Mechanisms of disassembly at the actin filament pointed and barbed ends

Abstract

Actin cytoskeleton dynamics power processes from cell motility to organelle trafficking, requiring rapid polymerization and depolymerization accelerated in cells by regulatory proteins. While mechanisms of accelerated polymerization are relatively well studied, those of depolymerization remain poorly understood. Here, we present twelve cryo–electron microscopy structures showing how cofilin, cyclase-associated protein (CAP), and capping protein (CP) coordinate their activities to accelerate depolymerization at both filament ends. Alone, CAP produces a ~4.0 Å lateral displacement of the first pointed-end subunit, whereas cofilin reverts terminal subunits at the pointed and barbed ends to a G-actin–like conformation and undertwists the filament short-pitch helix. When functioning together, these cofilin- and CAP-induced conformational changes are amplified to accelerate pointed-end disassembly. At the barbed end, the cofilin-induced changes trigger stepwise CP dissociation and favor depolymerization. These findings support end-specific mechanisms of filament disassembly through accelerated subunit dissociation, slowed subunit addition, and barbed-end uncapping.

Actin filament structures with cofilin, CAP, and CP capture depolymerization snapshots at pointed and barbed ends.

INTRODUCTION

Actin dynamics drive fundamental processes ranging from cell and organelle motility to the maintenance of cell shape and polarity (1). Central to these processes are actin’s adenosine triphosphatase (ATPase) activity and its interactions with numerous regulatory proteins. Actin’s ATPase activity is coupled to its transition between monomeric (G-actin) and filamentous (F-actin) forms. F-actin is a polar filament, described either as a single left-handed short-pitch helix or as two right-handed long-pitch helices (fig. S1 illustrates common definitions) (2). G-actin comprises outer (subdomains 1 and 2) and inner (subdomains 3 and 4) domains, occupying the outermost and innermost sides of F-actin, respectively. These domains are separated by two opposing clefts, which converge near the middle of the protein: the nucleotide-binding and hydrophobic clefts (fig. S1A). The hydrophobic cleft (H-cleft) mediates most interactions with actin-binding proteins (2), as well as intersubunit contacts along the long-pitch helix, with each subunit inserting its D-loop (P38-S52, so named because it binds deoxyribonuclease I) into the H-cleft of the subunit above (fig. S1). G-actin has minimal ATPase activity, but nucleotide hydrolysis occurs immediately upon polymerization, triggered by a ~20° rotation of the outer domain relative to the inner domain that primes the catalytic site and converts “twisted” G-actin into “flat” F-actin subunits (3). The critical concentrations for ATP-actin binding to the barbed and pointed ends of the filament differ (0.1 versus 0.6 μM), so that at steady state ATP-actin adds preferentially to the barbed end, followed by fast ATP hydrolysis, slow phosphate release, and ADP-actin dissociation from the pointed end. However, in cells, the monomer concentration is orders of magnitude higher than the critical concentrations at both ends, and filament turnover is much faster than actin’s ATPase alone would permit (4, 5). This acceleration is achieved by regulatory proteins that control the various steps of actin turnover, including nucleation, elongation, capping, severing, and depolymerization. Despite recent progress in understanding the structural bases of several of these processes (6–10), depolymerization remains poorly understood and is addressed here.

Cofilin, including three major proteins in vertebrates (cofilin-1, cofilin-2, and actin depolymerizing factor), is the primary filament disassembly factor in cells, involved in both severing and depolymerization (11–17). It binds F-actin with high positive cooperativity (18) and a preference for ADP-actin (11, 19, 20), forming clusters termed cofilactin. These clusters can grow symmetrically toward both filament ends or more rapidly toward the pointed end, depending on cofilin isoform and experimental conditions (14, 17, 21). Severing occurs preferentially at junctions between bare and cofilin-decorated regions, with ~80% of events localized to the pointed-end boundary of a cluster (14, 17, 20, 21), where structural perturbations are more pronounced (6). When a cofilin cluster reaches the barbed end, it displaces CP, promotes depolymerization, and inhibits recapping (14). Cofilin also modestly enhances pointed-end depolymerization (11, 14–16). Both severing and depolymerization are accelerated many-fold by partner proteins; actin-interacting protein-1 (AIP1) synergizes with cofilin to boost severing (22–26), while cyclase-associated protein (CAP) accelerates pointed-end depolymerization (15, 16). Although well studied biochemically, the structural basis of cofilin-CAP–mediated depolymerization remains poorly understood. Here, we describe 12 cryo–electron microscopy (cryo-EM) structures of actin filaments with cofilin, CAP, and CP, capturing snapshots of the multistep processes responsible for pointed-end depolymerization, barbed-end uncapping, and barbed-end depolymerization.

