuA4 (pNuA4) complex, including Esa1, Yng2, Eaf6, and Epl1, can maintain acetyltransferase activity and acetylate the substrates (25–27). The remaining NuA4 subunits can form recruitment modules that are involved in DNA repair, transcription initiation, and transcription elongation (22, 28). Unlike other histone modifiers, which can recognize specific residues on the histone tails, the targets of NuA4 involve multiple residues (29). Previous studies have revealed that NuA4 acetylates three lysine sites on histone H4 and two lysine sites on H2A (30). Moreover, mass spectrometric analysis identified that H2A.Z was a direct substrate for NuA4 (8). These four lysine sites on the N-terminal tail of H2A.Z, including K3, K8, K10, and K14, are substrates for NuA4 (2).

In 2016, the cryo-EM structure of pNuA4 in complex with the nucleosome aided the development of a model of how the histone-modifying enzyme recognizes its substrate (31). The recently reported high-resolution cryo-EM structure of NuA4 from Saccharomyces cerevisiae, bound to the nucleosome, shows that the HAT module of NuA4 recognizes the disk face of the nucleosome through the H2A-H2B acidic patch and the nucleosomal DNA (32). The structure demonstrated that the N-terminal tail of H4 extended to the catalytic pocket of Esa1 for the acetylation. H2A.Z shares approximately 60% sequence identity with canonical H2A. In the acidic region at the C-terminus, H2A.Z contains an additional acidic amino acid compared to H2A, which may facilitate the binding of certain complexes. Previous studies have shown that H2A.Z is as efficient a substrate for acetylation by NuA4 as canonical H2A in recombinant NCPs (33). Furthermore, the H2A tail has been demonstrated to be crucial for the action of pNuA4 (34). H2A.Z may also play a crucial role in the function of NuA4. Although the current cryo-EM structures of the NuA4 complex with mononucleosomes could reveal how the histone-modifying enzyme recognizes its substrate (31) and provide important information for understanding how NuA4 functions (21, 22, 35), the complete acetylation process by the NuA4 complex and how it acetylates both H4 and H2A.Z in the nucleosome remain largely unknown due to the lack of the related structures.

Here, we report seven cryo-EM structures, including five structures of the pNuA4–ZCN (H2A.Z-containing nucleosome) complex, of which three are at the harboring state, one at the pre-H4-acetylation state, and one at the resetting state, and two structures of the pNuA4–H4KQ complex (pNuA4 and H2A.Z-containing nucleosome with an H4 acetylation mimic mutation), of which one is at the pre-H2A.Z-acetylation state and one at the resetting state. We then performed single-molecule Förster resonance energy transfer (smFRET) studies on different pNuA4–nucleosome complexes in the absence or presence of Ac-CoA to reveal the dynamic acetylation process of pNuA4 on H4 and H2A.Z. Based on our cryo-EM analysis and smFRET results, we proposed a model illustrating the catalytic process and the mechanism by which pNuA4 sequentially acetylates both H4 and H2A.Z in an H2A.Z-containing mononucleosome.

Results

pNuA4–ZCN Complex at States Preparing for H4 Acetylation.

To investigate the structural basis of NuA4’s catalytic mechanism, we assembled the pNuA4 and H2A.Z-containing nucleosome from S. cerevisiae. pNuA4, which contains Esa1, Yng2, Eaf6, and Epl1, was coexpressed as described in the previous research (31) (SI Appendix, Fig. S1A). The SDS-PAGE result demonstrated that pNuA4 was acquired successfully, albeit the separation of Yng2 and Eaf6 is not clear due to their similar molecular weights (SI Appendix, Fig. S1B). According to previous studies (36, 37), we purified the S. cerevisiae histones H2A.Z, H2B, H3, and H4, along with 147-bp 601 sequence DNA. The histone octamer was then assembled with 147-bp 601 DNA into an H2A.Z-containing nucleosome (referred to as ZCN thereafter) in vitro following the previously established procedure (38) (SI Appendix, Fig. S1C). Before proceeding to assemble the pNuA4–ZCN complex, we first checked the quality of the assembled ZCN using metal-shadowing electron microscopy and 5% Native-PAGE analysis (SI Appendix, Fig. S1 D and E). As shown in SI Appendix, Fig. S1D, the result indicates that the assembly of ZCN is homogeneous. To obtain an optimized pNuA4–nucleosome complex, we mixed either wt ZCN or the nucleosome mutants with pNuA4 in various ratios in vitro by diluting pNuA4 (SI Appendix, Fig. S1E). To separate the free nucleosomes from the complex, we subjected the mixture to ultracentrifugation. As shown in SI Appendix, Fig. S1 F and G, the 5% Native-PAGE results for the Grafix (39) and uncrosslinked conditions demonstrate successful separation of the complex and free nucleosomes. To obtain a more stable complex, we chose a nucleosome:pNuA4 ratio of 1:8 and employed the Grafix method (39) to acquire the pNuA4 and nucleosome complex for structure determination. Both the negative-stain EM and cryo-EM micrographs indicated that the pNuA4–ZCN complex is suitable for the following cryo-EM studies (SI Appendix, Fig. S2 A and B). We first performed the cryo-EM single-particle 3D analysis on the pNuA4–ZCN complex without adding Ac-CoA (SI Appendix, Fig. S2C). Intriguingly, four distinct classes of 3D reconstruction maps were acquired, including one we referred to as the pre-H4-acetylation state (Pre-H4AS) and three we referred to as harboring states (Harb-S1, Harb-S2, and Harb-S3) (Fig. 1 and SI Appendix, Figs. S3 and S4 A and B). We consider the Harb-S1 and Harb-S2 states (Fig. 1 B and C) to represent the “searching” states of NuA4 when NuA4 recognizes the substrate nucleosome. In these states, the loop of Esa1 is flexible and can undergo conformational changes. Based on the structure in Fig. 1A, we hypothesize that once NuA4 identifies the correct substrate and begins acetylation, the N-terminal of Epl1 interacts with Esa1, stabilizing the Esa1 loop and preventing its movement, which allows efficient acetylation of the substrate. Therefore, we suggest that Harb-S1 and Harb-S2 represent the searching states of NuA4, while the Pre-H4AS state (Fig. 1A) corresponds to the state in which NuA4 is about to start working. The reported reconstruction resolutions for Pre-H4AS, Harb-S1, Harb-S2, and Harb-S3 are 3.8 Å, 4.1 Å, 3.9 Å, and 4.3 Å, respectively (SI Appendix, Fig. S3 and Table S1). The Pre-H4AS, Harb-S1, and Harb-S2 exhibit one pNuA4 binding to one ZCN (Fig. 1 and SI Appendix, Fig. S4 D–F), whereas Harb-S3 is featured with two pNuA4 binding on one ZCN (SI Appendix, Fig. S4 A and B). The pNuA4 double-bound ZCN shows a pseudo-2-fold symmetry, occupying approximately 13% of the total good particles, compared to 87% of the good particles for pNuA4 single-bound ZCN (SI Appendix, Table S1). We think the Harb-S3 might also be a state searching for substrates. Overall, the conformations of the single pNuA4-bound classes are very similar (Fig. 1 and SI Appendix, Fig. S4 D–F), all of which are anchored to the SHL 1.5 DNA (SI Appendix, Fig. S4 D–F), albeit rotations and translations of NuA4 were observed when the structures are aligned by the histone octamer (SI Appendix, Fig. S4 D–F).

