Structural basis of actin monomer re-charging by cyclase-associated protein

Actin polymerization powers key cellular processes, including motility, morphogenesis, and endocytosis. The actin turnover cycle depends critically on “re-charging” of ADP-actin monomers with ATP, but whether this reaction requires dedicated proteins in cells, and the underlying mechanism, have remained elusive. Here we report that nucleotide exchange catalyzed by the ubiquitous cytoskeletal regulator cyclase-associated protein (CAP) is critical for actin-based processes in vivo. We determine the structure of the CAP–actin complex, which reveals that nucleotide exchange occurs in a compact, sandwich-like complex formed between the dimeric actin-binding domain of CAP and two ADP-actin monomers. In the crystal structure, the C-terminal tail of CAP associates with the nucleotide-sensing region of actin, and this interaction is required for rapid re-charging of actin by both yeast and mammalian CAPs. These data uncover the conserved structural basis and biological role of protein-catalyzed re-charging of actin monomers.

T he actin cytoskeleton is critical for a wide range of cellular processes, including migration, morphogenesis, and endocytosis. Consequently, defects in the regulation of actin dynamics and actin network organization are linked to a number of diseases, including cancer metastasis, immune and neurological disorders [1][2][3] . The rapid polymerization of actin filaments, that provides force for the above-mentioned cellular processes, must be balanced by the disassembly of "aged" actin filaments, and recycling of actin monomers for new rounds of filament assembly. This process, called "treadmilling", consists of four phases: (1) incorporation of assembly-competent ATP-actin monomers to the rapidly growing actin filament barbed end; (2) ATP-hydrolysis, followed by Pi release, on actin subunits in the filament; (3) dissociation of ADP-actin monomers from the pointed end of filament; and (4) "re-charging" of ADP-actin monomers with ATP 4 . In vitro, these four phases, as well as nucleation of new actin filaments, are relatively slow, and thus a large collection of actin-binding proteins such as the Arp2/3 complex, formins, and ADF/cofilin, evolved to enhance the rate of actin dynamics [5][6][7][8][9][10][11][12][13] . While previous studies have demonstrated the in vivo importance of these proteins that catalyze actin nucleation, polymerization, and disassembly [14][15][16][17][18] , it has remained unclear whether additional protein machinery is also required to catalyze actin monomer "re-charging" (exchange of ADP for ATP) in cells.
Two evolutionarily conserved proteins, profilin and cyclaseassociated protein (CAP), can catalyze nucleotide exchange on actin monomers in vitro. However, CAP appears to be better suited for this function, because it binds the substrate (ADP-Gactin) with much higher affinity compared to profilin 19,20 .
Moreover, whereas all CAPs tested so far catalyze nucleotide exchange in vitro, only a subset of profilins accelerate nucleotide exchange in biochemical assays 21 . Finally, only CAP has been shown in vitro to effectively catalyze nucleotide exchange on cofilin-bound ADP-actin monomers 22 . These observations have called into question the popular view depicting profilin as the key driver of actin monomer recharging, and suggest instead that CAP may perform this conserved function. Until now, however, it has not been possible to rigorously test whether either or both proteins serve this function in vivo, due to an absence of mutants that disrupt nucleotide exchange activity without compromising actin binding.
CAPs are multi-domain, multifunctional proteins that oligomerize into hexamers and promote rapid actin filament turnover in vitro and in cells [22][23][24] . Whereas yeasts and invertebrates have only one CAP protein, vertebrates express two CAP isoforms: ubiquitously expressed CAP1 and muscle-specific CAP2 25 . The N-terminal half of CAPs binds ADF/cofilin-actin monomer complexes and actin filaments, and accelerates ADF/cofilin and twinfilin-mediated actin filament disassembly [26][27][28][29][30] . The Cterminal half of CAPs harbors two proline-rich regions, PP1 and PP2, which bind to profilin and SH3 domain proteins, respectively, and a Wiscott Aldrich Syndrome protein homology  Fig. 1 Crystal structure of the CARP domain from mouse CAP1 in complex with ADP-actin. a CARP domain forms a homodimer (the subunits are in blue and cyan) that binds two actin molecules (in green) through their subdomains 1, 2, and 3. b The CARP domain dimer covers a large surface (1944 Å 2 ) on each actin monomers. The two subunits of the CARP domain dimer employ two different interfaces (primary interface in yellow; secondary interface in red; and C-terminal tail of CARP from the secondary interface in magenta) for association with each actin monomer. c Structural comparison of actin monomerbinding mechanisms of CARP domain of CAP1 (6fm2), profilin (2btf) 44 , gelsolin segment-1 (1eqy) 45 , ADF-H domain of twinfilin (3daw) 46 , and WH2 domain of ciboulot (1sqk) 47 . Front and side views of the complexes are shown. PDB entries are indicated in brackets 2 (WH2) domain, which binds to both ADP-actin and ATP-actin monomers 20,21,[31][32][33][34] . At the C-terminus of CAPs is a homodimeric β-sheet domain, which displays structural similarity to other, functionally unrelated proteins, including X-linked retinitis pigmentosa 2 protein (RP2), and hence is referred to as a CAP and RP2 (CARP) domain 35 . The CARP domain of CAPs binds  specifically to ADP-G-actin, and together with the adjacent WH2  domain catalyzes nucleotide exchange on actin monomers 20,27,34 .
