Research Article

Structure of the RSC complex bound to the nucleosome

See allHide authors and affiliations

Science  15 Nov 2019:
Vol. 366, Issue 6467, pp. 838-843
DOI: 10.1126/science.aay0033

The architecture of the RSC complex

RSC is a Snf2-family chromatin remodeler complex that controls the promoter architecture of most of the genes in yeast. Using single-particle cryo–electron microscopy, Ye et al. determined the structure of RSC bound to the nucleosome. The structure reveals the modular architecture of RSC, shows how RSC engages the nucleosome, and explains the remodeling directionality. RSC shows strong similarities to homologous human complexes that are frequently mutated in cancers, and this structure provides valuable information for understanding these systems.

Science, this issue p. 838


The RSC complex remodels chromatin structure and regulates gene transcription. We used cryo–electron microscopy to determine the structure of yeast RSC bound to the nucleosome. RSC is delineated into the adenosine triphosphatase motor, the actin-related protein module, and the substrate recruitment module (SRM). RSC binds the nucleosome mainly through the motor, with the auxiliary subunit Sfh1 engaging the H2A-H2B acidic patch to enable nucleosome ejection. SRM is organized into three substrate-binding lobes poised to bind their respective nucleosomal epitopes. The relative orientations of the SRM and the motor on the nucleosome explain the directionality of DNA translocation and promoter nucleosome repositioning by RSC. Our findings shed light on RSC assembly and functionality, and they provide a framework to understand the mammalian homologs BAF/PBAF and the Sfh1 ortholog INI1/BAF47, which are frequently mutated in cancers.

In chromatin, DNA wraps around histone octamers to form nucleosomes. Access to the nucleosomal DNA is facilitated by Snf2-family chromatin remodelers, which are highly conserved from yeast to mammals (1). Snf2 and Sth1, two closely related paralogs, are the catalytic motor subunits of the two similar yeast SWI/SNF and RSC remodeling complexes, respectively. Although the motors remodel nucleosomes in vitro on their own, they function in complexes with multiple auxiliary subunits in vivo to provide regulation and targeting. RSC is enriched at nucleosomes that flank nucleosome-free regions (NFRs) and controls the promoter architecture of most genes in yeast (2). It makes use of DNA translocation to slide or eject the promoter +1 nucleosomes to extend NFRs, possibly to expose transcription start sites and promote transcription.

Low-resolution structures of SWI/SNF, RSC, and the human homolog PBAF complexes have been reported (3, 4). However, it remains unclear how the subunits assemble the complexes and engage nucleosomes. Here, we report the cryo–electron microscopy (cryo-EM) structure of RSC bound to the nucleosome, providing insights into RSC organization and function.

Overall structure of RSC bound to the nucleosome

Purified RSC was mixed with nucleosomes and the stable adenosine triphosphate (ATP) analog ADP-BeFx and subjected to cryo-EM analysis (fig. S1 and table S1). The RSC-nucleosome complexes were highly flexible, with two prominent classes identified. One class (class A) showed a slightly unwrapped linker DNA, with an overall resolution of 7.1 Å (fig. S2, A and B), whereas the other class (class B) showed EM density for a longer linker DNA, with an overall resolution of 7.6 Å (fig. S2, C to E).

We found that RSC in either class is organized into three modules (Fig. 1A and movie S1): the motor module (motor domain of Sth1), the actin-related protein (ARP) module (the HSA helix of Sth1 bound by an Arp7-Arp9-Rtt102 trimer), and the substrate recruitment module (SRM, including the N-terminal fragment of Sth1 and multiple additional subunits). The three modules, connected through two disordered loops of Sth1, show rigid-body motions relative to each other (fig. S3), explaining the flexibility. Because the biological implications of classes A and B are the same, we mainly focused on the 7.1 Å class A structure.

Fig. 1 Overall structure of the RSC-nucleosome complex.