Fig. 1.

Architecture of the pNuA4–ZCN complex in the absence of Ac-CoA. (A–C) Cryo-EM reconstruction maps of the pNuA4–ZCN complex in the pre-H4-acetylation state (Pre-H4AS) (A), and in the harboring states Harb-S1 (B) and Harb-S2 (C) respectively. Harb-S1 and Harb-S2 represent the searching states of NuA4, while the pre-H4AS state corresponds to the state in which NuA4 is about to start working. (D–F) The corresponding atomic models of Pre-H4AS (D), Harb-S1 (E), and Harb-S2 (F). The red arrows in panels (A) and (D) indicate Epl1RL in the N-terminal of Epl1. The pNuA4 regions in panels (A) and (D) are indicated by dashed lines. The subunits are colored in the same scheme in all of the panels, shown as DNA (sky blue), H3 (hot pink), H4 (lime), H2A.Z (sandy brown), H2B (dark cyan), Esa1 (dodger blue), Epl1 (salmon), Yng2 (violet), and Eaf6 (yellow) respectively.

We found that Pre-H4AS displays a clear feature of loop density in the EPcA domain of Epl1 (40), which is sandwiched by Esa1 and the ZCN in the cryo-EM map (Fig. 1 A and D). Guided by the high-resolution reconstruction features, we were able to trace and build the atomic model of this loop (from R58 to L74, Epl1RL) (Fig. 2A). Notably, only in the pre-H4-acetylation state, i.e., Pre-H4AS, could Epl1RL density be clearly traced (Fig. 1 A and D). Compared with the previously reported structure of NuA4-H2A nucleosome complex (PDB: 7VVU) (32), the NP domain of Epl1 in our structure was resolved with more residues (Fig. 2A), which adopts a distinctly different conformation (SI Appendix, Fig. S4H).

Fig. 2.

Intersubunit interaction and the role of Epl1 in the pNuA4–ZCN complex. (A) The interaction between Esa1 and Epl1 in the Pre-H4AS state. The density of the Epl1RL loop (from R58 to L74) is highlighted as a mesh, fitted with the corresponding atomic model. The atomic model of Esa1 is shown in a cartoon representation. (B) The conformational changes of the Esa1LE loop (from L391 to E407) as indicated in (A) at different states. The pink, yellow, and blue balls indicate the center of mass of the Esa1LE loop in Harb-S2, Harb-S1, and the Pre-H4AS state, respectively. The distance is given in Å. (C) The atomic model of the pNuA4–ZCN complex in the Pre-H4AS state, viewed from two angles. The boxed areas are enlarged in (D), (E), (F), and (G), respectively. (D–E) Enlarged views of the interfaces between Esa1 and Epl1. (F and G) Enlarged views of the interfaces between Epl1 and H2A.Z (F) and between Esa1 and H2A.Z (G). The key amino acid residues are shown in stick and ball representations. The potential hydrogen bonds are indicated by dashed lines. The color scheme is consistent with Fig. 1A.

Interestingly, in Esa1, a loop between L391 and E407 (Esa1LE as shown in Fig. 2A) was found to adopt different conformations among the pre-H4-acetylation state (Pre-H4AS) and the harboring states (Harb-S1, Harb-S2) (Fig. 2B). That is, the Esa1LE in Pre-H4AS adopts a distinctly different conformation from those in Harb-S1 and Harb-S2, while the center of mass of Esa1LE in Pre-H4AS shifts approximately 11.7 Å horizontally (Fig. 2B). The movement of Esa1LE was most likely driven by Epl1RL, which could push Esa1LE to the acidic patch on the H2A.Z-containing nucleosome. This implies the key role of Epl1 in regulating the conformation of Esa1 and its interaction with the acid patch on nucleosome, which is important in modulating higher-order chromatin structure (41). Upon aligning Harb-S1 and Harb-S2, the calculated RMSD is 1.079, indicating a conformational change in NuA4, as shown in SI Appendix, Fig. S4G. Additionally, we investigated the interactions between NuA4 and SHL1.5 DNA and identified a 5.5 Å shift in the center of mass of the DFL of Epl1 between Harb-S1 and Harb-S2 states (SI Appendix, Fig. S4G), suggesting flexibility on the binding mode of NuA4. These cryo-EM structures show that the enzyme active site Glu338 on Esa1 in Pre-H4AS is much closer to the N-terminal of H4 than to the N-terminal of H2A.Z (Fig. 1 A and D and SI Appendix, Fig. S4I). Similarly, the corresponding distances in the H4 acetylation state (PDB: 7VVU) (32) and the Pre-H4AS state exhibit the same pattern (SI Appendix, Fig. S4I). Therefore, we conclude that Pre-H4AS represents a state which is ready to acetylate H4, namely, the pre-H4-acetylation state, and the Harb-S1 and Harb-S2 represent the searching states of NuA4 prior to acetylation.

As indicated by previous studies (2, 30) and our western blot assay, pNuA4 can acetylate both H4 and H2A.Z (SI Appendix, Fig. S5B), but only the Pre-H4AS state was well resolved in our structures. We suspect that the pNuA4 binds to the H2A.Z-containing nucleosome in an H4-preferred manner. To further validate this speculation, we performed smFRET assays on the pNuA4–ZCN complex (SI Appendix, Figs. S5 and S6). We chose to add a SNAP tag–labeled with SNAP-Surface 649 dye to the N-terminal of Esa1 (SI Appendix, Fig. S5A). The structure of NuA4 fused with the SNAP tag was predicted using AlphaFold3 (42). According to the western blot results, both NuA4 and SNAP-tagged NuA4 exhibit similar acetylation patterns on H2A containing nucleosome and ZCN (SI Appendix, Fig. S5B). Meanwhile, an Alexa568 molecule was labeled close to SHL2.0 of the nucleosome (SI Appendix, Fig. S5C). An alternating excitation experiment validated that FRET can be observed between Alexa568 and SNAP-Surface 649 dye (SI Appendix, Fig. S6). The structural prediction of pNuA4 with the SNAP tag by AlphaFold3 (42) indicated that the predicted structure of pNuA4-SNAP-tag adopts a similar conformation to the previous structure of pNuA4 (PDB: 5J9U) (SI Appendix, Fig. S5A). After a rigid body fitting (SI Appendix, Fig. S5D), we proceeded to roughly estimate the distance between the Glu159 on SNAP-tag, which is involved in the binding of the fluorescein substrate (43), and the 96th base of the DNA, labeled with Alexa568 near SHL2.0 (SI Appendix, Fig. S5E). As demonstrated by our cryo-EM study, pNuA4 can also bind with another surface of the nucleosome (SI Appendix, Fig. S4A). However, when pNuA4 binds with another surface of the nucleosome(SI Appendix, Fig. S5F), the measured distance (approximate 12 nm, SI Appendix, Fig. S5E) indicated that smFRET is impossible to occur between them. Importantly, the structure-based measurements indicated that a higher smFRET value should be observed when pNuA4 accesses H4, compared to a lower value for H2A.Z. Consistently, the first non-zero FRET value observed in the smFRET assay, corresponding to the initial binding state where pNuA4 interacts with the nucleosome, is denoted as the “initiate binding” state. This yielded two major smFRET values: 0.4 and 0.2 (Fig. 3 A–F). Therefore, combined with our cryo-EM structures, we conclude that the smFRET value of 0.4, which occurs at the chance of 77%, should be contributed by the states with pNuA4 close to H4, while the smFRET value of 0.2 represents states with pNuA4 close to H2A.Z (Fig. 3 A–F). Because when pNuA4 is accessing H2A.Z, the distance between the SNAP tag and Alexa568 will be larger than that of H4, consequently producing a smaller smFRET value.

