Arp2/3 complex is bound and activated by two WASP proteins

Abstract

Actin related protein 2/actin related protein 3 (Arp2/3) complex nucleates new actin filaments in eukaryotic cells in response to signals from proteins in the Wiskott–Aldrich syndrome protein (WASP) family. The conserved VCA domain of WASP proteins activates Arp2/3 complex by inducing conformational changes and delivering the first actin monomer of the daughter filament. Previous models of activation have invoked a single VCA acting at a single site on Arp2/3 complex. Here we show that activation most likely involves engagement of two distinct sites on Arp2/3 complex by two VCA molecules, each delivering an actin monomer. One site is on Arp3 and the second is on ARPC1 and Arp2. The VCAs at these sites have distinct roles in activation. Our findings reconcile apparently conflicting literature on VCA activation of Arp2/3 complex and lead to a new model for this process.

The actin cytoskeleton is dynamically remodeled to provide the protrusive forces needed for cell migration, adhesion, polarization, and vesicle trafficking. The polymerization of actin monomers into polarized filaments generates part of the mechanical energy driving these processes (1). The actin related protein 2/actin related protein 3 (Arp2/3) complex plays a central role in cytoskeletal dynamics by controlling filament nucleation. This asymmetric seven-protein assembly nucleates new filaments from the side of existing filaments (2, 3). Two of these proteins, Arp2 and Arp3, are related to actin (4, 5), and serve as the first two subunits of the nascent filament (6). The complex also contains five additional proteins, ArpC1, ArpC2, ArpC3, ArpC4, and ArpC5.

Actin nucleation by Arp2/3 complex is greatly stimulated by the conserved C-terminal VCA element of proteins in the Wiskott–Aldrich syndrome protein (WASP) family (7, 8). Functionally, the VCA (also referred to as the WCA) can be divided into different regions, which contribute to different aspects of Arp2/3 activation. These are an actin binding V region (alternatively referred to as a WH2 region), a central connecting C region, which bears sequence similarity to the V region (9, 10) and a C-terminal A (acidic) region. Together, the C and A regions (referred to as CA) contribute most of the energy of association for Arp2/3 complex (11) and can drive a conformational change (12–14). The V region binds actin monomer (11, 15) and delivers an initial subunit to the nascent daughter filament (11, 16–18). In addition to CA-induced conformational changes and V region delivery of actin, engagement of the mother filament appears to be required for Arp2/3-dependent filament nucleation, thus ensuring the branched structure of the resulting network (19, 20).

Crosslinking studies have identified Arp3 (21–23), Arp2 (21–23), ArpC1 (21–24), ArpC3 (21, 22, 25), and ArpC5 (25) as potential binding surfaces of the VCA. It has been difficult to merge these studies into a coherent structural picture of VCA binding. For example, the A region has been suggested to bind both ArpC1 (26) and Arp3 (21), but these subunits do not contact each other in either the inhibited crystal structure (27) or in the EM reconstruction of the filament branch (6). Further, while initial studies emphasized the importance of the CA:Arp3 contacts and proposed that the first actin was delivered to Arp3 (11, 21, 28), a more recent SAXS study placed the first actin on the Arp2 (9). Resolving where the CA region binds Arp2/3 complex and where the first actin is delivered are essential components of a mechanistic description of the nucleation process.

Implicit in these conflicting models is that one VCA delivers one actin to Arp2/3 complex to trigger nucleation. However, we recently showed that dimerization of WASP-family VCA domains substantially increases their activity, suggesting the existence of multiple VCA-binding sites on Arp2/3 complex (29). Competition binding experiments suggested that dimeric VCAs engage two sites on Arp2/3 complex (29), one favored by VCA and one favored by a distinct Arp2/3 ligand, cortactin (8, 22). This raised the possibility that activation of Arp2/3 complex, even by monomeric VCAs, may occur through engagement of both binding sites. Recent technical advances have allowed us to directly explore this possibility (30, 31).