(A) Three different views of the structure of the complex. Arrows indicate the directionality of DNA translocation. (B) Domain architectures of the RSC subunits. The domains that are not structurally resolved in the current study are colored white.

Focused refinements led to a 3.8 Å map of the motor domain (fig. S2F) and a 3.4 Å map of the SRM, with the rigid core at ~3.0 Å (figs. S1G and S2, G to K). The high-quality map of the SRM allowed us to build atomic models of the N-terminal fragment of Sth1, Rsc2, Rsc3, Rsc4, Rsc6, Rsc7, Rsc9, Rsc30, Rsc58, Htl1, Sfh1, and two molecules of Rsc8 (termed Rsc8a and Rsc8b) (Fig. 1B). The interactions among the subunits were further validated by cross-linking mass spectrometry (CL-MS) analyses (fig. S4 and data S1).

The nucleosome is mainly bound by the motor domain of Sth1, with limited contacts from other subunits, except Sfh1 (see below). An open space embraced by the SRM and some weak density was found (fig. S2, B and D), which may relate to the hypothetical nucleosome-binding central cavity proposed previously (5). However, no nucleosome was found inside this open space.

The SRM comprises the majority of RSC (Fig. 1A), with the conserved subunits Rsc6, Rsc8, Rsc9, Rsc58, and the N-terminal fragment of Sth1 interweaving into a rigid core. The SRM is further organized into three lobes. Rsc3 and Rsc30, which contain DNA-binding Zn-cluster domains (ZnDs), decorate the conserved core at DNA-binding lobe (DB lobe) (6, 7). Rsc2 and Rsc4, which contain histone tail–binding bromodomains (BDs) and BAH domain (810), are located at the histone tail–binding lobe (HB lobe). Sfh1, which contains a nucleosome-binding C-terminal tail (CTT; see below), is located at the nucleosome-binding lobe (NB lobe). As a result of tethering to the SRM through flexible loops, the structures of the substrate-binding motifs (e.g., ZnD, BD, BAH, and CTT) could not be resolved under current conditions. Nonetheless, the SRM clearly positions these elements to colocalize along one side of the nucleosome, with the DB, HB, and NB lobes proximal to their binding epitopes, suggesting mechanisms for coordinated nucleosome recruitment or retention.

Structure of Sth1 in the context of the RSC complex

Sth1 is composed of five major domains (Figs. 1B and 2A): the N-terminal domain (NTD), the domain preceding the HSA helix (preHSA), HSA, motor, and the C-terminal extension and BD. A polyalanine model was built for the poorly conserved NTD because of the lack of defined features (fig. S5A). The modestly conserved preHSA (~25% identity between Sth1 and HsBRG1) threads through the DB and HB lobes. The HSA helix connects to the preHSA and motor domains through flexible loops. The motor domain interacts with the nucleosome in a manner very similar to the isolated adenosine triphosphatase (ATPase) fragment of Snf2 (fig. S6) (11, 12).

Fig. 2 Structure of Sth1.

(A) Locations of the different domains of Sth1 within the RSC-nucleosome complex. The boxed region is enlarged in fig. S5B. (B) SRM assembly through preHSA. Residues related to the cancer-associated mutations in the human homologs are shown as red spheres. The boxed regions are enlarged in fig. S5, C to E.

One surface of HSA binds to the ARP proteins, as observed in the structure of HSASnf2 (13), and the exposed surface, containing multiple conserved positively charged residues (fig. S5A), is proximal to the nucleosomal DNA (Fig. 2A and fig. S5B). The structure suggests that HSA may function to orient RSC by projecting the ARP proteins away from the nucleosome and by orienting HSA, the preHSA, and the associated SRM toward the exit DNA (Fig. 2A). As a result, we found one preferred nucleosome orientation with the linker DNA at the exit side. This contrasted with the isolated motor domain of Snf2, which binds to either of the two nucleosome orientations in the presence of ADP-BeFx (11).

