Histone acetylation readers Bdf1 and Yaf9 direct SWR1 remodeler to +1 nucleosome

Abstract

The histone variant H2A.Z marking permissive chromatin is deposited by the multicomponent SWR1 chromatin remodeler, which is targeted to nucleosome-free promoters by a DNA length–sensing module. How SWR1 is directed to the flanking acetylated +1 nucleosome, its physiological substrate, has been enigmatic. We show by live-cell, single-molecule tracking that SWR1 subunits Bdf1 and Yaf9 harboring histone acetylation reader domains differentially regulate chromatin binding: Bdf1 promotes SWR1 association, while Yaf9-YEATS slows its dissociation. Notably, single-molecule tracking and genome-wide chromatin immunoprecipitation combined with exonuclease treatment reveal Bdf1 and Yaf9 contributions to global SWR1 targeting and histone exchange at +1 nucleosomes. Our findings highlight the in-cell biochemistry of histone readers and suggest a generalizable, two-stage mechanism wherein acetylated nucleosome interactions initially constrain the three-dimensional diffusion of SWR1 to increase local concentration, followed by stochastic one-dimensional diffusion at nucleosome-depleted regions with directional capture by acetylated +1 nucleosomes.

Histone readers Bdf1 and Yaf9 direct SWR1 to +1 nucleosomes, enabling targeted H2A.Z deposition via a two-step mechanism.

INTRODUCTION

The genomes of eukaryotes are organized into nucleosomes, DNA-histone complexes forming the basic unit of chromatin organization in the cell nucleus. Each nucleosome consists of 147 base pairs (bp) of DNA wrapped around an octamer of core histones, comprising a tetramer of histones, (H3-H4)₂, with dimers of histones H2A-H2B associated on each side of the (H3-H4)₂ tetramer (1). There are compositionally variant nucleosomes containing genetically distinct versions of each core histone, particularly for histone H2A (2), among which H2A.Z is the most widely conserved from yeast to humans. Genome-wide mapping shows that H2A.Z is site-specifically enriched in nucleosomes at gene regulatory regions, including enhancers and promoters (3), and functional studies reveal the emergence of H2A.Z as a marker for transcriptionally poised genes (4) and additionally implicate this histone variant in DNA repair (5). Elevated levels of H2A.Z are also observed in different forms of cancer (3), and H2A.Z is further implicated in neurological dysfunction (6).

The yeast SWR1 chromatin remodeler binds to canonical nucleosomes and conducts H2A.Z deposition in an adenosine 5′-triphosphate (ATP) hydrolysis–dependent histone exchange reaction in which an H2A-H2B dimer is replaced by the H2A.Z-H2B dimer (7, 8). SWR1 activity is notably site-specific, with preferential enrichment of the SWR1 complex and resultant H2A.Z incorporation at the +1 nucleosome position adjacent to the nucleosome-depleted regions (NFRs) near yeast gene promoters. The 14-component, 1-MDa SWR1 complex is structurally assembled on the catalytic Swr1 adenosine triphosphatase (ATPase) protein scaffold (9), which is organizationally distinguished by N- and C-terminal halves (10), and the enzymatic regulator Swc5 (11) and one of the Act1 subunits interact with both regions. The C-terminal half of Swr1, which contains the ATPase domain necessary for the H2A.Z exchange, forms a “C-SWR1” subcomplex with several components responsible for the high-affinity [~6 nM dissociation constant (K_D)] DNA binding (12): the Swr1 ATPase domain (13) along with Swc6 (13), Swc2, and Swc3 (12) subunits. Mutant SWR1 complexes deficient for Swc2 (and the associated Swc3 subunit) show strong defective Swr1 binding to DNA and nucleosome substrates in biochemical assays and to +1 nucleosomes at gene promoters in vivo (12, 14). The wild-type (WT) SWR1 complex binds preferentially to nucleosomes with a long linker DNA, a feature that should facilitate SWR1 binding to NFRs in chromatin. Moreover, single-molecule tracking (SMT) experiments in vitro recently revealed that upon DNA binding, SWR1 undergoes rapid one-dimensional (1D) diffusion on naked DNA stretched between optical traps (0.024 μm² s⁻¹, or ~100 ms for traversal between NFR-flanking nucleosomes) (15), which would allow the efficient delivery of SWR1 to −1 or +1 nucleosomes.

While biophysical, biochemical, and structural studies have provided substantial insights into DNA binding mechanisms provided by C-SWR1, the functional contributions of subunits associated with the N-terminal half of Swr1 in the N-SWR1 subcomplex to target search—particularly Bdf1 and Yaf9, two histone readers of posttranslational modifications (PTMs)—in the context of cellular chromatin remain poorly understood. Yeast Bdf1 belongs to the bromodomain and extraterminal domain (BET) family of proteins containing two bromodomains (BD1 and BD2) followed by an extraterminal domain (16). BET family proteins have been implicated in human diseases, and BET inhibitors are intensively investigated as promising therapeutic leads (17). The Bdf1 protein binds to the tetracetylated H4 histone peptide (K_D = 8.4 μM) (18), and a structural study of the bromodomains in the closely related human Brdt indicates that the BD1 domain of Bdf1 should simultaneously bind to diacetylated lysines K5 and K8 of histone H4, while BD2 binds to a single acetyl lysine (e.g., H3K18) (19). Using biochemical approaches, we and others have previously shown that Bdf1 enhances the affinity of SWR1 to acetylated nucleosomes in vitro (12, 20). However, the in vivo role of Bdf1 in SWR1 targeting and histone H2A.Z exchange has heretofore been analytically inseparable from its role as a subunit of the general transcription factor TFIID (21).

The Yaf9 subunit of SWR1 contains the YEATS domain (represented by Yaf9, ENL, AF9, Taf14, and Sas5), a histone PTM reader module that is distinct from the bromodomain by its recognition of histone H3 lysine acylations (acetylation and crotonylation) (22). Biochemical and structural studies show that the YEATS domain of Yaf9 interacts with acetylated and crotonylated forms of histone H3K27 and H3K9 (22, 23) via an aromatic cage presenting key residues T69, Y70, and W89 that bind to the acetyl lysine of histone H3 (23). Yaf9 is a subunit of SWR1 as well as in the HAT complex NuA4; its human homolog, GAS41 (glioma amplified sequence 41), is implicated in cancer and neurological disorders (24).

These two highly conserved bromodomain and YEATS domain and other histone readers are found in numerous multicomponent complexes for chromatin remodeling, histone modification, and transcription preinitiation (25). With the seminal concept of a histone code for transcriptionally active and repressive chromatin states (26), pioneering studies of histone reader domains have established the biochemical and structural basis of histone PTM recognition (25, 27). However, the quantitative, mechanistic contributions of histone readers, separately and combined, to the in vivo chromatin targeting of the large multiprotein complexes in which they reside remain an open question. Furthermore, the quantitative inputs from histone PTM- and DNA-based recognition, two distinct recruitment pathways used by myriad multiprotein machineries for genome metabolism, have been obscure.

Here, we use SMT (28) in living yeast to directly measure the chromatin binding and diffusional behavior of SWR1 in the native nuclear environment at high spatiotemporal resolution (29). We applied SMT to analyze SWR1 mutants and quantify the contributions of Bdf1 and Yaf9 in the search for chromatin targets. The two histone readers confer distinct kinetic properties to chromatin association and disassociation by SWR1, and their combined inputs approach the measured values for DNA binding components of the complex. Notably, we also use chromatin immunoprecipitation combined with exonuclease treatment (ChIP-exo) to reveal a prominent role for these histone readers in the directional engagement of SWR1 at the +1 nucleosome, resulting in enriched deposition of H2A.Z immediately downstream of the site of transcription preinitiation complex formation. These findings point to a consolidated targeting mechanism that entails the recognition of histone PTMs at two diffusive levels in addition to DNA recognition, leading to the local, site-specific recruitment of SWR1 for the creation of variant nucleosome architecture poised for transcription.

