Mechanism of Chromatin Remodeling Revealed by the Snf2-Nucleosome Structure PDF

Article doi:10.1038/nature22036 Mechanism of chromatin remodelling revealed by the Snf2-nucleosome structure Xiaoyu Liu1,2,3*, Meijing Li1,2,3*,...

Article doi:10.1038/nature22036 Mechanism of chromatin remodelling revealed by the Snf2-nucleosome structure Xiaoyu Liu1,2,3*, Meijing Li1,2,3*, Xian Xia1,2*, Xueming Li1,2,3 & Zhucheng Chen1,2 Chromatin remodellers are helicase-like, ATP-dependent enzymes that alter chromatin structure and nucleosome positions to allow regulatory proteins access to DNA. Here we report the cryo-electron microscopy structure of chromatin remodeller Switch/sucrose non-fermentable (SWI2/SNF2) from Saccharomyces cerevisiae bound to the nucleosome. The structure shows that the two core domains of Snf2 are realigned upon nucleosome binding, suggesting activation of the enzyme. The core domains contact each other through two induced Brace helices, which are crucial for coupling ATP hydrolysis to chromatin remodelling. Snf2 binds to the phosphate backbones of one DNA gyre of the nucleosome mainly through its helicase motifs within the major domain cleft, suggesting a conserved mechanism of substrate engagement across different remodellers. Snf2 contacts the second DNA gyre via a positively charged surface, providing a mechanism to anchor the remodeller at a fixed position of the nucleosome. Snf2 locally deforms nucleosomal DNA at the site of binding, priming the substrate for the remodelling reaction. Together, these findings provide mechanistic insights into chromatin remodelling. Packaging of genomic DNA into chromatin in eukaryotic cells serves (20 bp) at one end and no linker DNA at the other end17. A truncated Snf2 to store and protect the genetic materials. However, this creates an (residues 666–1400) from S. cerevisiae (ScSnf2), which has been shown obstacle to accessing to the genetic information. Cells have evolved a to be highly active in ATP hydrolysis and chromatin remodelling7, large family of chromatin remodelling enzymes that alter the chromatin was used to form a complex with the NCP-167 in the absence of nucle- structure and allow access to the DNA as required1. otide. Three-dimensional reconstruction of the complex was deter- Chromatin remodellers are superfamily II (SF2) helicase/translocase- mined using cryo-EM (Fig. 1 and Extended Data Fig. 1), which revealed like proteins, which include Snf2, imitation switch (ISWI), Swr1 and that ScSnf2 binds the NCP-167 at two different positions: one around Chd1. These proteins share a common catalytic core, but each subfamily super-helical location 2 (SHL2) and the other around SHL6 (Extended member has distinct auxiliary domains that confer specific properties Data Fig. 2). A small fraction of the enzymes was found to bind the on different remodellers1,2. Their association with other subunits within nucleosomes simultaneously at both sites (Extended Data Fig. 2 inset). large complexes further increases the diversity of the remodellers. SHL2 of the NCP is a well-known strategic interaction site for The conserved catalytic core of each remodeller consists of two Snf2 and ISWI remodellers2,4,18–20. The structure of ScSnf2 bound RecA-like domains, which perform the basic function of coupling at SHL2 (the SHL2 complex) was reconstructed at a resolution of nucleic acid binding and ATP hydrolysis to chromatin remodelling. 4.69 Å and is discussed here. The structure of ScSnf2 bound at SHL6 The biochemical activity of the chromatin remodellers is remarkable (the SHL6 complex) was reconstructed at a higher overall resolution given the complexity of their substrate, the nucleosome. The nucleo­ (3.97 Å), which adopts essentially the same structure as that at SHL2 some core particle (NCP) contains about 146 base pairs (bp) of DNA (Extended Data Fig. 3), further supporting the assignment of the tightly wrapping in ~​1.7 turns around a histone octamer via many secondary structures of the enzyme. The binding of Snf2 at SHL6 DNA–histone contacts3. Chromatin remodellers are able to overcome was unexpected, which may be specific to the isolated Snf2 enzyme. these DNA–histone contacts, slide the histone octamer along the DNA, The biological significance of the SHL6 complex requires further catalyse histone exchange and alter the structure of the nucleosome validation. and even evict it1,2. A DNA wave/bulging model has been proposed Three-dimensional reconstruction of the samples shows that ScSnf2 to explain the sliding reaction4–6. However, the underlying structural seems to preferentially bind to the NCP at the side proximal to the basis of this process is largely unclear. Very few crystal structures of linker DNA, which is referred to here as the entry point of the nucleo­ real chromatin remodellers have been reported7–9, and the currently some (Fig. 1). Presumably owing to the proximity to the linker DNA, available structures of remodellers in complex with the nucleosome the entry point DNA shows higher flexibility than the distal DNA end substrate are at very low resolution10–16. (referred to as the exit point). Similarly, the entry point DNA of the free To understand how the remodellers interact with the nucleosome NCP-167, which was reconstructed at a resolution of 3.92 Å, also shows substrate, we determined the cryo-electron microscopy (EM) structures higher flexibility than the exit point DNA under the same conditions of Snf2 in complex with a mononucleosome. Our work illustrates the (Extended Data Fig. 1i). This asymmetric feature of NCP-167 helps structure of an active remodeller engaging with its substrate, shedding to orient the crystal structure of the ‘601’ nucleosome into the EM mechanistic light on the remodelling process. density map21, and the greater flexibility probably makes the entry point DNA the preferential site of binding. We cannot exclude the possibility Overall structure of the Snf2–NCP complex that Snf2 binding at the exit point of NCP-167 might fail to provide NCPs were assembled with a 167-bp DNA fragment (NCP-167), con- a defined structure that could be detected by the cryo-EM protocol taining a ‘601’ positioning sequence, an extranucleosomal linker DNA we used. 1 Ministry of Education Key Laboratory of Protein Science, Tsinghua University, Beijing 100084, China. 2School of Life Sciences, Tsinghua University, Beijing 100084, China. 