Here, we show that at μM concentrations two VCA peptides bind Arp2/3 complex simultaneously. Cross-linking data indicate that one acidic region binds ArpC1 and the other Arp3. Both of these VCAs can accommodate actin when bound to Arp2/3 complex. A VCA dimer lacking one of the two V regions shows very low activity, suggesting that delivery of actin by both V regions is needed for high-potency activation. Delivery of actin to Arp3 appears to have the greatest impact on activation. We propose that nucleation proceeds through an assembly of two VCAs, two actin monomers, and Arp2/3 complex and suggest a structural model for this intermediate.

Results

Two Distinct WASP VCA-Binding Sites on Arp2/3 Complex.

We used a multisignal sedimentation velocity analytical ultracentrifugation (SVAUC) approach to find the stoichiometry of an Alexafluor-488 labeled VCA (termed VCA*) binding to Arp2/3 complex (Fig. 1 and Fig. S1 A–C in SI Appendix; see also Supplementary Materials and Methods in SI Appendix). This approach is not based on measurement of mass and thus avoids the uncertainty associated with such measurements in complex systems. Instead, the VCA-Arp2/3 complex stoichiometry is directly determined from the quantity of VCA* cosedimenting with Arp2/3 complex. This is accomplished by integration of c_k(s) distributions derived from a multisignal sedimentation velocity experiment (Fig. 1A and Fig. S1 B and C in SI Appendix) (30, 31). In the presence of excess VCA*, two VCAs cosedimented with each Arp2/3 complex (Fig. 1A). Similar results were obtained under a range of conditions (Table S1 in SI Appendix), indicating that there are two binding sites for VCA on Arp2/3 complex.

Fig. 1.

Two VCAs bind Arp2/3 complex, at distinct sites. Sedimentation velocity analytical ultracentrifugation experiments of 0.54 μM Arp2/3 complex in 9.9 μM VCA* (see Fig. S1A in SI Appendix) were collected using absorbance at 496 nm and interference signals. See Fig. S1 B and C in SI Appendix for sedimentation data and fits. (A) The two signals were used to determine c_k(s) distributions for the VCA* and Arp2/3 complex. The shaded regions are integrated to find the quantity of Arp2/3 and VCA* that cosediment (indicated in panel). See Table S1 in SI Appendix for a summary of observed stoichiometries. (B) The observed ratio of bound VCA* to Arp2/3 complex when the indicated concentrations of cortactin NtA domain were added. A and B are adapted with permission from ref. 31.

The NtA domain of cortactin, a non-WASP-family Arp2/3 complex ligand, competes with VCA dimers for binding to Arp2/3 complex (22, 29). In the presence of NtA competitor, the VCA*-Arp2/3 ratio in the centrifugation experiments dropped from 2∶1 to an apparent saturating value of one VCA* per Arp2/3 complex (Fig. 1B) (31). This result is consistent with two distinct binding sites, one where NtA and VCA compete, and a second where VCA binds with higher affinity than NtA (29). NtA competition for one of two VCA-binding sites is also consistent with previous cross-linking data (22).

We next asked if both of the binding sites can accommodate VCA bound to actin monomers. To accomplish this, we used SVAUC to quantify the amount of VCA-actin cosedimenting with Arp2/3 complex. In the presence of excess VCA-actin, but not actin alone (with Latrunculin B to prevent actin polymerization), all of the Arp2/3 complex sedimented more rapidly than free or VCA-bound Arp2/3 complex (Fig. 2 A and B). Thus, actin associates with Arp2/3 complex in a VCA-dependent manner. Under these conditions, species that contain Arp2/3 complex sediment more rapidly than species that do not. Integrating the signals in the slowly (< 7 S) and rapidly (> 7 S) sedimenting peaks allowed us to determine that approximately two VCA-actins cosedimented with Arp2/3 complex (see Fig. 2C and Supplementary Materials and Methods in SI Appendix). Thus, both VCAs associated with Arp2/3 complex could accommodate an actin (Fig. 2C). In addition to these direct measurements of stoichiometry, we also determined the masses of these species using SVAUC data acquired at a slower speed (Fig. S2A in SI Appendix). A similar mixture of VCA-actin (plus Latrunculin B) and Arp2/3 complex yielded an apparent mass of 333 kDa (Fig. S2 A and B in SI Appendix). This finding is consistent with the mass predicted for two actins and two VCAs bound to one Arp2/3 complex (324 kDa) and is distinguishable from masses obtained from all binary mixtures (Fig. S2B in SI Appendix). Thus, both the stoichiometry and mass measurements indicate a VCA₂∶actin₂∶Arp2/3 assembly.

