Research Article

Structure and dynamics of the yeast SWR1-nucleosome complex

See allHide authors and affiliations

Science  12 Oct 2018:
Vol. 362, Issue 6411, eaat7716
DOI: 10.1126/science.aat7716

From DNA unwrapping to histone exchange

The yeast SWR1 complex, a member of the INO80 family of nucleosome remodelers, exchanges the H2A-H2B histone dimer for the Htz1 variant–containing dimer. Unlike all other remodelers, SWR1 does not translocate the nucleosome. Willhoft et al. applied structural and single-molecule analyses to show that the interaction between SWR1 and the nucleosome destabilizes the DNA wrapped around the histone core. This SWR1-catalyzed partial unwrapping of the DNA was regulated by adenosine triphosphate (ATP) binding but did not require ATP hydrolysis.

Science, this issue p. eaat7716

Structured Abstract

INTRODUCTION

Canonical nucleosomes contain two copies of each of four histone proteins: H2A, H2B, H3, and H4. However, variants of these histones can be inserted by adenosine triphosphate (ATP)–dependent chromatin-remodeling machines. The yeast SWR1 chromatin-remodeling complex, a member of the INO80 remodeler family, catalyzes the exchange of H2A-H2B dimers for dimers containing Htz1 (H2A.Z in human) in an ATP-dependent manner. However, the mechanism by which SWR1 exchanges histones is poorly understood. Despite having a DNA translocase subunit similar to that in the INO80 complex that slides nucleosomes, no net translocation of nucleosomes has been reported for SWR1. Consequently, the function of the ATPase activity, which is required for histone exchange in SWR1, has remained enigmatic.

RATIONALE

To obtain sufficient quantities for structural analysis, we generated the complete 14-subunit yeast SWR1 complex in insect cells. Binding of nucleosomes to SWR1 is stabilized in the presence of an ATP analog (ADP•BeF3), which we used to prepare a complex with a canonical yeast H2A-containing nucleosome. Structural analysis was undertaken by cryo–electron microscopy (cryo-EM). We also used single-molecule FRET (smFRET) techniques to probe the dynamics of nucleosomes bound to SWR1. Fluorescent probes were positioned on the H2A histones and the end of the DNA to monitor changes in nucleosome dynamics upon binding of SWR1 and ATP (or ATP analogs).

RESULTS

We determined the cryo-EM structure of the SWR1-nucleosome complex at 3.6-Å resolution. The architecture of the complex shows how the SWR1 complex is assembled around a heterohexameric core of the RuvBL1 and RuvBL2 subunits. The Swr1 motor subunit binds at superhelical location 2 (SHL2), a position it shares in common with other remodelers but not with its most closely related complex, INO80, which binds at SHL6-SHL7. Binding of ATP or ADP•BeF3 to the SWR1-nucleosome complex induces substantial unwrapping of the DNA wrap. Conformational changes in the motor domains of the Swr1 subunit drive a single–base pair translocation of the DNA wrap from the DNA entry site. The single–base pair DNA translocation accompanies conformational changes in the histone core that begin to destabilize the histone dimer interface. Using smFRET methods, we further probed these conformational changes to show how an increase in the dynamics of the SWR1-bound nucleosomes is dependent on binding of ATP but not hydrolysis.

CONCLUSION

The cryo-EM structure of the SWR1 complex bound to a nucleosome reveals details of the intricate interactions between components of the SWR1 complex and its nucleosome substrate. Interactions between the Swr1 motor domains and the DNA wrap at SHL2 distort the DNA, causing a bulge with concomitant translocation of the DNA by one base pair, coupled to conformational changes of the histone core that likely destabilize the dimer interface. Furthermore, partial unwrapping of the DNA from the histone core takes place upon binding of nucleosomes to the SWR1 complex. Single-molecule data monitor this unwrapping and show how the dynamics are altered by ATP binding prior to hydrolysis.

Structure of the SWR1-nucleosome complex.

(A) 3.6-Å SWR1-nucleosome map. (B) Binding of SWR1-ADP•BeF3 to the nucleosome induces multiple changes: (i) The DNA wrap is peeled away by ~2.5 turns, (ii) DNA is translocated by one base pair, and (iii) SHL2 is distorted as a consequence of motor domain closure. These distortions are a precursor to histone exchange and can be monitored by smFRET.