3Tsinghua-Peking Joint Center for Life Sciences, Beijing 100084, China. *​These authors contributed equally to this work. 4 4 0 | NAT U R E | VO L 5 4 4 | 2 7 A p r i l 2 0 1 7 © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH Dyad a Lobe1 Lobe2 Exit Entry Snf2 PostHSA Core1 SuppH Core2 Brace SnAc H3 H4 tail H4 b SnAc c SuppH NCP-167 H2A PostHSA Core2 H2B Brace-II Core1 L1254 L977 d SuppH Figure 1 | Overall structure of Snf2 bound to the nucleosome around V1235 Brace-II Brace-I SHL2. Two different views of the cryo-EM density map superimposed with the structures of ScSnf2 and the nucleosome. ScSnf2, green; histone PostHSA DNA Brace-I H4, gold; H3, blue; H2A, cyan; H2B, purple; 5′​-DNA, red; 3′​-DNA, yellow. The linker DNA is invisible under the contour level set to show high- quality maps of Snf2 and NCP. e Brace-I Brace-II HsBrg1 SVEEKILAAAKYKLNVDQKVIQAGMFDQKSSSHERRAFLQAILEHEEQDESRHCS 1264 The remodeller at SHL2 mainly contacts the nucleosome through DmBrm SVEERILAAARYKLNMDEKVIQAGMFDQKSTGSERQQFLQTILHQDDNEEEEENE 1281 the DNA components, with additional interactions with the proximal ScSnf2 SVEEVILERAYKKLDIDGKVIQAGKFDNKSTSEEQEALLRSLLDAEEERRKKRES 1270 MtSnf2 SVEEKILERARFKLDMDGKVIQAGRFDNKSSETDRDAMLRTLLETADMAESGEQE 1056 histone H4 tail (Fig. 1). A local resolution analysis indicated that the phosphate backbones of the nucleosomal DNA mostly keep their regi­ f 500 g ATPase activity (min–1) stry on the surface of the histone octamer, with some distortion round 400 0.8 the site bound by ScSnf2 (Extended Data Fig. 4a, b). The remodeller Fraction cut 0.6 part of the complex is at a resolution of 5.04 Å, which clearly defines the 300 secondary structural elements of ScSnf2 (Extended Data Fig. 4c). The 200 0.4 0.02 two RecA-like core domains of ScSnf2 maintain their overall structures 0.2 0 100 like those of the closely related homologue from the thermophilic yeast 0 20 40 60 Myceliophthora thermophila (MtSnf2) in the absence of the substrate7, 0 0 0 20 40 60 T e 77 D 5D 4D with some local conformational changes and a large rotation of their W rfac e L9 V1 23 L1 25 Time (min) relative orientation upon nucleosome binding. int Figure 2 | Structure of Snf2 in the nucleosome-bound state. a, Domain Structure of Snf2 in nucleosome-bound state organization of Snf2. PostHSA, core1, SuppH, core2, Brace helices and Previous study has indicated that ScSnf2 (666–1400) contains the SnAc are magenta, green, yellow, cyan, red and orange, respectively. postHSA, catalytic core and SnAc domains7 (Fig. 2a). The core1 and b, Overall structure of ScSnf2 in the nucleosome-bound state. Nucleosomal DNA bound within the core1–core2 cleft, grey circle. core2 domains of MtSnf2 stack together in the resting state, leading c, Conformational changes of the core1 domain. Structure of the to an inactive conformation. Upon NCP binding, the relative orienta- core1 domain of MtSnf2 in the resting state is grey (Protein Data tion of the two core domains of ScSnf2 is rotated by ~​80° (Extended Bank (PDB) accession number 5HZR)7. Arrows indicate the movement Data Fig. 5), which generates a new core1–core2 interface through of SuppH and postHSA upon nucleosome binding. d, Conformational direct contact between two newly formed Brace helices (Brace-I changes of the core2 domain. e, Sequence alignments around of the and Brace-II) of the core2 domain and the SuppH helix (also named Brace helix region of four Snf2 subfamily proteins. Residues mutated Protrusion 1) of the core1 domain (Fig. 2b). Helicase motif VI (arginine in this study are highlighted in yellow. f, DNA- (black bars) and NCP- fingers) is disordered in the resting state, but becomes ordered in the (white bars) stimulating ATPase activities. Error bars, s.d. (n =​  3). nucleosome-bound state. Moreover, motif VI is brought into close g, Chromatin remodelling activity. Wild type (WT) interface, black proximity with the ATP-binding element of motif I (P-loop), explaining square; L977D, yellow; V1235D, red triangle; L1254D, red diamond. The activities of the mutants were barely detectable and are further the activation of the ATPase activity of the enzyme upon nucleosome shown in the inset. Error bars, s.d. (n =​ 3). For gel source data, see binding (Extended Data Fig. 5). Supplementary Fig. 1. The individual lobes of the enzyme undergo local conformational changes. Within lobe1, postHSA and SuppH contact each other in the resting state, and they show coordinated movement upon nucleosome activity (Fig. 2g and Extended Data Fig. 6a). Likewise, the L977D muta- binding (Fig. 2c)7. PostHSA, which has been shown to downregulate the tion of SuppH dramatically reduced the remodelling activity, whereas activity of the enzyme22, is not directly involved in substrate binding or the ATPase activities of the mutant were reduced approximately two- core1–core2 communication, suggesting it may negatively regulate the fold. The SuppH–Brace helix interactions are unexpected and only remodelling activity through SuppH. Within lobe2, the Brace helices form in the substrate-bound state, playing an important role for the of MtSnf2 are mostly disordered in the resting state7. Upon nucleosome activities of Snf2, particularly in coupling ATP hydrolysis to chromatin binding, the equivalent sequence of ScSnf2 forms two helices remodelling. Interestingly, several mutations in the SuppH of Sth1 were (Fig. 2d). These two newly formed helices protrude from the core2 found to suppress the lethal phenotype of Arp7 and/or Arp9 deletion domain and contact the core1 domain, particularly through interac- in yeast23. Our model provides the structural basis for the increase tions with SuppH. of the remodelling activity of the Sth1 suppressor mutants, L681F in The structure suggests that the Brace–SuppH contact functions in particular, through enhancement of the hydrophobicity of SuppH22. core1–core2 communication in the substrate-bound state. The Brace The importance of the Brace-I helix is not specific to the Snf2 helices and SuppH, both of which are highly conserved among the subfamily proteins, and it extends to the remodeller ISWI. Mutation Snf2 subfamily of remodellers, interact with each other mainly through of the Brace helix of ISWI from M. thermophila (MtISWI) (V638D, hydrophobic residues (Fig. 2e). Mutations V1235D of Brace-I and corresponding to V1235D of ScSnf2; Extended Data Fig. 7) has also L1254D of Brace-II diminished the DNA- and nucleosome-stimulating been shown to decouple ATP hydrolysis from chromatin remodelling9, ATPase activities of the enzyme by two- to fivefold (Fig. 2f). More suggesting various chromatin remodellers use a conserved mechanism importantly, these mutations severely abrogated the remodelling of core1–core2 communication. However, the ATPase domains of Snf2 2 7 A p r i l 2 0 1 7 | VO L 5 4 4 | NAT U R E | 4 4 1 © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  RESEARCH Article a c 600 further, we reconstituted mutant nucleosomes without the H4 tail ATPase activity (gH4-NCP). Relative to the intact nucleosome, Snf2 showed an SnAc 400 (min–1) approxi­mate twofold reduction in the remodelling activity towards 200 gH4-NCP (Extended Data Fig. 8f). Taken together, these results Core2 suggest that the H4 tails of the nucleosome bind to Snf2, reinforcing its 0 T W 826D 113D 185A remodelling activity in vitro. Core1 5′ S T1 W1 Compared with the dramatic loss of the remodelling activity of Snf2 d 3′ 0.8 caused by the mutations of Brace and SuppH, the impact of H4 binding Fraction cut 0.6 is modest in vitro, and its significance within the holo-complex, PostHSA 0.4 particularly insides the cells, will require further examination. It is 0.2 interesting to find that the H4 tails bind to a similar surface patch in 0 Snf2 and ISWI, but play different regulatory roles. The H4-binding 0 20 40 60 Time (min) sites are critical for ISWI and protected by AutoN, yet the equivalent elements weakly modulate Snf2 and are exposed to solvent, consist- b 3′ 5′ 3′ ent with the conserved but not exchangeable catalytic cores of the 5′ 48 49 –49 Core2i remodellers7,9,19,24. 