Structure of Arp2/3 Complex in Its Activated State and in Actin Filament Branch Junctions

See allHide authors and affiliations

Science  28 Sep 2001:
Vol. 293, Issue 5539, pp. 2456-2459
DOI: 10.1126/science.1063025


The seven-subunit Arp2/3 complex choreographs the formation of branched actin networks at the leading edge of migrating cells. When activated by Wiskott-Aldrich Syndrome protein (WASp), the Arp2/3 complex initiates actin filament branches from the sides of existing filaments. Electron cryomicroscopy and three-dimensional reconstruction of Acanthamoeba castellanii andSaccharomyces cerevisiae Arp2/3 complexes bound to the WASp carboxy-terminal domain reveal asymmetric, oblate ellipsoids. Image analysis of actin branches indicates that the complex binds the side of the mother filament, and Arp2 and Arp3 (for actin-related protein) are the first two subunits of the daughter filament. Comparison to the actin-free, WASp-activated complexes suggests that branch initiation involves large-scale structural rearrangements within Arp2/3.

The Arp2/3 complex is the cellular factor that generates new filaments in a site-directed, signal-controlled fashion at the leading edge of motile cells (1) and forms identical branches in vitro (2). The complex binds the sides and slow-growing (pointed) ends of actin filaments (3). The dendritic actin networks generated by the Arp2/3 complex have been thought to be critical for producing the force that drives lamellipodia protrusion. Conserved among eukaryotes, this seven-subunit complex consists of two actin-related proteins, Arp2 and Arp3, and five novel proteins. Structural models of the Arps (4) suggested that the complex provides a nucleation site for a new actin filament. The COOH-terminal region of WASp family members containing the WASp homology 2 (W) and acidic (A) domains, together with preassembled actin filaments, activates the Arp2/3 complex, thus promoting barbed (fast-growing) end nucleation of new (daughter) filaments (5–8). Electron microscopy, chemical cross-linking, yeast genetics, sedimentation, and hydrodynamic modeling provided evidence for the organization of the complex and identified subunits that interact with actin, WA, and profilin [reviewed in (9)]. Crystallography has not yet provided direct information on the structure of the complex. We used electron cryomicroscopy and single-particle analysis to obtain the structure of WA-bound Arp2/3 complex at 3.2 nm resolution. Image analysis of actin filament branches formed by the Arp2/3 complex suggests that the Arps form the first two subunits in the daughter filament. A large conformational change, involving one of the Arps and at least one small subunit, appears to occur during branch formation.

Unstained, fully hydrated Acanthamoeba Arp2/3 complex bound to the WA domain of Scar, a WASp-like protein, was imaged in vitrified buffer (Fig. 1A) (10). Three independent three-dimensional (3D) reconstructions were calculated (11), and the high similarity among these reconstructions (>92% correlation) verified the reliability of the structure. The 0.5 Fourier shell correlation criterion (12) indicates a resolution of 3.2 nm (Fig. 1C). Gold labeling was used to localize the NH2-terminus of WA within the reconstruction (13). Single-particle analysis of Acanthamoeba Arp2/3 complex without the WA domain bound did not converge to a consistent 3D reconstruction. It is conceivable that the WA domain stabilizes the free complex and locks it into a narrow range of conformations.

Figure 1

Single-particle reconstructions ofAcanthamoeba and yeast Arp2/3 complex with WA. (A) Electron micrographs and reconstructions. Protein is displayed in bright shades. Column 1, raw images of yeast Arp2/3 complex in uranyl acetate; column 2, Acanthamoeba Arp2/3 complex in vitrified buffer; column 3, 2D class averages of theAcanthamoeba data set; column 4, corresponding projections of the final reconstruction; column 5, corresponding surface views. Three roughly orthogonal views are shown: row 1, side view; row 2, front view; and row 3, top view. (B) Angular distribution of the 88 class averages obtained from 4414 images ofAcanthamoeba complex used to generate the 3D structure in the first row of (D). Surface views for some angles are shown for reference. Each class average is represented as a color-coded circle, where red corresponds to larger number of images contributing to the average (maximum 306) and blue corresponds to a low number of images (minimum 8). The diagram shows that the reconstruction is free of artifacts caused by missing views. (C) Fourier shell correlation between two reconstructions calculated from two randomly selected half sets of the respective data. Red is theAcanthamoeba data set, blue is yeast. (D) 3D reconstructions of Acanthamoeba (upper row) and yeast (lower row) Arp2/3 complex. The wire mesh isosurface (teal) encloses the complete molecular volume of the complex (relative molecular mass of 220 kD). The solid representation encloses 50% of the molecular mass. The three columns show views from the side (left column), front (middle column), and top (right column). The straight arrows indicate the axis and the circular arrows indicate the turning direction that generate a particular view from the side view (row 1).