Despite the fundamental requirement of CAPs for actin cytoskeleton organization and function across the eukaryotic kingdom [36][37][38][39][40][41] , the underlying mechanism by which this protein regulates cytoskeletal dynamics in vivo has remained elusive. Moreover, the CARP domain does not display any structural homology to other known actin-binding domains, and despite extensive mutagenesis 20,42,43 the mechanism by which CAPs associate with actin monomers and catalyze nucleotide exchange has thus remained a mystery. Here, we determined the crystal structure of CAP1/ADP-G-actin complex. Combined with molecular dynamics (MD) simulations, biochemical experiments, and in vivo studies on budding yeast, we uncover the structural basis and biological role of CAP-catalyzed nucleotide exchange on actin monomers.

Results
Crystal structure of CAP1 317-474 /ADP-G-actin complex. To reveal the principles of CAP-actin interactions, we crystallized the CARP domain of mouse non-muscle isoform, CAP1 (CAP1 317-474 ) in complex with unmodified muscle ADP-G-actin. Crystals diffracted anisotropically to 2.3 Å in the c direction, and to 3.2-3.3 Å in the a and b directions (Supplementary Table 1, see Methods). The obtained crystal structure of a symmetric complex containing a dimer of the CARP domain bound to two ADP-actin monomers reveals several new and unexpected features (Fig. 1a,   Fig. 1a, and Supplementary Table 1). First, each actin monomer in the complex contacts each of the two subunits in the CARP homodimer, and each CARP monomer (within the homodimer) binds to two actin monomers using two distinct interfaces. This results in a sandwich-like structure, with the CARP homodimer squeezed between two ADP-actin molecules. The intertwined CARP domain dimer has an S-shape organization, in which the two ADP-actin molecules fit perfectly on both sides (Fig. 1a, b). Second, CAP interacts with actin unlike any other actin monomer-binding motifs that have been structurally characterized [44][45][46][47] . All of these other proteins bind to the "front side" or the barbed end interface of actin between subdomains 1 and 3, whereas the CARP domain binds to the "back side" on subdomains 1, 2, and 3 of actin (Fig. 1c). Third, the CARP domain binds G-actin through a much larger interface compared to other G-actin-binding proteins/domains. The large binding interface results from the two monomers within the CARP domain dimer interacting with actin monomers through different interfaces. The "primary interface" of the CARP domain on subdomains 1 and 3 of actin overlaps with the binding site of profilin on actin. However, the "secondary interface", formed between the second CARP monomer and actin subdomain 2, is different from the interaction sites of other actin monomerbinding proteins characterized so far (Fig. 1a-c and Supplementary Fig. 1b). Analysis of the structures also revealed steric clashes between the CARP domain and profilin, as well as between the CARP domain and the twinfilin's ADF-H domain ( Supplementary Fig. 1g), providing a structural explanation for why these protein domains compete with each other for G-actin binding 20,34,43 .
Although interactions with actin do not alter significantly the structure of the CARP domain ( Supplementary Fig. 1c), the CARP domain induces a conformational change in the actin monomers. In the crystal structure, the D-loop of actin subdomain 2 has a pulled-back conformation and associates with the secondary interface of the CARP domain ( Fig. 1a, b). This orientation is drastically divergent from all other D-loop conformations of actin reported so far 44,46,[48][49][50] (Fig. 2a, Table 3). The simulations demonstrate that the CARP homodimer stabilizes the D-loop of actin subdomain 2 in the extended conformation, while in uncomplexed ADP-actin this loop is dynamic and adopts a variety of conformations (Fig. 2c, d and Supplementary Fig. 2d). Compared to the structure of uncomplexed ADP-actin 48 , the CARP-bound ADP-actin monomer displayed also other small variations, which may be linked to the conformational change in the D-loop (Supplementary Fig. 1d-f, h). Together, the crystal structure and MD simulations reveal that the CARP homodimer binds to two ADP-actin monomers through a unique structural mechanism, and alters the conformation of the actin monomers.
