Interactions between the nucleosome histone core and Arp8 in the INO80 chromatin remodeling complex

Abstract

Actin-related protein Arp8 is a component of the INO80 chromatin remodeling complex. Yeast Arp8 (yArp8) comprises two domains: a 25-KDa N-terminal domain, found only in yeast, and a 75-KDa C-terminal domain (yArp8CTD) that contains the actin fold and is conserved across other species. The crystal structure shows that yArp8CTD contains three insertions within the actin core. Using a combination of biochemistry and EM, we show that Arp8 forms a complex with nucleosomes, and that the principal interactions are via the H3 and H4 histones, mediated through one of the yArp8 insertions. We show that recombinant yArp8 exists in monomeric and dimeric states, but the dimer is the biologically relevant form required for stable interactions with histones that exploits the twofold symmetry of the nucleosome core. Taken together, these data provide unique insight into the stoichiometry, architecture, and molecular interactions between components of the INO80 remodeling complex and nucleosomes, providing a first step toward building up the structure of the complex.

Chromatin remodeling complexes have recognized roles in remodeling of nucleosomes during transcription and DNA repair (1). In addition to the DNA translocase subunit, these complexes typically consist of several proteins, although the biochemical functions of these additional subunits remain a mystery in most cases. Actin-related proteins (ARPs) are a group of proteins with homology to actin (2). Several ARP family members (Arp4–Arp9) are components of nucleosome-modifying complexes, including remodelers (e.g., INO80, RSC), histone exchangers (e.g., INO80, Swr1), and acetylation complexes (e.g., NuA4, Tip60) (1, 3). ARPs with known structures are Arp2 and Arp3 within the cytoplasmic Arp2/3 complex (4) and nuclear Arp4 (5). Nuclear ARPs have diverged further from the actin progenitor than Arp2/3, and all contain large insertions in the actin-like fold that have unknown functions (2). A regulatory role has been suggested for nuclear ARPs (6, 7), but this remains poorly understood. More recently, roles for nuclear ARPs besides chromatin remodeling have been suggested (8).

Arp8 contains a conserved domain of ∼630 amino acids that includes a core of ∼390 amino acids with homology to actin and three large insertions of unknown function (SI Appendix, Fig. S1). Arp8 is essential for activity of INO80, and deletion of Arp8 results in loss of INO80 function, with multiple effects on such processes as double-strand break repair (9), homologous recombination (10), and chromosome alignment (11). Yeast strains lacking Arp8 fail to recruit Arp4 and actin to the INO80 complex, suggesting associations among these proteins within the complex (12).

EM studies have been carried out on SWI/SNF, RSC, and ACF complexes (13–17). Although these studies have revealed the overall architecture of the complexes, the mechanism of remodeling remains unclear. Crystal structures have been determined in very few of the individual protein subunits of chromatin remodeling complexes, and none in complex with nucleosomes to provide the context of interactions within the remodeler.

The INO80 chromatin remodeling complex comprises a nine-protein core, conserved from yeast to humans, along with several species-specific proteins that associate with the core (18–21). In addition to the INO80 subunit, the core complex comprises actin, three actin-related proteins (Arp4, Arp5, and Arp8), Ies2, Ies6, and two AAA+ proteins (Rvb1 and Rvb2) related to the bacterial hexameric DNA translocating motor protein RuvB.

INO80 appears to play a significant role in remodeling nucleosomes in response to DNA damage (9, 22, 23). The Ies4 subunit of the yINO80 complex becomes phosphorylated by Tel1 (ATM) and Mec1 (ATR) kinases as a part of the DNA damage response (24). INO80 is also recruited to sites of DNA damage and interacts with nucleosomes that contain histone γ-H2AX, a signal for DNA repair processes (25, 26). A direct interaction between Arp4 of the INO80 complex and γ-H2AX has been demonstrated (27). Arp8 also has been shown to interact with free histone cores (12); however, little is known about the molecular details of these interactions, particularly in the context of nucleosomes.