Fig. 3.

The smFRET analysis of pNuA4 binding with the wt H2A.Z, H4KQ, and H2AZKQ nucleosome in the absence of Ac-CoA and the structure comparison between pre-H4AS and pre-H2AZAS. (A–C) The FRET traces of pNuA4 binding to the wt H2A.Z nucleosome (A), H4KQ nucleosome (B), and H2AZKQ nucleosome (C), respectively, in the absence of Ac-CoA. (D–F) The corresponding distribution of initial binding for the pNuA4 binding to the wt H2A.Z nucleosome (D), H4KQ nucleosome (E), and H2AZKQ nucleosome (F). The “N” values which denote the total number of frames for the initial binding state under each condition are 1,071 (D), 1,885 (E), and 936 (F) respectively. The data were binned with a size of 0.03. (G) The reconstructed cryo-EM map of pNuA4–nucleosome complex in the pre-H2A.Z acetylation state, Pre-H2AZAS. The pNuA4 is indicated by dashed lines. (H) The pNuA4–nucleosome complexes cryo-EM map comparisons between Pre-H4AS (gray) and Pre-H2AZAS (blue) states, viewed from two angles. The pNuA4 orientation angle differences are given in degrees.

Interface Between pNuA4 and ZCN at Pre-H4 Acetylation State.

Previous studies have demonstrated that Epl1 plays critical roles in promoting the enzymatic activity of Esa1 (23, 26, 40). Because the Epl1RL density is well resolved in Pre-H4AS (Fig. 2A), we investigated the interactions related to Epl1RL in this state, and several interactions were observed, associated with Epl1RL. As shown in Fig. 1A, the Epl1RL is sandwiched between Esa1 and ZCN. Extensive interactions via electrostatic interactions and hydrogen bonds were formed among the subunits of pNuA4 and ZCN (Fig. 2 and SI Appendix, Fig. S7). The solvent-accessible interfaces between Epl1RL and ZCN, and Esa1 and ZCN, are 2,235 Ų and 3,318 Ų, respectively.

The dual function loop (DFL) in Epl1 (SI Appendix, Fig. S4C) is important for the catalytic activity of NuA4 (31, 44). In Pre-H4AS, the DFL of Epl1 is close to the nucleosomal SHL1.5 DNA (SI Appendix, Fig. S4 D–F) and the electrostatic surface indicates a strong electrostatic attraction between the DFL and DNA (SI Appendix, Fig. S7A). Notably, similar interactions were also observed in the harboring states, Harb-S1 and Harb-S2 (SI Appendix, Fig. S7A), suggesting that Epl1 is important for the recognition of DNA, consistent with the previous studies (32, 34). Moreover, through the electrostatic surfaces, we found that Epl1 has close interactions with Esa1 and H2A.Z (Fig. 2C and SI Appendix, Fig. S7 B–E). Two hydrogen bonds were formed between H65 on Epl1RL and I368 on Esa1 (Fig. 2D), as well as I68 on Epl1RL and F404 on Esa1 (Fig. 2E) were detected. Meanwhile, a salt bridge was formed between K59 on Epl1RL and D93, which is located on the acidic patch of the H2A.Z nucleosome (Fig. 2F). Furthermore, a cation–pi interaction was formed between Y396 on Esa1 and R90 on H2A.Z, while hydrogen bonds were found between Y396 and E66, as well as between Y397 and K73 (Fig. 2G). The interactions associated with Epl1 (Fig. 2) imply the important role of Epl1 in regulating NuA4’s function of acetylation. Consistently, previous research showed that the deletion of the NP domain of Epl1 can dramatically decrease the acetylation of NuA4 (40). In the harboring states Harb-S1 and Harb-S2, although the density of Epl1 was obscured, the interaction between Esa1 and H2A.Z appears similar (SI Appendix, Fig. S7 F and G). We speculate that in the harboring states, the flexibility of Epl1 allows pNuA4 to move freely until it finds the optimal substrate for acetylation. Previous studies have shown that the Tudor domain of Esa1 is critical for the HAT activity on nucleosomes (31, 34, 45, 46). However, in the cryo-EM structures we obtained, the density corresponding to the Tudor domain of Esa1 is missing, likely due to the high flexibility of Esa1. Additionally, Piccolo’s ability to acetylate nucleosomes requires the N-terminal sequences of Yng2. Moreover, the central region of Yng2 participates in the interaction with Epl1 domain II within EpcA. Although the C-terminal region of Yng2 is not essential for its activity or association with Epl1 EPcA domain, it is necessary for binding H3K4me3. Significantly, the EPcA homology region of Epl1 is sufficient for the formation of the piccolo NuA4 complex together with Yng2 and Esa1 (34, 45). In our structure, the interactions between the N-terminal regions of Yng2 and Epl1 with DNA reinforce the chromatin-specific function of the pNuA4 complex. These interactions are likely crucial for proper positioning of the complex on the nucleosome, promoting more efficient acetylation of histone tails by stabilizing the nucleosome binding interface.

We then compared the structures we obtained with those from the previous study (32). Both our study and that of others show that the HAT module of NuA4 binds to the H2A or H2A.Z nucleosome through Epl1 and Esa1, with the double-function loop (DFL) of Epl1 positioned near the SHL1.5 in both H2A and H2A.Z containing nucleosomes (SI Appendix, Fig. S8A). Due to the larger acidic patch on H2A.Z compared with that on H2A (SI Appendix, Fig. S8E), the interaction of NuA4 with ZCN differs from its interaction with H2A-containing nucleosomes. In the structure of the complex containing the H2A nucleosome, the EPcA-N domain of Epl1 anchors to the acidic patch on the H2A-H2B dimer via Arg56 and Arg58, stabilizing the interaction between the HAT module and the H2A nucleosome (SI Appendix, Fig. S8 B and C) (32). Additionally, the subunit Esa1 interacts with H2A through Tyr396 (SI Appendix, Fig. S8D) (32). For the H2A.Z nucleosome, our data show that, unlike the interactions between the acid patch on the H2A-H2B dimer and Arg56/Arg58, the EPcA-N domain of Epl1 anchors to the acidic patch on the H2A.Z-H2B dimer via Lys59 (Fig. 2F and SI Appendix, Fig. S8 B and C), while Tyr396 and Tyr397 of Esa1 interact with E66 and Arg90 on H2A.Z (Fig. 2G and SI Appendix, Fig. S8D). Western blot experiments suggest that, similar to canonical H2A in the recombinant NCPs, H2A.Z is a substrate that can be acetylated by NuA4 (SI Appendix, Fig. S5B), consistent with the previous studies (2, 8, 33, 47, 48). Additionally, smFRET experiments show that when H2A-containing nucleosomes are present as the substrates, 73% of the first non-zero FRET state, with a FRET value of 0.4, corresponds to a state where pNuA4 is close to H4, indicating that pNuA4 preferentially binds to H4 (SI Appendix, Fig. S11D). Therefore, while the differences between H2A and H2A.Z may influence the interaction mode between NuA4 and the nucleosome, they do not appear to significantly alter the substrate preference of NuA4.

pNuA4–ZCN Complex at the State Preparing for H2A.Z Acetylation.