RESULTS AND DISCUSSION

The transition from F-actin to cofilactin

To understand the interplay between cofilin, CAP, and CP, we determined cryo-EM structures of CPα1/β2-capped F-actin in the presence or absence of cofilin-2 and full-length CAP-1 (hereafter referred to as CP, cofilin, and CAP, respectively). CP not only produces shorter filaments through capping, helping overcome a technical hurdle by increasing the number of ends per micrograph (7), but also has independent functional significance, as it is displaced by cofilin to enable barbed-end depolymerization (14). In total, we obtained 12 structures, all with ADP bound to F-actin, the preferred nucleotide state for cofilin binding (11, 19, 20). These include five structures with cofilin and/or CAP at the pointed end, six with cofilin and/or CP at the barbed end, and one with cofilin bound to the middle of the filament (figs. S2 to S23 and tables S1 to S3).

At ~2.1 Å, the structure of cofilin bound to the middle of the filament represents the highest-resolution cryo-EM structure of this complex to date and serves as a reference for analyzing cofilin’s effects at the pointed and barbed ends (Fig. 1, A to C, and movie S1). Beyond providing a clearer visualization of interactions, including numerous water molecules (Fig. 1B), this structure agrees well with previous cofilactin structures (6, 27–29). Cofilin binds at the interface between two subunits of the long-pitch helix (Fig. 1C). It inserts a helix (L111-K127) and its N-terminal tail into the H-cleft of the subunit above (toward the pointed end) and displaces the D-loop of the subunit below (toward the barbed end), which normally occupies this cleft. The displaced D-loop becomes disordered and is not resolved. Although loss of this key interaction might be expected to destabilize the filament, cofilin compensates by acting as a buttress between subunits along the long-pitch helix. As a result, cofilactin remains structurally stable, albeit mechanically more compliant than bare F-actin, with severing occurring at boundaries between rigid F-actin and compliant cofilactin segments (30).

Fig. 1. Cofilactin structure.

(A) Cryo-EM map of cofilactin at ~2.1 Å resolution (actin, gray; cofilin, yellow). (B) Inset showing Mg²⁺-ADP and water molecules in the nucleotide-binding cleft. A secondary inset highlights the Mg²⁺-binding site. (C) Inset showing interactions of cofilin (yellow) with two actin subunits along the long-pitch helix (gray transparent surface with Cα traces in gray and blue). A secondary inset highlights cofilin’s helix L111-K127 and its N terminus bound in the H-cleft of the actin subunit above (toward the pointed end). Residue S3, whose phosphorylation inhibits cofilin binding and severing, is shown in red. (D) Three perpendicular views of superimposed Cα traces illustrating how actin’s conformation in the cofilactin structure (blue, this work) resembles that of other G-actin structures (gray; PDB: 1J6Z, 2PBD, 3A3Z, 1KXP, 4PKG, 2A42, 1IJJ, 1Y64, and 4M63) and differs from F-actin (red; PDB: 8F8P). Structures were superimposed on the inner domain (residues 145 to 336) to highlight outer-domain rotation. Cofilactin, however, differs from both G- and F-actin in having a more open H-cleft, as the binding of cofilin’s helix and N terminus separates subdomain-1 of the outer domain from subdomain-3 of the inner domain by ~5 Å (indicated by a blue arrow).

Another consequence of the loss of the D-loop/H-cleft interaction along the long-pitch helix is that actin’s outer domain becomes free to rotate back toward the twisted, G-actin conformation (Fig. 1D). Although this conformation has been described as distinct from both G- and F-actin, it falls within the broad spectrum of G-actin conformations, which, unlike F-actin, display varying degrees of outer domain rotation (Fig. 1D). The main difference between cofilactin and G-actin is the region spanning K336 to the C terminus (F375), which borders the outer domain side of the H-cleft and is displaced by ~5.0 Å in cofilactin, widening the cleft relative to G- or F-actin to accommodate cofilin’s helix and N terminus (Fig. 1, C and D). Binding of cofilin’s N-terminal tail in this cleft likely explains why phosphorylation at S3 inhibits binding and severing of F-actin (6, 31). The cofilin-induced cleft widening may be allosterically coupled to the nucleotide cleft, possibly explaining cofilin’s preference for ADP-actin (11, 19, 20). The G-actin–like transition in cofilactin alters the intersubunit twist of the filament: The left-handed short-pitch helix is undertwisted from −167° to −162°, while the right-handed long-pitch helices are overtwisted from 27° to 36° (fig. S1, B and C). This change in intersubunit twist affects only actin subunits that directly contact cofilin and does not propagate into the bare filament (6).