RESULTS

Major impact of Yaf9 and attendant Bdf1 depletion on H2A.Z incorporation

To study the contribution of individual SWR1 subunits to its search and engagement with canonical nucleosome targets for H2A.Z deposition (Fig. 1A), we first performed SMT of a functional histone H2A.Z-HaloTag fusion in WT and mutant cells (30). We applied a Bayesian inference method, saSPT (31), to classify molecules into one of several diffusional states, particularly the low-mobility (<0.1 μm² s⁻¹) chromatin-bound H2A.Z state, and the highly diffusive “free” state (Fig. 1, B and C). The relative abundance of chromatin-binding and free H2A.Z populations was also quantified by kinetic modeling of displacement histograms using “Spot-On” (32), which is particularly suited for limited datasets. The results show that 86% of total H2A.Z is incorporated in live dividing yeast, confirming previous findings (Fig. 1, C and D) (30). Both Spot-On and saSPT provide comparable estimates of the chromatin-bound H2A.Z fraction (Fig. 1D and fig. S1D), while saSPT yields further insights on the varied mobility of the free H2A.Z population. Diffusion analyses of nuclear HaloTag alone (Halo) and Halo-H2B provide controls for highly mobile and chromatin-bound molecules, respectively (Fig. 1C).

Fig. 1. Live yeast SMT shows the major impact of YAF9 gene deletion on H2A.Z incorporation.

(A) Representative cartoon showing SWR1-mediated ATP-dependent incorporation of one or two H2A.Z-H2B in nucleosomes. (B) Pipeline for the analysis of the fast-tracking imaging dataset. Spot-On kinetic modeling of displacement histograms (radial displacement r at time interval dt = 1 frame shown) was used to estimate the fractions of chromatin-bound and freely diffusing H2A.Z-Halo populations. State arrays (saSPT) modeling of trajectories provides the distribution and occupancy of diffusional states. (C) saSPT-modeled histogram of H2A.Z-Halo diffusion coefficients in WT and cells with SWR1 and YAF9 gene deletion. (D) Spot-On results showing fractions of chromatin-bound H2A.Z-Halo in WT and mutant cells with gene deletions of indicated SWR1 subunits. Standard deviations from two or three biological replicates are shown.

As expected, the loss of the catalytic Swr1 subunit, while not affecting the H2A.Z abundance (fig. S1, A to C), greatly reduces the steady-state fraction of H2A.Z incorporated into chromatin compared to WT when analyzed by saSPT (Fig. 1C) or Spot-On (29% versus 86%; Fig. 1D). The remaining incorporated H2A.Z is no longer enriched at promoter-proximal nucleosomes (14) and likely deposited from the “free” H2A/H2A.Z-H2B pool during nucleosome turnover. The loss of the Yaf9 subunit in the yaf9∆ strain with the attendant loss of interacting Bdf1 (described below) reduces chromatin incorporation of H2A.Z nearly to the level caused by the loss of Swr1 itself (34% versus 29%) (Fig. 1, C and D). Concomitant with the decrease in chromatin-bound H2A.Z, saSPT reveals an increase in several distinctive, fast-diffusing “free” populations (Fig. 1C), likely H2A.Z-H2B in association with histone chaperones (fig. S1E) (33). Other SWR1 subunits tested by gene deletion also display lower H2A.Z deposition in the corresponding strains but to a lesser extent than shown by yaf9∆ (Fig. 1D and fig. S1F). These results are consistent with a prominent role for Yaf9 and/or Bdf1 in depositing H2A.Z in living yeast.

Yaf9 and attendant Bdf1 depletion affects SWR1 binding kinetics

We extended the SMT analysis to examining mutant effects on the diffusivity of the SWR1 complex catalyzing H2A.Z deposition, as represented by an endogenous Swr1-Halo fusion labeled with Janelia Fluor dye (JF552). The Swr1 catalytic subunit provides the main scaffold for the modular assembly of cross-interacting subunits (Fig. 2A), and depletion of one subunit from the complex often results in the loss of additional interacting subunits (Fig. 2B, right) (9, 34, 35). In yaf9∆ cells, both Yaf9 and interacting Bdf1 are known to be lost and the Swc4 association is reported to be decreased, but all other components persist (9, 34, 35). Live-cell SMT shows that the total chromatin-bound SWR1 (Swr1-Halo) population exhibiting stable and transient interactions is substantially diminished in yaf9∆ cells (23%) compared to WT (51%) (Fig. 2C), indicating that Yaf9 and/or Bdf1 are partly responsible for targeting SWR1 to chromatin, likely via acetyl-histone interactions with the two bromodomains of Bdf1 and the YEATS domain of Yaf9. [Under our experimental culture conditions in regular rich nutrient media, the level of histone crotonylation is low (36)].

Fig. 2. Loss of Yaf9 and Bdf1 affects chromatin binding and residence time of SWR1.

(A) Subunit interactions on the Swr1 scaffold based on previous studies (34, 35). Fast tracking (10-ms frame interval) (B and C) and slow tracking (250-ms frame interval) (D and E) of Swr1-Halo. (B) saSPT-modeled histogram of Swr1-Halo diffusion coefficients in WT and SWR1 subunit mutants. Right, size of mutant SWR1 complexes. The asterisk (*) indicates other indirectly lost subunits. (C) Spot-On results showing fractions of chromatin-bound Swr1-Halo in WT and mutant cells with gene deletions of indicated SWR1 subunits. The stably bound fraction [from (D) and (E)] after normalization to the total chromatin-bound fraction is shaded (F_sb) in the plot. (D) Survival probability plots (1-CDF) of observed dwell times of single-bound molecules. Merged replicates are shown here; individual replicates along with double-exponential decay fit are shown in fig. S2B. (E) Fraction of stably bound F_sb (%) and corrected average residence times of stable binding τ_sb (in seconds) of Swr1-Halo in WT and mutant cells. Bootstrap resampling errors of merged replicates are shown.

For comparison, the loss of two DNA binding components of SWR1 (12, 14) [Swc2 depletion and the accompanying loss of Swc3 (9)] reduces the total chromatin binding of SWR1 in swc2∆ cells to the minimum level (17%) compared to WT (51%), as shown by saSPT (Fig. 2B) and Spot-On analyses (Fig. 2C and tables S1 and S2), consistent with earlier work (30). Moreover, depletion of Swc3 alone (swc3∆) and the Arp6 component (arp6∆) that anchors Swc2 and Swc3 to the ATPase subunit reveals a similar reduction of Swr1 binding (15 and 18%, respectively) (Fig. 2C), while analysis of swc5∆ shows a moderate reduction of SWR1 (Swr1-Halo) binding to chromatin (38%) (Fig. 2C). Thus, while the DNA binding module performs the expected chromatin targeting functions for SWR1, the quantitative contributions from the two histone readers to chromatin binding are sizeable. Control experiments indicate no substantive Swr1-Halo expression change in mutant cells (figs. S2D and S4C).