Abstract

The yeast SWR1 complex exchanges histone H2A in nucleosomes with Htz1 (H2A.Z in humans). The cryo–electron microscopy structure of the SWR1 complex bound to a nucleosome at 3.6-angstrom resolution reveals details of the intricate interactions between components of the SWR1 complex and its nucleosome substrate. Interactions between the Swr1 motor domains and the DNA wrap at superhelical location 2 distort the DNA, causing a bulge with concomitant translocation of the DNA by one base pair, coupled to conformational changes of the histone core. Furthermore, partial unwrapping of the DNA from the histone core takes place upon binding of nucleosomes to SWR1 complex. The unwrapping, as monitored by single-molecule data, is stabilized and has its dynamics altered by adenosine triphosphate binding but does not require hydrolysis.

The specific incorporation of histone variants into nucleosomes can trigger changes in chromatin structure (1) and the recruitment of cellular mediators (2). Histone variants can be inserted by histone chaperones during chromatin deposition (3), or at a later stage, by adenosine triphosphate (ATP)–dependent chromatin remodeling machines (4). Several variants exist for canonical H2A and H3 histones, such as the replication-independent yeast H3 variant H3.3, which is present at actively transcribed genes (5), or human H2A.X and H2A.Z, both of which have essential roles in DNA repair (6).

The yeast SWR1 chromatin-remodeling complex is a member of the INO80 remodeler family (7). SWR1 drives the exchange of H2A/H2B dimers for Htz1/H2B in an ATP-dependent manner (8). In humans, the equivalent histone exchange reaction is catalyzed by the SRCAP (9) and TIP60 (10) complexes, although the latter complex also catalyzes histone acetylation (10). However, the mechanism by which all of these complexes exchange histones has remained enigmatic. Despite having a superfamily 2 (SF2) helicase/translocase (Swr1 subunit) at its heart, and despite the dependence of histone exchange on the ATPase activity of this subunit (8), no net translocation of nucleosomes has been reported. By contrast, the related yeast INO80 remodeler [reported, albeit controversially, to catalyze the reverse exchange reaction by replacing Htz1/H2A.Z with H2A (1113)] appears to use a mechanism of limited translocation at the nucleosomal entry/exit region in order to exchange a histone dimer (14). Recent structures of the INO80-nucleosome complex place the ATPase motor at superhelical location SHL6-SHL7 (15, 16), where it could potentially expose the H2A(Z)-H2B dimer for histone exchange by translocating DNA toward the nucleosome dyad. By contrast, the SWR1 motor subunit (Swr1) binds at SHL2 (17), a position it shares in common with other remodelers (1820).

Structure of a complex of SWR1 with a bound nucleosome

Biochemical data have shown that SWR1 prefers nucleosome substrates with overhangs on both sides, with at least one of these needing to be long [>60 base pairs (bp)] (17, 21). We therefore used a canonical H2A-containing yeast nucleosome with overhangs on both sides (113N25), in the presence of the ATP analog ADP•BeF3, to prepare a complex suitable for structure determination by cryo-EM (Fig. 1 and figs. S1 to S3).

Fig. 1 Overview of the SWR1-nucleosome complex architecture.

(A) Linearized cartoon of Swr1 subunit showing domain boundaries. HD1 and HD2 are the two ATPase domains; HSA is the helicase/SANT-associated domain. (B) Overview of SWR1-nucleosome complex at 3.6 Å (left) and built coordinates (right). (C) View of Swr1 insert (ribbon) inside RuvBL1/RuvBL2 ring structure (surface). (D) Rotated view of SWR1, showing Swr1 insert following the inner contour of the RuvBL1/RuvBL2 ring, with the Swr1 insert protrusion emerging from the ring. (E) Close-up of Swr1 protrusion interacting with Arp6.

As shown previously (21, 22), and in common with the related INO80 complex (23), the SWR1 complex contains a single heterohexamer of the RuvBL proteins (Fig. 1B) that serves as a scaffold around which many of the other subunits are assembled (Movie 1). Each ATPase site is occupied by an adenosine diphosphate (ADP) molecule. If there are any additional functions of the RuvBL hexamer, these do not require ATP hydrolysis because catalytically dead ATPase subunits show wild-type histone exchange activity (fig. S4).

Movie 1.