Interactions of CAP with ADP-G-actin and ATP-G-actin. We next performed mutagenesis and MD simulations experiments to test the roles of different CARP domain surfaces, and to reveal the mechanism by which the adjacent WH2 domain, which is unresolved in our structure (see Methods), contributes to actin monomer binding. We introduced four groups of mutations to the conserved clusters of residues in the WH2 domain of CAP1, and six groups of mutations in clusters of residues located at the primary and secondary actin-binding interfaces in our co-crystal structure ( Table 1 Table 3) revealed a stable association between the isolated WH2 domain (CAP1 248-295 ) and an ATP-actin monomer ( Supplementary Fig. 5a).
While the CARP domain was dispensable for ATP-G-actin binding, mutations in both WH2 and CARP domains affected the ability of C-CAP to bind ADP-G-actin. Whereas mutations in the WH2 domain modestly decreased ADP-G-actin binding, some mutations in the CARP domain (e.g., K 365 A, N 367 A, and D 372 A in the primary interface) resulted in a complete loss of ADP-G-actin binding ( Table 1, Fig. 3c, and Supplementary  Fig. 3c). Importantly, mutations in both the primary (mutant 5 and mutant 6) and secondary interfaces (mutant 7 and mutant 10) were defective in ADP-G-actin binding, demonstrating that both interfaces of the CARP domain are required for interactions with actin. We then generated a MD simulation model of CAP1 248-474   Fig. 5d). Collectively, these results reveal the structural mechanisms by which CAP interacts with ATP-actin using its WH2 domain (Fig. 3b), and with ADP-G-actin using a combination of its WH2 and CARP domains (Fig. 3c).
Mechanism of CAP-catalyzed nucleotide exchange. One fascinating feature of the CARP domain-ADP-actin co-crystal structure is the close proximity of the C-terminal tail of the CARP domain with the "nucleotide state sensing region" of actin. The occupancy of third phosphate alters the hydrogen bonding network in the nucleotide-binding loops P1 and P2 of actin. These differences are relayed to methylated His 73 located in the "sensing loop" that changes conformation between the two nucleotide states of actin 51 . Our structure revealed that the Cterminal tail of the CARP domain forms a hydrogen bond between a backbone oxygen of CARP Ala 473 and the imidazole ring of actin His 73 (Fig. 4a, b). Moreover, two other possible contacts between the C-terminal tail of CAP and the nucleotide sensing region of actin were revealed in MD simulations of the ADP-actin-CARP domain complex (System 1, Supplementary Table 3, Supplementary Fig. 6a), proposing that the C-terminal tail of CAP may contribute to nucleotide exchange on actin.
We generated mutant versions of mouse and budding yeast C-CAPs lacking the four C-terminal residues (Δ4C). Importantly, this same mutation in both proteins caused severe defects in nucleotide exchange on actin without compromising ADP-Gactin binding. Instead, the Δ4C mutants bound ADP-G-actin  Table 3) from MD simulations. The regions important for ADP-G-actin binding are indicated in orange and red, and the C-terminal tail of CAP1 is in magenta. For detailed effects of the mutants on actin binding, see Table 1  with higher affinity compared to the wild-type proteins (Fig. 4c, d, f, g, Supplementary Fig. 6b). Moreover, gel filtration analysis demonstrated that also in the context of full-length protein, Δ4C mutant does not disrupt ADP-G-actin binding activity of yeast Srv2 ( Supplementary Fig. 6c). Importantly, deletion of the four Cterminal residues converted the C-terminal halves of mouse and yeast CAPs into actin monomer sequestering proteins (Fig. 4h), and severely compromised the ability of mouse C-CAP to enhance actin filament turnover in the presence of cofilin (Fig. 4e). These results suggest that deletion of the four C-terminal residues halts CAP's normal progression, leaving it bound to ADP-actin but unable to convert monomers to the ATP-bound state, and therefore severely impairing actin filament turnover.