Bdf1 and/or Yaf9 might regulate the chromatin-binding fraction of SWR1 by enhancing its association rate via transient interactions with acetylated histones. The diffusive behavior of free SWR1 molecules in the yaf9Δ strain analyzed by saSPT is consistent with this view. In the WT strain, we find one broad SWR1 peak of high diffusivity, while the yaf9Δ mutant shows two overlapping highly diffusing SWR1 peaks (Fig. 2B). Although a small size reduction and corresponding change in hydrodynamic radius should increase the diffusion coefficient of the mutant complex, the appearance of a second peak of even greater diffusivity suggests that in addition to their size contributions, the histone reader components have a mechanistic role in restraining the diffusion of a measurable portion of the free WT SWR1 pool. A similar effect is observed for mutants involving the DNA binding module (particularly swc2Δ and arp6Δ), which should also undergo transient interactions with nucleosomal and nucleosome-free DNA in chromatin. Consistent with this, the average diffusion rate of free SWR1 (D_free) population, as observed by Spot-On kinetic modeling, is enhanced in mutants with impaired DNA binding ability—swc3∆, swc2∆, and arp6∆ (0.56, 0.55, and 0.57 μm² s⁻¹, respectively)—compared to WT D_free (0.41 μm² s⁻¹), but the mobility of free SWR1 is most enhanced upon the loss of Yaf9 and interacting Bdf1 in yaf9∆ cells (D_free of 0.74 μm² s⁻¹) (fig. S2A).

Yaf9 and attendant Bdf1 depletion reduces SWR1 residence time

Live-cell SMT enables the measurement of the duration of chromatin-binding events, which are revealed by a “slow-tracking” imaging regime using longer camera exposure (250 ms) to motion blur freely diffusing molecules and lower laser power to reduce photobleaching. To estimate the upper temporal limit for SWR1 binding to chromatin, we plotted the survival probability (1-CDF) fit to a double-exponential function to estimate stable and transient SWR1 residence times (τ_sb and τ_tb) and corresponding stable and transient chromatin-binding fractions (F_tb and F_sb, respectively) (Fig. 2D; fig. S2, B and C; and tables S1 and S3). We find that WT SWR1 (Swr1-Halo) has a stable binding fraction (F_sb = 16%, normalized to total bound and free molecules; Fig. 2C) and—after correcting for photobleaching and chromatin movements using an H2B-Halo reporter (37)—a stable residence time (τ_sb = 5.1 s) that is similar to the times of other ATP-dependent chromatin remodelers (τ_sb = 4 to 7 s) (Fig. 2, C to E) (38).

Mutant SWR1 complexes (Swr1-Halo) imaged in deletion strains swc2∆, swc3∆, swc5∆, and arp6∆ exhibit stable residence times near the 5.1-s WT value (5.4, 4.5, 5.7, and 6.4 s, respectively), but we find markedly lower values for stable chromatin-binding fractions (F_sb = 3, 3, 9, and 3%, respectively) compared to WT (F_sb = 16%) (Fig. 2, C and E). However, unlike these mutants, the stable residence time of SWR1 for the yaf9∆ strain (τ_sb = 3.6 s) is reduced by 1.5-fold compared to that for WT (τ_sb = 5.1 s) (Fig. 2E), indicating an elevation of the SWR1 dissociation rate (1/τ_sb). In addition, we also observe a decrease in the SWR1 stable binding fraction (F_sb = 6%) (Fig. 2, C and E). In yaf9∆ cells, the reduction of SWR1 residence time and stable binding fraction increases search time, i.e., the interval between two stable binding events, from 18 s (WT) to 30 s (mutant) (table S1).

Structure-guided mutation of the Swr1-Bdf1 interface enables singular Bdf1 removal from SWR1

Biochemical studies have shown that the Bdf1 subunit of SWR1 contains two bromodomains that bind to acetylated histone H4 (18) and that the purified SWR1 complex has higher affinity for acetylated nucleosomes in vitro (12). However, the shared presence of Bdf1 in the general transcription factor TFIID precludes simple dissection of the in vivo functions of Bdf1 in H2A.Z exchange. The bdf1∆ mutant strain grows extremely poorly (fig. S3E), unlike swr1∆ cells that present near normal growth (fig. S4F, left), suggesting that the bdf1∆ phenotype is primarily due to TFIID malfunction. Mapping of domains within the Swr1 scaffold revealed that Bdf1 interacts with Swr1 residues 211 to 265 in the reconstituted SWR1 complex (35), and an AlphaFold2 (39) prediction narrows the Swr1-Bdf1 interaction surface to 11 residues (241 to 251) (Fig. 3A and fig. S3A) (40). Accordingly, we generated mutant complexes carrying Swr1-FLAG truncations of either 55 residues (Swr1∆55^-Bdf1) or 11 residues (Swr1∆11^-Bdf1), confirming by SDS–polyacrylamide gel electrophoresis that affinity-purified mutant complexes are only deficient only for the Bdf1 component (Swr1∆55^-Bdf1 and Swr1∆11^-Bdf1; Fig. 3B and fig. S3B) (41). This structure-guided mutational strategy enabled specific analysis of Bdf1 function in the chromatin-binding kinetics of SWR1.

Fig. 3. The structure-guided Swr1 mutant shows Bdf1 requirement for chromatin-binding kinetics of SWR1.

(A) Design of Swr1 truncation mutants for loss-of-Bdf1 experiments. Schematic representation of the Swr1 subunit and region of Bdf1 interaction. (B) Loss of Bdf1 from SWR1 shown by immunoaffinity FLAG purification of mutant complexes. (C) Spot-On results showing fractions of chromatin-bound Swr1-Halo after the loss of Bdf1. The stably bound fraction [from (E) and (F)] after normalization to the total chromatin-bound fraction is shaded (F_sb) in the plot. (D) saSPT-modeled histogram of diffusion coefficients of Swr1-Halo. (E) 1-CDF plot of Swr1-Halo in mutant cells. (F) F_sb along with corrected average residence time τ_sb for Swr1-Halo in mutant cells.

Bdf1 facilitates chromatin association of SWR1

The Bdf1 protein has high affinity for the tetra-acetylated histone H4 tail (18), consistent with the presence of bromodomains BD1 and BD2, with the wide BD1 structural pocket that simultaneously binds to H4 K5ac and K8ac. Fast tracking of SWR1 (Swr1-Halo) in Swr1∆11^-Bdf1 cells reveals substantial reduction of total chromatin binding (28%) compared to WT (51%) as shown by Spot-On kinetic modeling, demonstrating the singular importance of Bdf1 (Fig. 3C), and concurring analysis by saSPT also shows reduction of the slow-diffusing “bound” fraction for the Bdf1-depleted complex (Fig. 3D). Using the slow-tracking regime, the survival probability profile of the bound SWR1 shows a stable residence time of 4.8 s for the mutant SWR1 complex (Swr1∆11^-Bdf1), which is comparable to the WT value of 5.1 s (Fig. 3, E and F, and fig. S4, D and E). However, the stable binding fraction is substantially diminished (F_sb = 7%) compared to WT (F_sb = 16%). Given that steady-state chromatin binding is a function of the kinetic on- and off-rates (K_D = K_off/K_on), the results indicate that Bdf1, likely via the two bromodomains, has a primary role in enhancing the on-rate kinetics of SWR1. Moreover, saSPT analysis of free SWR1 diffusion for the Swr1∆11^-Bdf1 mutant reports the presence of a second peak of greater diffusivity (Fig. 3D and fig. S3G), indicating a role for Bdf1 in constraining the free diffusion of WT SWR1 for higher on-rate kinetics. Intriguingly, despite a decrease in the association of mutant SWR1 to chromatin, H2A.Z incorporation in the mutant strain remains indistinguishable from WT (86%) (fig. S3, C and D), suggesting that despite the impairment, SWR1 lacking Bdf1 but retaining normal Yaf9 has sufficient activity to achieve normal steady-state H2A.Z levels under culture conditions.