Rotating overview of SWR1-nucleosome complex coordinates fitted into 3.6-Å cryo-EM envelope. Subunits are colored as in Fig. 1B.

The motor subunit of INO80-family proteins is characterized by a large insertion in the second ATPase domain (7) (Fig. 1A). As with the INO80 motor subunit (23), the insert in Swr1 forms a large extended structure that makes intimate contacts with the RuvBL hexamer (Fig. 1, B to E). However, the structures of this insert are quite different, although largely planar in both cases. One notable difference between the insert structures is a two-helix extension that protrudes between two RuvBL subunits in SWR1 and makes contact with the Arp6 subunit, in a region that is an insert in the actin fold (Fig. 1E). A function for this interaction, other than architectural, is unknown.

As with the INO80-nucleosome complex (15, 16), the N-terminal region of the SWR1 complex [termed Subcomplex 1 (SC1), comprising the N-terminal region of Swr1 and the actin, Arp4, Swc4, Swc7, Yaf9, and Bdf1 subunits] is present but disordered in our structure.

Interactions between the Swr1 ATPase domains and the nucleosome

The nucleosome is bound in a cleft between two lobes that extend from the RuvBL core of the complex. Multiple contacts are made with the histone core and DNA wrap. The Swr1 subunit ATPase domains are located at the canonical SHL2 position seen in the structures of most other remodelers (19, 20) and consistent with SWR1-nucleosome footprinting studies (17). There is clear density for a bound nucleotide (ADP•BeF3) at the ATP-binding site (fig. S3) and, as a consequence, the domains are in the “closed” conformation observed in the ATP-bound state of other SF2 translocases (24) such as Chd1 (19) (fig. S5), placing the motor domains on the “tracking” strand consistent with 3′-5′ translocation toward the nucleosome dyad axis (Fig. 2, A and B). Although located at SHL2, the binding site is displaced by 1 bp toward the dyad relative to Chd1 and Snf2, resulting in a ~35° rotation of the complex around the bound DNA duplex toward the other gyre of the DNA wrap (Movie 2). The DNA of the bound nucleosome at the SHL2 site contacts the N-terminal ATPase domain (HD1), as observed in other remodelers (19, 20), but as a result of the 1-bp shift, it now makes much more extensive contacts with the other gyre, suggesting a much tighter interaction on the DNA (Fig. 2C). Furthermore, across the second ATPase domain (HD2), the bound DNA shows substantial distortion (Fig. 2, D and E, and fig. S5). Unlike DNA binding across the motor domains of Chd1, the DNA duplex makes a sharp kink, the result of which is lifting of the DNA from the histone surface to make a bulge in the wrap (fig. S5). The kink is a consequence of two α helices being rammed against the duplex (Fig. 2E). In order to create this bulge, the DNA has also been translocated by 1 bp (Movie 3). This suggests that the power stroke is ATP-binding, causing a translocation of 1 bp together with a distortion of the bound DNA. A step size of one base per ATP has been deduced for other SF1 and SF2 translocases (15, 2527).

Fig. 2 Details of the Swr1 motor and nucleosomal DNA distortions at SHL2.

(A) Top-down view onto the nucleosome showing the position of the Swr1 motor domain and the Arp6/Swc6 module. Inset: Swr1 motor domain bound to the tracking strand (gold). (B) Nucleosomal position of Swr1 motor domain at SHL2, highlighting approximate delineations of HD1 and HD2 motor domain lobes. (C) Details of the Swr1 HD1 interactions with nucleosomal DNA. HD1 wedges between the top and bottom gyres. (D) Details of the Swr1 HD2-DNA interactions. The DNA is distorted in the SWR1-nucleosome complex as it lifts onto the HD2 motor domain. (E) Comparison of Swr1 and Chd1 nucleosomal DNA trajectories, showing Swr1 motor domain helices α9 and α10 (yellow) pushing on DNA. Structures are aligned on the motor domains of Swr1 and Chd1. The SWR1 footprint (17) coincides well with the Swr1 binding site.

Movie 2.

Rotation of Chd1 motor domain relative to Swr1 motor. Chd1 (pink, PDB ID 5O9G) is shown moving from a superposition with the Swr1 motor domain to its position in the deposited Chd1-nucleosome structure. This shows that Swr1 is rotated by 35° about the DNA axis.

Movie 3.