The pointed end with bound cofilin, CAP, or both – Depolymerization mechanism

There are at least two pathways by which cofilin can decorate filament ends: (i) severing at the pointed or barbed end of mid-filament clusters, and (ii) growth of these clusters until they reach an end. This suggests that end decoration by cofilin is common, but its structural consequences are unknown. Here, we obtained pointed-end structures with bound cofilin alone, cofilin plus one or two CAPs, and one or two CAPs without cofilin, revealing how these proteins cooperate during pointed-end depolymerization (Fig. 2 and movie S2). While the mid-filament cofilactin structure obtained by helical reconstruction reached high resolution, end structures were derived from single-particle analysis and were generally resolved at intermediate resolution (3.0 Å or worse). This is because the number of filament ends is limited in micrographs, and these must be separated into pointed and barbed ends and further subdivided into distinct conformational states, reflecting the highly dynamic nature of the ends.

Fig. 2. Pointed-end structures with cofilin, CAP, or both.

(A) Domain diagram of CAP (OD, oligomerization domain; PR, proline-rich; WH2, WASP-homology-2). (B to F) Cryo-EM maps of the pointed end with bound cofilin (yellow), one HFD (orange), two HFDs, and cofilin with one or two HFDs, respectively (resolutions indicated). The first and second actin subunits are dark and light green, respectively. (G and H) The two pointed-end subunits with bound cofilin or HFD adopt a twisted G-actin–like conformation, similar to the free pointed end and distinct from the flat mid-filament F-actin conformation (gray) (7). The F- to G-actin transition involves a ~20° rotation of the outer domain relative to the inner domain (3). (I) Binding of the second HFD generates a steric clash with the first actin subunit, resolved by a ~4.0 Å lateral displacement (red arrow). The displacement is illustrated by comparison of the first subunit in the one-HFD (gray) structure. (J and K) Top and side views show that the cofilin-induced undertwist of the short-pitch helix amplifies this clash, increasing the lateral displacement of the first subunit to ~7.6 Å. The first subunit moves with its bound cofilin and HFD and likely dissociate as a unit. (L) Mass photometry of the cross-linked CAP-actin complex. Insets show SDS-PAGE of uncross-linked Expi293F-expressed CAP (copurifying with substoichiometric actin) and the fully actin-saturated CAP-actin complex obtained by glycerol-gradient centrifugation (gel lanes extracted from full gels shown in fig. S25). Histogram shows counts per 5 kDa bin. The experimental mass (364 ± 77 kDa) from a Gaussian fit supports a tetrameric complex (theoretical mass 376 kDa).

The pointed-end cofilactin structure is nearly indistinguishable from the mid-filament structure, with subunits adopting the same G-actin conformation and intersubunit twist (Fig. 2, B and G). Compared with the free pointed end, where terminal subunits also adopt a G-actin conformation (7), the main differences are the altered intersubunit twist and wider H-cleft characteristic of cofilactin (Fig. 1D and fig. S1C). The structure suggests that the modest acceleration of pointed end depolymerization observed with cofilin alone (11, 14–16) results from two factors: (i) a reduced on rate, due to unfavorable interactions between incoming monomers and terminal subunits with a G-actin conformation and altered twist, and (ii) an increased off rate, resulting from cofilin-induced rupture of the D-loop/H-cleft interaction, which leaves terminal subunits loosely bound to the filament.