To further dissect the role of the Bdf1 bromodomains in SWR1 recruitment, we introduced point mutations (T186F and T353F) within BD1 and BD2 to disrupt their acetyl-lysine binding pockets (18). Unlike the growth-defective bdf1∆ strain, the Bdf1-BD1,2^YF mutant exhibited normal growth under standard conditions (fig. S3E). Fast tracking of SWR1 in this mutant revealed a reduction in total chromatin binding to 38% (fig. S3F), intermediate between WT (51%) and the more severe defect observed upon the loss of interacting Bdf1 in the Swr1∆11^-Bdf1 mutant (28%). Concordantly, partial defects on the restrained diffusion of free SWR1 were observed in Bdf1-BD1,2^YF (fig. S3, G and H). These findings are consistent with a partial impairment in bromodomain-dependent ligand recognition in the Bdf1-BD1,2^YF mutant. Similar to observations in the Swr1∆11^-Bdf1 mutant, slow-tracking SMT revealed that the stable residence time of SWR1 in the Bdf1-BD1,2^YF mutant (4.8 s) was comparable to that of WT cells (5.1 s) (fig. S3I), suggesting a limited role of bromodomains in maintaining the residence time of stably bound SWR1.

The YEATS domain enhances SWR1 residence time, reducing dissociation

The YEATS domain of Yaf9 has been shown to have high affinity for acetylated H3 at H3K27 and H3K9 (22, 23). The high-resolution structure of the YEATS domain of yaf9 bound to the H3K27ac peptide highlights three key hydrophobic residues T69, Y70, and W89 critical for H3K27ac binding (Fig. 4A) (23). Although the yaf9∆ strain provides functional insight on the joint loss of Yaf9 and Bdf1, we assessed the specific importance of the acetyl interaction by the Yaf9-YEATS domain by introducing alanine substitutions at T69, Y70, and W89 (Yaf9-YEATS^3A) (23). Fast-tracking SMT analysis shows a minimal reduction of total chromatin binding by SWR1 (Swr1-Halo) in the Yaf9-YEATS^3A mutant strain (46%) compared to WT (51%) (Fig. 4B). However, the slow-tracking regime reveals that the stable residence time of SWR1 (τ_sb = 5.1 s) is reduced by nearly twofold in the Yaf9-YEATS^3A mutant (τ_sb = 2.7 s) (Fig. 4, D and E, and fig. S4D), while the stable chromatin-binding fraction (F_sb = 16%) shows no change from the WT level (Fig. 4B). Thus, in contrast to the role of Bdf1 in raising the SWR1 on-rate, the YEATS domain primarily functions to reduce the WT SWR1 off-rate (K_D = 1/τ_sb). The net consequence is a small decrease in H2A.Z incorporation (82%) compared to WT (86%) (fig. S4A).

Fig. 4. The YEATS domain mutant reduces SWR1 residence time on chromatin.

(A) Structure of the Yaf9 YEATS domain bound to the H3K27ac peptide [from (23)] with key hydrophobic residues interacting with H3K27Ac highlighted. (B) Spot-On results of Swr1-Halo tracking in the Yaf9-YEATS^3A (T69A, Y70A, and W89A) mutant and double mutant with Swr1 (241–251∆). The stably bound fraction [from (D) and (E)] after normalization to the total chromatin-bound fraction is shaded (F_sb) in the plot. (C) saSPT analysis of Swr1-Halo in mutant cells. (D) 1-CDF plot of Swr1-Halo in Yaf9-YEATS^3A mutants. (E) F_sb and corrected average residence times of stable binding τ_sb of Swr1-Halo in Yaf9-YEATS^3A mutants. The corrected stable residence time of Swr1-Halo shows a reduced residence time of SWR1 in Yaf9-YEATS^3A mutants.

Bdf1 and Yaf9 cooperate to boost SWR1 binding

Because the two histone readers are present in the same complex, we tested for cooperating activities by generating a double mutant combining Bdf1 loss (Swr1∆11^-Bdf1) with the Yaf9-YEATS^3A mutation. As expected, the total chromatin binding of SWR1 in double-mutant cells (23%) is diminished compared to WT (51%) (Fig. 4B), a lower level than that observed for the loss of Bdf1 alone (28%; Fig. 3C). Furthermore, the behavior of free SWR1 clearly shows a distinct peak of higher mobility for the double mutant (Fig. 4C and fig. S3G), and the free SWR1 molecules have enhanced mobility with a D_free of 0.64 μm² s⁻¹ compared to the WT value of 0.41 μm² s⁻¹ (fig. S3H). The observed behavior of free SWR1 in double-mutant cells is consistent with findings for the yaf9∆ strain with the attendant loss of interacting Bdf1. In addition, the SWR1 residence time in double-mutant cells (τ_sb = 3.1 s) is reduced compared to that in WT (τ_sb = 5.1 s), akin to the Yaf9-YEATS^3A single mutant (τ_sb = 2.7 s), and the stable chromatin-binding fraction is reduced in the double mutant (6%) compared to the Yaf9-YEATS^3A single-mutant and WT strains (16%) (Fig. 4, B, D, and E). Thus, both the residence time and chromatin-binding fractions of SWR1 in the double mutant match values obtained from yaf9∆ cells where Yaf9 and Bdf1 are jointly lost from the complex (Fig. 2, A and D).

H2A.Z incorporation is measurably reduced by 14% in the double mutant (72%) compared to WT (86%), in contrast to little or no change of H2A.Z deposition in single-mutant strains (figs. S3, C and D, and S4, A and B). These findings indicate that Bdf1 and the Yaf9-YEATS domain cooperate to promote steady-state SWR1 chromatin binding and H2A.Z incorporation. Consistently, the double mutant exhibited a synthetic growth defect compared to the single mutants (fig. S4F), indicative of a genetic interaction and functional cooperation between Bdf1 and Yaf9-YEATS in vivo. However, despite the 14% reduction, it is intriguing that the yaf9∆ strain exhibits even lower H2A.Z incorporation (34%) than the double mutant, suggesting that the Yaf9 subunit—beyond the three alanine-substituted residues important for acetyl-lysine interaction—has an additional role relevant to SWR1 activity.

Bdf1 and Yaf9 cooperate in genome-wide targeting of SWR1 to the +1 nucleosome

Compared to the DNA binding components of SWR1, histone readers Bdf1 and Yaf9 have a lesser influence on the global recruitment of SWR1 and incorporation of H2A.Z as measured above by SMT. However, because of the reported asymmetry of histone acetylation between −1 and +1 nucleosomes (42), we hypothesized that histone readers have a key role in preferential SWR1 targeting to the +1 nucleosome. Accordingly, we used ChIP-exo to examine the distribution of SWR1 (represented by Swr1-TAP) in histone reader mutants. ChIP-exo previously demonstrated the specific enrichment of Swr1-TAP at gene promoters, with little detection downstream of the +1 nucleosome (43).