Morph between canonical nucleosome (PDB ID 1ID3) and Swr1-bound nucleosome, which highlights the translocation event induced by SWR1 binding as well as the distortion of the histone octamer from the canonical positions.

The location of the Swr1 ATPase domains at the SHL2 position (Fig. 2, A and B) is quite different from that of the INO80 complex, which binds at SHL6-SHL7 (1416). However, SWR1 differs from INO80 in a number of important ways. Notably, a single complex of SWR1 is required for activity (21), whereas INO80 functions as a dimer (28). An INO80 dimer is able to slide nucleosomes along DNA, whereas a SWR1 monomer instead swaps histone dimers (21) and is unable to slide nucleosomes (17). Although yeast INO80 has been reported to also catalyze histone exchange (11, 14), this activity is controversial (12, 13) and it is not known whether exchange activity requires INO80 dimers. Furthermore, although SWR1 is unable to slide nucleosomes, the ATPase activity of the “motor” domains is nonetheless essential for activity (8). Limited sliding of DNA within the confines of the DNA wrap has recently been demonstrated for SWR1 (29). The tighter interactions between the N-terminal motor domain (HD1) and the other DNA gyre may act as a block to prevent net translocation by SWR1 but appears to allow limited local translocation of the DNA wrap leading up to that block.

Other contacts with the nucleosome

Arp6 and Swc6 form a tight heterodimer with Swc6 entwined around the actin fold of Arp6 that forms extensive contacts with two adjacent OB-folds of the RuvBL hexamer, with density for an ADP•BeF3 moiety at its center (fig. S3). This module provides a further tether to the nucleosome via contacts made with nucleosomal H2A and linker DNA on the opposite side to the Swr1 motor (Fig. 3, A to D). At this site, the DNA is unwrapped by ~65° from the canonical nucleosome path, exposing 2.5 turns of the wrap (Fig. 3B). This open conformation is stabilized via specific contacts with Swc6 (Fig. 3C) and may be further stabilized by Swc3 (fig. S6). In addition, a two-pronged interface of Swc6 with H2A, involving a hydrophobic core centered around the αC region (Fig. 3D), could contribute to the specificity of SWR1 for H2A-containing nucleosomes (30). Although the resolution of the interface between Swc6 and the nucleosome is insufficient to identify the Swc6 sequence, the region of contact with H2A involves residues 86 to 94 and from residue 109 to the C terminus—regions with differences between H2A and Htz1 that may contribute to the specificity. Interactions with Asp91 and Glu93 of the H2A acidic patch anchor this module firmly on the face of the octamer. The acidic patch has previously been shown to be important for SWR1 histone exchange (17). Similarly, we observe a loss in histone exchange activity when Arp6/Swc6 is deleted (Fig. 3E and fig. S6). Contacts made with Arp6/Swc6 are mirrored by Arp5/Ies6 in the INO80 complex (15, 16), which have been proposed to facilitate a ratchet-like mechanism for nucleosome sliding by INO80.

Fig. 3 Contributions of nonmotor subunits to structural changes in the nucleosome.

(A) Overview of SWR1-nucleosome structure showing the location of the Arp6-Swc6 module relative to Swr1 motor domains. (B) Close-up of SHL6 region bound by Arp6-Swc6. The canonical nucleosome (PDB ID 1AOI; gray) is superimposed onto a Swr1-bound nucleosome. The change in angle corresponds to approximately a ~65° rotation away from the nucleosomal wrap, exposing ~2.5 turns of the canonical wrap. (C) Swc6 interactions pinning DNA in the SHL6-SHL7 region. (D) Swc6 interactions with nucleosomal H2A showing specific interactions with the C-terminal tail and acidic patch (red). (E) The effect of subunit deletions on histone exchange activity of SWR1. Amino acid abbreviations: A, Ala; D, Asp; E, Glu; H, His; I, Ile; K, Lys; L, Leu; N, Asn; P, Pro; R, Arg; S, Ser; Y, Tyr. WT, wild type.