Another peculiar detail of the nucleotide exchange activity was revealed by the CARP domain mutants that bind to ADP-G-actin with weaker affinity. Whereas mutants in the CARP domain that completely disrupted ADP-G-actin binding lead to severe defects in nucleotide exchange, the CARP domain mutants displaying compromised, but still detectable, affinity for ADP-G-actin were [nM] C-Srv2 [nM] C-Srv2 [μM] C-CAP/C-Srv2 slightly more efficient in promoting nucleotide exchange compared to the wild-type protein (Table 1 and Supplementary  Fig. 3d). Thus, stable association between CAP and ADP-G-actin is not necessary for nucleotide exchange, at least under the in vitro conditions used in this assay.
Physiological role of CAP-catalyzed nucleotide exchange. The Δ4C mutant described above enabled us to test the in vivo importance of CAP's nucleotide exchange function, because the mutant does not disrupt actin binding. We integrated the srv2-Δ4C mutant at the SRV2 locus of budding yeast, and verified that it is expressed at levels similar to wild-type Srv2 ( Fig. 5a and Supplementary Fig. 7). Strikingly, the srv2-Δ4C mutant impaired cell growth, cell morphology, and actin organization as severely as a full deletion of the SRV2 gene, despite missing only four residues located at its C-terminus ( Fig. 5b-d). The srv2Δ and srv2-Δ4C mutants both caused a dramatic increase in the size of mother cells, a depolarization of cortical actin patches, and loss of normal actin cable staining. These results suggest that CAP's nucleotide exchange activity plays a critical role in recharging actin monomers in vivo.

Discussion
A model for CAP-catalyzed re-charging of actin monomers is presented in Fig. 6. In cells, ADP-actin monomers either dissociate spontaneously from the filament pointed ends or their depolymerization is enhanced by the ADF-H domain proteins ADF/cofilin and twinfilin 12,13,28 . The C-terminal half of CAP efficiently catalyzes nucleotide exchange on both uncomplexed and ADF/cofilin-bound ADP-actin monomers 22,23,34 . We demonstrate that the dimeric CARP domain associates with these newly depolymerized ADP-actin monomers through a unique structural mechanism that involves two separate CARP domain surface regions, which collectively interact with the "back side" of actin on subdomains 1, 2, and 3. The CARP domain specifically binds ADP-actin monomers, whereas the isolated WH2 domain of CAP displays no appreciable binding preference for ADP-actin vs. ATP-actin monomers 20,34,43 . Because ADP-actin and ATPactin monomers do not display drastic structural differences 48 , future work is required to reveal why CARP domain specifically associates with ADP-G-actin while the WH2 domain of CAP binds both ADP-actin and ATP-actin monomers. Nevertheless, our data suggest that CAP combines its WH2 and CARP domains to achieve high-affinity interactions with ADP-actin monomers, and to promote the dissociation of ADF/cofilin from ADP-actin monomers. Following the initial interaction of the CARP domain dimer and ADP-actin monomers, the WH2 domains "embrace" the two ADP-actin monomers leading to a formation of a compact complex, where nucleotide exchange of actin occurs. The "re-charged" ATP-actin monomers remain transiently bound to the WH2 domains of CAP, and then are released spontaneously, or transferred to profilin, which binds to the adjacent PP1 polyproline region in CAP 31 .
Our experiments provide evidence that three structural features of the CAP/ADP-G-actin complex are important for nucleotide exchange. First, we showed that the C-terminal tail of the CARP domain, which "penetrates" into the actin molecule and associates with its "nucleotide sensing region", is important for efficient "re-charging" of actin by CAP. Second, we found that the CARP domain induces a unique conformational change in the D-loop of actin. Importantly, subtilisin-cleavage of the D-loop of ADP-actin monomers results in~3-fold increase in the rate of ADP-to-εATP nucleotide exchange on actin ( Supplementary  Fig. 2c), similar to what was previously reported for ATP-to-εATP exchange on subtilisin-cleaved actin 52 , suggesting that the D-loop communicates allosterically with the nucleotide-binding pocket. However, in contrast to the mechanism by which Sos guanine-nucleotide-exchange-factor promotes GDP-for-GTP exchange on Ras GTPase, we did not detect major conformational changes in the nucleotide-binding pocket of actin 53 . Third, the presence of functional WH2 domain is important for efficient nucleotide exchange by mouse CAP1. Consistent with previous structural work 47,54,55 , our MD simulations suggest the WH2 domain of CAP binds to the cleft between actin subdomains 1 and 3. The precise mechanism by which the WH2 domain contributes to nucleotide exchange remains to be elucidated. However, MD simulations provided evidence that association with the CARP domain increases the dynamics of ADP-G-actin leading to an opening of the cleft between subdomains 2 and 4, and that these features are further augmented in the presence of the WH2 domain ( Supplementary Fig. 8). It is also important to note that the role of the WH2 domain in nucleotide exchange varies between different species. Although malaria parasite CAP, which is entirely composed of a CARP domain, efficiently catalyzes nucleotide exchange on actin 43 , mouse and budding yeast CAPs cannot efficiently promote nucleotide exchange without functional WH2 domain, especially in the presence of cofilin 27,56 .