The pointed end cofilactin structure enables analysis of the structural consequences of CAP binding, either alone or with cofilin. CAP comprises, from N to C terminus, oligomerization (OD), helical-folded (HFD), proline-rich (PR), WH2, and CAP-retinitis pigmentosa (CARP) domains (Fig. 2A). Although full-length CAP was used in our first cryo-EM data collection in hopes of visualizing the entire oligomer, only HFD was resolved (Fig. 2, C to F). This is consistent with the observation that N-CAP (residues 1 to 215), a fragment comprising OD and HFD, accounts for CAP’s full depolymerization activity (15, 16). Accordingly, in subsequent datasets we used N-CAP, which can be further concentrated to increase occupancy at the pointed end. One or two HFDs were observed bound to the first subunit or to the first and second subunits at the pointed end along the short-pitch helix (Fig. 2, C to F). In agreement with a previous crystal structure of monomeric actin bound to HFD and the C-terminal domain of twinfilin (a cofilin homolog) (15), HFD binds at the top of actin subdomains 2 and 4, with three consecutive antiparallel α-helices spanning residues P107-Y172 mediating the interaction (fig. S24A). The actin-binding surface of HFD is highly conserved among CAP-1/CAP-2 sequences, whereas the solvent-exposed side is not (fig. S24B). In actin subdomain 2, the interaction involves residues P38-L65, including the D-loop, which is stabilized upon HFD binding. In subdomain 4, HFD contacts residues T202-E207 and L242-G245. The HFD-binding surface on actin is exposed only at the pointed end, explaining why HFD was not observed bound on filament sides or at the barbed end.

Other than stabilizing the D-loop, HFD binding did not alter the conformation of pointed-end subunits, which remained in the G-actin state (Fig. 2H). The persistence of the G-actin conformation in terminal subunits at the free, cofilin-bound, and HFD-bound pointed end suggests that the D-loop/H-cleft interaction, absent at the pointed end, is a key determinant of the G- to F-actin transition. Consistently, at the barbed end, where this interaction is present, subunits adopt the F-actin conformation (7) (see below).

Changes in the relative position of pointed-end subunits depended on whether one or two HFDs were bound and whether cofilin was present. In the absence of cofilin, HFD effects were assessed relative to the free pointed-end structure (7). Binding of a single HFD to the first subunit did not alter its position and is therefore unlikely to enhance depolymerization. By contrast, binding of a second HFD to the second short-pitch subunit created a steric clash with the first subunit, which was resolved by lateral displacement of the latter by ~4.0 Å (Fig. 2I). Although this movement disrupts lateral contacts, the first subunit remained tethered to the long-pitch subunit immediately below through D-loop/H-cleft interaction, likely explaining why CAP only modestly increases depolymerization in the absence of cofilin (15, 16).

In the presence of cofilin, HFD effects were analyzed relative to the pointed-end cofilactin structure (Fig. 2B). As before, binding of one HFD to the first cofilactin subunit did not alter its conformation or position and is thus unlikely to accelerate depolymerization beyond cofilin alone. By contrast, binding of a second HFD to the second cofilactin subunit induced a pronounced displacement of the first subunit of up to ~7.6 Å (Fig. 2K). This displacement resolves a steric clash with the second HFD, further exacerbated by the undertwist of the short-pitch helix induced by cofilin, which simultaneously rotates the first and second HFDs toward one another (Fig. 2J). The first subunit is displaced together with its bound cofilin and HFD, suggesting that all three dissociate as a unit (Fig. 2K and movie S2).

Although the OD was not observed in the maps, the two HFDs seen at the pointed end likely belong to the same oligomer, as oligomerization is crucial for CAP-mediated pointed-end tracking and depolymerization (15, 16, 32). Substituting CAP’s OD with domains of known oligomerization state shows that depolymerization increases progressively from monomer to dimer, trimer, and tetramer (15, 32), indicating that while HFD drives depolymerization, oligomerization is required for full activity. The precise oligomeric state of CAP, however, remains unsettled, with some studies reporting a hexamer (16, 33, 34) and others a tetramer (32, 35). Given the central importance of oligomerization for CAP activity, we revisited this question taking advantage of our human cell–expressed, full-length CAP, which is natively folded. CAP co-purifies with substoichiometric amounts of actin (fig. S25A), likely bound to the high-affinity site within its C-terminal region (36). Upon addition of excess actin, we obtained a stoichiometric complex (fig. S25B), which was stabilized by GraFix (gradient fixation) (37) and analyzed by mass photometry, yielding a measured mass of 364 ± 77 kDa, consistent with a CAP-actin tetramer (theoretical mass: 376 kDa) (Fig. 2L).

These results support a depolymerization model in which conformational changes induced individually by cofilin and CAP are amplified when both are bound at the pointed end, resulting in unfavorable subunit addition and accelerated dissociation. Effectively, cofilin and CAP cooperate to alter the on- and off-rates of pointed-end subunits, shifting the equilibrium toward depolymerization. Subunits dissociate along with their bound HFD and cofilin and are presumably captured by C-CAP (CAP’s C-terminal domain, residues 216 to 475), which recycles monomers by releasing HFD and cofilin and promoting ADP-to-ATP exchange (15, 34, 36). Tetramerization may enable CAP to remain processively bound during depolymerization (16), with two arms engaging the two terminal subunits, one recycling the dissociating subunit, and the fourth poised to replace the departing one (movie S2).

