An Actin-Filament-Binding Interface on the Arp2/3 Complex Is Critical for Nucleation and Branch Stability

An actin-filament-binding interface on the Arp2/3 complex is critical for nucleation and branch stability

Erin D. Goleya,2, Aravind Rammohanb,3, Elizabeth A. Znameroskia, Elif Nur Firat-Karalara, David Septb,4, and Matthew D. Welcha,1

aDepartment of Molecular and Cell Biology, University of California, Berkeley CA 94720; and bDepartment of Biomedical Engineering and Center for Computational Biology, Washington University, St. Louis, MO 63130

Edited* by Thomas D. Pollard, Yale University, New Haven, CT, and approved February 24, 2010 (received for review October 8, 2009)

The Arp2/3 complex polymerizes new actin filaments from the branched filament geometry is proposed to be particularly suited sides of existing filaments, forming Y-branched networks that for harnessing actin polymerization to generate motile force (11). are critical for actin-mediated force generation. Binding of the Atomic-resolution structures of the Arp2/3 complex with and Arp2/3 complex to the sides of actin filaments is therefore central without bound nucleotide and inhibitors have been determined to its actin-nucleating and branching activities. Although a model (12–15). Moreover, structural models of the Y-branch junction of the Arp2/3 complex in filament branches has been proposed have been constructed using electron microscopy (16, 17), culmi- based on electron microscopy, this model has not been validated nating in a 2.6-nm resolution 3D model derived from docking using independent approaches, and the functional importance of crystal structures of Arp2/3 complex and actin into a reconstruc- predicted actin-binding residues has not been extensively tested. tion from electron tomography (18). In this model, Arp2 and Using a combination of molecular dynamics and protein-protein Arp3 interact with the pointed end of the daughter filament, docking simulations, we derived an independent structural model and all seven subunits contact the mother filament. ARPC2 of the interaction between two subunits of the Arp2/3 complex and ARPC4 comprise the major mother-filament-binding inter- that are key to actin binding, ARPC2 and ARPC4, and the side of face, consistent with earlier data from chemical cross-linking (19), an actin filament. This model agreed remarkably well with the pre- Arp2/3 complex reconstitution (20), and antibody-inhibition experiments (21). However, despite advances in our understanding vious results from electron microscopy. Complementary mutagen- BIOCHEMISTRY esis experiments revealed numerous residues in ARPC2 and ARPC4 of Y-branch structure, the functional importance of Arp2/3 that were required for the biochemical activity of the entire com- complex residues implicated in mother-filament binding has plex. Functionally critical residues clustered together and defined a not been extensively tested apart from an analysis of ARPC2 arc35 Saccharomyces cerevisiae surface that was predicted by protein-protein docking to be buried ( ) mutants in , which demonstrated in the interaction with actin. Moreover, key residues at this inter- an important role for ARPC2 residues in Arp2/3 complex nucle- face were crucial for actin nucleation and Y-branching, high-affinity ating activity in vitro and in growth and endocytosis in vivo (22). F-actin binding, and Y-branch stability, demonstrating that the In this study we used molecular dynamics and protein-protein affinity of Arp2/3 complex for F actin independently modulates docking simulations to generate an independent model of the interaction between the ARPC2 and ARPC4 subunits of the branch formation and stability. Our results highlight the utility Arp2/3 complex and the side of an actin filament. Using informa- of combining computational and experimental approaches to tion from this and previous models, we tested the role of amino study protein-protein interactions and provide a basis for further acid residues on the exposed surfaces of ARPC2 and ARPC4 by elucidating the role of F-actin binding in Arp2/3 complex activation examining the biochemical properties of mutant Arp2/3 com- and function. plexes. Using this approach we defined key residues that play a specific and critical role in F-actin binding, actin nucleation, and cytoskeleton ∣ actin branching Y-branch stability. he actin cytoskeleton plays an essential role in diverse cellular Results Tprocesses ranging from motility to division. A key control Protein-Protein Docking Simulations Yield an Independent Structural point in the cycle of actin filament (F actin) assembly is the Model of ARPC2/ARPC4 Bound to F Actin. Numerous lines of evidence rate-limiting nucleation step, which can be accelerated in a regu- suggest that the ARPC2 and ARPC4 subunits of the Arp2/3 lated manner by the action of nucleating factors. One of the complex constitute the primary F-actin side-binding interface major actin-nucleating factors in cells is the Arp2/3 complex, a (19–21). To generate an independent structural model of the in- protein complex that consists of seven subunits including the teraction between these proteins, we performed protein-protein actin-related proteins (Arp) Arp2 and Arp3 and the additional docking simulations using the crystal structures of an ARPC2/ Arp2/3 complex (ARPC) polypeptides ARPC1–ARPC5. The ARPC4 heterodimer (12) and an actin filament consisting of Arp2/3 complex has been conserved during the evolution of most eukaryotic cells and plays an important functional role in cell mi- gration, endocytosis, phagocytosis, and pathogen infection (1). Author contributions: E.D.G., A.R., D.S., and M.D.W. designed research; E.D.G., A.R., E.A.Z., E.N.F.-K., and D.S. performed research; E.D.G., D.S., and M.D.W. analyzed data; and On its own the Arp2/3 complex is inactive, but it is activated to E.D.G., D.S., and M.D.W. wrote the paper. polymerize actin by binding to proteins called nucleation-promot- The authors declare no conflict of interest. ing factors (NPFs) (1) as well as to ATP (2, 3). Moreover, the *This Direct Submission article had a prearranged editor. nucleating activity of the Arp2/3 complex is stimulated by binding 1To whom correspondence should be addressed. E-mail: [email protected]. to F actin (4, 5), a phenomenon that results in autocatalytic actin 2Present address: Department of Developmental Biology, Stanford University School of assembly (6). Once activated, the complex nucleates the polymer- Medicine, Stanford, CA 94305. ization of daughter filaments that emerge from the sides of 3Present address: Corning Incorporated, Modeling & Simulation, Corning, NY 14831. mother filaments in a stereotypical Y-branch orientation with 4Present address: Department of Biomedical Engineering and Center for Computational an approximate branch angle of 70° (7–9). Such Arp2/3-contain- Medicine and Bioinformatics, University