Our in vivo work using the yeast srv2-Δ4C mutant provides the first direct evidence that CAP-catalyzed nucleotide exchange is critical for actin cytoskeleton organization and function. Earlier genetic evidence from yeasts suggested that nucleotide exchange catalyzed by profilin may be important in vivo 57,58 . However, the  4) The CARP domain has little if any affinity for ATP-actin, leaving dimeric CAP molecules associated with ATP-G-actin solely through their WH2 domains. (5) Profilin has high affinity for ATPactin monomers, and binds directly to the PP1 domain of CAP 31 , adjacent to the WH2 domain. Thus, as ATP-actin monomers dissociate from CAP, they are rapidly bound by profilin, replenishing the pool of ATP-actin monomers available for assembly. This leaves CAP primed for the next round of nucleotide exchange mutants used in these studies all weakened actin-binding affinity, and thus the in vivo role of profilin in nucleotide exchange remains to be tested in the manner we have demonstrated here for CAP, with a mutation that disrupts nucleotide exchange without weakening actin affinity. Two other observations calling into question whether profilins catalyze nucleotide exchange in vivo are that profilins do not bind ADP-G-actin with high affinity 19 , and profilins are much less efficient than CAP in accelerating actin filament turnover in the presence of cofilin ( Fig. 4e and ref. 22 ). Whether profilin has partially redundant roles with CAP in promoting nucleotide exchange in cells, or instead is required primarily to maintain homeostasis in distributing actin monomers between formin-dependent and Arp2/3 complex-dependent actin assembly pathways, remains to be determined 59,60 .
Collectively, our study uncovers the molecular mechanism by which the WH2 and CARP domains of CAP associate with actin monomers and accelerate nucleotide exchange on ADP-actin. However, it is important to note that in most eukaryotic organisms, including yeast and mammals, CAP oligomerizes into hexameric complexes that have additional functions in accelerating ADF/cofilin-dependent and twinfilin-dependent actin filament disassembly, which depend on the N-terminal helical folded domain 21,[26][27][28] . Thus, in the future it will be important to reveal how the different activities of CAP are structurally and functionally coordinated between its N-terminal and C-terminal functional units.
For biochemical experiments, mutations to C-CAP were generated using primers shown in Supplementary Table 4. Expression of the wild-type and mutant proteins (both mouse and yeast) was performed as above in one liter scale. Cells were suspended to a lysis buffer containing a Complete Ultra protease inhibitor cocktail (Roche) and stored at −80°C by snap freezing. For protein purification, cell suspensions were thawed in a water bath and disrupted by sonication. Cell lysates were clarified by centrifugation and batch purified by adding 1 ml of Ni-NTA agarose beads (Qiagen) for 2 h at 4°C rotating. Beads were washed with wash buffer (50 mM Tris-HCl, 150 mM NaCl, 50 mM imidazole, pH 7.5) for 40xCV and suspended to cleavage buffer (50 mM Tris-HCl, 150 mM NaCl, 25 mM imidazole, 1 mM DTT, pH 8.0) for O/N cleavage with~0.01 mg/ml of 3C protease at 4°C. Supernatant was collected and concentrated to 0.5 ml volume with Amicon Ultra-4 10 kDa centrifugal filter (Merck) and the proteins were further purified by gel filtration (SD200 Increase 10/300, GE Healthcare) in 2 mM Tris-HCl, 100 mM KCl, 1 mM MgCl2, 0.1 mM CaCl2, 0.1 mM DTT, pH 8.0. Protein-containing fractions were concentrated as above stored by snap-freezing with liquid N 2 at −80°C. Mutants used for biochemical assays eluted in similar volumes in the gel filtration, those which eluted in void fraction were discarded from further analysis.
Cofilin and profilin were purified as previously described 20,43 . Muscle α-actin was prepared from rabbit muscle acetone powder (Pel Freez) as previously described 20 and stored at 2 mg/ml by snap-freezing with liquid N 2 at −80°C.