Here we present the crystal structure of the conserved C-terminal domain (CTD) of yeast Arp8 (yArp8CTD) at 2.7-Å resolution and the complex with ADP at 3.25-Å resolution. The protein contains an actin-like core with additional inserted regions that extend across the surface of the protein. Our biochemical data show that the protein binds nucleotides but has undetectable ATPase activity. They also show that yArp8CTD binds to free histones H3, H4, and H3/H4 tetramers, but not to H2A or H2B. Furthermore, we identify a major contact region between the H3/H4 histones and one of the insertions within yArp8CTD. A 3D EM image reconstruction of the yArp8CTD–nucleosome complex is consistent with these biochemical data, but also reveals that two yArp8CTD molecules bind to a nucleosome, hinting at a role for a dimeric form of the protein. Although yArp8CTD is monomeric, recombinant full-length yArp8 exists in monomeric and dimeric states. A 3D EM image reconstruction of the yArp8 dimer shows a bilobal structure containing a deep cleft with the yArp8CTDs positioned on both sides of this cleft, consistent with the arrangement of yArp8CTD monomers in the complex with a nucleosome. Although full-length Arp8 binds to H3/H4 tetramers, a dimeric form of the protein is required to make a complex of sufficient stability to be detectable by gel filtration. Taken together, these data suggest the presence of two Arp8 molecules in the INO80 complex.

Results

Crystal Structure of the yArp8 C-Terminal Domain.

Several fungal species, including Saccharomyces cerevisiae, encode an Arp8 protein that is longer at the N terminus compared with most other organisms (SI Appendix, Fig. S1). Arp8 proteins from most organisms, including humans, correspond to the C-terminal region of the S. cerevisiae protein, so this protein is more representative of the general class of Arp8 proteins than the full-length yArp8. Consequently, a C-terminal fragment comprising residues 248–881 was expressed and purified from Escherichia coli (hereinafter called yArp8CTD). The crystal structure of yArp8CTD was determined at a resolution of 2.7 Å (Fig. 1 and SI Appendix, Table S1). The protein fold contains a core with homology to actin (24% sequence identity with yeast actin), and 354 Cα atoms of this core superimposed on the yeast actin structure, with an rmsd of 2.4 Å. The actin fold comprises four subdomains with clefts between them that bind nucleotides (28). The fold is conserved in the structures of Arp2, Arp3, and Arp4 proteins (4, 5). There are several insertions within the actin-like domain of yArp8, which we named based on their location within the actin fold. Insert 2A (residues 301–390) extends from subdomain 2 and crosses over to subdomain 4 before returning to subdomain 2, forming a flap that closes off the nucleotide-binding cleft. This is in the contact region with p34 in the Arp2/3 complex (4) and is the DNase I binding loop of actin (28). Insert 3A (residues 621–699) also forms an extended structure that crosses the boundary between subdomains 1 and 3, blocking access to the nucleotide-binding site from the opposite face of the protein to insert 2A. Arp4 also has an insert 3A, but this insert is largely disordered in the crystal structure (5). Insert 3B (residues 760–823) forms a pair of additional helices that protrude from the surface of subdomain 3.

Fig. 1.

Crystal structure and nucleotide binding of yArp8CTD. (A) Overall structure of yArp8CTD. The color scheme used matches that in SI Appendix, Fig. S1, which summarizes the secondary structure. The bound ADP is shown in magenta. (B) Isothermal titration calorimetry data for ATP binding to yArp8CTD. (C) The ADP-binding site in yArp8CTD. ADP and selected residues are shown as sticks, with the Fo–Fc difference density superimposed at the 3σ contour level. (D) ATP-binding site of yeast actin (PDB ID code 1YAG; magenta), with corresponding residues of yArp8CTD (yellow) superimposed.

Nucleotide Binding to yArp8CTD.

Nucleotide binding has been demonstrated directly for Arp1–Arp4 (5, 29–32). We used a fluorescent ATP analog, 2′,3′-O-(2,4,6-trinitrophenylcyclo-hexadienylidene)-ATP (TNP-ATP), to explore the ability of yArp8 to bind nucleotides. A fluorescence enhancement and blue shift were observed when yArp8CTD was added to TNP-ATP, indicative of binding to the protein (SI Appendix, Fig. S2). These effects were diminished when ATP was added to the cuvettete. Using isothermal titration calorimetry, we determined that yArp8CTD binds ATP with a binding constant of 0.54 ± 0.13 μM and a stoichiometry of 1 per protein molecule (Fig. 1). Two mutants of Arp8 (H272A and S275A) have been suggested to have ATP-binding defects (12). However, the S275A mutant binds ATP only 2.5-fold less tightly than WT, and the H272A mutant actually binds ATP 10-fold more tightly than WT (SI Appendix, Fig. S2), likely explaining their WT phenotype (12). Although we tested the ability of yArp8CTD to hydrolyze ATP, we were unable to detect significant ATPase activity.