50 –50 W1185 W1185 Primary contacts between nucleosomal DNA and Snf2 Core2i 51 –51 Motif II ScSnf2 engages with its nucleosome substrate extensively through the K900 (II) 50 Core2i 52 –52 Motif II DNA components via a cleft formed by the central core1 and core2 52 Motif I Motif IV 53 –53 Motif II domains (Fig. 3a). Unlike a previous hypothesis19, ScSnf2 does not T1113 (IV) K905 (II) Motif IV 54 –54 Motif II intercalate between the DNA and the histone octamer. Instead, it Motif VI Motif V associates with the exposed DNA surface. The core1 domain wedges R1142 55 –55 between the two gyres of the nucleosomal DNA, and the core2 domain R1164 (V) 54 Motif Ia 56 –56 binds the DNA close to the dish-face of the nucleosome. The remod- R1142 S826 (Ia) Motif Ib 57 –57 eller interacts with the phosphate backbones along the minor groove 56 K878 (Ib) of the two strands of the nucleosomal DNA18 (Fig. 3b), consistent with Motif Ib 58 –58 the lack of sequence specificity in DNA binding by the remodeller. Figure 3 | Binding of the nucleosomal DNA by Snf2. a, Overall The binding of DNA by ScSnf2 involves multiple conserved helicase interaction between ScSnf2 and the nucleosomal DNA. b, Interaction and non-helicase motifs within the core1–core2 cleft. For simplicity, between the nucleosomal DNA and the core1–core2 cleft. Elements of the DNA strand with its 5′​end located at the entry point of the nucleo­ ScSnf2 involved in DNA contact are coloured blue and the corresponding some is called the ‘5′​-strand’, and the complementary strand is the helicase motifs are indicated in parentheses. The DNA bases are numbered ‘3′​-strand’. The core1 domain of ScSnf2 binds to the nucleosomal DNA on the basis of the 5′​-strand starting from the ‘601’ sequence. Right, in a manner similar to that of SsoRad54 in complex with naked DNA26 schematic diagram of the primary DNA interactions. c, DNA- (black (Extended Data Fig. 9). Helicase motifs Ia centring on S826 and Ib cen- bars) and NCP- (white bars) stimulating ATPase activities. Error bars, tring on K878 are close to the 5′​-strand. Consistent with the structure, s.d. (n =​  3). d, Chromatin remodelling activities. Wild type, black square; S826D, green circle; T1113D, cyan triangle; W1185D, cyan diamond. Error S826D mutation of motif Ia severely attenuated the ATPase (Fig. 3c) bars, s.d. (n =​ 3). For gel source data, see Supplementary Fig. 1. and chromatin remodelling activities (Fig. 3d and Extended Data Fig. 6b). The structure is also supported by previous biochemical ana­ lyses, which showed that mutation of the equivalent residue of K878 and ISWI are not interchangeable19,24. One characteristic element of of motif Ib in MtSnf2 abrogated the remodelling activity7. Likewise, the Snf2 subfamily proteins is the Brace-II helix, which is absent in most of motif II of MtSnf2 is disordered in the resting state, but the ISWI. Instead, ISWI contains NegC (Extended Data Fig. 7), which pro- equivalent region of ScSnf2 becomes ordered upon binding to the NCP, trudes from the core2 domain and negatively regulates the enzyme9,25. making extensive contacts with the 3′​-strand. Mutation of motif II of In contrast, the Brace-II helix interacts with the Brace-I helix, playing MtSnf2 has also been shown to disrupt the activities of the enzyme. a positive role in supporting the activity of Snf2. The core2 domain contacts the nucleosomal DNA through canonical helicase motifs IV and V. This structure is in agreement with the previous Snf2–hisone H4 contacts study showing the cross-linking of the region between motifs IV and One feature of ScSnf2 at SHL2 of the nucleosome is the interaction with V to nucleosomal DNA19, but differs from the model of SsoRad54, in the amino (N)-terminal tail of H4, which extends from the dish-face which motifs IV and V are distal to the DNA-binding cleft (Extended of the nucleosome and contacts a highly negatively charged surface Data Fig. 9)26. Motif IV centring on T1113 and motif V centring of the core2 domain (Extended Data Fig. 8a). This surface of ScSnf2 on Arg1164 of ScSnf2 interact with the backbone of the 5′​-strand. corresponds to the acidic pocket of ISWI that binds to the basic patch of Supporting our model, T1113D mutation of motif IV diminished the H4 (Extended Data Fig. 8b)9,25. The conserved sequence and structure activities of the remodeller (Fig. 3c, d). Similarly, mutation of motif V suggest that the acidic surface of ScSnf2 may bind to the basic patch of has been shown to abolish the remodelling activity of the enzyme7,27. the H4 tail. Consistent with this notion, disruption of the acidic pocket In addition to the canonical helicase motifs, other elements of ScSnf2 of Snf2 (KK, E1069K D1121K) markedly weakened the binding of the within the core1–core2 cleft are also involved in nucleosome binding. protein to the H4 tail (Extended Data Fig. 8c). The structural integrity We found a tryptophan residue upstream motif VI (W1185), which is of the KK mutant was maintained, as suggested by the intact ATPase conserved not only in the Snf2 subfamily proteins, but also in Chd1 activity (Extended Data Fig. 8d). Likewise, mutations of the basic patch and ISWI remodellers9. The bulky side chain of W1185 inserts into the of H4 reduced the binding interaction (Extended Data Fig. 8e). minor groove of the nucleosomal DNA, and packs against the back- The H4 tails of the nucleosome are important for the activation of bone of the 3′​-strand (Fig. 3b and Extended Data Fig. 4d). Whereas ISWI remodellers, but seem to be less so for the regulation of Snf2 the W1185A mutation modestly modulated the ATPase activity, the enzymes1. In fact, the H4-binding KK mutation of Snf2 slightly but mutant dramatically lost its remodelling activity (Fig. 3c, d). W1185 reproducibly reduced the remodelling activity approximately two- to of ScSnf2 is located at the N terminus of motif VI, which senses the threefold (Extended Data Fig. 8f). To validate the Snf2–H4 interaction γ​-phosphate group of ATP through the ‘arginine fingers’ (R1196 and 4 4 2 | NAT U R E | VO L 5 4 4 | 2 7 A p r i l 2 0 1 7 © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH a a Entry b 5′ 3′ Core1 Core2 β-Hairpin SHL2 R464 R1196 5′ R467 P-loop R1199 3′ R880 K885 Figure 5 | Directionality of DNA translocation driven by Snf2. K855 a, Superimposition of the structure of ScSnf2 bound to the NCP and that SHL-6 of NS3 (grey, PDB accession number 3KQH)29 bound to single-stranded DNA (orange) in the absence of nucleotide. The core1 domains of the proteins are aligned. Arrow indicates the proposed direction of DNA b HsBrg1 GSPAARRAFVPQLRSGKFNVLLTTYEYIIKDKHILAKIRW 875 translocation. b, Comparison of the structures of the ATPase active sites DmBrm GSPQGRRLLQNQMRATKFNVLLTTYEYVIKDKAVLAKIQW 894 of ScSnf2 and NS3 (grey). P-loop (motif I) of Snf2 is blue. The ‘arginine ScSnf2 GSPNERKAKQAKIRAGEFDVVLTTFEYIIKERALLSKVKW 888 MtSnf2 GPPNARKMQQEKIRQGKFQVLLTTYEYIIKDRPLLSKIKW 675 fingers’ of ScSnf2 and NS3 are shown as stick models. Arrow indicates the proposed movement of the core2 domain relative to the core1 domain c d upon ATP binding. 