of Michigan, Ann Arbor, MI 48109. ing branched structures have been observed in the actin network This article contains supporting information online at www.pnas.org/cgi/content/full/ within lamellipodia at the leading edge of motile cells (10). This 0911668107/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.0911668107 PNAS Early Edition ∣ 1of6 Downloaded by guest on October 4, 2021 eight monomers corresponding to the Holmes F-actin model A actin 1 (23). To capture the conformational flexibility of the two docking actin 2 partners, we used Nanoscale Molecular Dynamics (NAMD) (24) R190 to perform a 100-ns molecular dynamics simulation of the full K58 Arp2/3 complex and a 30-ns simulation of the actin filament. D80 Structures were extracted every 10 ns for the ARPC2/ARPC4 E100 D81 heterodimer and every 5 ns for F actin, resulting in 10 and 6 R95 structures, respectively. For each combination of structures, we performed 100 docking simulations using Rosetta-Dock software R158 (25, 26), resulting in a total of 6,000 docked structures. D143 Using the best-scoring model from this initial protein-protein ARPC4 ARPC2 docking run as a starting point, we carried out a perturbation run actin 3 in which the two docking partners were subjected to an additional 10,000 rounds of random translation/rotation and redocking. B ARPC1 When we examined the resulting energy landscape using the final ARPC5 best-scoring structure as a reference, we observed a funnel of Arp2 docking scores converging on the lowest-scoring structure at 0 Å rmsd (Fig. S1). Score funnels such as this one were found 90 to be a strong indicator of the robustness of the docking model, ARPC4 as the low-scoring model in such funnels often corresponded with atomic-level accuracy to the structure determined experimentally by x-ray crystallography in the Critical Assessment of Predicted Interactions (CAPRI) protein-protein docking experiment (27). ARPC3 Thus, the model with the lowest docking score from the pertur- ARPC2 Arp3 bation run, shown in Fig. 1A and B and Movie S1, represents a robust computationally derived model of the ARPC2/ARPC4-F- C actin-binding interaction. Strikingly, when we overlaid our model of the ARPC2/ 90 90 ARPC4-F-actin interaction on the one derived from electron microscopy (Movie S1), we found they were very similar, with aCα rmsd of only 5.9 Å. The similarity between these models provides independent support for the results from electron mi- Arp2 Arp3 ARPC1 ARPC2 ARPC3 ARPC4 ARPC5 croscopy studies. Moreover, the robustness of the protein-protein docking model is highlighted by the fact that it was derived Fig. 1. Protein-protein docking simulations identify a putative mother independently of information from electron microscopy. filament binding site on the Arp2/3 complex. (A) Surface rendering of the Encouraged by the similarity of our protein-protein docking structures of an ARPC2/ARPC4 heterodimer and a portion of the actin fila- model with models from electron microscopy, we identified ment, with residues predicted to form salt bridges highlighted in red (acidic) the interaction surface on ARPC2/ARPC4 that is predicted to and blue (basic), and the surface predicted to be within 4 Å of the other bind- lie within 5 Å of F actin based on the protein-protein docking ing partner in yellow, based on molecular dynamics and protein-protein docking simulations. Each structure is rotated 90° away from the other to analysis. This interface resides primarily on ARPC4 (residues show the binding interfaces. (B) Structures of the whole Arp2/3 complex with M1, A3, T4, L5, R55, N56, E57, K58, E59, K77, Q78, A79, the predicted mother filament binding interface highlighted in yellow. D80, E81, E83, D143, K144, S147, K150, L151, S152, N154, (C) Structure of the inactive Arp2/3 complex (12) docked onto a mother A155, R158, I159, E162, E163, and K166) but extends along filament in the orientation dictated by the position of ARPC2/ARPC4 in ARPC2 (residues D159, R160, E187, R189, R190, A191, the top-scoring model from docking simulations. H193, F228, Y261, and R265) at the junction between the two proteins. The surface on F actin primarily resides on subdomains virus expression system and purified to homogeneity. None of 1 and 2 of a single monomer (residues E2, D3, E4, T5, Q41, M44, the mutations had an effect on the stability or assembly of the Q49, K50, D51, S52, K68, N78, D80, D81, K84, H87, H88, Y91, complex, as each purified complex was obtained with similar yield N92, R95, A97, E99, E100, H101, F127, and N128 in rabbit mus- and had the appropriate subunit stoichiometry. cle α-actin). Most of the residues on ARPC2/ARPC4 predicted We assessed each purified complex for its ability to promote by protein-protein docking to lie within 5 Å of actin are also actin polymerization using the pyrene-actin assembly assay predicted to lie within 5 Å of actin by electron microscopy (Fig. 3). Because Arp2/3-mediated actin nucleation is autocata- (18) (Table S1). Interestingly, some of these residues were also lytic, we expected that mutations affecting F-actin binding would proposed to be involved in mother-filament binding by a homol- decrease activity. Of the 11 mutants tested, 8 exhibited reduced ogy modeling study that identified highly conserved amino acids activity and 3 exhibited near-wild-type activity (Fig. 3A and within the Arp2/3 complex (28) (Fig. 2A and Table S1). Together, Table S2). To quantify these defects, we used the polymerization the protein-protein docking, electron microscopy, and homology data to calculate the concentration of actin filament barbed ends modeling approaches made complementary and testable predic- generated by each mutant complex as a function of Arp2/3 tions about the potential importance of specific Arp2/3 complex concentration (Fig. 3B). Mutations causing reduced activity surface residues in its nucleation and actin-binding activities. diminished the production of barbed ends from 2-fold (2DR, 4RE) to 12-fold (4DKK) (Table S2). These data indicate that Charged Surface Residues on ARPC2 and ARPC4 Are Critical for Arp2/3 numerous charged residues on the surface of ARPC2 and Complex Activity. We set out to test the functional importance of ARPC4 play a critical role in Arp2/3 complex function. charged surface residues on ARPC2 and ARPC4 by mutating We next examined the relationship between the location of clusters of these residues to alanine. We generated a total of each mutation on the surface of ARPC2/ARPC4 and the severity 11 mutant complexes that contained substitutions in 21 different of the biochemical defect. Strikingly, when the mutations were amino acid residues (Table S2 and Figs. S2 and S3). Each of the mapped onto the surface, there was a clear spatial clustering mutant complexes was expressed in insect cells using the baculo- by activity level (Fig. 2B). Mutations that caused moderate to