Previous studies suggest that both H4 and H2A.Z can be acetylated in the presence of Ac-CoA (8, 49). However, the structural information about the acetylation of H2A.Z by pNuA4 is very limited. In order to obtain the state associated with H2A.Z acetylation, we assembled the ZCN carrying four lysines (K5, K8, K12, and K16) with glutamine mutations in the N-terminal tail of H4 (referred to as H4KQ thereafter) to mimic the acetylation (50, 51). Because H4 has already been “acetylated” when carrying four lysine-to-glutamine mutations, we anticipated that pNuA4 would quickly skip the acetylation of H4 and allow more particles to be trapped in a state to acetylate H2A.Z. We obtained the pNuA4–H4KQ complex using Grafix following a similar process to the wild-type H2A.Z nucleosome. The western blot results confirmed that H4 was not acetylated under such circumstances, whereas H2A.Z was successfully acetylated in the presence of Ac-CoA (SI Appendix, Fig. S9A). Given that H4 is mimicking acetylation in H4KQ, which means no acetylation will be required for H4, the success of the acetylation of H2A.Z proves that the K-to-Q mutation in the H4 tail cannot abort the acetylation of H2A.Z.

We found that H4KQ nucleosomes yielded a lower apparent unimolecular rate constant (Km) with pNuA4 than wild-type ZCN (SI Appendix, Fig. S9B), implying the affinity of pNuA4 for the substrate is slightly increased. We then performed the cryo-EM structure analysis on the pNuA4–H4KQ complex without adding Ac-CoA and performed a thorough 3D classification (SI Appendix, Fig. S10A). Among the classified density maps, one class appears very similar to the pre-H4 acetylation state (SI Appendix, Fig. S11A). Meanwhile, one class in which pNuA4 adopts a totally different conformation was identified (SI Appendix, Fig. S11B), and the enzyme site of pNuA4 in this class (SI Appendix, Fig. S11B) was much closer to H2A.Z than in the pre-H4 acetylation state Pre-H4AS (SI Appendix, Fig. S11A). We term the class the Pre-H2A.Z acetylation state, i.e., Pre-H2AZAS (Fig. 3G). Compared with the pre-H4 acetylation state Pre-H4AS, pNuA4 in Pre-H2AZAS rotated about 87° laterally and 140° horizontally (Fig. 3H). After the rotation, pNuA4 in Pre-H2AZAS was anchored nearby SHL 4.5, instead of SHL 1.5 in Pre-H4AS, bringing it closer to the N-terminal of H2A.Z, with the N-terminal of Yng2 potentially helping position the N-terminal of H2A.Z into the enzyme’s active site on Esa1 (SI Appendix, Fig. S11B). However, this hypothesis will require further validation through high-resolution structural analysis.

The cryo-EM structure and 3D classification results of the pNuA4 and H4KQ-H2A.Z nucleosome complex suggest that pNuA4 acetylates H2A.Z in a dynamic way (Fig. 3 G and H and SI Appendix, Figs. S10A and S11B). We then performed smFRET studies on the pNuA4–H4KQ complex in the absence of Ac-CoA to analyze this dynamic process. The smFRET assays indicated that the chance for pNuA4 to access H4 significantly decreased from 77% in the wild-type ZCN (Fig. 3D) to 55% in H4KQ-H2A.Z nucleosome (Fig. 3E), suggesting that the “acetylated” H4 promotes pNuA4 to search for other substrates, i.e., H2A.Z. Moreover, we observed that pNuA4 exhibits a much more dynamic pattern in the H4KQ nucleosome (Fig. 3 B and E) than in the wild-type ZCN (Fig. 3 A and D) during the search for other substrates when the N-terminal of H4 has been “acetylated,” which is consistent with our cryo-EM study (Fig. 1 and SI Appendix, Fig. S11 A and B).

Furthermore, similar to the mutations on H4 lysines, to explore whether the acetylation of H2A.Z would affect pNuA4’s activity on H4, we introduced lysine (K3, K8, K10, and K14) with glutamine mutations on H2A.Z (H2AZKQ) to mimic the acetylation of H2A.Z and performed the smFRET experiments. Remarkably, since pNuA4 will skip the acetylation on H2A.Z, as expected, the smFRET results showed that the dynamic range of pNuA4 had decreased (Fig. 3C) and the chance of pNuA4 first binding to H4 was boosted to 87% (Fig. 3F). We then analyzed the dwell time of these initial binding states with different types of nucleosomes (wt H2A.Z, H4KQ, and H2AZKQ), which averaged around 2 s, as shown in SI Appendix, Fig. S11C. This suggests pNuA4 does not bind longer to H4 than H2A.Z, and the subpopulations (Fig. 3 D-F), calculated using Gaussian fitting, indicate that pNuA4 binds to H4 first with higher probability. Furthermore, when we replaced H4 with H4KQ, the binding probability at the H4 site decreased. Similarly, replacing H2A.Z with H2A.ZKQ resulted in a decreased binding probability at the H2A.ZKQ site. Therefore, these observations suggest that pNuA4 has an initial preference for binding at H4, and this preference is influenced by acetylation. Overall, the smFRET assays suggest that not only does the acetylation modification on the substrates lower the chance for pNuA4 to bind to them, but pNuA4 also prefers to bind to H4 to initiate the acetylation.

pNuA4–ZCN Complex After H4 and H2A.Z Dual-Acetylation.

In order to capture the state while pNuA4 is in the process of acetylation, we incubated pNuA4 with the wild-type ZCN in the presence of Ac-CoA. Surprisingly, the cryo-EM reconstruction results showed that only one stable conformation of the pNuA4–ZCN complex was captured (SI Appendix, Fig. S10B), which we named the resetting state (referred to as RS-CoA) (SI Appendix, Fig. S12A). Interestingly, RS-CoA adopts a conformation almost identical to that of Pre-H4AS with an RMSD of 0.62Å (SI Appendix, Fig. S12B). It is worth noting that the density of Epl1RL was well resolved in the RS-CoA state (SI Appendix, Fig. S12 A and B).

Our western blot results indicate that both H4 and H2A.Z are acetylated after mixing pNuA4 with ZCN and Ac-CoA for 10 min (SI Appendix, Fig. S9A). This result suggests that RS-CoA might represent a state after both H4 and H2A.Z are acetylated. Combining the states of Pre-H4AS (pre-H4 acetylation state) and RS-CoA (the resetting state), we speculate that the process of acetylation by pNuA4 probably starts with H4 and ends with H4. To verify this, we performed similar cryo-EM studies on the pNuA4–H4KQ complex in the presence of Ac-CoA (SI Appendix, Fig. S10C). Surprisingly, after adding Ac-CoA to the pNuA4–H4KQ complex, instead of multiple states of the pNuA4–H4KQ complex observed previously in the absence of Ac-CoA (SI Appendix, Fig. S11 A and B), only one state referred to as KQ-RSCoA (the resetting state of the pNuA4–H4KQ complex) was observed (SI Appendix, Fig. S12C). It is worth noting that KQ-RSCoA exhibits high similarity to RS-CoA (SI Appendix, Fig. S12 A and C). Moreover, similar to Pre-H4AS (Fig. 1 A and D) and RS-CoA (SI Appendix, Fig. S12A), the density of Epl1RL was also well resolved in KQ-RSCoA (SI Appendix, Fig. S12C). These results suggest that, after the acetylation of H2A.Z, the pNuA4–nucleosome complex returns to RS-CoA (the resetting state), which is ready for the next cycle.