A region of extended density, which we assign as Swc2, forms a large interface crossing the main body of the complex, interacting with two consecutive subunits of the RuvBL hexamer before traversing the motor domains, crossing the DNA wrap and then making contacts with the distal face of the nucleosome on the acidic patch (fig. S6). A strikingly similar region of density is observed in INO80 corresponding to the Ies2 subunit (16). Although these subunits do not share recognizable sequence similarity, both their structures and locations share much in common. Ies2 plays an important role in regulating ATPase and sliding activities (3133). However, although there are similarities in the interaction (such as interaction with the acidic patch), the actual binding site is quite different, so it is unclear whether the role of this interaction is similar or different in the two complexes.

SWR1 causes unwrapping of the DNA at the nucleosome entry and exit sites

By creating nucleosomes with two overhangs (113N25), we satisfied the requirement of the SWR1 complex for at least a 60-bp overhang on at least one side for optimal activity, although a substrate with overhangs on both sides appears to bind even better (34). The structure shows interactions with both DNA overhangs (Fig. 3 and fig. S7). The DNA wrap at one end, corresponding to the SHL6-SHL7 region, is peeled back from the histone surface where it interacts with Arp6/Swc6 (Fig. 3). DNA unwrapping is seen in many other remodeling systems (INO80, Chd1), so it may be a common way to reduce friction as the DNA wrap slides around histones (15, 16, 19). At the other DNA overhang, we also see interactions with the SWR1 complex (figs. S6 and S7). However, this interaction is rather variable and several different conformational classes could be distinguished in the data set, suggesting a high degree of mobility in this region (figs. S1 and S7).

Single-molecule assay to monitor SWR1-nucleosome unwrapping

To characterize SWR1-induced nucleosome unwrapping observed in the cryo-EM structure, we developed a single-molecule Förster resonance energy transfer (smFRET) assay to monitor nucleosome conformation upon SWR1 binding (Fig. 4A) (35, 36). Surface-immobilized fluorescently labeled nucleosomes were imaged by FRET in the absence and presence of SWR1. Nucleosomes alone exhibit three main peaks (Fig. 4B), corresponding to bilabeled (~0.7 FRET), proximal-only (~0.9 FRET), and distal-only (~0.5 FRET) labeling, based on their photobleaching pattern (fig. S8). Counting the fraction of donor-acceptor labeled nucleosomes as a function of time (Fig. 4C) shows that SWR1 remains as active on surface-immobilized nucleosomes as in bulk (21). In the presence of the slowly hydrolyzable ATP analog ATPγS, exchange was not observed (Fig. 4C). This single-molecule histone exchange assay enabled us to quantify whether either of the H2A/H2B dimers (distal or proximal) was exchanged preferentially (fig. S8). The fraction of proximal-only fluorescent nucleosomes (Dp) remained constant over time (Fig. 4D).

Fig. 4 Single-molecule nucleosome dynamics induced by SWR1.

(A) Schematic diagram of the experimental setup: Bilabeled nucleosomes (AF555 on each H2A, blue), bound to biotinylated (orange) and labeled (AF647, red) 257-bp DNA, are surface-immobilized on neutravidin-coated (yellow) biotin-PEG slides. Nucleosomes comprise canonical H2A/H2B (beige discs) and H3/H4 dimers (gray discs). SWR1-nucleosome interactions are monitored via FRET between the donors and the acceptor. (B) FRET histogram of nucleosomes (in the absence of SWR1) shows three major populations: bilabeled (~0.7 FRET), proximal-only (~0.9 FRET), and distal-only (~0.5 FRET) fluorescent nucleosomes (N = 103). (C) Single-molecule histone exchange assay shows that SWR1 is active on immobilized nucleosomes at levels comparable to bulk experiments (21) in the presence of ATP (red, observed rate constant kobs = 0.05 ± 0.01 min−1) but not ATPγS (blue). Histone exchange removes fluorescently labeled H2A; therefore, activity is determined as the fraction of remaining fluorescent nucleosomes as a function of time (N = 290 at t = 0 min in ATP or ATPγS). (D) The fraction of proximal-only (Dp) fluorescent nucleosomes remains constant with time, indicating that proximal and distal H2A/H2B dimers are exchanged without preference (N = 74, 79, 36, and 40). (E) Percentage of dynamic traces in the presence of nucleosome alone (NCP, N = 55); nucleosome and SWR1 complex (SWR1, N = 67); nucleosome, SWR1, and ATP (N = 85); ATPγS (N = 121); or ADP•BeF3 (N = 102). Error bars in (C) to (E) denote SE.