500 Several conserved basic residues (K855, R880 and K885 of ScSnf2) were ATPase activity (min–1) 400 0.8 identified (Fig. 4b), suggesting that they may function as the secondary DNA-binding elements. This notion is supported by the higher- Fraction cut 300 0.6 resolution structure at SHL6, with the side chain of R880 contacting 200 0.4 the 3′​-strand (Extended Data Fig. 4e). Consistent with the structure, 0.2 R880E K885E double mutation reduced the nucleosome-dependent 100 ATPase by about twofold (Fig. 4c), whereas it had less impact on the 0 0 20 40 60 DNA-dependent ATPase activity, suggesting a specific role for these 0 T e 5E 0E E 5E 880 E Time (min) residues in nucleosome recognition. More importantly, relative to W fac er K 85 R 88 K 88 R 885 the protein with an intact interface, the initial rate of the remodelling t in K reaction catalysed by the R880E K885E mutant was reduced by over Figure 4 | Binding of nucleosomal DNA to the secondary DNA-binding 30-fold (Fig. 4d and Extended Data Fig. 6c). sites of Snf2. a, Binding of ScSnf2 to both DNA gyres of the nucleosome. The relatively mild defects in the ATPase activity of these mutants are Electrostatic surface is calculated with Pymol. The secondary DNA- consistent with the idea that the enzyme binds tightly to and is activated binding sites of ScSnf2, yellow dots. b, Multiple sequence alignments of Snf2 subfamily remodellers around the secondary DNA-binding sites. by the nucleosome through the primary nucleosome-binding surface. c, DNA- (black bars) and NCP- (white bars) stimulating ATPase activities. The secondary nucleosome-binding surface is not essential for ATP Error bars, s.d. (n =​  3). d, Remodelling activities. Wild type, black square; hydrolysis per se, but is important in coupling ATP hydrolysis to chro- K855E, green circle; R880E, red triangle; K885E, blue diamond; R880E matin remodelling. ScSnf2 embraces one DNA gyre of the nucleosome K885E, magenta inverted triangle. Error bars, s.d. (n =​ 3). For gel source through its primary DNA-binding sites and binds to the other DNA data, see Supplementary Fig. 1. gyre through its secondary DNA-binding sites, which would prevent rotation of the enzyme along the primary DNA gyre. Thus, disrup- R1199)7. Thus, W1185 is involved in dual connections both to the ATP tion of the secondary interactions would not perturb ATP hydrolysis sensing elements and to the nucleosome substrate, which may enable much, but probably disfavour the anchorage of the motor, leading to it to link ATP hydrolysis to the associated conformational change loss of nucleosome remodelling. More studies are needed to examine to perturb the DNA–histone binding, coupling ATP hydrolysis and the dynamics of this system further. nucleo­some remodelling. In ScChd1, R750 located between motif IV and motif V of the core2 Directionality of DNA translocation domain has been suggested to bind DNA8, and the equivalent residue The mechanism of nucleosome engagement by ScSnf2 suggests the of ScSnf2 (R1142) is close to the 5′​-strand (Fig. 3b). A previous study directionality of DNA translocation. The Snf2 subfamily remodellers showed that the insertion sequence within the core2 domain (core2i) is are SF2 translocases that move along free double-stranded DNA, show- implicated in nucleosome binding7. Consistent with this study, core2i is ing 3′​-5′​polarity by tracking one DNA strand (the tracking strand)28. in close proximity to the minor groove near dG(−​49) of the 3′​-strand This translocase activity is essential for chromatin remodelling, because and dG52 of the 5′​-strand. the enzyme binds the nucleosome at a fixed position and draws in Most of the nucleosomal DNA-binding elements described above, the extranucleosomal linker DNA4. As ScSnf2 binds both strands of including the helicase motifs (Ia, Ib, II, IV and V), W1185 and R1142 of the nucleosomal DNA, the tracking strand could not be identified ScSnf2, are conserved in the Snf2, ISWI and Chd1 protein subfamilies9. easily. To ascertain the directionality of DNA translocation, we com- Our model may be used as a prototype to investigate the mechanism of pared the structure of the ScSnf2–nucleosome complex with that of nucleosome binding by the catalytic cores of other remodellers. helicase/translocase NS3 (Fig. 5a)29, which also shows a 3′​-5′​ polarity. It has been proposed that the core1–core2 cleft of NS3 undergoes an Secondary nucleosomal DNA contacts of Snf2 open-to-close transition during the ATPase cycle, which is essential for ScSnf2 at SHL2 not only binds the nucleosomal DNA via the core1– propelling DNA translocation. core2 cleft, but also contacts the adjacent DNA gyre at SHL-6 through Superposition of the core1 domains of ScSnf2 and NS3 in the a highly positively charged surface patch of the core1 domain (Fig. 4a). nucleotide-free state showed that their core2 domains adopt similar 2 7 A p r i l 2 0 1 7 | VO L 5 4 4 | NAT U R E | 4 4 3 © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  RESEARCH Article a Dyad b the remodelling reaction32. Compared with the structure of NCP with 145-bp DNA33, the NCP with the ‘601’ positioning sequence contains 1 bp of extra DNA around SHL ±​ 2, leading to bulging/looping out of 58 the DNA backbone at these positions21,31 (Fig. 6a). SHL1.5 56 Snf2 engages with the NCP at the strategic site of SHL2, leading to SHL2 54 further bulging out of the DNA components of the nucleosome sub- 52 strate at the site of engagement (Fig. 6c, d). Our structure suggests 50 that the interactions of Snf2 with the adjacent DNA gyre through its SHL2.5 48 secondary DNA-binding surface may help to anchor the enzyme at SHL3 SHL2, which is further stabilized by the binding to the nearby H4 tail. Substrate engagement triggers conformational changes of the enzyme, bringing motifs I and VI closer in space, and poses Snf2 for ATP hydrolysis. Thus, NCP binding primes both the remodeller and its c d chromatin substrate for the subsequent reaction. It is conceivable Core1 Dyad 2° DNA that ATP binding and hydrolysis would then induce closure of the DNA-binding core1–core2 cleft, and deliver 1 bp of the DNA at SHL2 SuppH Ia, Ib II H3 H4 SHL2 towards the dyad, initiating the remodelling reaction. I 1° DNA Snf2 This mechanism of DNA translocation by Snf2 is consistent with the VI H2A H2B notion that the ATPase domain of the remodeller is an autonomous V IV Core2i machine, and it remodels the nucleosome 1–2 bp at a time6,30. A ‘wave– Brace Core2 ratchet–wave’ model for chromatin remodelling has been proposed, in which a tracking subdomain remains bound at a fixed position on the Figure 6 | Model of nucleosome sliding by Snf2. a, Superimposition of histone octamer, and a torsion subdomain undergoes a conformational the structures of nucleosomal DNA bound by ScSnf2 at SHL2 (red and