The ARP proteins adopt a conformation largely similar to that found in the crystal structure of the HSASnf2-Arp7-Arp9-Rtt102 complex; they make limited direct interaction with the nucleosome or the Sth1 motor domain. The structure suggests that the ARPs may stabilize the helical conformations of HSA and/or affect the conformation of the adjacent post-HSA, which interacts with the motor (14, 15), and hence modulates the structural integrity of the complex and the remodeling efficiency (16).

The preHSA interacts with several conserved auxiliary subunits (Fig. 2B). Specifically, Arg292 of Sth1 packs against Rsc9 at the DB lobe (figs. S2I and S5C). Likewise, Arg258, Arg198, and Arg202 interact with Rsc8 and Rsc58 at the HB lobe (fig. S5, D and E). These residues are highly conserved (fig. S5A), with mutations of the equivalent residues recurrently found in human cancers (17).

Structure of the NB lobe and its implications for nucleosome binding

Sfh1 is the major component of the NB lobe (Fig. 3A). It is highly conserved with Snf5 in budding yeast and INI1/BAF47 in animals (fig. S7A). The conserved domain contains two imperfect repeats, RPT1 and RPT2 (repeat 1 and 2), which interact with the SWIRM domains of Rsc8a and Rsc8b, respectively (Fig. 3B and fig. S2H). This is in line with the crystal structure of the RPT1 domain of INI1 bound to the SWIRM domain of BAF155 (homolog of Rsc8) (18). The INI1-BAF155 interaction is highly similar to that of Sfh1-Rsc8 (fig. S7B), supporting a conserved assembly mechanism.

Fig. 3 Structure of the NB lobe.

(A) Sfh1 (cyan) is proximal to the nucleosome. (B) Structure of the NB lobe. The boxed region is enlarged in (H). The positions of RPT1 and RPT2 of Sfh1 are indicated by ovals. (C) The CTT of Sfh1 is close to the H2A (purple)–H2B (orange) surface. (D) Multiple sequence alignments of the CTTs of Sfh1-like proteins. The residues mutated in this study are highlighted in green. (E) Complementation assays of Sfh1 derivatives on synthetic dropout medium with 5-fluoroorotic acid (5FOA) (left) and at high-salt conditions (right). p519, control. (F) Relative ATPase activities of the wild-type and Sfh1-ΔCTT RSCs. (G) Comparative nucleosome ejection on plasmid arrays by wild-type and Sfh1-ΔCTT RSCs. Left: Schematic of the ejection assay, with supercoiled plasmid (topoisomer) distribution revealed by a 2D gel. Lk, linking number; N, nicked; L, linear. A representative gel from multiple experiments is shown. (H) Interactions among Rsc7, Rsc8, and Sfh1. Amino acid abbreviations: A, Ala; D, Asp; E, Glu; F, Phe; G, Gly; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; W, Trp.

One prominent feature of the NB lobe is the CTT of Sfh1, which is close to the H2A-H2B dimer and the H4 tail of the nucleosome but is largely disordered in the structure (Fig. 3C). The CTTs of Sfh1 homologs are enriched in positively charged residues (Fig. 3D), which suggests that these residues may bind to the acidic pocket of H2A-H2B and/or the nearby DNA strands. In line with this idea, we identified a block of EM density proximal to the CTT above the H2A-H2B surface (fig. S8A). Similar to Snf5 (19, 20), the CTT of Sfh1 forms cross-links with H2A and H2B (fig. S4B), supporting close proximity in space.