Full-length Srv2 proteins were expressed as GST-tagged fusion proteins in pGAT2 vector. Plasmids were transformed into BL21(DE3) E. coli and protein expression was performed at 37°C in 2xLB media by IPTG induction at OD 600 of 0.5-0.7. Expression was continued at 16°C for 24 h. Cells were disrupted by sonication and supernatants clarified by centrifugation. Equilibrated glutathione agarose beads (Thermo Scientific) were added to the sample and incubated 2 h at 4°C. Beads were washed with wash buffer (50 mM Tris-HCl, 150 mM NaCl, pH 8.0) for 20xCV and protein was eluted with 4xCV in a gravity column with elution buffer (100 mM Tris-HCl, 100 mM NaCl, 20 mM reduced glutathione, pH 8.8). Elution fractions were pooled and loaded to HiTrap Q HP (GE Healthcare) anion exchange column equilibrated in Buffer A (100 mM Tris, 100 mM NaCl, pH 8.8). Column was washed with 5xCV of Buffer A and then 5xCV with 20% Buffer B (100 mM Tris, 400 mM NaCl, pH 8.8). Finally, protein was eluted with a salt gradient, peak fractions were pooled and concentrated with Amicon Ultra-4 50 kDa centrifugal filter (Merck) and stored as described above. . For data collection, crystals were snap frozen in liquid N 2 by fishing directly from 96-well plates and soaking for cryo-protection in 25% glycerol (v/v) containing mother liquid. Diffraction data were collected at Diamond Light Source synchrotron (Didcot, UK) at 100 K on I04 beamline with Pilatus3 6M detector. A complete data set was obtained at 0.97 Å wavelength with 0.1°oscillation angle per diffraction image with a total of 1400 images collected. Data were integrated using X-ray Detector Software 61 , merged and scaled with AIMLESS (CCP4 62 ). The initial molecular replacement solution of 1:1 complex CARPdomain and actin was obtained with BALBES 63 program, part of CCP4 suit, giving best result (Q factor = 0.6809) using 1k4z model for CARP domain and 3tpq for the actin with Rwork/Rfree = 0.3470/0.4470, after single REFMAC 64 refinement round. Rounds of manual model building in COOT 65 , introduction of translationlibration-screw grouping and refinement with BUSTER 66 lead to final model with R work = 0.186 and R free = 0.234. It is important to note that we used a CAP1 construct (CAP1 242-474 ) composed of the WH2 domain and the CARP domain in the crystallization trials, but electron density was observed only for the CARP domain (residues 317-473) and actin, whereas the first 74 residues of the CAP1 242-474 were absent from the structure. Mass spectrometry analysis of the crystallization drops confirmed that the flexible N-terminal region of CAP1 242-474 construct, containing the WH2 domain, had a tendency to degrade thus resulting in a product of a size of the CARP domain that was seen in the final electron density maps. The diffraction data was strongly anisotropic (CC 1/2 > 0.3 at 2.3 Å along l-axis, 3.2 Å along hk plane) with ΔB of 56.16 Å 2 between the axes which might explain a high overall B-factor (Supplementary Table 1) when using individual B-factor refinement in the final model. Despite high average B, distribution of B-factors along the structure was normal. We considered anisotropic treatment of the merged data using ellipsoidal cut (STARANISO) which in our model lowers the average B-factor from~129 to~75 Å 2 (Wilson B from 86 to 55 Å 2 ), normal for the resolution range. However, ellipsoidal cut decreased slightly the completeness of our data in the 3.0-3.6 Å resolution range. This caused the electron density maps to be slightly discontinuous in some parts of CARP, but allowed us to observe more high-resolution details overall in the electron density maps. For these reasons, anisotropically treated data was only used to polish and finalize the final model, especially in the placement of waters and some side chains. Finally, we refined the model against either of these data sets yielding nearly identical models (RMSD 0.5). Before deposition to PDB, the model was refined to R work /R free 18.6%/23.4% using non-treated data at 2.8 Å with I/σI = 1.3 as resolution cutoff criteria.