SWR1 unwrapping is dynamic and stepwise

The single-molecule FRET trajectories reveal a large increase in conformational dynamics in the presence of SWR1 and ATP. In the absence of SWR1, almost all observed trajectories appear static within our time resolution (250 ms; Fig. 4E and fig. S8), in agreement with previous studies (37, 38). In the presence of SWR1 and ATP, the fraction of dynamic traces increases to almost half (Fig. 4E). Dynamic FRET trajectories (Fig. 5A and fig. S9) show rapid excursions to lower FRET states, consistent with SWR1-induced unwrapping of nucleosomal DNA as observed in the structure (Fig. 3 and fig. S7). In the absence of ATP, the fraction of static traces is comparable to that of nucleosomes alone (Fig. 4E), indicating that ATP binding or hydrolysis is required for unwrapping.

Fig. 5 Analysis of unwrapped states of nucleosomal DNA.

(A) FRET time trajectory (gray) of a proximal-only (Dp) fluorescent nucleosome in the presence of SWR1 and ATP, with resulting hidden Markov model fit (HMM, black). (B) Zoom of gray box in (A) showing direct and reversible transient excursions into multiple mid-FRET unwrapped states ({Ui}). (C) Calculated transition density plot (TDP) of unwrapping of nucleosomal DNA in the presence of ATP (2000 transitions from 60 trajectories). TDP confirms that multiple partially unwrapped states are accessed from the high-FRET wrapped state (W). (D) FRET trajectory (gray) of a proximal-only fluorescent nucleosome in the presence of SWR1 complex and ATPγS with resulting HMM fit (black). Fewer and slower transitions are observed than in (A). (E) Zoom of the gray box in (D) showing direct and reversible transient excursions into multiple mid-FRET unwrapped states ({Ui}). (F) Calculated TDP of unwrapping nucleosomal DNA in presence of ATPγS (1404 transitions from 65 trajectories). (G) Schematic of various nucleosomal unwrapping states.

A zoom into these excursions (Fig. 5B) reveals the presence of various unwrapped states with FRET ratios ranging from 0.5 to 0.8. To analyze the observed transitions, and to identify the various observed states (Fig. 5A and fig. S9), we used hidden Markov modeling (38). To simplify the analysis, we focused exclusively on bona fide proximal-only fluorescent nucleosomes, although distal-only fluorescent nucleosome trajectories exhibited similar behavior (fig. S10). A transition density plot (TDP; Fig. 5C) summarizes the observed FRET transitions in all trajectories analyzed. SWR1 induces multiple, partially unwrapped states with the most frequent transition being between the fully wrapped state (W, ~0.9 FRET) and the second unwrapped state (U2, ~0.7 FRET). Additional transitions to ~0.8, ~0.6, and ~0.5 FRET indicate that unwrapping is a stepwise process involving multiple intermediates. However, most observed transitions occur directly without stopping at each intermediate state, likely indicating that stepwise transitions between intermediate states are faster than our 250-ms time resolution. The symmetric shape of the TDP about the diagonal shows that these transitions are reversible. Lowering the ATP concentration to 100 μM results in similar but less frequent excursions (fig. S11), indicating that SWR1 can unwrap the DNA multiple times per encounter.

SWR1 unwrapping requires ATP binding but not hydrolysis

In the presence of a slowly hydrolyzable ATP analog (ATPγS) that does not support histone exchange over the time scale of these experiments (Fig. 4C), the fraction of dynamic traces remains high (Fig. 4E). The corresponding smFRET trajectories (Fig. 5D and fig. S12) exhibit similar rapid excursions into lower FRET states, indicating that nucleosome unwrapping requires ATP binding but not hydrolysis. These excursions (Fig. 5E) involve similar transitions into various unwrapped states; however, steps into multiple intermediate states are now apparent even at 250-ms time resolution, likely due to slower transitions in the presence of ATPγS. Likewise, the resulting TDP (Fig. 5F) reveals additional transitions to intermediate states (Fig. 5, C and G). Taken together, these data show that ATP binding (but not hydrolysis) is required for SWR1 unwrapping on the pathway to H2A/H2B histone exchange. Similar results are observed with ADP•BeF3 (Fig. 4E and fig. S13).