2of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0911668107 Goley et al. Downloaded by guest on October 4, 2021 A ARPC4 A B

1.00 actin 5 WT molecular docking 0.75 WT 4 4RED homology modeling 4RED 2K4E 0.50 2K4E 3 2EE both 2 0.25 2EE 1 0.00 0 0 250 500 750 1000 [barbed ends] (nM) 0 255075100 polymerization (arbitrary) time (sec) Arp2/3 (nM)

1.00 actin 5 WT WT 0.75 4 2DR 2DR 4RE 0.50 4RE 3 2 0.25 1 0.00

[barbed ends] (nM) 0 0 250 500 750 1000 0 255075100 ARPC2 polymerization (arbitrary) time (sec) Arp2/3 (nM)

B ARPC4 1.00 actin 5 WT E59 WT 0.75 4 2EER severe defect 2EER 2ER R55 2ER 3 K295 0.50 4KD moderate defect 4KD K58 2 2E4K little defect 0.25 2E4K K84 4K3 1 4K3 R6 0.00 4DKK 0 4DKK K77 E184 0 250 500 750 1000 [barbed ends] (nM) 0 255075100 D80 polymerization (arbitrary) E204 time (sec) Arp2/3 (nM) Fig. 3. Mutations in charged surface residues on ARPC2 and ARPC4 cause a E140 range of actin nucleation defects. (A) Pyrene-actin polymerization assays D143 E208 with wild-type (WT) and mutant Arp2/3 complexes. Reactions contain BIOCHEMISTRY K144 E148 μ D159 3 M actin, either alone (actin) or with 20 nM WT or the indicated mutant E145 K150 complexes (see Table S2 for abbreviations) and 200 nM GST-WASP-WCA. E187 Mutant complexes with similar activity levels are grouped together and R190 R160 colored according to the scheme in Fig. 2B.(B) Concentration of barbed K166 ends generated by WT and mutant Arp2/3 complexes as a function of Arp2/3 concentration.