The barbed end with bound cofilin, CP, or both–Uncapping and depolymerization mechanisms

CP is the most abundant and ubiquitous barbed-end capping protein and an essential regulator of actin-based motility (38). It is a mushroom-shaped heterodimer composed of structurally related CPα and CPβ subunits (39). Of the six barbed-end structures obtained here (Fig. 3 and movie S3), one shows CP bound to the two terminal actin subunits in the absence of cofilin (Fig. 3A). At 2.6 Å, this represents the highest-resolution filament-end structure to date and serves as a reference for analyzing the cofilin-bound barbed-end structures. The two terminal actin subunits at the CP-capped barbed end adopt an F-actin conformation, similar to the free barbed end (7). CP spans the two long-pitch helices, inserting its helically folded α- and β-tentacles, corresponding to the C-terminal region of each heterodimer subunit, into the H-cleft of the second-to-last and last subunits, respectively. As described below, CP’s binding site either overlaps with or is reshaped by cofilin’s interaction with barbed-end subunits.

Fig. 3. Barbed-end structures with CP, cofilin, or both.

(A to F) Cryo-EM maps of the barbed end with bound CP (CPα, magenta; CPβ, pink), cofilin (yellow), and structures of CP with cofilin or cofilin alone. (G) In all cofilin-containing structures, the two barbed-end subunits (dark and light blue) adopt a twisted G-actin–like conformation distinct from the flat conformation of mid-filament subunits, the free barbed end, or the CP-bound barbed end (gray) (7). (H) Bottom-view showing how the cofilin-induced undertwist of the short-pitch helix rotates CP, which weakens its interaction with the barbed end. (I) CP retains its barbed-end–bound conformation (7), independent of cofilin binding and of the F- to G-actin transition of barbed-end subunits, distinct from the more curved conformation of CP’s mushroom head in the unbound state (gray) (39). (J) CP β-tentacle (magenta in transparent cryo-EM map density) inserted into the H-cleft of the last subunit (dark blue map) from part A. (K) Left to right, position of the α-tentacle in the structures from parts (A) to (C), showing progressive displacement by cofilin, first due to the F- to G-actin transition and undertwist and then through direct competition.

In the next structure, cofilin is bound along the filament sides but not at the barbed end itself, which remains occupied by CP (Fig. 3B). Cofilin induces the characteristic cofilactin conformation in all filament subunits, including the two terminal ones (Fig. 3G). The only distinction is that the H-cleft of the terminal subunits adopts a G-actin–like conformation, less open than in cofilactin (since cofilin is not bound at their barbed end) and also distinct from the F-actin–like H-cleft observed at both the free and CP-capped barbed end (7) (Fig. 1D). The F- to G-actin transition of barbed-end subunits likely results from cofilin displacing their D-loop from the H-cleft of the subunit above, underscoring the importance of this interaction in stabilizing the F-actin conformation. As observed with CAP’s HFD at the pointed end (Fig. 2J), CP rotates by ~9° due to the cofilin-induced undertwist of the short-pitch helix (Fig. 3H and movie S3). The F- to G-actin transition of terminal subunits and the associated remodeling of their H-cleft weaken CP’s interaction, particularly through the α-tentacle and, to a lesser extent, the β-tentacle (Fig. 3, J and K). Despite these changes, CP retains the same overall conformation as in the absence of cofilin, characterized by a flatter mushroom head (7) and distinct from the more bent conformation observed in isolation (39) (Fig. 3I).

In another structure, cofilin binds not only to the filament sides but also to the H-cleft of the second-to-last subunit (Fig. 3C), fully displacing the α-tentacle of CP from this position (Fig. 3K, right), a critical element of CP’s interaction with the barbed end (40). While CP is further displaced relative to its position in the previous two structures (movie S3), its overall conformation remains unchanged (Fig. 3I). CP remains loosely bound to the barbed end through its β-tentacle, which changes little across the three CP-containing structures (Fig. 3, A to C and J), and through CPα’s loop R260-D270, preceding the α-tentacle (Fig. 3I), which inserts in the groove formed between the two long-pitch helices at the barbed end. The conformation of the second-to-last barbed-end subunit changes little upon cofilin binding, except that its H-cleft becomes slightly more open to accommodate cofilin’s N terminus (Fig. 1, C and D).