To compare SWR1 distributions revealed by ChIP-exo, we aligned TFIID-dominated tandem genes at the respective midpoints of their NFRs or at the center of +1 and −1 nucleosomes. TFIID-dominated tandem genes, representing nearly half the genome, allow for the distinction of the −1 and +1 nucleosomes and an assessment at genes where H2A.Z incorporation predominates (44, 45). Analyzed in this way, we confirmed the high enrichment of SWR1 at the +1 nucleosome, with moderate enrichment at the −1 nucleosome compared to flanking chromatin (Fig. 5A, upper panel, and fig. S5D). Notably, in single mutants with either the loss of interacting Bdf1 (Swr1∆11^-Bdf1) or the YEATS domain (Yaf9-YEATS^3A), we observe a reduction in SWR1 enrichment at the +1 nucleosome [upper panels—composite plot analysis of all genes (fig. S6, A and B) or tandemly oriented genes (Fig. 5, A and B, and figs. S5D and S7, A and B)]. Comparing WT and mutant cells reveals that genes exhibiting higher levels of SWR1 at the +1 nucleosome are primarily affected in both mutants (lower panels; Fig. 5, A and B, and fig. S6, A and B). The absence of histone PTM readers in these mutants does not affect the level of SWR1 at the −1 nucleosome; if anything, we observe miniscule enrichment (Fig. 5, A and B, upper panel; and figs. S5D and S7A).

Fig. 5. Histone readers direct SWR1 binding to and H2A.Z deposition at +1 nucleosomes.

Enrichment of SWR1 (A to C) and H2A.Z (D to F) at 2328 tandem genes [ribosomal protein (RP) genes excluded]. (A) Loss of Bdf1 leads to reduced SWR1 enrichment at the +1 nucleosome. The top panel shows ChIP-exo of Swr1-TAP in WT and Swr1∆11^-Bdf1 mutant. The bottom panel shows the SWR1 level at the +1 nucleosome of individual genes with WT levels plotted against the Swr1∆11^-Bdf1 mutant. AU, arbitrary units. (B) SWR1 enrichment at the +1 nucleosome in the Yaf9-YEATS^3A mutant. (C) Binding of SWR1 to the +1 nucleosome is severely reduced in the double mutant. (D and E) ChIP-exo of H2A.Z-3Flag shows that levels of H2A.Z are similar to WT in Swr1∆11^-Bdf1 and Yaf9-YEATS3A single mutants. (F) H2A.Z levels are reduced at +1 in the double mutant. The effect on genes with intermediate and high levels (red square) and a smaller subset with low levels (blue square) of H2A.Z at +1 is highlighted (bottom panel). [Plots in (A) to (F) represent averages of three replicates for SWR1 and two replicates for H2A.Z.]

Next, we combined the loss of interacting Bdf1 (Swr1∆11^-Bdf1) with the YEATS^3A mutation and observed substantial reduction of SWR1 at the +1 nucleosome and a small increase at the −1 nucleosome (Fig. 5C, upper panel, and fig. S5D). For the double mutant, the level of SWR1 at the −1 nucleosome approaches that of the +1 nucleosome, indicating a major failure of directional engagement by the SWR1 complex at NFR-flanking nucleosomes. Given that the genome-wide distribution of cognate histone PTMs for Bdf1 (acetylated H4 K5, K8, K12, and K16) and the Yaf9 YEATS domain (acetylated H3K27) (42) is substantially higher at the +1 nucleosome relative to the −1 counterpart (fig. S5, A and B), we propose that SWR1 enrichment at the +1 nucleosome is caused by its capture on acetylated histones via bromodomain and YEATS domain interactions during the 1D diffusion of SWR1 on NFR DNAs (15) and, less likely, via three-dimensional (3D) nucleoplasmic diffusion (15).

Partial functional redundancy of histone readers for H2A.Z deposition

We further examined the functional impact of histone reader mutants on H2A.Z deposition by the ChIP-exo analysis of strains harboring H2A.Z-Flag. Compared to WT cells, where H2A.Z exhibits characteristic enrichment at +1 over −1 nucleosomes, single-mutant strains for either Bdf1 loss (Swr1∆11^-Bdf1) or the Yaf9-YEATS^3A mutation show only a marginal effect on the genome-wide enrichment of H2A.Z at the +1 nucleosome (Fig. 5, D and E, and figs. S5D and S6, D and E), consistent with partial redundancy between the two readers for H2A.Z deposition. However, these two histone reader mutants together result in a notable reduction of H2A.Z at the +1 nucleosome (Fig. 5F and figs. S5D and S6F, top). Further examination of individual genes reveals that Yaf9 and Bdf1 enrich H2A.Z at the +1 nucleosome for those exhibiting moderate to high levels of the histone variant (Fig. 5F and fig. S6F, bottom, red rectangle; i.e., the density of genes inside the red rectangle is higher below the line than above). In addition, for the double-mutant strain, the SWR1 complex relocates to sites of low occupancy, boosting H2A.Z levels at those sites. We observe such boosting for genes typically low in H2A.Z at the +1 nucleosome (Fig. 5F and fig. S6F, bottom, blue rectangles; i.e., the density of genes inside the blue rectangle is higher above the line than below) and for ribosomal protein genes known to have low H2A.Z (fig. S5C) (46).

We conclude that the histone reader components of SWR1 not only assist in 3D association-dissociation kinetics on acetylated chromatin but also perform a pivotal function in the directional capture of SWR1 as it navigates the NFR at gene promoters to preferentially bind the lysine-acetylated +1 nucleosomes, with the consequential enrichment of H2A.Z at the +1 location. Differential H2A.Z incorporation is not wholly dependent on these histone readers because the +1 nucleosome still retains elevated H2A.Z in the double-mutant strain (Fig. 5F), suggesting that other mechanisms also make a substantial contribution, including the DNA sequence composition underlying +1 nucleosomes (47) and histone PTMs (48) affecting the intrinsic stability of the canonical nucleosome, both of which are likely to influence the histone exchange reaction.

DISCUSSION

In this work, we have measured by live-cell SMT the quantitative contributions of Bdf1 and Yaf9 components of the yeast SWR1 remodeler to H2A.Z deposition at promoter-proximal +1 nucleosomes. SWR1 binding to NFRs is directed by its free DNA length–sensing module (12, 40), but Bdf1 and Yaf9 components also make substantial contributions to the kinetics of SWR1 association and dissociation on chromatin. Previous biochemical studies of the two bromodomains of yeast Bdf1 (BD1 and BD2) established that these domains specifically recognize acetylated histone tail peptides, suggestive of contributions by Bdf1 to chromatin binding (18, 49), and the YEATS domain of Yaf9 is known to bind acetyl-lysine and crotonyl-lysine (22). Our SMT analysis of the singular Bdf1 depletion mutant (Swr1∆11^-Bdf1) and the Yaf9-YEATS^3A mutant shows that Bdf1 makes a sizeable contribution to the association rate of SWR1, while the YEATS domain slows the dissociation of SWR1 from chromatin. Thus, the two distinct histone readers influence steady-state SWR1 binding by different kinetic mechanisms, which may be related to the structural architectures of the bromodomain and YEATS domain interactions with acetylated histone tails. A wide and shallow bromodomain pocket suggests greater access to acetylated histone tails to increase interaction frequency, whereas a less exposed YEATS domain pocket may affect dissociation kinetics by limiting the degrees of freedom for the rupture of intermolecular interactions (Fig. 6, E and F).

Fig. 6. 3D and 1D diffusion models for site-specific SWR1 targeting and H2A.Z deposition.

(A) 3D diffusion model of SWR1. Acetylated histone tail binding by the bromodomain and YEATS domain in Bdf1 and Yaf9, respectively, constrains the mobility of free SWR1 in regions of genome with high levels of histone acetylation, such as +1 nucleosomes and gene bodies of active or transcriptionally poised genes. For nucleosomes, the central H3-H4 tetramer is depicted in gray, which is flanked by H2A-H2B dimers in blue. The H3 tail (cyan) and H4 tail (pink) are shown extending outward, with red stars indicating acetylated lysine residues. For SWR1 (yellow), subunits containing DNA and histone PTM reader domains are indicated, along with the bound H2A.Z-H2B dimer. (B) The DNA binding module of SWR1 facilitates stable association with NFR at gene promoters. (C) In line with recent in vitro studies (15), the promoter-associated SWR1 could rapidly diffuse on DNA at NFR and reach flanking nucleosomes. (D) The bromodomain and YEATS domain in SWR1 facilitate preferential SWR1 binding to the acetylated +1 nucleosome. (E) SWR1 mediates one or two rounds of H2A.Z incorporation. Consistent with in vivo H2A.Z mapping (63), the model depicts preferential NFR-distal H2A.Z deposition.