change and pulls the DNA4. Our findings suggest the core1 and core2 yellow), free NCP-145 (cyan, PDB accession number 2NZD)33 and ‘601’ domains of Snf2 may function as the tracking and torsion subdomains, NCP bound by RCC1 (grey, PDB accession number 3MVD)21. Boxed respectively, and their relative movement during the ATPase cycles region is analysed further in b. c, An enzyme-centred view of DNA would then drive translocation of the nucleosomal DNA. translocation mediated by Snf2. The primary (1°) and secondary (2°) DNA In summary, our findings illustrate how Snf2 engages with the gyres of the nucleosome bound by Snf2 are indicated as circles. The red nucleo­some substrate, which provides the structural basis of chromatin arrow indicates the relative contraction motion of the core1 and core2 domains during the ATPase cycle. d, A substrate-centred view of DNA remodelling. This basic mechanism of nucleosome sliding probably translocation. Snf2, green circle; H4 tail, gold. The bulging/looping out also applies to the ISWI and Chd1 subfamilies of remodellers. of 1 bp of the DNA at SHL2 is schematically illustrated. The red arrow Online Content Methods, along with any additional Extended Data display items and indicates the direction of DNA loop propagation driven by contraction of Source Data, are available in the online version of the paper; references unique to the remodeller. these sections appear only in the online paper. positions (Fig. 5b). The β​-hairpin of NS3, which is essential for the received 1 August 2016; accepted 28 February 2017. unwinding of double-stranded DNA29, is absent in ScSnf2, consistent Published online 19 April 2017. with the lack of helicase activity of the remodeller1. The 5′​-strand of the 1. Clapier, C. R. & Cairns, B. R. The biology of chromatin remodeling complexes. nucleosomal DNA overlays well with the single-stranded DNA bound Annu. Rev. Biochem. 78, 273–304 (2009). by NS3 with the same polarity, suggesting the 5′​-DNA strand functions 2. Narlikar, G. J., Sundaramoorthy, R. & Owen-Hughes, T. Mechanisms and as the tracking strand when ScSnf2 binds to double-stranded DNA. By functions of ATP-dependent chromatin-remodeling enzymes. Cell 154, 490–503 (2013). analogy to NS3, ATP binding and hydrolysis would lead to closure of 3. Luger, K., Mäder, A. W., Richmond, R. K., Sargent, D. F. & Richmond, T. J. Crystal the core1–core2 cleft. Because ScSnf2 is anchored at a fixed position structure of the nucleosome core particle at 2.8 Å resolution. Nature 389, on the nucleosome, this relative movement of the core domains would 251–260 (1997). pump the linker DNA towards the dyad (Fig. 5a), consistent with the 4. Saha, A., Wittmeyer, J. & Cairns, B. R. Chromatin remodelling: the industrial revolution of DNA around histones. Nature Rev. Mol. Cell Biol. 7, 437–447 directionality of DNA translocation during a chromatin remodelling (2006). reaction4,30. The assignment of the directionality of chromatin remod- 5. Becker, P. B. & Hörz, W. ATP-dependent nucleosome remodeling. Annu. Rev. elling is also supported by the higher-resolution structure at SHL6 Biochem. 71, 247–273 (2002). 6. Mueller-Planitz, F., Klinker, H. & Becker, P. B. Nucleosome sliding mechanisms: (Extended Data Fig. 3). new twists in a looped history. Nature Struct. Mol. Biol. 20, 1026–1032 (2013). 7. Xia, X., Liu, X., Li, T., Fang, X. & Chen, Z. Structure of chromatin remodeler Swi2/ Distortion of the nucleosomal DNA Snf2 in the resting state. Nature Struct. Mol. Biol. 23, 722–729 (2016). 8. Hauk, G., McKnight, J. N., Nodelman, I. M. & Bowman, G. D. The Owing to the extensive interactions with the catalytic core of ScSnf2, chromodomains of the Chd1 chromatin remodeler regulate DNA access to the both strands of the nucleosomal DNA around SHL2 are locally dis- ATPase motor. Mol. Cell 39, 711–723 (2010). torted, with the surrounding sequence lifted off from its canonical path 9. Yan, L., Wang, L., Tian, Y., Xia, X. & Chen, Z. Structure and regulation of the chromatin remodeller ISWI. Nature 540, 466–469 (2016). on the surface of the histone octamer and the phosphate backbone of the 10. Leschziner, A. E. Electron microscopy studies of nucleosome remodelers. 3′​-strand showing a dramatic displacement as far as ~​5 Å (Fig. 6a, b). Curr. Opin. Struct. Biol. 21, 709–718 (2011). Given the greater flexibility of the proximal entry-point DNA of 11. Watanabe, S. et al. Structural analyses of the chromatin remodelling enzymes INO80-C and SWR-C. Nature Commun. 6, 7108 (2015). NCP-167, the local DNA distortion near SHL2 would probably cause 12. Tosi, A. et al. Structure and subunit topology of the INO80 chromatin a relatively smaller energy penalty, which provides a rationale for the remodeler and its nucleosome complex. Cell 154, 1207–1219 (2013). observed preference of Snf2 binding at this DNA segment over the 13. Nguyen, V. Q. et al. Molecular architecture of the ATP-dependent chromatin- distal DNA end. remodeling complex SWR1. Cell 154, 1220–1231 (2013). 14. Yamada, K. et al. Structure and mechanism of the chromatin remodelling factor ISW1a. Nature 472, 448–453 (2011). Discussion 15. Racki, L. R. et al. The chromatin remodeller ACF acts as a dimeric motor to Nucleosomal DNA displays intrinsic plasticity31. Different structures space nucleosomes. Nature 462, 1016–1021 (2009). 16. Leschziner, A. E., Lemon, B., Tjian, R. & Nogales, E. Structural studies of the of the NCP show variable DNA stretching around SHL±​2 and SHL±​5. human PBAF chromatin-remodeling complex. Structure 13, 267–275 The gain or loss of 1 bp at these sites has been proposed to facilitate (2005). 4 4 4 | NAT U R E | VO L 5 4 4 | 2 7 A p r i l 2 0 1 7 © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH 17. Lowary, P. T. & Widom, J. New DNA sequence rules for high affinity binding to 31. Tan, S. & Davey, C. A. Nucleosome structural studies. Curr. Opin. Struct. Biol. 21, histone octamer and sequence-directed nucleosome positioning. J. Mol. Biol. 128–136 (2011). 276, 19–42 (1998). 32. Mueller-Planitz, F., Klinker, H., Ludwigsen, J. & Becker, P. B. The ATPase domain 18. Zofall, M., Persinger, J., Kassabov, S. R. & Bartholomew, B. Chromatin of ISWI is an autonomous nucleosome remodeling machine. Nature Struct. Mol. remodeling by ISW2 and SWI/SNF requires DNA translocation inside the Biol. 20, 82–89 (2013). nucleosome. Nature Struct. Mol. Biol. 13, 339–346 (2006). 33. Ong, M. S., Richmond, T. J. & Davey, C. A. DNA stretching and extreme 19. Dechassa, M. L. et al. Disparity in the DNA translocase domains of SWI/SNF kinking in the nucleosome core. J. Mol. Biol. 368, 1067–1074 and ISW2. Nucleic Acids Res. 40, 4412–4421 (2012). (2007). 20. Dechassa, M. L. et al. Architecture of the SWI/SNF-nucleosome complex. Mol. Cell. Biol. 28, 6010–6021 (2008). Supplementary Information is available in the online version of the paper. 