To learn more about the acetylation process of pNuA4, we further investigated what happens if H4 or H2A.Z cannot be acetylated. To do this, we first mutated the four lysines (K5, K8, K12, and K16) in the N-terminal of H4 to arginine (H4KR), which prevents the acetylation of H4 from occurring (8). With these mutations, the Km value was measured to increase about 2-fold compared with the wild-type complex (SI Appendix, Fig. S9B), which suggests that the mutations on H4 could impact the enzyme affinity of pNuA4. Interestingly, the western blot results indicated that H2A.Z can still be successfully acetylated with these H4KR mutations (SI Appendix, Fig. S9A). Subsequently, in contrast to H2AZKQ, we introduced four lysines (K3, K8, K10, and K14) to arginine (H2AZKR) mutations in the H2A.Z tail to disable the acetylation on H2A.Z (8). Notably, the measured Km values on H2AZKQ and H2AZKR are decreased approximately 35-fold and 2-fold, respectively, compared to that of wild-type H2A.Z (SI Appendix, Fig. S9C), implying that the affinity of the enzyme to the two mutation substrates is increased. Moreover, as suggested by the western blot analysis, these mutations in H2A.Z have no noticeable effect on pNuA4’s ability to acetylate H4 (SI Appendix, Fig. S9A). We thus conclude that the acetylation of H4 or H2A.Z by pNuA4 is independent of one another, while H4 is preferred over H2A.Z.

We then prepared the samples of pNuA4–H4KR, pNuA4–H2AZKR, and pNuA4–H2AZKQ complexes for cryo-EM analysis (SI Appendix, Fig. S9 D–G). Interestingly, regardless of the mutations, after adding Ac-CoA, pNuA4 always adopts a nearly identical conformation to RS-CoA (SI Appendix, Fig. S9 E and G). Surprisingly, only one conformation is obtained for the pNuA4–H4KR complex in the absence of Ac-CoA (SI Appendix, Fig. S9D), which is different from the pNuA4–H4KQ complex (SI Appendix, Fig. S11 A and B). Meanwhile, consistent with the pNuA4–H4KR complex without adding Ac-CoA, only one stable conformation is obtained from the pNuA4–H2AZKQ complex in the absence of Ac-CoA (SI Appendix, Fig. S9F). Therefore, NuA4 might preferentially acetylate H4. Once H4 acetylation is complete, NuA4 may begin searching for the next substrate. If H4 has not been acetylated, NuA4 may stay on the H4 side, waiting for acetylation to initiate.

To further explore the dynamic acetylation process, we performed smFRET experiments on the pNuA4–ZCN (wt or H4KQ) nucleosome complexes in the absence or presence of Ac-CoA (SI Appendix, Fig. S12 D–G). Interestingly, the smFRET values in the presence of Ac-CoA (SI Appendix, Fig. S12 E and G) display a larger dynamic change compared with the results in the absence of Ac-CoA (SI Appendix, Fig. S12 D and F). In the presence of Ac-CoA, the smFRET values bounced between higher and lower smFRET values for both WT (SI Appendix, Fig. S12E) and H4KQ (SI Appendix, Fig. S12G), suggesting larger dynamic conformational changes in pNuA4. Notably, during the dynamic transitions, the smFRET value begins with 0.4, suggesting that H4 is the first substrate to be acetylated.

As indicated by the dynamic smFRET values, there are several intermediate states when pNuA4 acetylates nucleosome substrates (SI Appendix, Fig. S12 D–G). For the H4KQ nucleosome, the smFRET transitions in the absence of Ac-CoA show a similar pattern to the smFRET transition of the WT nucleosome in the presence of Ac-CoA (SI Appendix, Fig. S12 H and I), suggesting H4 is the preferred substrate of pNuA4. Our cryo-EM analysis revealed multiple states of the pNuA4–H4KQ nucleosome complex in the absence of Ac-CoA (SI Appendix, Fig. S11 A and B), which is consistent with the dynamic transitions observed in the smFRET analysis (SI Appendix, Fig. S12 F and I). Although the cryo-EM structures of the pNuA4–H4KQ complex were not resolved at high resolutions, we were able to fit the structure of the pNuA4 unambiguously and measure the distances between the SNAP tag and the Alexa568 molecule (SI Appendix, Fig. S11 A and B). Taking advantage of this procedure, we measured the distances in the pNuA4–H4KQ complex in the absence of Ac-CoA. The distances were measured within a significantly varied range (SI Appendix, Fig. S11 A and B), which is highly consistent with the dynamic transitions of the smFRET values (SI Appendix, Fig. S12 F and I).

Discussion

Many posttranslational modification enzymes, e.g., histone methyltransferases, are highly site-specific and able to precisely work on their target residues (52). Consequently, various methyltransferases have been identified to function at different sites, e.g., K4, K9, of histone H3 (53, 54). In contrast, an acetyltransferase can typically acetylate multiple substrates (55).

NuA4 is the only essential acetyltransferase in yeast and can catalyze the acetylation of the histones H2A, H2A.Z, and H4, thereby affecting gene transcription (20). However, the complete acetylation process of NuA4 and how it acetylates H4 and H2A.Z differently remains largely elusive.

In this study, we focus on how NuA4 synergistically acetylates different substrates, e.g., H4 and H2A.Z, in an H2A.Z-containing nucleosome. Based on our cryo-EM and smFRET results, we propose a model for how NuA4 acetylates both H4 and H2A.Z in the H2A.Z-containing nucleosome (Fig. 4 and Movie S1).

Fig. 4.

A model for pNuA4 acetylating H2A.Z containing nucleosome (ZCN). In Pathway 1, H4 is the preferred substrate and is acetylated first. This pathway includes the following states: (A) Positioning state: pNuA4 is recruited to the ZCN to initiate the acetylation process; (B) H4 harboring state: pNuA4 adopts different conformations while searching for its preferred substrate, H4; (C) Pre-H4 acetylation state: Epl1 acts as a regulator to stabilize Esa1, strengthening its interaction with the nucleosome; (D) Acetylation of H4: In the presence of Ac-CoA, H4 is acetylated; (E) Repositioning: pNuA4 is repositioned to target other substrates, such as H2A.Z; (F) Acetylation of H2A.Z: H2A.Z is captured and acetylated; (G) Dual-acetylated state: Both H4 and H2A.Z are acetylated; and (H) Resetting state: After all substrates (H4 and H2A.Z) are acetylated, NuA4 resets in preparation for another cycle of acetylation (I). Alternatively, in Pathway 2, H2A.Z could be acetylated first, though with lower probability. This pathway includes the positioning, harboring, pre-acetylation, and acetylation states of H2A.Z. After H2A.Z is acetylated, pNuA4 is repositioned and continues to search for other substrates, such as H4. Once both H2A.Z and H4 are acetylated, pNuA4 resets for the next round of acetylation. The subunits are colored in the same scheme as in Fig. 1.