Two-stage consolidated model for SWR1 targeting

We further revealed by ChIP-exo analyses that the combined inputs of Bdf1 and Yaf9-YEATS have a considerable influence on the directional recruitment of SWR1 to +1 nucleosomes for H2A.Z exchange, a unique property not observed for its DNA binding components. These findings suggest a consolidated two-stage model describing how histone readers guide multicomponent complexes in which they reside during the search for chromatin targets. The model is based in part on recent in vitro studies demonstrating that on initial DNA binding from 3D diffusion in solution, SWR1 undergoes rapid 1D diffusion (sliding and hopping) on free DNA until blockage by a nucleosome in its path (15). Accordingly, we propose that a diffusive search in at least two regulated stages is conducted on chromatin in living cells (Fig. 6). At the initial stage, the 3D diffusion of the enzyme complex in the nucleoplasm is locally constrained by transient, weak, multivalent Bdf1-BD1-BD2 and Yaf9-YEATS interactions with acetylated-histones enriched at +1 nucleosomes and active gene bodies (Fig. 6A). (We have not investigated the two-dimensional surface scanning of nucleosome arrays in vitro, but similar diffusive constraints would apply.)

The constraints on 3D diffusion cause a rise in local SWR1 concentration, leading to increased high-affinity binding at the NFR followed by primarily 1D diffusion to flanking nucleosomes. NFR binding from 3D diffusion can initiate anywhere on exposed DNA lengths and is directed by the DNA recognition module of SWR1 (Swc3, Swc2, and Swr1 ATPase) (Fig. 6B) (40). Binding directly from 3D to NFR-proximal nucleosomes is less probable because of DNA compaction. In the second, 1D-diffusive stage, stochastic, bidirectional 1D sliding and hopping of SWR1 on NFR DNA are expedited by an open structural conformation of SWR1 stabilized by ATP binding (Fig. 6C) (15, 40). 1D diffusion at the NFR ends with blockage and preferential capture by the flanking +1 nucleosome enriched for acetylated histone tails (Fig. 6D). Capture and retention are facilitated by the on-rate increase because of histone tail acetyl-lysine interactions with Bdf1 and the off-rate decrease because of interactions by Yaf9-YEATS, both contributing to a higher frequency of stably bound SWR1 at the +1 nucleosome for H2A.Z exchange. Such a consolidated search strategy that combines the recognition of accessible DNA and histone lysine acetylation at several diffusive dimensions provides distinct opportunities for biological regulation of remodeler recruitment and H2A.Z enrichment.

Histone acetylation boosts SWR1 recruitment

SMT combined with genetic manipulation allows dissection of the individual roles of histone PTM reader proteins Bdf1 and Yaf9 in regulating the on- and off-rates of SWR1. It is of interest that sizeable changes of mutant SWR1 complex kinetics on chromatin do not correlate in a linear fashion with changes of enzymatic activity and resultant deposition of H2A.Z at +1 nucleosomes. Notably, the singular depletion of Bdf1 substantially reduces both the stable binding and transient binding of SWR1 (Fig. 3, C and D) but has no detectable effect on global H2A.Z incorporation (fig. S3E) and little effect on +1 nucleosomes (Fig. 5D), indicating that the presence of Yaf9-YEATS in the mutant complex provides sufficient recognition of acetylated histone tail to yield normal steady-state H2A.Z levels. Likewise, the single Yaf9-YEATS^3A mutant exhibits increased off-rate (Fig. 4, D and E) but little or no reduction of H2A.Z incorporation either globally (fig. S4, D and E) or on +1 nucleosomes (Fig. 5E), suggesting that the Bdf1 bromodomain function can mostly compensate for damage to Yaf9-YEATS. Thus, there is considerable functional redundancy provided by the inclusion of two histone readers in the SWR1 complex.

However, a deficiency of both histone readers results in clear impairment of local and global SWR1 binding and H2A.Z deposition at +1 nucleosomes (Fig. 6, A to E). Bdf1 and Yaf9 are biochemically located within the N-terminal module of SWR1 (34), and cryo–electron microscopy structures and models of the SWR1-nucleosome complex illustrate the spatial proximity of the two histone readers oriented with the engaged nucleosome via DNA binding components of SWR1 (40). This configuration allows BD1-BD2 bromodomains and the YEATS domain to reach acetylation marks on the H3 and H4 tails of the same nucleosome with ease, providing a physical explanation for our kinetic observations demonstrating functional redundancy between histone readers in living cells. Thus, the interaction of either one and possibly both histone reader domains with acetylated H3 and H4 tails from the same nucleosome would stabilize the enzyme-nucleosome complex in an optimally engaged orientation for histone dimer exchange. Although this enhancement can be bypassed in vitro using unmodified nucleosomes at sufficient enzyme concentrations (12), our current results and proposed model warrant renewed investigations at the single-molecule level in vitro to elucidate the influence of histone PTMs and the Yaf9 and Bdf1 components of SWR1 during 1D and 3D diffusion to its nucleosomal target for histone exchange.

Notwithstanding redundancies between two histone readers of SWR1, a mutant of the Yaf9-YEATS domain by itself has been reported to exhibit a defective phenotype in timely H2A.Z deposition (22). Thus, an individual histone PTM reader could have a larger functional role in the deposition of H2A.Z in response to changing cellular environments; in metazoans, these scenarios could include tissue differentiation and pathological states. In addition, our studies were performed under cell culture in rich nutrient media where the level of lysine-crotonylation is low (36). It would be of much interest to apply SMT and ChIP-exo in single- and double-mutant strains to assess how SWR1 binding kinetics change when conditions favor histone crotonylation over acetylation in cells.

Human bromodomain proteins constitute a large family of 61 members, many of which have shown specificity for lysine acetylation of histone tails and are frequently mutated in cancers or form part of oncogenic fusions because of chromosomal rearrangements (50). Likewise, there are more than 100 YEATS domain proteins conserved among numerous species from yeast to humans (51). The YEATS protein family is also often mutated in cancers (52) and associated with oncogenic protein fusions (53). Both histone readers are included in numerous chromatin remodeling, modification, and transcription complexes that are endowed with a multiplicity of DNA binding and histone PTM binding modules (54). It is tempting to speculate that histone readers in general and acyl-lysine histone readers in particular function similarly to SWR1 by boosting the kinetics of chromatin interactions during 3D and 1D diffusion throughout two-stage targeting to optimize their physiological activities on the posttranslationally modified epigenome. Last, the dependence on differential histone acetylation for the asymmetric capture of SWR1 poses the intriguing question of the underlying basis for asymmetric histone acetylation at NFR-flanking nucleosomes. This provides a challenge for future studies on the complex interplay between chromatin remodeling, histone modification, and transcription.

MATERIALS AND METHODS

Generation of yeast strain

The C-terminal Halo-tagged Swr1 and H2A.Z proteins were expressed as sole copies in yeast cells, with a flexible (SG₄)₂ linker connecting the C terminus of Swr1 or H2A.Z to the Halo tag (30). H2B with an N-terminal Halo tag was expressed as a sole copy using a short (GA₃) linker between the Halo tag and H2B (30). Free Halo was expressed as a fusion to the bipartite SV40 nuclear localization signal (30).