Saccharomyces cerevisiae Arp2/3 complex bound to the WA domain of yeast Bee1, a member of the WASp family, was imaged in uranyl acetate stain after air-drying (Fig. 1A) (14). Processing of this yeast data set converged to a 3D reconstruction similar to the Acanthamoebastructure (Fig. 1D). The Fourier shell correlation indicates a resolution of about 3.9 nm for the yeast reconstruction (Fig. 1C). The correlation between the yeast and the Acanthamoebastructures at 3.9 nm resolution is 90%. Thus, at this resolution, the complexes are nearly identical as anticipated from the conserved sequences of their subunits.

When contoured at a threshold containing 100% of the molecular mass (220 kD), WA-bound Arp2/3 complexes are oblate ellipsoids with dimensions of 13 nm by 9 nm by 5 nm (Fig. 1D). Contouring at a higher threshold reveals a region of lower density between two large lobes, giving the reconstructions a horseshoe shape. This is similar to images of shadowed complexes, which indicate a cleft between two domains of the complex (2, 15).

Electron cryomicroscopy was used to image frozen-hydrated actin filament branches produced by Acanthamoeba Arp2/3 complex activated by Scar-WA (16). Individual actin monomers and extra density at branch sites were clearly visible in filtered images (Fig. 2A). From the unfiltered images, 167 branches were selected and used to generate a two-dimensional (2D) image of the projection view. This 2D reconstruction clearly reveals subunits in both actin filaments (Fig. 2B) and is fully compatible with quick-frozen, deep-etched, rotary-shadowed samples of the same preparation (Fig. 2, C and D). Fourier-space analysis suggests a resolution of ∼2.8 nm for the branch reconstruction. Myosin decoration of the branches formed by the yeast Arp2/3 complex in the presence of Bee1-WA (17) verified that the pointed ends of both filaments are directed toward the branch (Fig. 2E), as they are in extracted cells (18). This confirms that Arp2/3 complex nucleates actin filaments that grow at their barbed ends (2). In the 2D reconstruction, the mother filament continues unaffected through the branch, and an analysis of the diffraction pattern of the mother filament alone shows an undisturbed helical actin symmetry. These two findings indicate that the Arp2/3 complex is not inserted into the mother filament. The helical structure of the daughter filament starts about 6 nm from the side of the mother filament. Three bridges of density connect the pointed end of the daughter filament to three actin subunits in the mother filament. The density distribution in the connecting region indicates a thickness of about 5 nm. The quick-frozen, shadowed samples provide evidence that the contact occurs slightly out of plane (Fig. 2, C and D).

Figure 2

Electron micrographs of actin filament branches mediated by Arp2/3 complex. (A) Frozen hydrated actin branches made by Acanthamoeba Arp2/3 complex. For visualization purposes, a Gaussian real-space filter was applied to the image. (B) 2D reconstruction using 167 images of actin filament branches. (C) A gallery of quick-frozen, deep-etched, rotary-shadowed actin branches mediated byAcanthamoeba Arp2/3 complex. Bar is 7 nm. Note that the daughter filament appears to be elevated in respect to the mother filament for branches pointing to the left (lower two) and on a lower plane than the mother filament when branches point to the right (upper two). (D) A quick-frozen, shadowed actin filament with a branch pointing to the left. Note the visibility of the 37.5-nm right-handed half-twist crossover repeats. (E) Negatively stained actin branches mediated by yeast Arp2/3 complex, decorated with expressed myosin catalytic domain (no light chains). The arrows follow the direction of the bound myosin heads that show the appearance of arrowheads indicating the polarity of the actin filaments.

The helical symmetry of actin filaments and their defined polarity allowed us to fit atomic models of actin (19) unambiguously into the 2D reconstruction of both the mother and daughter filaments (Fig. 3A). If one assumes that all of the subunits of the mother and daughter filaments are actin, the remaining projection density at the branch junction is insufficient to accommodate the entire Arp2/3 complex. On the other hand, if two of the filament subunits are Arps [as postulated (20)], the complete complex can be accommodated within the volume. This is reasonable on structural grounds, because Arps share high sequence identity to actin and are about the same size as actin, thus are likely to be folded similarly (4). In our model (Fig. 3A), the Arps are the first two subunits of the daughter filament. The available structural data cannot distinguish whether Arp2 is at the pointed end of the daughter filament or in the second position. However, the pattern of conserved surface residues between actin and Arp2 (4) leads us to speculate that Arp2 makes the majority of contacts with actin at the growing, barbed end of the daughter filament and Arp3 is positioned at the pointed end of the branch (Fig. 3A). In this model, three subunits (p40, p35, and p14) of the complex make contact with subdomain 1 of actin subunits in the mother filament. This arrangement is fully consistent with previous cross-linking (2, 15), biochemical (8), and genetic (21) studies. After adding Arp2/3 complex to preformed actin filaments, we saw—in both frozen-hydrated and freeze-etched, quick-frozen samples—masses attached to the sides of the filaments whose size and shape (Fig. 3B) correspond very closely to those of the Arp2/3 complex model built from the 2D branch-junction reconstruction (Fig. 3, A and C). Because the filaments were polymerized before addition of Arp2/3 complex, our observation of the complex bound to their sides is consistent with the hypothesis that branches form on the side of mother filaments (22), and it does not support the hypothesis that one of the Arps is incorporated into the mother filament and the other into the daughter filament (23).