As shown in Fig. 4, once pNuA4 positions the ZCN, it will start searching for its substrates on the disk surface of ZCN. As suggested by our cryo-EM and smFRET results, pNuA4 has a much higher chance of binding to its preferred substrate, the N-terminal of H4, and initiating the acetylation process, which we term pathway 1. In the harboring states, the pNuA4 interacts with the H2A.Z-containing nucleosome via multiple regions (Figs. 1 and 2). With the help of Epl1, Esa1 interacts with the acidic patch of H2A.Z. pNuA4 is subsequently immobilized to ZCN and ready for the process of acetylation, i.e., pNuA4 arrives at the pre-H4-acetylation state. After completing the acetylation on H4, pNuA4 will search for other substrates along the surface of the nucleosome. At the pre-H2A.Z-acetylation state, pNuA4 finds its substrates, the N-terminal of H2A.Z, to perform the acetylation. After the acetylation of H2A.Z, pNuA4 will return to RS-CoA (the resetting state) and prepare for the next rounds of acetylation.

Alternatively, pNuA4 could, with a lower probability, also bind to the N-terminal of H2A.Z to start the acetylation process, which we referred to as pathway 2 (Fig. 4). In this way, pNuA4 will bind to H2A.Z first and initiate the acetylation process. After the acetylation of H2A.Z, pNuA4 will move to the pre-H4 acetylation state for the acetylation of H4. When both H4 and H2A.Z are acetylated, pNuA4 will return to RS-CoA (the resetting state) for the next cycle.

With a smaller chance, pNuA4 may start with H2A.Z acetylation. It is worth noting that our cryo-EM structures were resolved at a lower resolution and showed more intermediate states in H2A.Z acetylation than in H4 acetylation, suggesting a more dynamic process of H2A.Z acetylation. Our smFRET experiments are highly consistent with this notion. The smFRET results implied that when the acetylation of H2A.Z is finished, pNuA4 may not fall off, but keeps rotating along the disk surface of the nucleosome, as if to check whether all the substrate has been acetylated (SI Appendix, Fig. S12 H and I). In summary, these results suggest that the acetylation process of the H2A.Z-containing nucleosome by pNuA4 is a dynamic and complex process. In the acetylation process, both H4 and H2A.Z can be acetylated by pNuA4. Among these substrates, H4 appears to be the preferred target, as pNuA4 binds to H4 with higher probability. However, further validation is required to confirm this preference.

In vivo, the acetylation of H4 and H2A.Z by the NuA4 complex is a coordinated process that plays a central role in regulating chromatin structure, gene expression, and cellular processes (10–13). Both H4 and H2A.Z acetylation are involved in maintaining chromatin in a transcriptionally active state, but they also function in distinct biological contexts, often working together to facilitate gene activation, DNA repair, and cell cycle regulation (10–13).

Previous studies have shown that NuA4 promotes the deposition of H2A.Z, which in turn prevents the unintended activation of stress response genes (56–58). In our model, we propose that NuA4 preferentially acetylates the N-terminal of H4, and this acetylation may serve as a marker for H2A.Z deposition in vivo. Acetylation of H4 enhances gene activation by modifying promoter regions, while acetylation of H2A.Z plays a key role in transcriptional regulation and chromatin remodeling. Together, these acetylation marks facilitate gene expression, regulate metabolic pathways, and support optimal plant growth under fluctuating environmental conditions (59). The preferential acetylation by NuA4 might provide a potential mechanism for gene expression regulation in vivo.

In yeast, the NuA4 and SWR1 complexes can work synergistically to regulate chromatin (33, 60). NuA4 acetylates H4, promoting an open chromatin structure and facilitating the recruitment of SWR1, which deposits the histone variant H2A.Z at heterochromatin boundaries. NuA4 further acetylates H2A.Z, stabilizing these boundaries and enhancing antisilencing activity. The final nucleosomal acetylation of both H4 and H2A.Z is crucial for defining chromatin boundaries, preventing the spread of heterochromatin, and ensuring proper gene expression in yeast. This coordinated acetylation process plays a central role in regulating chromatin dynamics and safeguarding active gene regions from inappropriate silencing (61).

H4 and H2A.Z acetylation are involved in DNA damage response and repair (62). Upon DNA damage, the NuA4–Tip60 complex can acetylate H4, promoting chromatin relaxation and facilitating the deposition of the histone variant H2A.Z at the damage site. Following its incorporation, H2A.Z is further acetylated by NuA4–Tip60, which stabilizes the open chromatin state, allowing for the recruitment of DNA repair factors like BRCA1 and 53BP1. The synergistic acetylation of H4 and H2A.Z ensures chromatin remodeling, repair factor accessibility, and efficient DNA damage repair (62). Therefore, the acetylation of both H4 and H2A.Z is tightly linked in the regulation of chromatin structure and function. While H4 acetylation is often associated with gene activation and chromatin remodeling, H2A.Z acetylation serves a complementary role in maintaining chromatin flexibility and promoting transcriptional activation. In DNA repair, the dual acetylation of H4 and H2A.Z not only facilitates chromatin opening at damage sites but also ensures that the chromatin structure is restored after repair. Therefore, the coordinated acetylation of both histones is essential for the regulation of pluripotency and differentiation pathways.

The structures we obtained provide detailed insights into the acetylation process of H2A.Z-containing nucleosomes by NuA4, helping to further our understanding of the potential mechanisms in vivo. However, due to the low resolution of the structure in which NuA4 acetylates H2A.Z, additional structural details remain inaccessible. Our western blot results suggest that both NuA4 and SNAP-tagged NuA4 exhibit similar acetylation patterns on H2A containing nucleosome and ZCN (SI Appendix, Fig. S5B), although we could not completely rule out the potential influence of SNAP-tag. Furthermore, in vivo, the presence of linker DNA or other posttranslational modifications (PTMs) may influence the behavior of NuA4 in accessing its substrates (33). Further research will be required to address these issues and fully elucidate the acetylation mechanism of NuA4 in vivo.

Method Details

Preparation of the NuA4 Core Complex and NCP.

The pNuA4 core complex (Esa1, Epl1, Yng2, and Eaf6) and the octamer were expressed in Escherichia coli following the previous protocols (1, 31, 36). The N termini of Esa1, Epl1, and Eaf6 were tagged with 6× His, GST, and MBP, respectively. Esa1 (140 to 445 amino acids) and Yng2 (1 to 120 amino acids) were inserted into the plasmid RSF-Duet. Epl1 (50 to 400 amino acids) was inserted into the plasmid PGEX. Eaf6 (1 to 130 amino acids) was inserted into the plasmid pMAL. The Esa1-Yng2 and Epl1 were cotransformed into BL21 competent cells. The expression of the proteins was induced by 0.5 mM IPTG at an OD of 0.8. The E. coli cells were cultured in LB medium at 37 °C for 4 h. The E. coli cells transformed with the Eaf6 plasmid were cultured in LB medium at 16 °C overnight. The expression of the proteins was induced by 0.2 mM IPTG at an OD of 0.8. The cells were collected by centrifugation at a speed of 4,000 rpm using a fixed-angle centrifuge rotor of Bakeman for 20 min. The supernatant was discarded. The pelleted cells were washed and resuspended in the lysis buffer (20 mM HEPES, pH 8.0, 300 mM NaCl, 10% glycerol, and 5 mM Imidazole). To obtain the pNuA4 complex, the cells expressing Esa1-Yng2-Epl1 were mixed with the cells expressing Eaf6 at a weight ratio of 1:1. Then, E. coli cells were broken at 750 MPa for 10 min and centrifuged at 15,000 rpm using a fixed-angle centrifuge rotor of Bakeman for 30 min. The supernatant was collected. The pNuA4 complex was first purified in three steps with nickel, glutathione S-transferase (GST), and amylose affinity chromatography. The tags were cleaved by Tobacco Etch virus (TEV) protease at 4 °C for 12 h, and the complex was further purified by the MonoQ and size-exclusion column chromatography using Superdex 200 in a running buffer containing 10 mM HEPES, pH 7.0, 300 mM NaCl, and 10 mM DTT.