In the following two structures, CP is fully displaced from the barbed end, and cofilin occupies the H-cleft of either the second-to-last subunit or both terminal subunits (Fig. 3, D and E). The cofilin molecules bound to the H-cleft of the two terminal subunits are poorly resolved in the cryo-EM maps, presumably because they interact with only one actin subunit rather than the two engaged along the filament sides (Fig. 1A). Thus, cofilin binding to the two barbed-end subunits appears transient. Consistently, in the final structure, the barbed end is unoccupied, with cofilin bound only to the filament sides (Fig. 3F). In this structure, the two terminal subunits have a G-actin–like conformation and H-cleft.

Collectively, these structures show that when cofilin reaches the barbed end, either through cluster growth or severing at the barbed-end boundary of a cluster, actin subunits adopt a G-actin–like conformation and the short-pitch helix becomes undertwisted. This has several consequences, including CP dissociation, inhibition of CP re-capping, transient capping by cofilin, unfavorable subunit addition, and enhanced subunit dissociation due to loss of the key D-loop/H-cleft interaction. Together, these effects are expected to slow elongation or promote depolymerization, consistent with biochemical observations (14).

Concluding remarks

The proteins studied here lie at the heart of actin dynamics. Several factors underscore CAP’s central importance, from its conservation across eukaryotes, including Leishmania, whose ancient cytoskeleton contains fewer than ten proteins (41), to its abundance in mammalian cells, where it is present at a ~1:4 ratio to actin (42). Its role in filament disassembly is also well documented, with studies showing that CAP depletion leads to F-actin accumulation and reduced filament turnover (42–45). Likewise, cofilin and CP are two of five essential components of actin-based in vitro motility, together with actin, Arp2/3 complex, and one of its WASP-family activators (46). While in the 25 years since that landmark study the field has focused primarily on mechanisms of filament assembly, here, we have addressed the equally important mechanisms of disassembly that replenish the monomer pool for new rounds of polymerization. At the pointed end, we have shown how CAP cooperates with cofilin to induce conformational changes that promote subunit dissociation while preventing reincorporation, consistent with biochemical findings (15, 16). At the barbed end, we have shown how cofilin drives a transition toward the G-actin conformation and alters the intersubunit twist to promote CP and monomer dissociation while limiting their rebinding, also consistent with biochemical evidence (14). A recent study shows how coronin, cofilin, and AIP1 cooperate to accelerate filament severing (26). Together, the two studies provide a unified framework for understanding how actin filaments are disassembled and recycled for new rounds of polymerization.

MATERIALS AND METHODS

Protein cloning, expression, and purification

Human CAP-1 (UniProt Q01518) was cloned from a human embryonic kidney 293 cDNA library using primers designed to add an N-terminal 6×His tag and a C-terminal Strep tag (primers listed in table S4). The polymerase chain reaction product was inserted into vector pJC7 (Addgene no. 177872) for expression in mammalian Expi293F cells. This construct served as the template to generate construct N-CAP (residues 1 to 220) cloned into vector pRSFDuet-1 (Novagen) for bacterial expression. The cDNA for cofilin-2 (UniProt Q9Y281) was similarly obtained and cloned into vector pTYB11 (New England Biolabs).

Full-length human CAP (construct 6×His-CAP-Strep) was expressed in Expi293F cells using PEI MAX transfection reagent (Polysciences) at a 3:1 (v/w) PEI:DNA ratio. Cell suspensions were incubated with shaking for 72 hours at 37°C in 8% CO₂, pelleted at 3000g for 10 min, and stored at −80°C. Pellets were resuspended in 20 mM Hepes (pH 7.4), 300 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and cOmplete Mini protease inhibitor cocktail (Roche). Cells were lysed with a Dounce homogenizer, and lysates were clarified by centrifugation at 20,000 rpm for 30 min at 4°C. The supernatant was loaded onto a Strep-Tactin column (IBA Lifesciences) and eluted with 40 mM biotin. The eluate was concentrated using an Amicon ultrafiltration device and further purified on a 26/60 Superdex 200 size exclusion column (Thermo Fisher Scientific).