To correlate the observed FRET values to the number of unwrapped base pairs, we determined an empirical distance calibration linking the distance between the end of the DNA and the labeled H2A histones (fig. S14). On the basis of this calibration, which we regard as qualitative rather than quantitative, we estimate that SWR1 unwraps 10 to 12 bp from the nucleosome, which is consistent with the degree of unwrapping observed in the structure (Fig. 3 and fig. S7).

Exchange of histone dimers is distributive with no preference for linker-proximal or -distal dimers

Although dimer exchange by SWR1 is sequential (30), it is not known whether this process is distributive or processive, nor is it known whether there is any preference for which dimer is exchanged first. We looked at histone dimer exchange under catalytic conditions (1:10 ratio of SWR1 to nucleosomes, i.e., 20 exchange reactions) at various ratios of dimer to nucleosome (fig. S15). At a 1:1 ratio, we observed around 31% doubly exchanged nucleosomes, although the majority (69%) have a single dimer exchanged, indicating a distributive rather than processive mechanism. At lower dimer/nucleosome ratios, the proportion of single exchanges is even higher (fig. S15). Furthermore, the proportion of dye-proximal and -distal dimer exchanges was approximately equal, suggesting little preference for which dimer is exchanged first. Together, these data imply that SWR1 exchanges dimers one at a time, most likely binding in two orientations (probably at each SHL2 site) to achieve this process. The exchange reaction, therefore, is better considered in terms of which dimer is proximal or distal to the SWR1 motor domain binding site rather than whether it is linker-distal or -proximal, because in vivo the nucleosomes will be flanked by linker DNA on both sides. As discussed above, the structure strongly suggests that the dimer that is exchanged is the one proximal to the SHL2 site at which SWR1 is bound, independent of which side any linker DNA might be attached.

Implications for histone exchange

The mechanism for histone exchange by SWR1 is still poorly understood. Very early studies showed that not only was the Swr1 subunit a member of the superfamily 2 translocases, like other chromatin remodelers that slide nucleosomes, but also that the ATPase activity of this subunit was essential for activity (8); these findings suggested that DNA translocation at some level might be a component of the histone exchange mechanism. However, no group has been able to demonstrate net nucleosome sliding by SWR1, although limited sliding within the nucleosome has recently been reported (29).

Experiments with nucleosomes containing two-base gaps on the tracking strand within the SHL2 sites on either side of the nucleosome dyad have interesting effects on histone exchange (17). A two-base gap on one strand, at positions located between base pairs ~17 to 22 from the nucleosome dyad, permits exchange of one dimer but prevents exchange of the other. Furthermore, a nucleosome in which gaps are placed at both SHL2 regions (in both cases on the tracking strand) prevents exchange of both histone dimers, even though binding to this nucleosome is unaffected. The region of SHL2 that is sensitive to the presence of a gap (bases 17 to 22 from the dyad) corresponds very nicely with the contact regions of the motor domains with the DNA wrap (Fig. 2E). However, placing a gap on either side of this contact region, but on the same DNA strand, has only a modest effect on the rate of exchange and both dimers can be exchanged. This latter observation suggests that, at best, there can be only limited translocation of DNA by the motor domains; similar experiments with bona fide nucleosome sliders show that a single-base gap ahead of the motor domain binding site will block translocation, but only when placed in the tracking strand (18). By contrast, for SWR1, a gap in the tracking strand close to the SHL2 site is sufficient to prevent histone exchange, which suggests a requirement for some form of ATP-driven conformational changes of the DNA at the binding site. Although footprinting shows that there is no net sliding of nucleosome position as a consequence of exchange even at base pair resolution (17), limited transient translocation (i.e., a few base pairs) of part of the DNA wrap has been proposed (29).

In our structure, we do indeed observe limited translocation of the DNA wrap from the entry site to the Swr1 binding site by a single base pair to create a bulge. We also observe some distortion of the histone core (Fig. 6, fig. S16, and Movie 3). The histone core “flexes” and the upper tier, comprising a histone dimer and part of the H3/H4 tetramer, twists relative to the lower tier. These conformational changes result in changes at the interface between the H2A/H2B dimer and H3/H4 (Fig. 6 and fig. S16). The changes in the nucleosome core likely also explain the increased dynamics of the DNA tails, as observed in our single-molecule studies (Figs. 4 and 5).

Fig. 6 Distortions of core histones as a consequence of DNA translocation.