ARPC2 induced conformational changes were observed for mutants with the most (4DKK) and least (4RED) severe nucleation defects Fig. 2. Residues on ARPC2 and ARPC4 identified by molecular docking and (Fig. 4A), suggesting that deficiencies in nucleating activity are homology modeling lie on a surface that is critical for Arp2/3 complex activity. not due to a failure to bind nucleotide or NPF. Next, we used (A) Surface rendering showing the location of amino acid residues selected cosedimentation at a range of F-actin concentrations to measure for mutagenesis (opaque and color coded) based on molecular docking B (blue), homology modeling (green), or both (magenta). (B) Surface rendering binding to actin filaments (Fig. 4 ). The affinity of the severely showing the correlation between the location of mutations on the surface of defective 4DKK mutant for F actin (Kd ¼ 7.8 3.7 μM; mean ARPC2/ARPC4 and the severity of the actin nucleation defects. The surface standard error of the mean) was 6 times lower than wild type that is predicted to be within 4 Å of actin by protein-protein docking is (Kd ¼ 1.3 0.6 μM). This strongly suggests that the reduced opaque and yellow, and the side chains of mutated amino acids are shown polymerizing activity of 4DKK, and likely the other mutants, is in space filling representation and are color coded as follows: orange, due to a reduced affinity for actin filaments. severely defective; purple, moderately defective; green, unaffected. Finally, we compared the activity of the wild-type complex and the 4DKK mutant with regard to the formation and subsequent severe biochemical defects clustered on the predicted binding stability of Y branches. After controlling for the 12-fold differ- interface with F actin, and the least severe mutations scattered ence in actin-nucleating activity, wild type and 4DKK formed at the periphery or outside the predicted interface. Notably, 12 a similar percentage of branched filaments (10 nM wild type of the 14 mutations that caused moderate to severe defects formed 15 3% branched filaments (mean SEM); 120 nM are predicted to be within 5 Å of the mother filament, whereas 4DKK formed 13 1% branched filaments), and branch mor- only 1 of 7 mutations that caused little defect are within close phology was similar between wild type and mutant (Fig. 4C). This proximity to actin (TableS1). Thus the locations of mutations that suggests that the lower affinity of the 4DKK mutant results in influence actin polymerization support the electron microscopy lower nucleation and Y-branching activity, but nevertheless its and protein-protein docking models for the interaction of nucleating and Y-branching activities remain tightly coupled as ARPC2 and ARPC4 with the mother actin filament. has been shown previously for wild-type Arp2/3 complex (4–6). In addition to measuring Y-branch formation, we also measured A Surface on ARPC2/ARPC4 Is Critical for F-Actin Binding and Y-Branch Y-branch stability by initiating nucleation/branching reactions, Stability. We hypothesized that the mutations in residues in fixing branches with rhodamine-phalloidin at various times post ARPC2 and ARPC4 that compromised actin polymerization initiation, and quantifying branch frequency as a function of time activity did so by reducing the affinity for F actin, but not by per- (Fig. 4D). Wild-type branches dissociated with a t1∕2 (half life) of turbing the interaction of the Arp2/3 complex with other activa- 28 min, similar to previous reports (30). In striking contrast, tors such as ATP or NPFs. To test this, we first used wild-type branches formed by the 4DKK mutant dissociated with a much and mutant Arp2/3-FRETcomplexes (containing ARPC1 tagged shorter t1∕2 of <10 min. This indicates that residues within the with YFP and ARPC3 tagged with CFP) (29) as conformational ARPC2/ARPC4 heterodimer that affect the affinity of the sensors of ATP and NPF binding. No defects in ATP- or NPF- Arp2/3 complex for F actin are crucial for Y-branch stability.