Bacterial expression of His-N-CAP and cofilin was performed in ArcticExpress (DE3) RIL cells (Agilent Technologies) grown in terrific broth at 37°C for 7 to 9 hours to an optical density of 1.0. Expression was induced with 0.5 mM isopropylthio-β-d-galactoside and continued for 24 hours at 10°C. Cell pellets were lysed with a microfluidizer (Microfluidics) and clarified by centrifugation at 20,000 rpm for 20 min at 4°C. His-N-CAP was purified on a Ni–nitrilotriacetic acid affinity column and frozen for later use. Cofilin was purified through a chitin affinity column (New England Biolabs) in 20 mM Hepes (pH 7.5), 500 mM NaCl, and 1 mM EDTA. Self-cleavage of the intein was induced with 50 mM dithiothreitol (DTT) overnight at room temperature, and the protein was collected in the lysis buffer. Cofilin was then dialyzed in 20 mM MES (pH 6.0), 25 mM NaCl, 1 mM EDTA, and 1 mM DTT and further purified using a Mono S ion-exchange column (Pharmacia). The cofilin-containing peak was dialyzed in 20 mM Hepes, 50 mM NaCl, 1 mM EDTA, and 1 mM DTT, concentrated to 580 μM, and frozen for future use. Rabbit alpha skeletal muscle actin (UniProt P68135) and human CP (isoforms α1 and β2, UniProt P52907 and P47756-2) were obtained as we have described (7).

Cryo-EM grid preparation and data collection

Actin (25 μM) and CP (7.5 μM) were incubated for 1 hour in F-buffer [20 mM Hepes (pH 7.5), 50 mM KCl, 1 mM EGTA, 1 mM MgCl₂, 1 mM ATP, and 1 mM DTT]. Five minutes before freezing, full-length CAP (20 μM) was added ± cofilin (20 μM), yielding final concentrations of actin (15 μM), CP (3 μM), ± cofilin (10 μM), and CAP (7 μM). Samples were immediately applied to glow-discharged (PELCO easiGlow) 300-mesh R1.2/1.3 Quantifoil holey-carbon grids (Electron Microscopy Sciences), blotted for 3.5 s with Whatman 41 filter paper, and plunge-frozen in liquid ethane using a Vitrobot Mark IV.

After noticing that HFD was the only CAP domain observed in cryo-EM maps, we switched to N-CAP, which can be concentrated further to enhance the population of N-CAP–occupied pointed ends. Vitrification was performed as described above, with final protein concentrations of actin (35 μM), CP (5.8 μM), ± cofilin (30 μM), and N-CAP (31 μM).

Cryo-EM datasets were collected on three FEI Titan Krios microscopes (UPenn, NCEF, and NCCAT), each operated at 300 kV and equipped with a Gatan K3 direct electron detector and energy filter. Images were recorded in counting mode with pixel sizes of 0.41 to 0.54 Å/pixel and a defocus range of −2.5 to −0.5 μm.

Cryo-EM data processing

A total of 39,142 movies with cofilin and 24,913 without were imported into cryoSPARC version 4.7 (47). Patch motion correction (2× binning) and patch CTF estimation were performed with default settings. Micrographs were curated based on CTF fit resolution, mean intensity, relative ice thickness, total frame-motion distance, and manual inspection, yielding 37,100 accepted images with cofilin and 22,733 without. Figures S2 and S3 present the particle-picking and processing workflows for all structures, and tables S1 to S3 summarize the final data and model statistics.

For barbed- and pointed-end reconstructions, ~7000 filament ends were manually picked to train an initial Topaz model (48). These particles were used in ab initio reconstruction and three-dimensional (3D) classification to separate barbed and pointed ends. The resulting subsets were used to train a refined Topaz model that identified 3,129,053 end particles with cofilin and 3,259,536 without. These particle stacks were subjected to ab initio refinement with multiple classes. The best classes underwent heterogeneous refinement against a junk volume to remove spurious particles. The resulting cleaned stacks were subjected to 3D classification to separate barbed and pointed ends from mid-filament particles.

For both ends, 3D classification focused on the two terminal actin subunits was used to separate filament ends misaligned by one or more subunits. The selected particles were then realigned, re-extracted, and combined using non-uniform refinement to generate initial barbed- and pointed-end maps with weak density for end-bound proteins.

At the pointed end, 3D classification with masks focused on the HFD bound to the first or second subunit produced maps (± cofilin) of the pointed end lacking HFD or containing HFD on the first subunit or on the first and second subunits. Each class underwent nonuniform refinement, yielding the final pointed-end particle stacks.