To investigate the molecular details of interactions between Arp8 and nucleotide, we determined the structure of a complex of yArp8CTD with ADP at 3.25-Å resolution. ADP binds at a similar position to that observed for actin, within a cleft between the two domains (Fig. 1). The adenine moiety is located in a hydrophobic pocket, similar to actin and Arp4, although the identity of the residues is different (Fig. 1 and SI Appendix, Fig. S2). Furthermore, E563 in Arp8 replaces the function of R210 in actin by contacting the 2′-hydroxyl of the ribose ring. The sugar ring packs tightly against a conserved glycine residue in Arp8 that is conserved in actin, Arp2, Arp3, and Arp4 (G735 in Arp8; G302 in actin). The diphosphate tail of the nucleotide sits over the P2 loop in actin comprising residues 156–159 (GDGV), a structure that is conserved in Arp2, Arp3, and Arp4 but is GAAE in Arp8. In actin, a catalytic role has been proposed for H161 as the residue that polarizes a water molecule for attack of the γ-phosphate (33). This residue is conserved in actin, Arp2, and Arp3 but is S166 in Arp4 and R509 in Arp8. The apo protein and nucleotide-bound complex are superimposed with an rmsd of 0.31 Å over 623 Cα atoms.

yArp8CTD Binds to Histones.

Previous studies have shown that Arp4 and Arp8 interact with histones (12, 25, 27). Arp4 binds to all four core histones (33) but shows a marked preference for yeast H2A, particularly when it is phosphorylated (γ-H2A) (27). For Arp8, interaction with histones H3 and H4 within histone octamers has been demonstrated by pull-down experiments (12). To further analyze the interaction between yArp8CTD and histones, we performed pull-down experiments using GST-tagged yArp8CTD with histones (Fig. 2). These studies revealed interactions between yArp8CTD and histones H3, H4, and H3/H4 tetramers (Fig. 2). We then probed the details of these interactions using 10 residue peptide arrays spanning each of the yeast histones, probed with yArp8CTD (Fig. 2). Several specific regions of interaction with H3 and H4 histones were revealed. For histone H3, three strong interaction sites were located, centered around K56, L71, and the C-terminal region, and a fourth (weak) binding site was identified around F84. For histone H4, all of the interactions were weaker, but three peptides centered around L23, A39, and the C-terminal region yielded measurable signals. These data were mapped onto the crystal structure of the nucleosome (34) to identify an interaction surface (Fig. 2). To further characterize the interface between Arp8 and H3/H4, we prepared fragments spanning the Arp8 inserts 2A, 3A, and 3B as GST fusion proteins for pull-down experiments. These experiments revealed that insert 3A was able to interact with H3/H4 tetramers, but neither of the other inserts nor GST was able to interact (Fig. 2). GST-tagged insert 3A also was able to interact with free histones in a manner indistinguishable from GST-tagged yArp8CTD (SI Appendix, Fig. S3). Attempts to create deletion mutant yArp8CTD proteins in which any of the inserts were truncated resulted in proteins that were insoluble and/or incorrectly folded. Consequently, we cannot determine whether insert 3A is sufficient for the interaction, or whether other regions of yArp8CTD are also involved at the Arp8–nucleosome interface.

Fig. 2.

yArp8CTD interacts with histones. (A) Pull-down experiments for GST-yArp8CTD with histones H2A, H2B, H3, H4, and H3/H4. Each pair of lanes corresponds to GST control (Left) and GST-yArp8CTD (Right). (B) Decamer peptide arrays for histone H3 (Upper) and H4 (Lower) probed with yArp8CTD. For the H3 array, row L1 corresponds to 20 sequential peptides from residues 1–10 (A), then residues 2–11 (B), through to residues 20–29 (T). Each row down presents the next 20 peptides. The H4 array follows a similar pattern beginning at row L8 (A). (C) Data from B overlaid on the nucleosome structure. Histones H3 (yellow) and H4 (green) are highlighted in red to indicate positive peptides from the array data. (D) Pull-down experiments with GST-tagged inserts 2A, 3A, and 3B with H3/H4 tetramers. Lanes are as follows: GST, GST control; Arp8, GST-yArp8CTD plus H3/H4; 2A, GST insert 2A plus H3/H4; 3A, GST insert 3A plus H3/H4; 3B, GST insert 3B plus H3/H4; and H3/H4, H3/H4 alone.