We applied standard antibiotic cassette–based methods for gene deletions and used a modified version of CRISPR-Cas9 (55) for amino acid substitutions in the YEATS domain of Yaf9, as well as for Swr1 truncations. In this method, a plasmid construct was designed for the expression of guide RNA (gRNA). A 20-nucleotide-long CRISPR guide sequence was selected using Benchling software, ensuring an on-target score of at least 50 and an off-target score between 70 and 100. Additional sequences, as shown below, were added to the ends of the 20-nucleotide CRSIPR guide for cloning into the pSW082 plasmid: 5′ GACTTTnnnnnnnnnnnnnnnnnnnn 3′; AAnnnnnnnnnnnnnnnnnnnnCAAA 5′.

The 5′ ends of oligos were phosphorylated with T4 polynucleotide kinase, followed by annealing and cloning into pSW082 via golden gate assembly using a polymerase chain reaction (PCR) cycler. The reaction mix contained Bsm B1 (a type IIS restriction enzyme) and T4 DNA ligase. Insertion of the gRNA sequence disrupted the green fluorescent protein expression cassette in pSW082. After bacterial transformation, white colonies were selected, and cloning was confirmed by DNA sequencing. The pSW082 with the inserted sequence was then digested with Eco RV, and the gRNA expression cassette was gel purified for use in yeast transformation. In addition, the Cas9 expression plasmid (pWS158) was digested with Bsm B1, and the linearized plasmid was gel purified before use in yeast transformation.

Donor DNA was synthesized as a double-stranded DNA (gene block) from IDT. The gRNA binding site in the donor DNA was modified (without altering the encoded amino acids) to prevent binding gRNA to the edited genome. The donor DNA contained ~200 bp of overlap with the DNA flanking the Cas9 cut site for homologous recombination. This gene block was further amplified by PCR and purified using a Qiagen PCR purification kit.

For genome editing, yeast cells were transformed with three DNA fragments: 100 ng of Bsm B1–linearized pWS158 plasmid, 200 ng of Eco RV–digested gRNA cassette from pSW082, and 2 μg of donor DNA containing the mutation and overlapping sequences for recombination. Transformed cells were selected on a CSM-URA plate and then transferred to a yeast extract, peptone, and dextrose (YPD) plate. Typically, three yeast clones were selected and verified by DNA sequencing. This CRISPR-Cas9 method enabled efficient, scarless gene editing. For a detailed description of the method, refer to Shaw et al. (55) and Benchling’s CRISPR tools. Yeast strains used in this work are listed in table S4.

FACS analysis of H2A.Z abundance in mutants

Overnight cultures in the early log phase [OD₆₀₀ (optical density at 600 nm) < 1] were diluted to an OD₆₀₀ of 0.2 and labeled with JF646 dye for 2 hours. Free dye was removed through five washes, after which the cells were resuspended in phosphate-buffered saline containing 4% formaldehyde (Thermo Fisher Scientific, 16% formaldehyde ampules) and incubated for 15 min at room temperature with gentle mixing.

The cells were then washed again and stored in phosphate-buffered saline at 4°C in the dark. Just before performing fluorescence-activated cell sorting (FACS) analysis, the cells were briefly sonicated at 4°C using a Biorupter 300 (Diagenode). Flow cytometry analysis was carried out using a BD FACS Canto flow cytometer.

Fluorescent labeling of cells for SMT

All yeast cells were grown and imaged in complete supplement mixture (CSM) media with additional adenine hemisulfate (40 mg/l). Overnight early log phase cultures (OD₆₀₀ < 1) were diluted to an OD₆₀₀ of 0.1 and labeled with JF552 dye for 4 hours at 30°C. For fast tracking, H2A.Z and Swr1 were labeled with 10 and 30 nM JF552 dye, respectively. For slow tracking, Swr1 was labeled with 10 nM JF552 dye. Free dye was removed by performing five 1-ml washes, after which the cells were immobilized on concanavalin A–coated coverslips (no. 1.5, ~0.17-mm thickness, 25-mm diameter, Electron Microscopy Services). Before immobilization, the coverslips were assembled in metal chambers (35-mm diameter, Attoflour, Invitrogen). Any floating cells were washed way to achieve a near-monolayer of yeast cells. Live cells were then imaged for SMT at room temperature while submerged in 1 ml of CSM + adenine media.

Single-molecule imaging in live cells

The wide-field single-molecule imaging setup used in this study has been described previously (30, 56). Briefly, a Zeiss Axio Observer Z1 microscope was custom fit with external lasers and cameras. Images were acquired using the Hamamatsu Photonics C9100-13 and its newer model C9100-23B, both of which are electron-multiplying charge-coupled device (EM-CCD) cameras. The cameras were operated at ~−80°C cooling and an electron microscopy gain of 1200×. The size of each pixel was 107 nm. For fast tracking, images were captured every 10 ms using a laser power of ~1 kW/cm², whereas for slow tracking, images were captured every 250 ms using a lower laser power of ~50 W/cm², with continuous laser exposure for both imaging regimes. A Coherent OBIS LS 552-nm, 150-mW laser was used for all single-molecule imaging using JF552 dye, while a Vortran Stradus 637-nm, 140-mW laser was used for imaging JF646, which offers higher cell permeability. The latter was used to assess Swr1-Halo expression levels under saturating conditions with 20 nM JF646 dye labeling.

Localize and track single molecules

Movies obtained by the microscope camera were preprocessed where initial frames with a higher density of molecules were removed. A substack of movie frames, typically containing fewer than two molecules per nucleus, was saved as 16-bit files for the analysis of single-molecule trajectories. A nuclear mask was created manually from the maximum-intensity z projection of the substack movies, and these binary masks were later used to eliminate any single-molecule tracks that were located outside the nuclei.

We used DiaTrack version 3.05 (57) for localization of single molecules in each frame and for the curation of single-molecule tracks. The “remove blur” value was set to 0.08 (slow tracking) or 0.1 (fast tracking), while the “remove dim” value ranged from 5000 to 7000. The maximum allowable distance between two consecutive localizations within the same track was 6 pixels for fast tracking and 3 pixels for slow tracking. For each imaging session, the DiaTrack parameters were adjusted to ensure robust SMT for one movie (or the first 200 frames of the movie). After that, all movies from that imaging session were processed in batch mode using DiaTrack.

Analysis of fast-tracking single-molecule data

We used the R language–based Sojourner (https://github.com/sheng-liu/sojourner) for the analysis of single-molecule track data. Tracks shorter than three frames (with a minimum of two displacements) were removed from the analysis pipeline to avoid false detections of single molecules and issues with meaningful analysis of very short tracks. Tracks located outside the nuclear region were also removed. Data from different movies within the same imaging session were merged into a single data file and analyzed using Spot-On kinetic modeling (32) of two model displacements for two states: bound and free molecules. The Spot-On scripts (Python) were run on our local server with the following settings: bin width, 0.01; number of time points, 6; jumps to consider, 4; use entire trajectories, No; max jump, 1 to 1.2 μm; dZ, 0.6 μm; D_bound, 0.001 to 0.1 μm² s⁻¹, D_free, 0.15 to 5 μm² s⁻¹; and three interactions of model fitting were performed on the cumulative distribution function (CDF) of displacements. In our experience, Spot-On kinetic modeling provides a robust estimate of the chromatin-bound population even from a limited number of trajectories (as low as ~2500 trajectories; see table S3).