Regarding CTT function, whereas a plasmid encoding wild-type SFH1 fully complemented the lethal phenotype of sfh1Δ (21), a plasmid encoding the sfh1 CTT truncation mutant (amino acids 1 to 384, named Sfh1-ΔCTT) restored viability but displayed multiple strong conditional phenotypes, likely resulting from defects in the transcription of stress response genes (22) (Fig. 3E and fig. S8B). Furthermore, the loss of individual arginine or lysine residues (replacement by alanine) in the CTT did not confer phenotypes, suggesting the functional involvement of multiple positively charged residues. To further explore the function of the CTT, we examined its nucleosome binding in vitro (fig. S8, C and D). The CTT bound the nucleosome, and mutation of the individual conserved positively charged residues Arg401, Arg404, Lys403, or Arg400 to alanine modestly weakened the binding. Interestingly, mutations of all four residues together largely diminished the binding, which suggests that the CTT may bind to the nucleosome through multivalent interactions.

To determine the role of the Sfh1 CTT in nucleosome remodeling, we purified (fig. S8E) and compared the biochemical activity of wild-type RSC to that of mutant RSC lacking the Sfh1 CTT (Sfh1-ΔCTT). The nucleosome-dependent ATPase activity and the nucleosome sliding activity (on yeast 5S nucleosomes) of the mutant complex were robust and comparable to the wild type (Fig. 3F and fig. S8F). Notably, RSC Sfh1-ΔCTT was unable to eject yeast nucleosomes residing in closed circular arrays, whereas wild-type RSC was fully capable of ejection (Fig. 3G). These results suggested that the Sfh1 CTT helps RSC bind to the dish face of the nucleosome and stabilize RSC association during DNA translocation, enabling nucleosome ejection (16, 23, 24). Taken together, these results show that the CTT of Sfh1, by homology INI1/BAF47 in human cells, functions to stably anchor RSC on the nucleosome, thereby facilitating nucleosome ejection in vitro and fitness in vivo.

The NB lobe contains one additional component, Rsc7. Whereas the N-terminal region of the Rsc7 is not essential for the RSC assembly, the C-terminal domain (CTD) confers full Rsc7 function (22). Consistent with these studies, we could not detect the NTD in the structure, whereas we could identify the CTD (fig. S9A), which staples Sfh1 and Rsc8b together (Fig. 3H). The Sfh1-Rsc8 binding sequence of Rsc7 is conserved in animal BAF45a (α1, fig. S10). In agreement, BAF45a forms cross-links to the Sfh1 homolog in the PBAF complex (25).

Structure of the DB lobe and its implications for exit DNA binding

We found that the core of the DB lobe comprises Rsc9, Rsc6, the coiled-coil (CC) domain of Rsc8, the C-terminal half of the preHSA, and a middle segment of Rsc58 (Fig. 4A). On the surface of the DB lobe reside the yeast-specific subunits Rsc3, Rsc30, and Htl1 (Fig. 4A).

Fig. 4 Structures of the DB lobe and HB lobe.

(A) Structure of the DB lobe. The boxed region is enlarged for analysis in fig. S14B. (B) Schematic illustration of the interactions of Rsc8 with other subunits of RSC. (C) Two different views of the Rsc8 structure. (D) Positions of Rsc3 and Rsc30. (E) Rsc2, Rsc4, and the BD of Rsc58 are located close to the N-terminal tails of histone H3. (F) Two different views of the structure of the HB lobe.

Rsc9 is composed of armadillo repeats (fig. S11A) (26), referred to as the Rsc9 homology domain (R9HD). The R9HD binds to Rsc6 and the preHSA at the DB lobe, with the less conserved N-terminal tail extending to the HB lobe (fig. S9B). The R9HD shows some conservation to BAF200 in animals, with ~30% sequence similarity to human BAF200 (fig. S11B), which is the characteristic component of the PBAF complex (1).

Rsc8 is a multidomain protein containing SWIRM, zinc finger (ZnF), SANT, and CC domains (Fig. 4B). All these structural elements except the ZnF domain are highly conserved (fig. S12). Consistent with a previous study (27), Rsc8a and Rsc8b dimerize mainly through the CC domains. The Rsc8 dimer connects the three lobes of SRM together, with SWIRM at the NB lobe, ZnF and SANT at the HB lobe, and CC at the DB lobe (Fig. 4C and fig. S9C), suggesting a key role of Rsc8 in RSC assembly. These findings provide the structural basis for earlier studies suggesting the formation of a core complex by BRG1-BAF155-BAF170-INI1 tetramer (28) and the central role of BAF155-BAF170 in BAF/PBAF assembly (25).