Nucleosomes were reconstituted in vitro following the previous protocol (1, 37). Briefly, the S. cerevisiae histone octamers and their mutants were coexpressed in E. coli BL21 (DE3) competent cells. The octamers were purified through SP Sepharose HP chromatography (GE Healthcare) in buffer A (20 mM Tris–HCl, pH 7.5, 0.5 M NaCl, 1 mM EDTA, and 5 mM 2-mercaptoethanol) and buffer B (20 mM Tris–HCl, pH 7.5, 2 M NaCl, 1 mM EDTA, and 5 mM 2-mercaptoethanol). According to the result of 15% SDS-PAGE, we combined the target fractions for further purification, dialyzed them in buffer B (20 mM Tris–HCl, pH 7.5, 2 M NaCl, 1 mM EDTA, and 5 mM 2-mercaptoethanol), and then loaded them onto a HiLoad Superdex 200 gel filtration column in buffer B. The fractions containing the histone octamers were then combined and concentrated to 1 mg/ml. Next, the histone octamers were mixed with the Widom 601 DNA at a molar ratio of 1:1 to form the reconstituted nucleosomes according to the previous protocol (1, 37).

Cryo-EM Sample Preparation and Data Collection.

Based on the EMSA assay, in order to reduce the ratio of free mononucleosome and obtain the stable pNuA4–NCP complex in vitro, the pNuA4 complex was mixed with NCP at a molar ratio of 8:1, followed by incubation at 4 °C in 10 mM HEPES, 50 mM NaCl, pH 7.0 for 30 min. The pNuA4–NCP complex was subsequently separated by Grafix (39). Briefly, a continuous 10 to 40% gradient of glycerol, supplemented with 0 to 0.15% gradient of glutaraldehyde, was generated on a BioComp gradient master. Then, the pNuA4–NCP mixture was loaded onto the top layer of the glycerol gradient. The ultracentrifugation was then performed at a speed of 40,000 rpm using a Beckman MSL-50 rotor for 21 h at 4 °C. Fractions were collected and checked by 5% Native-PAGE gel electrophoresis. Guided by the Native-PAGE result, the fractions containing the homogeneous complex were combined, the glycerol was removed, and the complex was concentrated to 80 μg/mL.

Vitrobot Mark IV (FEI) was used to prepare the cryo-EM grids. 3.5 µL samples were applied to the plasma-cleaned Quantifoil R1.2/1.3 holey gold grids with a blot force of 2 for 3 s in 100% humidity. The grids were then plunged into liquid ethane, which was cooled by liquid nitrogen. The cryo-EM data were collected using a 300 kV Titan Krios equipped with a K2 direct electron detector (Gatan) with a pixel size of 1.04 Å in a defocus range of −1.5 to −2.5 µm. Each micrograph was dose-fractioned to 32 frames. The total accumulated dose is 60.0 e⁻/Ų. The preparation procedures for the other complexes in the article followed the same steps.

Image Processing and Model Building.

The data analysis procedure for the wild-type NCP dataset with or without Ac-CoA, the H4KQ NCP with or without Ac-CoA, and the other complexes mentioned followed a similar procedure. Take the wild type NCP without Ac-CoA for example; the collected 13,905 movies were subjected to patch motion correction and patch CTF estimation in cryoSPARC (63). Those micrographs with large astigmatism (>0.5 μm), heavy ice contamination, or serious aggregation were discarded. Particles were picked by a template picker using a nucleosome, which resulted in 1,931,184 particles. These particle subsets were then subjected to several rounds of 2D classification followed by multiple rounds of heterogeneous refinement. In order to obtain relatively accurate alignment parameters, we initiated the alignment process by focusing on the nucleosomes in cryoSPARC (63). Then, the particles were transferred to RELION 4.0, followed by a round of 3D classification without alignment (64). We applied a spherical mask to isolate the region of the nucleosome plate in conjunction with the pNuA4 complex. Three good classes were selected and then further refined using CryoSPARC’s local refinement tools to yield the final maps (63). For the pNuA4 two-side occupied class, two masks focusing on both sides of the NCP were used to isolate the region of the nucleosome plate in conjunction with the pNuA4 complex. The class with clear density on both sides of the NCP was selected, followed by local refinement in cryoSPARC (63). Similarly, the data analysis for wild-type and H4KQ NCP supplemented with Ac-CoA datasets followed the same procedure. All the maps were subsequently postprocessed by EMReady (65).

For the H4KQ dataset, the preprocessing steps were similar to the ones mentioned above. Briefly, we first aligned the nucleosome and then applied a spherical mask to isolate the region of the nucleosome plate in conjunction with the pNuA4 complex. Subsequently, the classes that exhibited pNuA4 density on the nucleosome plate were selected and were subjected to 3D classification using CryoDRGN (66) to divide the entire dataset into 25 classes. The class with strong pNuA4 density was selected and further refined in cryoSPARC (63).

To perform the model building, the structures of NCP (PDB codes: 1ID3) and NuA4 (PDB codes: 5J9U) were rigid body fitted into the cryo-EM density maps using UCSF Chimera (67). The high quality of cryo-EM maps facilitated the model building of the complex by iterative manual adjustment in Coot (68). The structures were then refined in Phenix (69) for one round of real space refinement with secondary structure restraints.

HAT Assay.

Western blot was used for the initial characterization of HAT activities. Proteins in the reaction mixture were loaded onto a 15% SDS-PAGE gel and then transferred to a nitrocellulose membrane. An anti-acetyl lysine antibody (Abcam, ab190479, 1:2000 dilution) was used to detect the acetylation of the substrates, and quantification of the loading control was performed using an anti-histone H3 antibody (Abcam, ab1791, 1:2000 dilution). The Enhanced ECL luminescence detection kit (Thermo, 34577) was used for visualization. The blots were quantified by image scanning, and three independent blots were performed. Quantitative HAT activity was employed a continuous fluorometric HAT assay (70). In brief, free CoA produced in the acetyltransfer reaction reacts with N-[4-(7-diethylamino-4-methylcoumarin-3-yl)phenyl]maleimide (CPM) and forms a fluorescent CoA-CPM adduct that can be quantified by fluorescence intensity. The compound is excited by 400 nm light and emits fluorescence at 488 nm. The measured fluorescence intensity was converted to the concentration of CoA using a predetermined standard curve (71). For a 30-µL HAT reaction, 0.75 nM NuA4, 0.25 µM acetyl-CoA, 2-20 nM NCP, 1 µM CPM, and 5% DMSO were mixed in a buffer containing 10 mM HEPES, pH 7.0, 50 mM NaCl. The HAT activity was evaluated using a HITACHI F-7000, and the fluorescence signal corresponds to the production of CoA. Each reaction was run in triplicate and the results were reported as the mean ± SD. The fluorescence intensity of each substrate concentration was measured for 1,000 s. The stable linear growth phase was used to calculate the reaction’s initial velocity. The data were first analyzed using Excel and then fitted to the Michaelis–Menten equation using GraphPad Prism (GraphPad Software, Inc.), with the Km values derived from varying the concentration of the substrate.