For the trajectory-based Bayesian inference method saSPT (31), data from different imaging sessions were combined and the track lengths were trimmed to 12 frames (≤11 displacements per track). The following parameters were chosen for the saSPT analysis: 200 states of diffusion coefficient (μm² s⁻¹) log scale, −4 to 2; 100 states of localization errors, 0 to 0.3 μm; likelihood type, RBME; pixel size, 0.0107 μm; frame interval, 0.01 s; dZ, 0.58.

Analysis of slow-tracking single-molecule data

We selectively imaged a slower-moving molecule (using motion blur) with a camera integration time of 250 ms and analyzed trajectories to extract apparent dwell times. Similar to the fast-tracking data, we analyzed tracks that were composed of minimum two displacements and that were located in the nucleus. For the slow-tracking data, to account for blinking and localization error, tracks with a gap of one frame and that were located within less than 3 pixels were linked as a single trajectory. 1-CDF histograms were generated from the frequency plot of tracks with different temporal lengths, and 1-CDF curves were fitted to a double-exponential decay model

$$1-\text{CDF}=F_{\text{sb}}{e}^{-k_{\text{sb}}t}+F_{\text{tb}}{e}^{-k_{\text{tb}}t}$$ (1)

where the dissociation rates for stable-binding and transient-binding events are represented as k_sb and k_tb, respectively. The fractions of time the molecule spends at stable and transient binding sites are represented as F_sb and F_tb, respectively. The calculated k_sb and k_tb values are affected by artifacts such as photobleaching and chromatin movements. As described before (37), we used the decay kinetics of histone H2B, imaged under similar conditions, to correct these as follows

$${\mathrm{\tau }}_{\text{sb}}=\frac{1}{k_{\text{sb}}-k_{\text{sb},\mathrm{H}2\mathrm{B}}}$$ (2)

where ${\mathrm{\tau }}_{\text{sb}}$ is the corrected stable residence time.

To extract stable and transient events, we used a double-exponential fit as a simple and practical approach. However, we do not exclude the possibility that additional kinetic components may exist. Upon the availability of larger datasets, more sophisticated methods such as GRID (58) will be used to resolve these finer details.

Calculation of search kinetics

The search time is the average time it takes for a molecule to leave a specific binding site and find the next specific site. During this time, the molecule interacts with nonspecific sites multiple times (N_trials) with an average transient binding time τ_tb, and τ_free represents the average free time between nonspecific interactions. Assuming the random sampling of sites with equal probability of binding to all specific and nonspecific sites, the average search time between two consecutive specific binding events can be calculated as follows

$${\mathrm{\tau }}_{\text{search}}=N_{\text{trials}}\times {\mathrm{\tau }}_{\text{tb}}+(N_{\text{trials}}+1)\times {\mathrm{\tau }}_{\text{free}}$$ (3)

N_trials depends on the ratio of the number of specific sites (N_s) to nonspecific (N_ns) sites, and this ratio is represented as r_s

$$N_{\text{trials}}=\frac{N_{\mathrm{s}}+N_{\text{ns}}}{N_{\mathrm{s}}}=1+r_{\mathrm{s}}$$ (4)

As described previously (56), we determined $r_{\mathrm{s}}$ as

$$r_{\mathrm{s}}=\frac{1}{2}(\frac{1}{F_{\text{sb}}}-1)(\frac{{\mathrm{\tau }}_{\text{sb}}}{{\mathrm{\tau }}_{\text{tb}}}+1)$$ (5)

In fast tracking, F_bound is the percentage of the time a molecule spends bound to chromatin, either specifically or nonspecifically

$$F_{\text{bound}}=\frac{N_{\text{trials}}\times {\mathrm{\tau }}_{\text{tb}}+{\mathrm{\tau }}_{\text{sb}}}{[N_{\text{trials}}\times {\mathrm{\tau }}_{\text{tb}}+{\mathrm{\tau }}_{\text{sb}}]+[(N_{\text{trials}}+1)\times {\mathrm{\tau }}_{\text{free}}]}$$ (6)

Thus, ${\mathrm{\tau }}_{\text{free}}$ can be expressed in terms of $r_{\mathrm{s}}$ as

$${\mathrm{\tau }}_{\text{free}}=\frac{\frac{(1+r_{\mathrm{s}})\times {\mathrm{\tau }}_{\text{tb}}+{\mathrm{\tau }}_{\text{sb}}}{F_{\text{bound}}}-(1+r_{\mathrm{s}})\times {\mathrm{\tau }}_{\text{tb}}-{\mathrm{\tau }}_{\text{sb}}}{2+r_{\mathrm{s}}}$$ (7)

Using the values derived for $r_{\mathrm{s}}$ and ${\mathrm{\tau }}_{\text{free}}$ , we then calculated the search time as shown in the formula above.

Purification of the SWR1 complex

SWR1 complexes were purified from WT and mutant cells expressing Rvb1-MBP and Swr1-3XFlag using a sequential affinity purification method (41). Cells were grown in YPD media to an OD₆₀₀ of 1.5, harvested, and lysed using a freezer mill (SPEX) with liquid nitrogen. Protein extraction was performed in the presence of 0.2 M NaCl, and the lysate was cleared by high-speed centrifugation. The whole-cell extract was then incubated with amylose resin (New England Biolabs) for 1 hour at 4°C. After four washes, the bound proteins were eluted with maltose. For a second round of affinity purification, Anti-FLAG M2 agarose (Sigma-Aldrich) was added. The complex containing Swr1-Flag was eluted using the FLAG peptide, and the purified proteins from WT and mutant cells were resolved by SDS–polyacrylamide gel electrophoresis.

ChIP-exo 5.0 assay of H2A.Z and Swr1

Yeast strains containing C-terminal FLAG-tagged H2A.Z or TAP-tagged Swr1 were grown in YPD media to an OD₆₀₀ of 0.6 to 0.8 at 25°C. Cells were cross-linked with formaldehyde [final concentration of 1% (v/v)] at room temperature for 15 min, and the reaction was quenched with 125 mM glycine. The ChIP-exo assay and bioinformatics were performed as described previously (59). The identity of each mutant and epitope tag was confirmed within each ChIP-exo dataset by aligning to mutant sequences and epitope tags using the GenoPipe suite (60), available at https://github.com/CEGRcode/GenoPipe.

Analysis of ChIP-exo data

ChIP-exo data analysis was conducted using CLI ScriptManager version 0.13 (61), available at https://github.com/CEGRcode/scriptmanager. A normalization factor for each bam file was calculated by setting the total occupancy in the NFRs to be equal using ScriptManager–Sequence Read Analysis–Calculate Scaling Factor. To generate occupancy matrices, we used ScriptManager–Sequence Read Analysis–Tag Pileup to obtain the forward- and reverse-strand matrix cdt files from each bam file, which were then multiplied by the corresponding normalization factor. Using scripts in R, we combined the strands by first shifting the forward strand 50 bp to the right and the reverse strand 50 bp to the left and then summing the two stands. Mean occupancies of specified gene subsets were calculated and smoothed in composite occupancy plots. Occupancies at the +1 nucleosome of each gene were calculated as the sum of occupancies over 75 bp on both sides from the center of the nucleosome (total 150 bp). Gene subsets and coordinates for the midpoints of NFR, +1, and −1 nucleosomes were obtained from supplementary data 1 of Rossi et al. (45). Data from Weiner et al. (42) (GSE61888) were used to calculate histone acetylation occupancies using deepTools computeMatrix and plotProfile (62).

Supplementary Materials

This PDF file includes:

Figs. S1 to S7

Tables S1 to S4