Biochemical experiments. To determine binding of C-CAP proteins to actin with 4-chloro-7-nitrobenzofurazan (NBD) assay, NBD-actin was prepared as previously described 43  100,000 rpm with TLA-100 rotor (Beckman) for 5 min at 4°C. The affinities of mouse C-CAP constructs for ADP-actin and ATP-actin were determined as previously described 43 . Affinity of C-Srv2 proteins for ADP-actin was determined as previously described 20 . Samples were measured on a 96-well microtiter plate for fluorescence at 482/520 nm using Varioskan Lux plate reader (Thermo Scientific) with 1000 ms exposure at +25°C. The fluorescence data were fitted assuming one site competition for ADP-actin assuming 0.03 μM affinity for twinfilin 43 and one site saturating binding for ATP-actin and C-Srv2 proteins. Equations used for the fitting of the data are described in Graphpad 7 manual (https://www.graphpad. com/guides/prism/7/curve-fitting/index.htm?reg_binding-saturation.htm). For nucleotide exchange assays, Ca 2+ -ATP-actin was first exchanged to Mg 2 + -ATP-actin by using exchange buffer described above. ADP-actin was prepared in dialysis at +4°C for 3 h (5 mM Tris For cleavage of ADP-actin with substilisin, actin was prepared as described above in dialysis. Substilisin (Sigma-Aldrich, P5380) was added in 1:100 molar ratio to ADP-actin. Cleavage was continued for 2 h on ice and measured for nucleotide exchange as above. Completeness of the cleavage confirmed on SDS-PAGE was >90%.
Actin filament turnover assay was performed by following the release of inorganic phosphate with 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG) and purine-nucleoside phosphorylase (PNP) (EnzChek, Thermo Scientific). Rabbit muscle actin was mixed with 0.2 mM MESG and 6 units of PNP in G-buffer and polymerization was initiated with 20 mM KCl/NaCl and 2 mM MgCl 2 . After 10-min incubation, reactions with 5 μM protein(s) and 20 μM actin were prepared and transferred on a 96-well microtiter plate. The release of phosphate was followed for 30 min at 360 nm using Varioscan Lux multimode microplate reader (Thermo Scientific) at 25°C. The assay was repeated two times for each combination of proteins.
For actin cosedimentation assay, actin was polymerized in 20 mM Tris pH 7.5, 100 mM KCl, 0.2 mM EGTA, 1 mM MgCl2, 1 mM ATP, 0.2 mM DTT for 30 min at 22°C. Different amounts of C-CAP proteins were added to the mixture of 2.5 μM actin and incubated 30 min at 22°C. Samples were centrifuged at 100,000 rpm for 30 min with TLA-100 rotor (Beckman) to pellet the F-actin. Samples from supernatant and pellet were analyzed on SDS-PAGE, imaged with ChemiDoc XRS + imaging system (BioRad) and quantified using Image Lab (BioRad).
The thermal stability of the proteins was measured by Thermofluor assay (known also as Thermal Shift Assay) in the nucleotide exchange buffer. Proteins were diluted to 25 µM, and the SYPRO Orange dye (Thermo Scientific, #S6650) was diluted 300 times with the nucleotide exchange buffer prior to reaction. Reactions of 25 µl were prepared in 96-well PCR plate, consisting of 2.5 µl diluted protein, 2.5 µl of diluted SYPRO orange dye, and 20 µl of the nucleotide exchange buffer. The change of fluorescence as an indication for protein denaturation was measured on Mx30005P qPCR instrument (Agilent) every 30 s with a rate of 1°C/ 30 s. The midpoint values for unfolding were taken as a value for thermal stability.
Binding of full-length Srv2 proteins to rabbit muscle ADP-actin was analyzed by size-exclusion chromatography. Preparation of ADP-actin was performed as described for nucleotide exchange assays. SD200 Increase 10/300 (GE Healthcare) was equilibrated in 5 mM HEPES, 100 mM NaCl, 0.05 mM ADP, 0.05 mM MgCl2, 0.5 mM β-mercaptoethanol, and retention of ADP-actin was first analyzed by loading 300 μl of 10 μM ADP-actin to the column. Then, retention was analyzed for 10 μM ADP-actin mixed with 10 μM GST-Srv2, or with 10 μM GST-Srv2-Δ4C. Peak fractions were collected and analyzed by SDS-PAGE.
Atomistic MD simulations. The WH2 domain of the CAP1 was obtained by homology modeling based on three structures (PDB entries 1sqk 54 , 2a3z 47 , and 4pl7 67 ) using the comparative modeling protocol of Rosetta (RosettaCM) 68 . This initial model of WH2 was used for MD simulations of ATP-actin-WH2 (System 3, Supplementary Table 3). The model was further extended by adding PP2 and connecting it to the CARP domain from our structure. There were two possible ways of connecting WH2 to CARP domain: in cis (Supplementary Fig. 5b) or in trans (Supplementary Fig. 5c). The cis configuration requires PP2 to stretch across a distance that is at least 11 Å longer than in the trans configuration, making it both sterically and entropically unfavorable. The trans configuration has no steric clashes and in addition allows the regulatory region PP2 to be accessible, and thus represents more plausible model (Supplementary Fig. 5c). One thousand decoys of the ADP-actin-CAP1 248-474 dimeric complex were generated. The models were ranked based on their total score, and the top ranking five models were selected to initiate atomistic MD simulations.