Our attempts to demonstrate interactions between nucleosomes and Arp8 fragments were unsuccessful using the techniques described above. We rationalized that this was because the interactions were weaker, and thus we concentrated yArp8CTD with nucleosomes to raise their concentrations, thereby stabilizing even weak complexes, and then looked for direct evidence of complex formation by EM.

EM Reconstruction of the yArp8CTD/Nucleosome Complex.

To facilitate positioning of the nucleosome within the reconstruction, we used nucleosomes prepared using a 167-bp DNA fragment containing the strong positioning “Widom 601” sequence (35) (provided by D. Rhodes, Cambridge, UK) embedded within it (36). The extended ends allowed us to identify the location of the entry and exit points of the DNA in the nucleosomes. Our raw EM micrographs showed particles larger and clearly distinct from those of nucleosome alone, with approximately 80% particles of the complex and 20% free nucleosomes. From these micrographs, we obtained a 3D reconstruction at 21-Å resolution (Fig. 3 and SI Appendix, Fig. S4), which revealed an elongated shape when viewed from the top, with three distinct regions of density (Fig. 3). The central density forms a relatively flat disk flanked by regions of density on either side. Viewed from the end, the density flanking each side of the central disk has a shape similar to that of the yArp8CTD crystal structure; thus, we assigned this flanking density to yArp8CTD and the central disk to the nucleosome. The crystal structure of a nucleosome [PDB ID code 1KX5 (34), but with 10 extra residues of DNA at each end] could be fitted unambiguously into the central density (Fig. 3). The interactions between yArp8CTD and the nucleosome suggested by our model are consistent with our biochemical data (Fig. 2). Interestingly, although the overall shape of the central density accommodates the crystal structure of the nucleosome unambiguously, in the reconstruction there are cavities in some of the regions that correspond to the histones (arrow in Fig. 3). Furthermore, insert 3A of the crystal structure, shown to interact with H3/H4 (Fig. 2), is located outside of the density. However, additional, unaccounted for density is located on the exposed histone face of the nucleosome (red ellipse in Fig. 3) and may correspond to the location of insert 3A, consistent with biochemical data. Although we cannot rule out the possibility that these features are related to stain artifacts and/or limited resolution, it is also possible that changes have occurred within the complex; however, detailed analysis will require a higher-resolution reconstruction from cryo-EM.

Fig. 3.

EM 3D image reconstruction of the yArp8CTD–nucleosome complex. Bottom view (A) and side view (B) showing yArp8CTDs (magenta) binding to each face of the nucleosome (blue). The arrow indicates H3/H4 subunits that do not fit into the density. The red ellipse shows additional density that is unaccounted for by the crystal structures.

Biochemical Characterization of Full-Length yArp8 Reveals a Dimeric Species.

Full-length yArp8 protein was expressed and purified from E. coli (Fig. 4 and SI Appendix, Fig. S5). On heparin-Sepharose, the protein separated into two fractions. After gel filtration, protein that did not bind to the column gave molecular mass estimates of 95 kDa from both multiangle light scattering (MALS) and analytical ultracentrifugation (AUC) (Fig. 4 and SI Appendix, Materials and Methods and Fig. S6), consistent with the predicted molecular mass of full-length yArp8 (100 kDa), indicating that this form of the protein is monomeric. In contrast, the fraction of the yArp8 protein that was bound to heparin-Sepharose exhibited a broader peak on gel filtration. Both MALS and AUC analyses showed a mixture of species. MALS gave molecular mass estimates for the major peak at 231 kDa, with minor peaks at 108 and 446 kDa, whereas AUC showed a major peak at 217 kDa and smaller peaks at 380, 572, and 788 kDa (Fig. 4 and SI Appendix, Fig. S7). Notably, in each case the major species were 231 kDa (MALS) and 217 kDa (AUC), consistent with a dimer (molecular mass 200 kDa). Of note, the higher molecular mass forms were tetramers and hexamers, with no sign of trimers or pentamers, indicating (presumably) nonspecific association of dimers rather than a general aggregation of monomers.

Fig. 4.