21. Makde, R. D., England, J. R., Yennawar, H. P. & Tan, S. Structure of RCC1 chromatin factor bound to the nucleosome core particle. Nature 467, 562–566 Acknowledgements We thank J. Lei at the Center for Structural Biology (2010). (Tsinghua University) and the staff at the Tsinghua University Branch of the 22. Clapier, C. R. et al. Regulation of DNA translocation efficiency within the National Center for Protein Sciences Beijing for providing facility support. chromatin remodeler RSC/Sth1 potentiates nucleosome sliding and ejection. This work was supported by the National Key Research and Development Mol. Cell 62, 453–461 (2016). Program to Z.C. (2014CB910100) and to X.L (2016YFA0501102 and 23. Szerlong, H. et al. The HSA domain binds nuclear actin-related proteins to 2016YFA0501902), the National Natural Science Foundation of China to regulate chromatin-remodeling ATPases. Nature Struct. Mol. Biol. 15, 469–476 Z.C. (31570731, 31270762, 31630046) and to X.L. (31570730), Advanced (2008). Innovation Center for Structural Biology, Tsinghua-Peking Joint Center for Life 24. Fan, H. Y., Trotter, K. W., Archer, T. K. & Kingston, R. E. Swapping function Sciences, and the ‘Junior One Thousand Talents’ program to Z.C. and X.L. of two chromatin remodeling complexes. Mol. Cell 17, 805–815 (2005). Author Contributions X.Liu and X.X. prepared the proteins and performed the 25. Clapier, C. R. & Cairns, B. R. Regulation of ISWI involves inhibitory modules biochemical analyses; M.L. collected the EM data with help from X.Liu and X.X.; antagonized by nucleosomal epitopes. Nature 492, 280–284 (2012). M.L. and X.Li performed the EM analysis; Z.C. wrote the manuscript with help 26. Dürr, H., Körner, C., Müller, M., Hickmann, V. & Hopfner, K. P. X-ray structures of from all authors; Z.C. directed and supervised all the research. the Sulfolobus solfataricus SWI2/SNF2 ATPase core and its complex with DNA. Cell 121, 363–373 (2005). 27. Smith, C. L. & Peterson, C. L. A conserved Swi2/Snf2 ATPase motif couples Author Information Reprints and permissions information is available at ATP hydrolysis to chromatin remodeling. Mol. Cell. Biol. 25, 5880–5892 www.nature.com/reprints. The authors declare no competing financial (2005). interests. Readers are welcome to comment on the online version of the paper. 28. Saha, A., Wittmeyer, J. & Cairns, B. R. Chromatin remodeling through Publisher’s note: Springer Nature remains neutral with regard to jurisdictional directional DNA translocation from an internal nucleosomal site. Nature Struct. claims in published maps and institutional affiliations. Correspondence and Mol. Biol. 12, 747–755 (2005). requests for materials should be addressed to Z.C. (Zhucheng_chen@tsinghua. 29. Gu, M. & Rice, C. M. Three conformational snapshots of the hepatitis C virus edu.cn) or X.Li ([email protected]). NS3 helicase reveal a ratchet translocation mechanism. Proc. Natl Acad. Sci. USA 107, 521–528 (2010). Reviewer Information Nature thanks B. Bartholomew, T. Owen-Hughes and 30. Harada, B. T. et al. Stepwise nucleosome translocation by RSC remodeling the other anonymous reviewer(s) for their contribution to the peer review of complexes. eLife 5, 5 (2016). this work. 2 7 A p r i l 2 0 1 7 | VO L 5 4 4 | NAT U R E | 4 4 5 © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  RESEARCH Article Methods a defocus range of 1.8–3.2 μ​m and nominal magnification of ×​22,500, correspond- No statistical methods were used to predetermine sample size. The experiments ing to pixel size 0.66 Å of super-resolution counting mode35. Each micrograph were not randomized. The investigators were not blinded to allocation during was dose-fractionated to 32 frames with 0.25 s exposure time in each frame. The experiments and outcome assessment. dose rate was 8.2 counts per physical pixel per second, and the total dose was ~​50 Protein expression and purification. ScSnf2 (residues 666–1400) and the related electrons per square ångström. mutant proteins were expressed and purified similarly as described before7. The Image processing and model building. For negative-staining micrographs, nucleosomes were constituted with 167 bp DNA containing the ‘601’ positioning CTFFIND3 was used to estimate the defocus parameters36. A total of 13,863 sequence (5′​-­st­ra­nd: A­TC­GTACTTCTCGACAAGCTTCAGGATGTATATATCT particles were picked using the ‘e2boxer.py’ subroutine in the EMAN2 suit37. Two- GACACGTGCCTGGAGACTAGGGAGTAATCCCCTTGGCGGTTAAAACGC dimensional classification was performed with RELION 1.4 to screen and remove GGGGGACAGCGCGTACGTGCGTTTAAGCGGTGCTAGAGCTGTCTACGA most bad particles38. Then a sphere generated by SPIDER was used as the initial CCAATTGAGCGGCCTCGGCACCGGGATTCTCGAT; the ‘601’ positioning model for the first round of three-dimensional classification39. After two rounds sequence is underlined). of three-dimensional classification, 9,824 particles were selected, and subjected H4 tail (residues 1–21) from S. cerevisiae was cloned into a modified pET- to the final three-dimensional reconstruction. The cryo-EM super-resolution 28b vector with the His-tag deleted and a glutathione S-transferase (GST)-tag micrographs were 2 ×​ 2 binned, yielding an image stack with a pixel size of 1.32 Å. added to the carboxy (C) terminus. The H4 tail point mutants were generated Motion correction was performed using the MotionCorr program, which output by QuikChange mutagenesis. Recombinant proteins were overexpressed in motion-corrected integrated images for further processing40. CTFFIND3 was used the Escherichia coli expression strain Rosetta(DE3) and induced with 0.5 mM to determine defocus parameters. A total of 630,847 particles were picked using isopropyl-β​-d-thiogalactoside (IPTG). Proteins were purified by GST column, an automated in-house software suit. All subsequent two- and three-dimensional followed by an ion-exchange column (Source-15S, GE Healthcare) with HEPES image analyses were performed with RELION 1.4. After several rounds of two- buffer, pH 7.0, then subjected to gel-filtration chromatography (Superdex-200, GE dimensional classification, 462,321 particles were selected and divided into free Healthcare) in buffer containing 10 mM HEPES, pH 7.0, 50 mM NaCl and 5 mM NCP (254,777 particles), NCP–Snf2 complex (200,625 particles) and 2Snf2–NCP dithiothreitol (DTT). The purified protein was concentrated to 10 mg ml−1 and complex (6,919 particles) for further three-dimensional classification. A cylinder stored at −​80 °C. initial model generated with SPIDER and the negative-staining model were used ATPase and remodelling activities. ATPase and remodelling activities were as initial models for the first round of three-dimensional classification for the measured as described before7. To measure the basal DNA-independent ATPase free NCP and the NCP–Snf2 complex, respectively. After two rounds of three- activities, 1 μ​M proteins were used. To measure the ATPase activities of the dimensional classification, 63,311 free NCP particles were selected and subjected proteins in the activated state, 25 nM protein was used in the presence of 125 nM to three-dimensional auto-refinement. Then particle polishing and further three- double-stranded DNA (147 bp) or 125 nM NCP147. dimensional auto-refinement were applied, which resulted in a final free NCP map To compare the remodelling activities, 