Figure 3

Molecular model of the actin-bound Arp2/3 complex at the branch junction. (A) Model of actin filament branches mediated by Acanthamoeba Arp2/3 complex. The 2D reconstruction from Fig. 2B is shown in the background for reference. The backbone of the molecular model of filamentous actin, fitted to the 2D reconstruction, is shown in pink. The first two subunits of the daughter filament, shown in red and green backbone presentation, are assigned to be Arps. The other five subunits of the complex are assigned to the rest of the projection density using the proximity information from cross-linking, genetic, and yeast two-hybrid experiments [summarized in (9)]: p40 (purple); p35 (pink); p18 (yellow); p19 (light blue); p14 (orange). The size of the regions was chosen to approximate the respective molecular weights, assuming a thickness of about 5 nm. The barbed ends of the filaments are toward the top of the figure. (B) Images of Arp2/3 complex from Acanthamoeba bound to the side of mother filaments. All samples in the actin experiments were prepared in the presence of the activator (WA). Upper row: in vitrified buffer, a Gaussian real-space filter was applied to the images for visualization. Lower row: quick-frozen, deep-etched, rotary-shadowed specimen. Note the similarity of the position and shape of the complex to that seen at the branch junction shown in (A). Bar is 7 nm. (C) Average of six aligned images of Arp2/3 complex bound to the side of filaments in vitrified buffer overlaid with backbone presentation of an actin filament (pink) and backbone presentations of Arp2 (red) and Arp3 (green) in positions similar to those in (A). The remaining density was assigned to the other subunits and colored as in (A). (D) Density representations of the models of actin-bound (green) and the free, WA-activated (as shown in Fig. 1D, gray) Arp2/3 complex. The density for the branched model was calculated using a filament-like configuration for the two Arps and using the remaining projection density assuming a thickness of ∼5 nm. The view was generated by turning the model of the complex in (A) by 90° counter-clockwise. This view corresponds to the front view depicted inFig. 1. On the right, the best fit of the density representing the branched model (green) into the 3D reconstruction of the free Arp2/3 complex (gray) is shown. Note that the density corresponding to Arp2 cannot be accommodated by the reconstruction of the free complex. A possible large-scale conformational change of the free, activated complex upon binding to actin, a rearrangement in the position of Arp2, is indicated by an arrow. (E) The density for the branched model (green, left) and the best fit (right) of the branched model density (green) into that of the free Arp2/3 complex (gray). The binding site of the WA NH2-terminus (gold), as assigned by the labeling. With this fit the WA NH2-terminus is in close proximity to the F-actin interface of the Arp2/3 complex. The orientation of this view matches that of the projection densities in (A) and (C) and corresponds to the side view depicted in Fig. 1. A possible rearrangement [as in (D)] of Arp2 upon binding to the filament is indicated by an arrow. (F) The best fit of the density of the branched model (green) into that of the free Arp2/3 complex (gray). WA NH2-terminus is shown in gold. This view corresponds to the top view depicted in Fig. 1. The straight and circular arrows on the right indicate the axis and turning direction that generate this view from the view in (E) (the side view in Fig. 1).

If the Arps form the first two subunits of the daughter filament and have the helical symmetry of actin filaments, the Arp2/3 complex must change shape during formation of the branch. The reason for this is that two Arps arranged along the genetic helix of the actin filament cannot fit into the planar structure of the free Arp2/3 complex (Fig. 3, D through F). Thus, at least one of the Arps must rearrange to fit into the projection view of the actin-bound structure. Although it is possible to postulate reasonable subunit rearrangements that are consistent with the available data, molecular details of this conformational change will require an atomic structure of the complex and additional analysis.

A conformational change coupled to branching nucleation is consistent with the observation that preformed actin filaments are secondary activators of nucleation (8, 20, 2325). The reconstructions support the model (9) that activation of Arp2/3 involves rearrangement of the subunits in the complex to bring the Arps together in a filamentous arrangement. Interestingly, gold labeling places the WA domain close to the interface between the mother filament and Arp2/3 complex (Fig. 3, E and F), suggesting that WA and the mother filament may cooperate to facilitate the rearrangement of subunits in the complex, bringing the Arps together in a conformation suitable for barbed-end nucleation of the branch.

  • * To whom correspondence should be addressed. E-mail: dorit{at}


View Abstract

Stay Connected to Science

Navigate This Article