Purification and analysis of full-length his-tagged yArp8 protein. (Upper) Elution profile from heparin-Sepharose column and SDS gel of wash (W) and eluate (E) fractions. (Lower) Analysis after gel filtration by MALS of protein fractions that passed through the column (wash, Left) or were bound and eluted with a salt gradient (elution, Right). MALS analysis of unbound protein gave a molecular mass estimate of 94.6 kDa, corresponding to monomers (100.2 kDa) of yArp8. Bound protein contained a more complex mixture, with MALS analysis showing a major peak at 231 kDa (dimer) and minor peaks at 108 kDa (monomer) and 446 kDa.

We separated fractions corresponding to the monomeric and dimeric forms of the protein, which appear to be stable and to undergo very slow exchange, for further analysis. Recent work suggested that recombinant yArp8 is monomeric (5). However, the purification protocol did not include a heparin-Sepharose step, and, consequently, the dimeric species (which is more proteolytically sensitive) would not have been separated early on, and so might have been lost at different stages of the purification in favor of the monomeric species.

EM Reconstruction of yArp8 Dimer.

We used negative-stain EM to obtain a 3D image reconstruction of the yArp8 dimer at 22-Å resolution. During processing, clear twofold symmetry was observed in eigenimages and class averages, so C2 symmetry was imposed on the reconstruction (SI Appendix, Fig. S7). The EM reconstruction reveals an elongated, bilobed molecule (Fig. 5). Looking from the side, each lobe forms a distinctive arc curving inward to form a saddle-like structure, with a deep cleft separating the two lobes. The density for each lobe forms a three-pronged fork shape that matches similar aspects of the crystal structure; thus, we assigned the lobes as the yArp8CTDs. This arrangement places insert 2A mainly facing outward, contributing to the outer contour of the arc. Insert 2A partially blocks the nucleotide-binding pocket and is located where other proteins have been shown to bind to actin (28, 37), raising the possibility that insert 2A could act as a protein-binding site and regulate the ATPase activity and/or nucleotide exchange of Arp8. Insert 3A does not fit into the molecular boundary, but is located adjacent to additional unaccounted-for density corresponding to the middle prong of the fork. Conformational variation of this region was also suggested in the yArp8CTD–nucleosome reconstruction (Fig. 3). Insert 3B consists of two helices that fit nicely into the third prong of the fork. A comparison of the EM reconstructions of the yArp8 dimer and the yArp8CTD monomers seen in the nucleosome complex reveals that the CTDs are well aligned with those in the dimer and would require only a small swing open and outward (of approximately 30 degrees) anchored at the junction between the CTD and the N-terminal domain to accommodate a nucleosome within the cleft (Fig. 5).

Fig. 5.

Negative-stain EM reconstruction of the yArp8 dimer. (A) View showing the channel flanked by the yArp8CTDs. The three inserts—2A (blue), 3A (green), and 3B (red)—line the channel walls. (B) “End” view; the yArp8CTD crystal structure closely fits the density, with the exception of insert 3A (green), which sits above and adjacent to unaccounted-for density (red ellipse). (C) Overlay of yArp8 dimer (cyan) with yArp8CTD–nucleosome (gray). (D) Enlarged view of density corresponding to yArp8CTD in the two reconstructions, showing the rotation required to open the cleft to accommodate a nucleosome. For clarity, density corresponding to the nucleosome is omitted.

Functional Role for an Arp8 Dimer in the INO80 Complex.

To investigate the biological relevance of dimerization, we tested whether the interaction between Arp8 and H3/H4 tetramers was sufficiently stable for analysis by gel filtration. We first examined the interaction of monomeric and dimeric full-length His-tagged yArp8 proteins with H3/H4 tetramers and analyzed the complexes by gel filtration on a Superdex S200 column (GE Healthcare) (Fig. 6). Dimeric full-length yArp8 formed a complex with H3/H4 tetramers that runs at high molecular weight. Stoichiometric quantities of H3/H4 tetramers were complexed completely with yArp8 under these conditions (SI Appendix, Fig. S8). Varying the quantity of H3/H4 tetramers with a fixed amount of yArp8 dimers showed that one yArp8 dimer was able to bind a single H3/H4 tetramer tightly, with excess H3/H4 running at the expected position. In contrast, monomeric yArp8 showed no interaction with H3/H4 tetramers, with discrete peaks observed for each species (Fig. 6).

Fig. 6.