5 nM cy5-labelled 347-bp mono­ at 3.92 Å resolution estimated with the gold-standard Fourier shell correlation nucleosomes were incubated with 1 nM protein, 3 mM ATP and 40 U of HhaI in (FSC) 0.143 criterion41. remodelling buffer (20 mM Tris-HCl, pH 8.0, 20 mM KAc, 30 mM NaCl, 5 mM A total of 200,625 NCP–Snf2 complex particles were also subjected to three- MgCl2 and 0.1 mg ml−1 bovine serum albumin). Gels were analysed by a Typhoon dimensional classification based on the low-resolution signals. Although the Trio+​imager and quantified with the Quantity One program. density of linker DNA was weak, the low-resolution signal of this linker was still GST pull-down assay. Pull-down assays were performed at 4 °C in the binding strong and could be visualized as solid density after applying a low-pass filter to buffer containing 50 mM NaCl, 20 mM HEPES, pH 7.5 and 3 mM DTT. GST- 10 Å resolution (Extended Data Fig. 2 inset). The presence of linker DNA and the tagged proteins (7.5 μ​M) and GST alone (control) were pre-incubated with GST asymmetrical nature of the NCP-167 allowed two different binding conformations beads, and then mixed with 8 μ​M ScSnf2 (666–1400) proteins. After gentle rotation to be identified, with Snf2 binding to SHL6 (the SHL6 complex) and SHL2 (the for 60 min, the GST beads were washed four times with the binding buffer SHL2 complex) of the NCP, respectively. Another round of three-dimensional and eluted in 200 mM NaCl, 20 mM Tris-HCl, pH 8.0, 3 mM DTT and 30 mM classification was performed for the SHL6 complex subset with 127,232 particles glutathione. The samples were mixed with SDS loading buffer and electrophoresed and the SHL2 complex subset with 73,393 particles, respectively. Then, 90,725 and at 240 V for 35 min on 12% SDS–polyacrylamide gel electrophoresis. Gels were 42,383 particles were selected from these two subsets, and subjected to ‘polishing’ analysed by Coomassie blue staining. and ‘auto-refinement’ procedures, respectively. Finally, three-dimensional recon- Sample preparation and EM data collection. The ScSnf2 (666–1400)–NCP structions of the SHL6 and SHL2 complexes were calculated at resolutions of 3.97 Å complexes were obtained by mixing 15 μ​M protein with 5 μ​M NCP-167. We purified and 4.69 Å, respectively. To improve the resolution of the Snf2 protein part, NCP and stabilized the complex using the GraFix method34. To form the gradient, 6 ml was subtracted from the experimental particle images, and the remaining particles top solution containing 50 mM NaCl, 10 mM HEPES, pH 7.0 and 10% glycerol were processed following a focused classification procedure42. Local resolutions of (Sigma) was added to a tube (Beckman, 331372). Bottom solution (6 ml) containing all maps were estimated using Resmap37. After three-dimensional classification, 50 mM NaCl, 10 mM HEPES, pH 7.0, 30% glycerol and 0.15% glutaraldehyde 5,597 particles of 2Snf2–NCP were selected for auto-refinement and to acquire a (Polysciences) was then injected to the bottom of the tube using a syringe with a map at 11.26 Å resolution. blunt-ended needle. The tubes were placed into a gradient master (BioComp) to The structural model was built first by fitting the crystal structures of the NCP form a continuous density and glutaraldehyde gradient. Finally, 200 μ​l of sample (PDB accession number 3MVD)21, the individual core domains of MtSnf2 (PDB were loaded. The sample tubes were ultracentrifuged at 4 °C for 20 h at a speed of accession number 5HZR)7 and the H4 tail bound by MtISWI (PDB accession 35,000 r.p.m. (Beckman, Rotor SW-41Ti). Fractions were collected every 200 μ​l, number 5JXT)9 to the EM density map using UCSF Chimera43. The model was and were examined by electrophoresis at 70 V for 90 min on 4.5% native TBE poly­ further refined using phenix.real_real_space44 with secondary structure constraints acrylamide gels on ice. The gels were stained with SYBR Gold and analysed on a and rebuilt manually in Coot. Chemi-Doc XR+​system (Bio-Rad). The best fraction was selected and dialysed Data availability. Coordinates and EM maps have been deposited in the Electron to 50 mM NaCl, 10 mM Tris-HCl, pH 8.0 and 3 mM DTT and concentrated for Microscopy Data Bank and PDB under accession numbers EMD-6699 and 5X0X EM sample preparation. (the SHL6 complex), and EMD-6700 and 5X0Y (the SHL2 complex), respectively. Negative-staining samples were prepared using 0.75% uranyl formate. Holey All other data are available from the corresponding authors upon reasonable grids coated with continuous carbon film were glow discharged, and then 4 μ​l of request. sample were loaded. The grids were first blotted using filter paper, then washed with water and uranyl formate, sequentially. The samples were observed using 34. Stark, H. GraFix: stabilization of fragile macromolecular complexes for single a 200 kV Tecnai F20 microscope (FEI) equipped with a Gatan Ultrascan 4000 particle cryo-EM. Methods Enzymol. 481, 109–126 (2010). 35. Li, X., Zheng, S., Agard, D. A. & Cheng, Y. Asynchronous data acquisition and camera at a magnification of ×​62,000, corresponding to pixel size of 1.35 Å on the on-the-fly analysis of dose fractionated cryoEM images by UCSFImage. images. Defocus ranging from −​1.0 to −​2.5 μ​m and a total dose of ~​40 electrons J. Struct. Biol. 192, 174–178 (2015). per square ångström were used. 36. Mindell, J. A. & Grigorieff, N. Accurate determination of local defocus For cryo-EM sample preparation, a drop of 4 μ​l sample was applied to a and specimen tilt in electron microscopy. J. Struct. Biol. 142, 334–347 glow-discharged Quantifoil holey carbon grid (R1.2/1.3, 300 mesh). After waiting (2003). for 60 s, the grid was blotted for 4.0 s (under 100% humidity and 8 °C) and plunged 37. Ludtke, S. J. 3-D structures of macromolecules using single-particle analysis in EMAN. Methods Mol. Biol. 673, 157–173 (2010). into liquid ethane cooled by liquid nitrogen using FEI Vitrobot IV. The samples 38. Bharat, T. A., Russo, C. J., Löwe, J., Passmore, L. A. & Scheres, S. H. Advances in were observed using a Titan Krios microscope (FEI) operated at 300 kV, equipped single-particle electron cryomicroscopy structure determination applied to with a Gatan K2 Summit camera. UCSFImage4 was used for data collection under sub-tomogram averaging. Structure 23, 1743–1753 (2015). © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH 39. Frank, J. et al. SPIDER and WEB: processing and visualization of images in 42. Bai, X. C., Rajendra, E., Yang, G., Shi, Y. & Scheres, S. H. Sampling the 3D electron microscopy and related fields. J. Struct. Biol. 116, 190–199 conformational space of the catalytic subunit of human γ​-secretase. eLife 4, (1996). 4 (2015). 40. Li, X. et al. Electron counting and beam-induced motion correction enable 43. Pettersen, E. F. et al. UCSF Chimera—a visualization system for near-atomic-resolution single-particle cryo-EM. Nature Methods 10, 584–590 exploratory research and analysis. J. Comput. Chem. 25, 1605–1612 (2013). (2004). 