Rsc6, which is highly similar to Swp73 in yeast and BAF60s in animals (fig. S13A), bundles to the CC domain of Rsc8 (Fig. 4A and fig. S9D). In support of the conservation, the SWIB domain of Rsc6 adopts a structure homologous to that of mouse BAF60a (fig. S13B). The domain following SWIB does not fold into a rigid conformation but extends out from the DB lobe, contacting the weak EM density that probably results from Rsc3 and Rsc30 (fig. S2, B and D). We speculate that Swp73 and BAF60s may fold into the same structure as Rsc6, with a domain extending to bind to different transcription factors (29, 30); this would support the function of the DB lobe in linker DNA binding.

The middle Rsc6-binding domain (R6BD) of Rsc58 binds to Rsc6 and Rsc8 (fig. S14B). The interacting residues are highly conserved (figs. S11 to S14), which suggests that this structure may serve as a model for the human complexes. In support of this notion, the Rsc58 homolog BRD7 is a unique component of the PBAF complex (25, 31).

Rsc3 and Rsc30 are yeast-specific transcription factors that can heterodimerize and show a binding preference for GC-rich DNA sequences (6, 7, 32). Rsc3 and Rsc30 each contain an N-terminal DNA-binding ZnD, a dimerization domain (DD), and a CTD (Fig. 1B). We identified short β sheets and helices of the DDs, which form a dimer binding to the surface of Rsc9 (Fig. 4A). The interaction between the dimerization helices was also detected by CL-MS analysis (fig. S4B). Weak EM densities near the DDs were found (fig. S2E), which may result from the flexible ZnDs and CTDs of Rsc3 and Rsc30.

The DDs of Rsc3 and Rsc30 are located on the DB lobe at a position distal to the DNA entry side but close to the exit side; thus, they position the N-terminal ZnDs to bind the exit DNA (Fig. 4D). The N termini of DDs are ~60 Å away from the exit DNA and are connected to the ZnDs through spacer sequences with a length of ~100 residues, which would allow a large degree of flexibility to sample the exit DNA. Htl1, a yeast-specific small subunit of RSC, binds to the CC domain of Rsc8 (Fig. 4A and fig. S9F), in agreement with interactions between Htl1 and the C terminus of Rsc8 (33).

Structure of the HB lobe and its implications for histone tail binding

RSC and PBAF both contain subunits carrying multiple BDs, such as Rsc2 and Rsc4 in RSC, and their ortholog BAF180 in PBAF, which are partly responsible for recruitment of the complexes through binding to histone tails (8, 10). Interestingly, these histone tail–binding elements cluster at the HB lobe (Fig. 4E), containing one additional histone tail–binding element, the BD of Rsc58 (Fig. 4F).

Whereas the structures of the BDs and BAH of Rsc2 and Rsc4 could not be detected, we were able to identify the CTDs of Rsc2 (residues 741 to 883) and Rsc4 (residues 364 to 625), which map to the tip of the HB lobe (Fig. 4E). This structure is consistent with studies showing that the CTD of Rsc2 is necessary and sufficient for RSC assembly (34). Similarly, the C terminus of Rsc4 mediates RSC assembly (8). The CTDs of Rsc2 and Rsc4 are exposed at the surface and ~70 Å away from the H3 tails. They connect to the N-terminal BAH and BDs through long spacer sequences with a length of >100 and 50 residues, respectively, which would poise these histone tail–binding elements to access their substrates, in agreement with the functions of BDs and BAH in histone tail binding (9, 10). The BD of Rsc58 loosely packs at the periphery of the HB lobe (Fig. 4F), with its histone-binding pocket exposed to solvent and ~60 Å away from histone H3 (class B complex, fig. S3A), suggesting that Rsc58 might still bind to the histone H3 tails. The spatial proximity of the BDs and BAH domains to the histone tails was also supported by the CL-MS analysis (fig. S4B).