Superdex S200 gel filtration of a molar excess of H3/H4 tetramers mixed with either monomeric or dimeric yArp8. Fractions from each gel filtration run were analyzed on SDS/PAGE gels, as shown below the trace.

In light of the apparent importance of dimers for the interaction of H3/H4 tetramers and Arp8, we compared the interaction of H3/H4 tetramers and either monomeric His-tagged yArp8CTD or yArp8CTD artificially dimerized by the addition of an N-terminal GST tag. The behavior of this complex mirrored that of the full-length proteins. Monomeric yArp8CTD showed no interaction with histones and ran at the location expected for a monomer, but the dimeric protein induced a shift in the position of H3/H4 tetramers, albeit in a more transient (weaker) interaction than with the natural protein dimmer, because the H3/H4 peak was broadened rather than shifted entirely (SI Appendix, Fig. S8). GST protein alone was unable to interact with histone tetramers (SI Appendix, Fig. S8). The fact that even artificially induced dimers of yArp8 formed a more stable complex with H3/H4 tetramers provides conclusive evidence of the importance of a dimer in stabilizing interactions with the H3/H4 tetramer.

Discussion

Although all ARPs contain homology with actin of the nuclear ARPs, only Arp4 has been shown to bind nucleotide (5, 33). ATPase activity has not been detected, however. Similarly, we found that Arp8 is also able to bind ATP, but with nonmeasurable ATPase activity. The ATP-binding site of Arp8 contains a number of substitutions compared with actin, including several residues known to be important for ATPase activity (33). Monomeric actin has a catalytic rate, k_cat, of 7 × 10⁻⁶ s⁻¹, but is stimulated more than 40,000-fold on filament formation (38). By analogy, because we are studying Arp8 out of the context of its protein partners in the INO80 complex, the ATPase rate probably should be expected to be low. This calls into question the role of ATP binding and/or hydrolysis in Arp8, and indeed in other ARPs and actin, in chromatin remodeling complexes. At present we do not have an answer to this question, but a role in regulation has been suggested by others (6, 7).

In vitro, Arp4 and Arp8 interact with core histones (12, 39). For Arp4, insert 3A was shown to be responsible for the interaction (39), and the interaction with the full-length protein in yeast has shown selectivity for γ-H2A (27). However, the INO80 complex lacking actin, Arp4, and Arp8 still appears to interact specifically with histones containing γ-H2AX (27), suggesting that there may be additional determinants within the INO80 complex contributing to histone specificity. For Arp8, pull-down experiments with nucleosome cores have shown that the H3/H4 component is retained but H2A and H2B are released, suggesting that the interaction between Arp8 and H3/H4 destabilizes histone cores (12). Consistent with this idea, we identified a tight interaction between yArp8CTD and H3/H4 proteins, but no interaction with H2A or H2B.

Interestingly, the interaction with H3/H4 tetramers was weaker in the context of an intact nucleosome, suggesting that the interaction is somehow less favorable. There are two possible explanations for this observation. One explanation is that the interaction interfaces between Arp8 and the histones are obscured when nucleosomes assemble, which is possible if Arp8 interacts with an exposed surface created during H2A/H2B histone dimer exchange. The interaction sites seen in the peptide arrays suggest one such (albeit weak) interaction between Arp8 and the C-terminal region of H4 at the interface with H2A, which is consistent with this interpretation. However, several other interaction sites are located at the DNA–histone interface, and full interaction with these sites appears restricted in the nucleosome compared with H3/H4 tetramers. Consequently, an alternative explanation is that interaction between Arp8 and histones induces a conformational change that disfavors binding to nucleosomes compared with H3/H4 tetramers. Such a conformational change associated with binding might include disruption of the DNA–histone interface such as that observed for RSC remodeling complex binding to nucleosomes (40). Although we want to avoid overinterpreting our low-resolution structural studies and biochemical data based on the interaction of a single subunit of INO80 with nucleosomes, our findings are at least consistent with such a proposal.