Goley et al. PNAS Early Edition ∣ 3of6 Downloaded by guest on October 4, 2021 ABphenotype, protein-protein docking predicted that 12 were at the F-actin-binding interface, compared with 13 for electron 1.30 WT: Kd = 1.3 ± 0.6 µM 1.25 100 microscopy, and 9 for homology modeling. Conversely, of the 7 1.20 80 residues for which mutations caused little phenotype, protein- 1.15 60 protein docking predicted that 6 were not at the interface, com- 1.10 40 1.05 pared with 4 for electron microscopy, and 3 for homology model- 20 1.00 % Arp2/3 bound 4DKK: K = 7.8 ± 3.7 µM ing. Thus all of these approaches have predictive power, and each 0 d normalized FRET ratio 0 2.5 5 7.5 10 12.5 15 arrives at a complementary solution. The combination of these ++ CA ++ CA ++ CA Mg Mg Mg [F actin] (µM) methods provides a larger measure of confidence and makes a Mg-ATP Mg-ATP Mg-ATP broader set of predictions that can be tested experimentally. WT 4DKK 4RED It is important to note that the results of our mutagenesis CDexperiments agree with a previous analysis of the phenotypes caused by mutating evolutionarily conserved and solvent-exposed 5 nM WT residues in S. cerevisiae ARPC2 (22). In particular, mutating WT 1.00 10 nM WT several residues that are conserved between yeast and human 0.75 60 nM 4DKK 120 nM 4DKK ARPC2 caused correspondingly mild (yeast arc35-107; human 0.50 2EE; E204A), moderate (yeast arc35-104; human 2DR; D159A 0.25 4DKK R160A), and severe (yeast arc35-106; human 2ER, 2EER; 0 E187A R190A) defects in actin nucleation. Our results demon- 010 2030405060 strate that the functional roles of these residues are conserved time (min) across species and extend this analysis to examine the role of fraction of initial branch frequency of initial branch fraction adjacent residues in ARPC4. Fig. 4. Mutant Arp2/3 complexes undergo expected conformational Detailed examination of the biochemical activity of a severely changes on activator binding, but fail to bind F actin with high affinity or defective mutant in ARPC4 (4DKK) suggests that this surface of form long-lasting branches. (A) Box and whiskers plot of normalized FRET/ CFP ratio of WT, 4DKK, and 4RED Arp2/3-FRET complexes in the presence ARPC4 (and by extension the adjacent surface on ARPC2) is of Mg2þ, Mg-ATP, or Mg-ATP and GST-WASP-CA. The middle line of each crucial for actin nucleation and high-affinity binding to F actin, box indicates the median, and the top and bottom lines represent the third consistent with the observations that F-actin binding plays a and first quartiles (n ¼ 6). Whiskers indicate maximum and minimum mea- key role in activating the complex (4–6). Although the precise surements. (B) Percentage of WT (black) or 4DKK (orange) Arp2/3 complex mechanism of Arp2/3 complex activation by F actin remains found in the pellet fraction after high-speed centrifugation over a range unclear, mathematical simulations suggest that an activation re- of actin concentrations. Dissociation constants (Kd s) calculated from the reaction occurs after actin binding and that this is the rate-limiting sulting curves are indicated. Data are the mean SEM (n ¼ 6). (C) Images of μ D step leading to branch formation (31). In addition to their role in branches formed by WT and the 4DKK complexes. Scale bar 2 m. ( ) Graph promoting high-affinity actin binding, residues on the surface of of the fraction of the initial branching frequency (normalized such that the branch frequency at t ¼ 0 is 1) vs. time after initiating nucleation/branching ARPC2/ARPC4 may also participate in this reaction. Once for WT (black) and 4DKK (orange). Data are the mean SEM (n ¼ 3). activation occurs, our data suggest that Y branches form with nor- mal geometry even when these residues are mutated. However, Discussion our results also indicate that residues in ARPC2/ARPC4 are cru- By employing a combination of molecular dynamics and protein- cial for maintaining the stability of the Y branch, as the 4DKK protein docking simulations, we derived an independent structur- mutant undergoes much more rapid branch dissociation than the al model of the interaction between the ARPC2 and ARPC4 wild-type Arp2/3 complex. Together these data suggest that resi- subunits of the Arp2/3 complex and the side of the mother dues on this surface of ARPC2/ARPC4 are crucial for F-actin binding and that the affinity for F actin independently affects both filament. Our model corresponds remarkably well to previous Arp2/3 complex activation and Y-branch stabilization. models of the Arp2/3 complex in the Y branch derived from In addition to the key role played by ARPC2 and ARPC4, re- electron microscopy (16–18). Together these models provide can- sults from electron microscopy suggest that each of the remaining didate contact sites between the binding partners. By mutating five subunits makes contact with the mother filament (18), indi- residues within the predicted actin-binding surface on ARPC2/ cating that other interactions are also likely to be relevant in the ARPC4, we defined sites on this surface that are required for intact complex. When we added the structures of the remaining high-affinity binding to F actin, efficient actin nucleation and subunits (12) to the model of ARPC2/ARPC4 docked on the Y branching, and stability of Y branches. mother filament to generate a model of the Arp2/3 complex Our results suggest that the F-actin-binding region on ARPC2/ bound to F actin (Fig. 1C), we found that ARPC5 is very close ARPC4 occupies much of the exposed surface of ARPC4 and to the filament (4 Å), whereas other subunits are more distant extends onto ARPC2 near the interface between the two (Arp2, 23 Å; ARPC1, 11 Å; ARPC3, 22 Å). However, the crystal proteins. The surface is very similar to that obtained by fitting structure we used for subunit placement is presumed to be an a crystal structure of the Arp2/3 complex into the density of inactive conformation, and significant conformational changes Y-branch junctions observed by electron tomography (18) and are thought to occur during activation (12) that bring other encompasses some of the evolutionarily conserved residues that subunits into closer proximity with F actin (18). Insight into the were suggested to be involved in F-actin binding by homology potential impetus for these conformational changes comes from modeling (28). Mutating residues that span this surface caused our observation that, when all of the subunits are added to the moderate to severe defects in actin polymerization, supporting docking model, there is a steric clash between Arp3 (subdomain the notion that this surface is critical for Arp2/3 complex activity. 2) and F actin. This suggests that binding of the inactive Arp2/3 When we compared the relative success of protein-protein complex to F actin may result in a repositioning of Arp3, leading docking (this study), electron microscopy (18), and homology to the large-scale rearrangements in the complex that have been modeling (28) in predicting functionally critical residues on this suggested to accompany the transition between the inactive con- surface of ARPC2/ARPC4, we found that protein-protein formation and the structure in the branch junction (12, 18). docking and electron microscopy were comparably successful, Our results also suggest that the ARPC2/ARPC4 binding and both were more successful than homology modeling. Of interface on F actin primarily encompasses a surface on subdo- the 14 residues for which mutation caused a moderate/severe mains 1 and 2 of a single actin monomer, similar to the binding