At the barbed end, 3D classification with masks focused on weak CP density yielded classes with or without CP. Particles lacking CP were further classified with masks centered on cofilin bound to the last or second-to-last subunit, producing classes with cofilin absent or bound to the second-to-last subunit or both terminal subunits. Among CP-containing particles, classes either lacked cofilin or showed cofilin bound to the second-to-last subunit. Focused refinement of the latter produced two subclasses: one in which CP’s α-tentacle occupied the H-cleft of the second-to-last subunit and another in which the α-tentacle was fully displaced and cofilin occupied that cleft. Each class was subjected to nonuniform refinement to generate the final barbed-end particle stacks.

To obtain the mid-filament cofilactin map, actin filaments were manually selected to generate templates for a cryoSPARC Filament Tracer task with a starting filament diameter of 90 Å and a segment separation of 30 Å (0.3× diameter). The 1,951,405 initially picked particles underwent ab initio reconstruction followed by helical refinement. Heterogeneous refinement against a junk volume removed spurious particles, yielding 1,560,565 particles for final helical refinement (order 8) using initial helical parameters of −162° twist and 27.6 Å rise, producing the final volume for postprocessing.

For all structures, except those without CP, final particle stacks underwent global and local CTF refinement followed by reference-based motion correction. End structures were then locally refined with masks focused on the terminal subunits, end-bound proteins, or the filament core. The resulting maps were combined in ChimeraX (49) for model building and refinement.

Model building and refinement: Model building and refinement were performed in Coot (50) and Phenix (51), respectively. The highest-resolution mid-filament cofilactin structure was built first and served as the reference for all cofilin-containing structures. An AlphaFold (52) model of actin-cofilin was rigid-body fit into the cryo-EM map, and nonvisualized regions, such as actin’s D-loop, were removed. Water molecules were added automatically in Phenix using a conservative density cutoff of 5 σ and minimum/maximum distances of 2.4 Å and 3.2 Å from protein atoms. To regularize stereochemistry and optimize the map fit, real-space and temperature-factor refinements were carried out in Phenix with symmetry imposed. For structures without cofilin, our previous F-actin structure served as the starting model (7). An AlphaFold model of CAP’s HFD was fitted into pointed-end maps containing this domain. The refinement of CP-containing barbed-end structures used our prior structure of CP at the barbed end (7) as the starting reference model. A small number of water molecules observed in the CP-bound barbed-end map were added manually. Figures were prepared in PyMOL (Schrödinger, LLC) and ChimeraX (49). Final model quality and refinement statistics are provided in tables S1 to S3.

Measuring the mass of the CAP–actin complex

Expi293F cell–expressed CAP (~1 μM), which copurifies with substoichiometric amounts of actin, was incubated with excess actin (5 μM) on ice for 1 hour to ensure full saturation. The complex, in 20 mM Hepes (pH 7.5) and 100 mM NaCl, was loaded onto a 5 to 30% glycerol gradient prepared with a Gradient Master (BioComp Instruments), centrifuged at 40,000 rpm for 16 hours at 4°C using a Beckman SW 60 Ti rotor, and fractionated with a Piston Gradient Fractionator (BioComp Instruments). Gradients were run in duplicate, with and without fixation (37). For fixation, 0.125% (v/v) glutaraldehyde was added to the 30% glycerol solution, and the reaction was quenched after centrifugation with 40 mM glycine hydrochloride (pH 7.5). SDS–polyacrylamide gel electrophoresis (PAGE) of the nonfixed gradient fractions identified those containing stoichiometric CAP-actin complex (fig. S25B), and the corresponding fractions from the fixed gradient were concentrated threefold using a 50 kDa Amicon Ultra centrifugal filter (Millipore) for mass photometry analysis (Fig. 2L).

Mass photometry was performed on a Refeyn TwoMP-0220 instrument, calibrated with bovine serum albumin (monomer, 66.4 kDa; dimer, 133 kDa) and thyroglobulin (660 kDa). Before measurements, samples were dialyzed for 20 min against 20 mM Hepes (pH 7.5) and 100 mM NaCl to reduce the glycerol content and then diluted sixfold in the same buffer. Data were acquired with AcquireMP and analyzed with DiscoverMP in histogram and mass-plot modes (bin width, 5 kDa) using automated Gaussian peak detection to estimate molecular mass, particle count, and sigma values. Final histograms and Gaussian fits were generated in GraphPad Prism version 10.2.2.

Supplementary Materials

The PDF file includes:

Figs. S1 to S25

Tables S1 to S4

Legends for movies S1 to S3

References

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S3