(A) Summary of distortions in the histone core. Histones are shown as cylinders, with coloring to distinguish between canonical and SWR1-bound nucleosome positions. Upper- and lower-tier histones were defined on the basis of the views shown here. Canonical yeast nucleosome (PDB ID 1ID3) and SWR1-bound nucleosome were superimposed on residues 67 to 74 of H4, 108 to 126 of H2B, and 56 to 98 of H3. The general direction of histone movement is toward the motor-bound side of the nucleosome. (B) Edge-on view of the nucleosome, highlighting upper and lower histone tiers as well as upper and lower DNA gyres. Coloring is as in (A). Inset: Close-up of the motor-bound nucleosome region, showing a widening of the gap between the two gyres and distortion of the lower gyre in the presence of SWR1. (C) Change in the upper-tier H2A-H3 interface. As a consequence of SWR1-nucleosome interactions, the upper H2A-H3 interface decreases by approximately 180 Å2, as determined in a comparison with a canonical nucleosome (PDB ID 1ID3) (470 Å2 versus 650 Å2). By contrast, the lower-tier interface remains largely the same (660 Å2 versus 650 Å2). These changes are further detailed in fig. S16.

Consequently, our structural and single-molecule data suggest that the role of the Swr1 ATPase activity is to induce local distortion in the DNA at SHL2 that results from limited translocation against the HD1 domain of Swr1, and that this begins to destabilize the DNA wrap and histone dimer contacts within the nucleosome. Such a distortion of the dimer interface with H4 as a consequence of limited, and transient, DNA translocation has been suggested previously on the basis of biochemical data (17, 29), and our structure now provides mechanistic insights into that process. Our structure likely represents an initial stage in the reaction cycle, and the distortions we observe in both the DNA and the histone core could be amplified by additional translocation steps. Although ADP•BeF3 induces conformational changes similar to those likely accompanied by ATP binding, we note that nonhydrolyzable ATP analogs do not support histone exchange; thus, at least one, possibly several, rounds of ATP binding and hydrolysis may be required. As a consequence of histone exchange, there need not be any net translocation of the DNA wrap beyond the Swr1 motor binding site at SHL2, if any translocation of the DNA up to that point were transient until histone dimer release and/or exchange.

Materials and methods

Briefly, nucleosomes and SWR1 complex were prepared as described (21). The SWR1-nucleosome complex was assembled by incubating SWR1-ADP•BeF3:Htz1/H2B dimer complex with nucleosomes at a 1:1 molar ratio. Cryo-EM data acquisition, image acquisition, and structure reconstruction followed a similar procedure as described (15). Data processing and refinement statistics for the two cryo-EM structures are summarized in figs. S1 to S3 and table S1. See the supplementary materials for further details, including details of single-molecule experiments.

Supplementary Materials

www.sciencemag.org/content/362/6411/eaat7716/suppl/DC1

Materials and Methods

Figs. S1 to S16

Table S1

References (3948)

References and Notes

Acknowledgments: We thank Diamond for access and support of the Cryo-EM facilities at the UK national electron bio-imaging center (eBIC), funded by the Wellcome Trust, MRC, and BBSRC. Funding: Supported by the Wellcome Trust [095519/Z/11/Z and 209327/Z/17/Z (D.B.W.), Cancer Research UK (C6913/A21608 (D.B.W.)], the Medical Research Council [MR/N009258/1 and MR/R009023/1 (D.B.W.)], a core grant from the MRC London Institute of Medical Sciences (D.S.R.), and start-up funds from Imperial College London (D.S.R.). Author contributions: O.W., M.G., D.S.R., and D.B.W. designed the studies; O.W., E.Y.D.C., M.W., and R.A. performed the cryo-EM analysis; C.-L.L. and Y.C. conducted initial preliminary studies; O.W., E.A.M., and L.O. prepared the samples; M.G. conducted and analyzed the single-molecule experiments; C.-L.L. and O.W. conducted the biochemical experiments; and D.B.W. and D.S.R. analyzed the data and wrote the manuscript with input from all the authors. Competing interests: Authors declare no competing interests. Data and materials availability: Density maps are deposited at the Electron Microscopy Database (accession code EMD-4395) and protein coordinates are deposited at the Protein Data Bank (PDB ID codes 6GEJ and 6GEN).
View Abstract

Subjects

Navigate This Article