4of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0911668107 Goley et al. Downloaded by guest on October 4, 2021 surface predicted by electron microscopy (18). Importantly, our has been observed in control dockings with native complexes protein-protein docking model likely represents an encounter (25–27). The best-scoring structure from this perturbation analy- complex, as structural rearrangements in the mother filament be- sis was used for all subsequent work. yond what was sampled in our simulations are seen by electron microscopy (18). Interestingly, one of the residues in actin (R95) Baculovirus Strain Construction. Baculovirus strains expressing that was predicted by our analysis to form a salt bridge with untagged recombinant human Arp2/3 subunits, as well as strains ARPC2/ARPC4 was also implicated in binding to tropomyosin expressing ARPC1-Flag-His6, ARPC3-CFP, and ARPC1-YFP based on chemical modification studies (32), consistent with were generated as described previously (20, 29). Point mutations the observation from electron microscopy that the binding sites in ARPC2 and ARPC4 were made using the QuikChange Site- on F actin for ARPC2/ARPC4 and tropomyosin overlap (18). Directed Mutagenesis Kit (Stratagene) following the manufac- This offers an explanation for the findings that tropomyosin in- turer’s protocol. The DNA sequence of each construct was hibits actin nucleation and branching by the Arp2/3 complex in verified, and then baculovirus strains were prepared according to vitro (33) and that tropomyosin and Arp2/3 complex localize to the procedures supplied with the Bac-to-Bac system (Invitrogen). distinct zones of the actin network in cellular lamellipodia (34). The success of modeling a structure of the ARPC2 and Protein Expression and Purification. Recombinant Arp2/3 com- ARPC4 subunits bound to F actin highlights the utility of complexes were expressed by infecting Hi5 cells with baculovirus bining molecular dynamics simulations, protein-protein docking strains expressing all Arp2/3 subunits including ARPC1-Flag- simulations, and biochemical analyses to model protein-protein His6, as described previously (29). Arp2/3 complexes were puri- interactions. Future modeling efforts will require information fied by Ni-NTA chromatography (QIAGEN) followed by cation from additional electron microscopy and x-ray crystallography exchange chromatography on HiTrap SP (GE Biosciences) for studies aimed at determining both lower and higher resolution most complexes, or by anion exchange chromatography on structures of Arp2/3 complex in various conformations and HiTrap Q (GE Biosciences) for those complexes containing bound to its activators. The structural predictions from this ARPC1-YFP. Ion exchange chromatography was followed by and future studies will be invaluable for guiding investigation into gel filtration (Superdex 200, GE Biosciences) into gel filtration both the mechanism by which F-actin binding promotes Arp2/3 buffer [20 mM 3-(N-morpholino)propanesulfonic acid (MOPS) complex activation and the functional importance of Y branching pH 7.0, 100 mM KCl, 2 mM MgCl2, 5 mM EGTA, 1 mM EDTA, by Arp2/3 complex in cells. 0.5 mM DTT, 0.2 mM ATP, and 10% glycerol]. When purifying

complexes for use in FRETexperiments, ATP,EGTA, EDTA, and BIOCHEMISTRY Experimental Procedures MgCl2 were omitted from gel filtration buffer (FRET gel filtra- Molecular Dynamics and Docking Simulations. The crystal coordi- tion buffer) and from buffers used in other chromatography steps. nates for Arp2/3 complex (12) and ADP-actin (35) were obtained GST fusions of the Wiskott-Aldrich syndrome protein (WASP) from the Protein Data Bank (PDB ID codes 1K8K and 1J6Z, WASP homology 2, connector, and acidic region (GST-WASP- respectively). The Arp2/3 complex structure was missing a consid- WCA, residues 422–502 of human WASP) and the connector erable portion of Arp2, and smaller portions of the other six and acidic region (GST-WASP-CA, residues 441–502) were ex- subunits. The missing portion of Arp2 was rebuilt and the pressed in Escherichia coli BL21 (DE3) as described previously nucleotides were added to Arp2 and Arp3 using Protein Local (38). These proteins were purified by glutathione affinity chroma- Optimization Program (PLOP) (36). The structure of the actin tography (glutathione sepharose 4B, GE Biosciences) and then filament was constructed by replacing monomers in the Holmes by gel filtration chromatography (Superdex 200, GE Biosciences) model (23) with the ADP-actin crystal structure. Using the mole- into FRET gel filtration buffer. Protein was frozen in liquid cular dynamics package NAMD (37), simulations of Arp2/3 com- nitrogen and stored at −80 °C. plex and actin filaments were performed using the CHARMM27 force field, an NpTensemble at 1 atm pressure, a temperature of Arp2/3 Complex Activity Assays. Rabbit skeletal muscle actin (39) 300 K, a 10-Å cutoff for van der Waals interactions with a switch- and pyrene-labeled actin (40) were prepared as described ing distance of 8.5 Å, and Particle Mesh Ewald for long-range elsewhere. Pyrene-actin polymerization assays were performed electrostatics. Covalent bonds with hydrogens were held rigid and barbed end concentrations were calculated as described allowing us to take 2-fs time steps. Following equilibration runs previously (20) with the following modifications. Pyrene-actin of 10–20 ns, the production F-actin simulation was 30 ns and and unlabeled actin were mixed in G buffer (5 mM Tris pH the Arp2/3 complex simulation was 100 ns long. Structures were 8.0, 0.2 mM CaCl2, 0.2 mM ATP, 0.2 mM DTT) to generate a extracted every 5 ns for F actin and every 10 ns for the ARPC2/ 3-μM G-actin solution with 7% pyrene-actin. A 14-μL gel filtra- ARPC4 complex, resulting in 6 and 10 structures, respectively. tion buffer or Arp2/3 complex, NPF, or other protein factor was Analysis of the rmsd and root mean square fluctuations of each mixed with 6 μL10× initiation buffer (20 mM MgCl2,10mM simulation indicated that we were achieving reasonable sampling EGTA, 5 mM ATP), which was then mixed with a 40-μL G-actin of the backbone and surface loops of each complex, and the in- mix to initiate polymerization. Polymerization reactions were clusion of such flexibility was helpful in identifying the best-bound allowed to reach steady state, and curves were normalized for complex. differences in steady state fluorescence. Using each combination of F actin and ARPC2/ARPC4 struc- Actin copelleting assays were performed as described pre- tures, we performed 100 docking runs using Rosetta-Dock (25), viously using 50-nM Arp2/3 complexes and the indicated concen- resulting in a total of 6,000 predicted complexes. Because of the trations of F actin (20). Curve fitting and Kd determination was multiple binding sites available to the ARPC2/ARPC4 complex performed using Prism software (GraphPad Software). on the actin filament, cluster analysis of the docking results Debranching assays were performed essentially as described yielded limited information. Therefore, we chose instead to base previously (30). Briefly, pyrene-actin assembly assays were further analysis on the docking scores. performed as described above with wild-type (5 or 10 nM) or Starting with the best-scoring docked complex from the initial DKK mutant Arp2/3 complex (60 or 120 nM) and 200 nM set of runs, we performed an additional 10,000 perturbation runs GST-WASP-WCA. Samples were extracted at the indicated time using the default dock_pert parameters in Rosetta-Dock. When points, and filaments were immediately stabilized by addition of we plotted the docking score vs. the rmsd, using the best structure rhodamine phalloidin (Invitrogen Molecular Probes). The first from these perturbation runs as the reference, we saw the emer- time point (t ¼ 0) was taken when actin polymerization reached gence of a funnel-shaped energy landscape, reminiscent of what steady state. Rhodamine-phalloidin-stabilized filaments were