In this work, we determined the structure of RSC bound to the nucleosome. Within RSC, Sth1 functions as a global organizer orchestrating the complex into three modules (Fig. 5A), and Rsc8 works as a local organizer delineating the SRM into three substrate-binding lobes (Fig. 5B).

Fig. 5 Model of the assembly and action of RSC.

(A) Schematic of the structure of RSC bound to the nucleosome. Only the major structural elements are shown. DNA translocation results in the movement of the nucleosome away from the NFR. (B) Organization of the SRM.

RSC binds to the nucleosome primarily through the motor domain of Sth1, with the majority of the histone-DNA contacts maintained. This model is different from the previous notion of nucleosome embracement (5), which was proposed to result in extensive rearrangement of histone-DNA contacts to facilitate DNA translocation. Our findings support a unified mechanism of chromatin that involves the action of the motor domain itself (12, 35). This is consistent with DNA translocation by the different remodelers involving 1– to 2–base pair movements of DNA along the octamer surface (11, 24, 3638).

RSC repositions the promoter +1 nucleosome away from the NFR, whereas ISWI slides the nucleosome toward the open DNA (39, 40). Our structure suggests that Rsc3-Rsc30 binds to the open NFR DNA such that the motor of Sth1 is loaded onto the +1 nucleosome at the side distal to the NFR (Fig. 5A). RSC would then push the +1 nucleosome away from the NFR. In contrast, the HSS domain of ISWI binds to the NFR DNA and positions the motor domain to the nucleosome at the proximal side, and thus slides the nucleosome toward the NFR (39). Therefore, Rsc3-Rsc30 in RSC and HSS in ISWI both recognize the NFR DNA, yet the specific architectures of these remodeling complexes position their motors differently, enabling the enzymes to respond to the chromatin cues distinctly and conferring specific functions at promoters.

The structure of RSC also sheds light on the assembly of the mammalian homologs PBAF and BAF. In addition to the well-known conserved components (1), our work provides evidence that yeast Rsc2 and Rsc4 together represent the mammalian polybromo protein BAF180, as Rsc2 and Rsc4 are adjacent, and their BAH and BDs combine to largely constitute polybromo. Furthermore, our work shows that BAF45a, BAF200, and BRD7, which are the characteristic subunits of the PBAF (25), are homologous to Rsc7, Rsc9, and Rsc58 in RSC, respectively.

Supplementary Materials

Materials and Methods

Figs. S1 to S14

Table S1

Movie S1

Data S1

References (4157)

References and Notes

Acknowledgments: We thank the Electron Microscopy Laboratory, CryoEM platform, and High-performance Computing Platform of Peking University, and the Bio-Computing Platform at Tsinghua University. Funding: The National Key Research and Development Program (2017YFA0102900 to Z.C., 2016YFA0500700 to N.G., 2019YFA0508900 to Z.C. and N.G.), the National Natural Science Foundation of China (31570731 and 31630046 to Z.C., 31725007 and 31630087 to N.G.), HHMI (B.R.C.), NCI R01 CA 201396 (C.R.C. and N.V.), and NCI CA042014 to Huntsman Cancer Institute for core facilities. Author contributions: Y.Y. prepared the samples for EM and MS analyses; W.Z., K.C., and H.D. performed MS analyses; H.W., K.C., N.G., and Z.C. performed EM analysis and model building; C.R.C., N.V., and B.R.C. conducted genetic and biochemical analyses; Z.C. oversaw the project and wrote the manuscript with help 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-9905, 0777, 0778) and protein coordinates are deposited at the Protein Data Bank (PDB 6K15, 6KW3, and 6KW4) for the SRM, class A complex, and class B complex, respectively.

Stay Connected to Science

Navigate This Article