Importantly, the binding of yArp8 to H3/H4 tetramers is dependent on dimerization of Arp8, which can be mediated either through the N-terminal domain of the full-length yArp8 protein or via a GST dimer tagged to monomeric yArp8CTD. This is important, because the N-terminal dimerization interface is unique to Arp8 protein from S. cerevisiae and related fungi. However, although it has been shown that Arp8 (along with actin and Arp4) associates with the HSA domain of INO80 (7), the stoichiometry of Arp8 subunits in the complex has not been determined. The HSA domain of the RSC complex binds two separate ARPs (Arp7 and Arp9), which have been shown to form a heterodimer (7, 41). Indeed, many ARPs interact with one another or with actin, often in a pairwise fashion (4, 41). Arp8 is essential for recruitment of Arp4 and actin to the INO80 complex (12), but the mode of interaction with these proteins remains unknown. Our EM data reveal that the actin-like domains of the yArp8 dimer do not interact with one another in the same way as Arp2/3 or actin monomers within a filament; however, this does not preclude canonical interactions between Arp8 and actin/Arp4 or Arp5 within the INO80 complex.

Dimerization of yArp8 suggests the presence of two Arp8 molecules in the yINO80 complex and raises questions about the stoichiometry of other proteins. The only previous study that addressed this question used gel scan analysis of INO80 complex prepared from tagged yINO80 pulldowns (18), which suggested a 6:1 stoichiometry of the Rvb1 and Rvb2 proteins compared with actin and the INO80 subunits. However, this technique is not sufficiently quantitative to define the stoichiometry of subunits within the complex, and further investigation is needed to resolve this question. Although hArp8 lacks the N-terminal dimerization domain, and the oligomeric structure of the protein is unknown, it seems unlikely that the stoichiometry of the complex will differ between human and yeast, suggesting that there may also be two Arp8 molecules in the hINO80 complex. Nucleosomes contain a pseudomolecular dyad axis, and a doubling-up of at least some of the proteins within the INO80 complex could give rise to a molecular symmetry that is exploited when binding nucleosomes.

The present study provides an initial step toward understanding the interactions within the INO80–nucleosome complex. Our findings begin to address some of the basic functions of individual subunits within INO80 and provide the basis for future high-resolution studies of ARPs and other components of chromatin remodeling complexes that will shed light on the mechanics of these complex, multisubunit machines.

Materials and Methods

Cloning, Expression, and Purification of S. cerevisiae Arp8 Proteins.

All proteins were expressed and purified from E. coli cell extracts. yArp8CTD (residues 248–881) was purified in three steps using Ni-NTA, Mono-Q, and Superdex S200 resins (GE Healthcare). For yArp8, an additional heparin-Sepharose step was included between the Ni-NTA and Mono-Q resins.

Nucleotide Binding.

The fluorescent ATP analog TNP-ATP was used as described previously (42). For isothermal titration calorimetry measurements, a series of 5-μL injections of ATP solution (180–300 μM) were made into a cell containing 1.6 mL of protein solution equilibrated at 25 °C in an isothermal titration calorimeter (model VP-ITC; MicroCal). Data analysis was performed with the software originally supplied with this instrument.

Crystallization of yArp8CTD, Data Collection, and Structure Determination.

Crystals were obtained at 12 °C and protein concentrations of 13–16 mg/mL after 3–4 d from 17–20% wt/vol PEG 8000 or PEG 5000 monomethyl ether, 0.1 M buffer (Mes, Na cacodylate, or ADA; pH 6.5), and 0.2–0.3 M Li₂SO₄. X-ray diffraction data were collected at the Diamond Light Source (Didcot, UK), processed with MOSFLM (43), and scaled with SCALA (44). The structure was solved by molecular replacement using the actin component of yeast actin–gelsolin segment-1 complex (PDB ID code 1YAG), together with a multi-wavelength anomalous dataset on a single selenomethionine-substituted crystal with an iridium derivative dataset. The yArp8CTD model was built using Coot (45), and refined with REFMAC (46) and CNS (47). The resulting model was used for refinement of the ADP-bound structure. Data processing and refinement statistics are provided in SI Appendix, Table S1. Coordinates and structure factors have been deposited in the Protein Data Bank (PDB ID codes 4AM6 and 4AM7).

EM.

yArp8 or yArp8CTD–nucleosome complex was applied to glow-discharged continuous carbon grids, blotted, and stained with 2% (wt/vol) uranyl acetate. Data were collected at 50,000× magnification using a Philips CM200 FEG electron microscope operating at 200 kV under low-dose conditions. Particles were picked interactively using the boxer program in EMAN (48). All subsequent processing was completed in Imagic-5 (49), except where noted otherwise in SI Appendix, Materials and Methods.