The structures of protein molecules (actin, CARP domain, WH2, and CAP1 248-474 ) were prepared using Propka 69 (pK a estimation based on the crystallized complex and determination of protonation states at pH = 6.8), Chimera 70 (placing hydrogens), VMD 71 (protein structure building and visualization), and PyTopol (protein topology conversion from CHARMM to the GROMACS format). In all systems, E454 and H416 of the CARP domain were protonated; N-terminus of actin was acetylated; the His73 residue of actin was methylated (parameters were obtained by analogy); and the crystal water molecules and the bound Mg ion were kept. In each system containing ATP-actin, ATP was docked into the binding pocket based on the ADP position. Each system containing the WH2 domain was initiated from the top scoring five Rosetta models as described above.
All simulations were carried out using GROMACS 5.1 72 employing the Charmm36 force field 73 for the proteins and the TIP3P model 74 for water. The equations of motion were integrated using a leap-frog algorithm with a 2 fs time step. All bonds involving hydrogens were constrained using the LINCS algorithm 75 . Long-range electrostatic interactions were treated by the smooth particle mesh Ewald scheme 76 with a real-space cutoff of 1.2 nm, a Fourier spacing of 0.12 nm, and a fourth-order interpolation. A Lennard-Jones potential with a force-switch between 1.0 and 1.2 nm was used for the van der Waals interactions.
Each protein complex was placed in a rhombic dodecahedral simulation box maintaining at least a distance of 15 Å to the box sides and solvated in 0.15 M NaCl solution, ensuring neutrality of the system. Before production runs, steepest descent minimization and successive equilibration simulations (in total~500 ps) in the NVT and NPT ensembles using the Berendsen thermostat and barostat 77 were performed. In these simulations, initially all protein heavy atoms, and later only the C α atoms were restrained with a force constant of 1000 kJ/mol/Å 2 ; the time step was then ramped up from 1 to 2 fs. The number of water molecules and the average box volume at the start of the production runs are given in Supplementary Table 3.
For production runs, each system was simulated in the NPT ensemble for~1.2 µs. Five independent repeats were performed for each system (Supplementary Table 3). The total time scale covered in the simulations was >24 μs. Protein-ADP/ ATP-Mg complex, and solvent (water and NaCl) were coupled to separate temperature baths at 310°K using the Nosé-Hoover thermostat 78,79 with a time constant of 1.0 ps. Isotropic pressure coupling was performed using the Parrinello-Rahman barostat 80 with a reference pressure of 1 atm, a time constant of 5 ps, and a compressibility of 4.5 × 10 −5 bar −1 . All analyses were performed separately for each simulation repeat. The averages and standard deviations over independent repeats are reported.
To visualize the actin cytoskeleton in cells, yeast strains were grown in YEPD to log phase and fixed with formaldehyde (4% final) for 30 min at 25°C. Fixed cells were washed with 1× PBS and stained with AlexaFluor TM -488-Phalloidin (Thermo Scientific) overnight at 4°C. Cells were then washed three times with 1× PBS, and mounted on slides immediately before imaging. Cells were imaged on a Nikon N-SIM (Structured Illumination Microscopy) instrument (Nikon Instruments, Melville, NY) equipped with a SR Apo TIRF AC 100xHx1.49 N.A. oil immersion objective, a LU-N3-SIM laser unit, and an ORCA-flash4.0 CMOS camera (Hamamatsu Photonics, Boston, MA). 3D-SIM image stacks were acquired with a Z-interval of 0.1 µm for a 0.9 µm section at the central plain of the cell. Images were captured with 500 ms exposure time. Fifteen raw images were acquired per Z-position, and reconstruction of images were performed using the reconstruction slice method from NIS-Elements software (Nikon Instruments). Final images were then analyzed in ImageJ (NIH) to quantify actin patch number in mother cells and mother cell size, and the data were plotted in Graphpad Prism 6.
Quantification and statistical analysis. The statistical details, number of experiments and statistical analyses, are described in the figure legends or the Methods in all relevant cases. Softwares for quantification and data analysis are indicated in experimental details in the Methods section.
Data availability. The WH2-PP2-CARP model(s) obtained from atomistic simulations, together with the raw data, are available in Zenodo. The crystal structure has been deposited to the Protein Data Bank (PDB) under access code 6fm2. All relevant experimental data are available upon a reasonable request from the corresponding author.