Goley et al. PNAS Early Edition ∣ 5of6 Downloaded by guest on October 4, 2021 applied to poly(L-lysine)coated cover slips and imaged using an Tris pH 8.1, 40 mM MgCl2)or3-μL20× Mg-ATP (20 mM Tris pH Olympus BX51 microscope equipped with a 60× objective and a 8.1, 40 mM MgCl2, 10 mM ATP). A Fluorolog-3 spectrofluorom- Hamamatsu Orca-ER camera. Images were captured in TIFF eter (Jobin Yvon) was used for fluorescence measurements. The format using μManager software, and the frequency of branching solution was excited at 435 nm (CFP excitation) and scanned for was quantified by manual counting using Metamorph software emission from 450 to 560 nm. FRET/CFP ratios were determined (Molecular Devices). Images were prepared for presentation by dividing peak YFP emission (523–525 nm) by peak CFP emis- by adjusting brightness and contrast using Adobe Photoshop. sion (473–475 nm). Ratios were normalized by setting the average Curve-fitting and Y-branch t1∕2 calculations were performed ratio in the absence of ATP and GST-WASP-CA to 1 and then using Prism software (GraphPad Software). determining the ratios þATP and þGST-WASP-CA relative to FRET Experiments. FRET experiments were carried out as that baseline. described previously (29). Briefly, CFP and YFP tagged rArp2/ 3 complex (60 nM final concentration) was mixed with FRET ACKNOWLEDGMENTS. We thank Xiange Zheng for assistance with molecular dynamics simulations of Arp2/3 complex and F actin, Dorit Hanein and gel filtration buffer (20 mM MOPS pH 7.0, 100 mM KCl, Neils Volkmann for providing structural coordinates for their model of the 0.5 mM DTT, 10% glycerol) or GST-WASP-WCA in a total of Arp2/3 complex bound to F actin, and Ken Campellone and Erin Benanti 9 μL. This was combined with a 48-μL FRET buffer (50 mM Tris for comments on the manuscript. This work was supported by NIH/NIGMS pH 8.0, 112 mM KCl, 0.1 mM DTT) and 3-μL20× MgCl2 (20 mM grants GM059609 (M.D.W.) and GM067246 (D.S.).