EMSA.

pNuA4 was mixed with 5 nM H2A-NCP or H2A.Z-NCP at a molar ratio of 1:1, 2:1, 4:1, or 8:1 in 10 mM HEPES, 50 mM NaCl, pH 7.0 for 30 min with a total volume of 20 µL. After incubation on ice, 10 µL of each mixture was loaded onto a 5% Native-PAGE gel. The same amount of NCP and each ratio of pNuA4 were loaded onto a 15% SDS–PAGE gel and detected by Coomassie blue staining to show the correct ratio of NCP and pNuA4 for the input of the EMSA assays.

Nucleosome Reconstitution in smFRET Assays.

Double-stranded (ds) DNA constructs were created by annealing and ligating a set of overlapping oligonucleotides as described previously (72). All nucleosomes for the smFRET experiments were reconstituted with the following DNA fragments:

K1:5’Biotin/GGTACCCGTAGATCCTCTAGAGTGGGAGCTCGGAACACT

K2:5’ATCCGACTGGCACCGGCAAGGTCGCTGTTCAATACATGCACAGGATGTATATATCTGACACGTGCCTGGAGACTAGGGAG

K3:5’TAATCCCCTTGGCGGTTAAAACGCGGGGGACAGCGCGTACGTGCGTTTAAG

K4Alexa568:5’CGGT(Alexa568)GCTAGAGCTGTCTACGACCAATTGAGCGGCTGCAGACCGGGATTCTCCAGGGC

R1:5’GCCCTGGAGAATCCCGGTCTGCAGCCGCTCAATT

R2:5’GGTCGTAGACAGCTCTAGCACCGCTTAAACGCACGTACGCGCTGTC

R3:5’CCCCGCGTTTTAACCGCCAAGGGGATTACTCCCTAGTCTCCAGGCACGTGTCAGATATATACATCCTGTGCATGTATTGAACA

R4:5’GCGACCTTGCCGGTGCCAGTCGGATAGTGTTCCGAGCTCCCACTCTAGAGGATCTACGGGTACC

The nucleosomes were reconstituted with this “601” DNA (80 bp linker DNA) and histones as described previously (72).

NuA4 Labeling.

SNAP-Surface 649 (New England Biolabs) was dissolved in DMSO at a 1 mM concentration for stock. SNAP-tagged NuA4 was mixed with 5-fold excess SNAP-Surface 649 in NuA4-SNAP storage buffer (20 mM HEPES pH 7.0, 300 mM NaCl, and 10 mM DTT). The mixture was incubated at 4 °C overnight and the free dye was separated by dialysis.

smFRET Sample Preparation and Data Analysis.

Both slides and coverslips were washed with acetone, methanol, and a mixture of sulfuric acid and hydrogen peroxide in a volume ratio of 7:3, treated with aminosilane and sodium ethoxide, and then coated with a mixture of 99% mPEG (m-PEG-5000, Laysan Bio) and 1% of biotin-PEG (biotin-PEG-5000, Laysan Bio). The imaging buffer was composed of 50 mM HEPES, pH 7.0, 50 mM NaCl, 1 mM DTT, and an oxygen scavenging system (0.8% d-glucose; 1 mg/ml glucose oxidase; 0.4 mg/ml catalase; 1.5 mM Trolox). The experiments were performed on an objective-based total-internal-reflection fluorescence microscope (IX71, Olympus) at room temperature. During the recording period, NuA4 was introduced into the imaging chamber. The recording lasted approximately 30 s, with an exposure time of 50 ms per frame. However, we utilized alternating-laser excitation (ALEX) between 532 nm and 640 nm lasers, achieving a temporal resolution of 100 ms (SI Appendix, Fig. S6A). As the intensity excited by 640 nm increased 2-fold when two NuA4 proteins were bound on the nucleosome, we used this phenomenon to exclude data from the double NuA4 bound state. The signals of the Alexa568 and SNAP-Surface 649 channels were separated by a 630 nm dichroic mirror and detected by an EMCCD camera. 16-bit video with a time resolution of 50 ms was recorded and further analyzed. 100 pM nucleosome was anchored to coated quartz slides via a streptavidin–biotin linkage and then washed with imaging buffer. Following data collection, smFRET time traces were analyzed with an alternative step-finding algorithm to detect the FRET states (72, 73). In brief, we first fit a single large step to the data, determining the size and location based on a chi-squared calculation. We then search for new steps by fitting to the plateaus found in the previous cycles, each time selecting the most prominent one. This eventually leads to a series of “best” fits differing by only one step. We then compare each best fit in the series to a “counterfit” that has an equal number of steps as the original one but with step locations displaced randomly. A “step-indicator” S, defined as the ratio between the chi-squared of the counterfit and the chi-squared of the best fit, was then used to evaluate the quality of the step fits. The “step-indicator” S is large when the number of steps in the best fit is very close to the real number of steps. And initial binding state refers to the first binding state in which NuA4 first interacts with the nucleosome upon its introduction from solution. In our single-molecule FRET (smFRET) assay, this state is characterized as the first non-zero FRET state observed, prior to any subsequent FRET changes triggered by the addition of NuA4 to the imaging chamber. The “N” values are used to denote the total number of frames for the initial binding state under different conditions. The data were binned with a size of 0.03. The subpopulations were calculated using Gaussian fitting.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

A Simplified Model of How NuA4 Acetylates Both H4 and H2A.Z in the H2A.Z-containing Nucleosome. When pNuA4 positions the H2A.Z-containing nucleosome (ZCN), it has a much higher chance to bind to its preferred substrate, the N-terminal of H4, and initiate the acetylation process. In the H4-harboring states, pNuA4 interacts with the H2A.Z-containing nucleosome through multiple regions. With the assistance of the Epl1 NP domain, pNuA4 becomes immobilized on ZCN, preparing for acetylation, i.e., pNuA4 reaches the pre-H4-acetylation state. After completing the acetylation of H4, pNuA4 searches for other substrates, i.e., H2A.Z, along the nucleosome surface. In the pre-H2A.Z-acetylation state, pNuA4 locates the substrate, specifically the N-terminal of H2A.Z, and performs the acetylation. After the acetylation of H2A.Z, pNuA4 returns to the resetting state, preparing for the next round of acetylation.

Data, Materials, and Software Availability

The cryo-EM density map and atomic coordinates have been deposited in the Electron Microscopy Data Bank (EMDB) (https://www.emdatabank.org) and Protein Data Bank (PDB) (https://www.rcsb.org), respectively. The structure of pNuA4-ZCN complex at pre-H4-acetylation state [PDB: 8X2X (74), EMD-38021 (75)]; The structure of pNuA4-ZCN complex in class1 of the harboring state [PDB: 8X2Y (76), EMD-38022 (77)]; The structure of pNuA4-ZCN complex in class2 of the harboring state [PDB: 8X2Z (78), EMD-38023 (79)]; The structure of pNuA4-ZCN 2:1 complex at harboring state [PDB: 8X30 (80), EMD-38024 (81)]; The structure of pNuA4-ZCN complex with Ac-CoA at resetting state [PDB: 8X31 (82), EMD-38025 (83)]; The structure of pNuA4-ZCN H4KQ complex with Ac-CoA at resetting state [PDB: 8X32 (84), EMD-38026 (85)]; The structure of pNuA4-ZCN H4KQ complex without Ac-CoA at pre-H2A.Z-acetylation state [EMD-60587 (86)]. All study data are included in the article and/or supporting information.