41. Scheres, S. H. & Chen, S. Prevention of overfitting in cryo-EM structure 44. Afonine, P. V. et al. Towards automated crystallographic structure refinement determination. Nature Methods 9, 853–854 (2012). with phenix.refine. Acta Crystallogr. D 68, 352–367 (2012). © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  RESEARCH Article Extended Data Figure 1 | Negative-staining and cryo-EM structure complex. i, Cryo-EM density map of free NCP coloured on the basis of the analysis. a, Representative micrograph of negative staining. b, Two- local resolution and angular distribution of particle projections. j, Cryo-EM dimensional class averages of characteristic projection views of negative- density map of the Snf2 part of the SHL6 complex coloured on the basis of staining particles. c, Negative-staining density map of SHL6 complex. the local resolution. k, Cryo-EM density map of the Snf2 part of the SHL2 d, Representative micrograph of cryo-EM sample. e, Fast Fourier transforms complex coloured on the basis of the local resolution. l, The ‘gold-standard’ of image in d, with the Thon rings extending to ~​3.5 Å. f, Two-dimensional FSC curve calculated between two halves of data sets for the SHL6 complex, class averages of characteristic projection views of cryo-EM particles of the Snf2 (SHL6), the SHL2 complex, Snf2 (SHL2) and free nucleosome. m, Two- Snf2–NCP complex. g, Angular distribution of particle projections of the dimensional class averages of characteristic projection views of cryo-EM SHL6 complex. h, Angular distribution of particle projections of the SHL2 particles of the 2Snf2–NCP complex. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH Extended Data Figure 2 | Flow chart of cryo-EM data processing. Inset: low-resolution models to show the linker DNA that helped to orient the complex. For the SHL6 complex, the model was acquired from the yellow class (14.8%); SHL2 complex, grey class (20.0%); 2Snf2–NCP complex, structure with two copies of Snf2 bound to the same NCP at SHL2 (pink) and SHL6 (red). Scale bar, 2 nm. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. RESEARCH Article Extended Data Figure 3 | Comparison of the structure of Snf2 bound at SHL2 and SHL6. Snf2 bound at SHL2 is coloured as in Fig. 2, and that at SHL6 is coloured grey. Only the primary DNA is shown, with 5′​- and 3′​-strands bound by the Snf2 at SHL6 coloured magenta and orange, respectively. Red arrow indicates the proposed direction of DNA translocation driven by Snf2. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH Extended Data Figure 4 | Superposition of the structure and the EM c, Segmented map of ScSnf2. d, Region around W1185 of the SHL2 density map. a, Local resolution of the cryo-EM density map of the SHL2 complex. Segmented maps of Snf2 and the 3′​-DNA are coloured green and complex. b, Segmented map of the nucleosome part of the SHL2 complex. yellow, respectively. e, Region around R880 of the SHL6 complex. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. RESEARCH Article Extended Data Figure 5 | Conformational changes of Snf2 upon nucleosome binding. The structure of core1 domains of MtSnf2 in the resting state (grey, PDB accession number 5HZR)7 was aligned with that of ScSnf2 (green) in complex with the NCP. For clarity, only the structure around the central β​-sheets of the remodellers is shown. The core2 domains in the resting state and in the substrate state are coloured blue and cyan, respectively. The elements for ATP hydrolysis (motifs I and VI) are in red. Motif VI is disordered in the resting state, and becomes a helical structure in the nucleosome-bound state. The arrow indicates the movement of core2 domain relative to the core1 domain upon the binding of the nucleosome. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH Extended Data Figure 6 | Chromatin remodelling activities of mutant ScSnf2 (666–1400) containing DNA-binding mutations in the various constructs used in this study. a, Gels of the restriction enzyme- secondary DNA-binding sites. The cut fractions were quantified and are accessibility assays of ScSnf2 (666–1400) with wild-type interface and shown in Fig. 4d. d, Gels of the restriction enzyme accessibility assays of three core1–core2 interface mutants. The cut fractions were quantified ScSnf2 (666–1400) with wild-type interface towards gH4-NCP (left), H4- and shown in Fig. 2g. Three independent assays were performed and one binding KK mutation of Snf2 towards intact NCP (middle) and KK mutant was shown. b, Gels of the restriction enzyme accessibility assays of three Snf2 towards gH4-NCP (right). The cut fractions were quantified and are DNA-binding mutant ScSnf2 (666–1400). The cut fractions are shown shown in Extended Data Fig. 8f. in Fig. 3d. c, Gels of the restriction enzyme accessibility assays of three © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. RESEARCH Article Extended Data Figure 7 | Superposition of the structures of the core2 domains of ScSnf2 (cyan) and MtISWI (grey). The Brace helices of ScSnf2 are shown in red; NegC of MtISWI (PBD code 5JXT)9 is in orange. V1235 of ScSnf2 is equivalent to V638 of MtISWI. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.  Article RESEARCH Extended Data Figure 8 | Interactions between the histone H4 tail and intact interface and the H4-binding KK mutant. The experiments were ScSnf2. a, Binding of the histone H4 tail to a highly negatively charged repeated at least three times, and the representative gel shown. GST alone surface of ScSnf2. Electrostatic surface of ScSnf2 was calculated with was used as a negative control. d, ATPase activities of ScSnf2 (666–1400) Pymol. Red, negative electrostatic potential; blue, positive electrostatic with wild-type interface (black) and KK mutation (red) in the resting potential. b, Superimposition of the structure of the ScSnf2–NCP complex, state (-), and in the presence of DNA and NCP. e, GST pull-down assays the EM density map around the H4 tail (filtered to a resolution of 7.0 Å, of ScSnf2 (666–1400) with wild-type and four mutant H4 tail peptides. gold) and the crystal structure of the core2 domain of MtISWI (grey, PDB f, Chromatin remodelling activities of ScSnf2 (666–1400) with wild-type accession number 5JXT)9 in complex with the H4 tail (magenta). Acidic interface (black) and KK mutation (red) towards intact (filled symbols) residues of MtISWI surrounding the H4-binding pocket are shown as and mutant gH4 (open symbols) NCPs. For gel source data, sticks and labelled (grey), and the corresponding residues of ScSnf2 are see Supplementary Fig. 1. also labelled (cyan). c, GST pull-down assays of ScSnf2 (666–1400) with © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. RESEARCH Article Extended Data Figure 9 | Superposition of the structures of ScSnf2 and SsoRad54 (grey). The structures of the core1 domains of ScSnf2 and SsoRad54 (PDB accession number 1Z63)26 are aligned. The DNA bound by SsoRad54 is coloured light blue. The six DNA-binding elements of ScSnf2 (motifs Ia, Ib, II, IV, V and core2i) are labelled and coloured blue. Motifs IV and V of SsoRad54 are coloured magenta. © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.