1. Goley ED, Welch MD (2006) The ARP2/3 complex: An actin nucleator comes of age. 21. Bailly M, et al. (2001) The F-actin side binding activity of the Arp2/3 complex is essential Nat Rev Mol Cell Biol 7:713–726. for actin nucleation and lamellipod extension. Curr Biol 11:620–625. 2. Dayel MJ, Holleran EA, Mullins RD (2001) Arp2/3 complex requires hydrolyzable ATP 22. Daugherty KM, Goode BL (2008) Functional surfaces on the p35/ARPC2 subunit of for nucleation of new actin filaments. Proc Natl Acad Sci USA 98:14871–14876. Arp2/3 complex required for cell growth, actin nucleation, and endocytosis. J Biol 3. Le Clainche C, Didry D, Carlier MF, Pantaloni D (2001) Activation of Arp2/3 complex by Chem 283:16950–16959. Wiskott-Aldrich Syndrome protein is linked to enhanced binding of ATP to Arp2. J Biol 23. Holmes KC, Popp D, Gebhard W, Kabsch W (1990) Atomic model of the actin filament. Chem 276:46689–46692. Nature 347:44–49. 4. Machesky LM, et al. (1999) Scar, a WASp-related protein, activates nucleation of actin 24. Phillips JC, et al. (2005) Scalable molecular dynamics with NAMD. J Comput Chem filaments by the Arp2/3 complex. Proc Natl Acad Sci USA 96:3739–3744. 26:1781–1802. 5. Higgs HN, Blanchoin L, Pollard TD (1999) Influence of the C terminus of Wiskott- 25. Gray JJ, et al. (2003) Protein-protein docking with simultaneous optimization of J Mol Biol – Aldrich syndrome protein (WASp) and the Arp2/3 complex on actin polymerization. rigid-body displacement and side-chain conformations. 331:281 299. Biochemistry 38:15212–15222. 26. Schueler-Furman O, Wang C, Baker D (2005) Progress in protein-protein docking: 6. Pantaloni D, Boujemaa R, Didry D, Gounon P, Carlier MF (2000) The Arp2/3 complex atomic resolution predictions in the CAPRI experiment using RosettaDock with an Proteins – branches filament barbed ends: Functional antagonism with capping proteins. Nat improved treatment of side-chain flexibility. 60:187 194. Cell Biol 2:385–391. 27. Schueler-Furman O, Wang C, Bradley P, Misura K, Baker D (2005) Progress in modeling Science – 7. Mullins RD, Heuser JA, Pollard TD (1998) The interaction of Arp2/3 complex with actin: of protein structures and interactions. 310:638 642. Nucleation, high affinity pointed end capping, and formation of branching networks 28. Beltzner CC, Pollard TD (2004) Identification of functionally important residues of Arp2/3 complex by analysis of homology models from diverse species. J Mol Biol of actin filaments. Proc Natl Acad Sci USA 95:6181–6186. 336:551–565. 8. Blanchoin L, et al. (2000) Direct observation of dendritic actin filament networks 29. Goley ED, Rodenbusch SE, Martin AC, Welch MD (2004) Critical conformational nucleated by Arp2/3 complex and WASP/Scar proteins. Nature 404:1007–1011. changes in the Arp2/3 complex are induced by nucleotide and nucleation promoting 9. Amann KJ, Pollard TD (2001) Direct real-time observation of actin filament branching factor. Mol Cell 16:269–279. mediated by Arp2/3 complex using total internal reflection fluorescence microscopy. 30. Martin AC, Welch MD, Drubin DG (2006) Arp2/3 ATP hydrolysis-catalysed branch Proc Natl Acad Sci USA 98:15009–15013. dissociation is critical for endocytic force generation. Nat Cell Biol 8:826–833. 10. Svitkina TM, Borisy GG (1999) Arp2/3 complex and actin depolymerizing factor/cofilin 31. Beltzner CC, Pollard TD (2008) Pathway of actin filament branch formation by Arp2/3 in dendritic organization and treadmilling of actin filament array in lamellipodia. complex. J Biol Chem 283:7135–7144. J Cell Biol 145:1009–1026. 32. Johnson P, Blazyk JM (1978) Involvement of an arginine residue of actin in tropomyo- 11. Mogilner A (2006) On the edge: Modeling protrusion. Curr Opin Cell Biol 18:32–39. sin binding. Biochem Biophys Res Commun 82:1013–1018. 12. Robinson RC, et al. (2001) Crystal structure of Arp2/3 complex. Science 294:1679–1684. 33. Blanchoin L, Pollard TD, Hitchcock-DeGregori SE (2001) Inhibition of the Arp2/3 13. Nolen BJ, Littlefield RS, Pollard TD (2004) Crystal structures of actin-related protein 2/3 complex-nucleated actin polymerization and branch formation by tropomyosin. Curr Proc Natl Acad Sci USA – complex with bound ATP or ADP. 101:15627 15632. Biol 11:1300–1304. 14. Nolen BJ, Pollard TD (2007) Insights into the influence of nucleotides on actin family 34. DesMarais V, Ichetovkin I, Condeelis J, Hitchcock-DeGregori SE (2002) Spatial Mol Cell – proteins from seven structures of Arp2/3 complex. 26:449 457. regulation of actin dynamics: A tropomyosin-free, actin-rich compartment at the 15. Nolen BJ, et al. (2009) Characterization of two classes of small molecule inhibitors of leading edge. J Cell Sci 115:4649–4660. Nature – Arp2/3 complex. 460:1031 1034. 35. Otterbein LR, Graceffa P, Dominguez R (2001) The crystal structure of uncomplexed 16. Volkmann N, et al. (2001) Structure of Arp2/3 complex in its activated state and in actin actin in the ADP state. Science 293:708–711. Science – filament branch junctions. 293:2456 2459. 36. Jacobson MP, et al. (2004) A hierarchical approach to all-atom protein loop prediction. 17. Egile C, et al. (2005) Mechanism of filament nucleation and branch stability revealed Proteins 55:351–367. by the structure of the Arp2/3 complex at actin branch junctions. PLoS Biol 37. Kale LEA (1999) NAMD2: greater scalability for parallel molecular dynamics. J Comput 3:1902–1909. Phys 151:283–312. 18. Rouiller I, et al. (2008) The structural basis of actin filament branching by the Arp2/3 38. Yarar D, D’Alessio JA, Jeng RL, Welch MD (2002) Motility determinants in WASP family complex. J Cell Biol 180:887–895. proteins. Mol Biol Cell 13:4045–4059. 19. Mullins RD, Stafford WF, Pollard TP (1997) Structure, subunit topology, and actin-bind- 39. Spudich JA, Watt S (1971) The regulation of rabbit skeletal muscle contraction: ing activity of the Arp2/3 complex from Acanthamoeba. J Cell Biol 136:331–343. biochemical studies of the interaction of the tropomyosin-troponin complex with 20. Gournier H, Goley ED, Niederstrasser H, Trinh T, Welch MD (2001) Reconstitution of actin and the proteolytic fragments of myosin. J Biol Chem 246:4866–4871. human Arp2/3 complex reveals critical roles of individual subunits in complex structure 40. Cooper JA, Walker SB, Pollard TD (1983) Pyrene actin: documentation of the validity of and activity. Mol Cell 8:1041–1052. a sensitive assay for actin polymerization. J Muscle Res Cell Motil 4:253–262.

6of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0911668107 Goley et al. Downloaded by guest on October 4, 2021