Fig. 2.

Two actins are delivered by two VCAs to Arp2/3 complex. Sedimentation velocity analytical ultracentrifugation experiments with Arp2/3 complex, N-WASP VCA (second V-region only), and actin. (A and B) c(s) distributions were derived for the indicated combinations of materials. Data were collected at 50,000 rpm. Actin stoichiometry estimates made use of integrated signal intensities for the peaks at 3.8 S in “VCA + Actin” and “Arp2/3 + Actin + VCA,” at 9.2 S in “Arp2/3,” and 11.4 S in “Arp2/3 + Actin + VCA.” “Arp2/3” and “Arp2/3 alone” in A and B are the same data. (C) Calculations for the determination of stoichiometry. Signals are derived from integration of data shown in A and B. See also Fig. S2 in SI Appendix for determination of stoichiometry through estimation of the VCA-actin-Arp2/3 assembly molecular weight.

Identification of the VCA-Binding Sites on Arp2/3 Complex.

Previous studies have found that VCA can be cross-linked to Arp3, Arp2, ArpC1, and ArpC3 (21–25). Here we divide these contacts into two distinct VCA-binding sites and identify the regions of VCA that cross-link to distinct Arp2/3 subunits at each site. To achieve this refined VCA:Arp2/3 binding information, we used two different approaches to introduce labeling agents at specific locations in VCA.

First, we coupled a photo-activatable biotin label transfer agent to individual cysteine residues introduced by mutagenesis into CA peptides. Upon photo-activation this reagent can transfer biotin to Arp2/3 complex subunits within approximately 17 Å of the donor cysteine (Fig. 3A and Fig. S3 A and C in SI Appendix). Adjacent subunits can then be identified by probing blots of SDS/PAGE resolved samples for biotinylated species. Qualitative estimates (Fig. S3B in SI Appendix) of the relative labeling intensity for N-WASP CA donors A462C, E473C, A481C, D486C, E491C, E498C are summarized in Fig. 3B. In general, Arp3 is labeled from sites throughout the C and A regions. Arp2 is labeled from sites in the N terminus of the C region through the beginning of the acidic region. ArpC1 and ArpC3 are both labeled from sites at the C terminus of the A region. Labeling of ArpC3 was identified using a slight mobility shift identified in anti-ArpC3 Western blotting (Fig. S3 D and E in SI Appendix). The ArpC2, ArpC4, and ArpC5 subunits were not labeled by any positions in CA. Similar results were obtained with analogous cysteine mutants in N-WASP VCA (Fig. S3C in SI Appendix) and WAVE1 VCA (Fig. S3 D and E in SI Appendix).

Fig. 3.

Label transfer from CA to Arp2/3 complex. (A) Arp2/3, E491C and E498C N-WASP CA label transfer donors were mixed as shown and illuminated. Where indicated, cortactin NtA domain was added as a competitor. Biotinylated Arp2/3 subunits were visualized using Neutravidin-HRP and identified by comparison to the Memcode Blue stained blot (lane marked MB). Nonspecific cross-reaction with ArpC3 is indicated with “ns.” Label transfer to ArpC3 results in a mobility shift, see Fig. S3 D and E in SI Appendix. (B) Label transfer from N-WASP CA mutants (listed beneath the CA sequence) was scored qualitative for biotinylated band intensity (see Fig. S3B in SI Appendix). Cortactin NtA dependent changes are indicated at right. No labeling of ArpC2, ArpC4, or ArpC5 was observed. (C) Label transfer experiments using Cys-NtA and NtA-Cys as donors. “MB” and “ns” as described in A. (D) N-WASP CA, where W503 was substituted with BPa, was mixed with Arp2/3 complex, and illuminated when indicated. Coomassie brilliant blue staining of products, new bands are indicated by an asterisk. (E) Cross-linking products from D are identified by blotting. Lane 1, molecular weight markers. Lane 2, 4, and 6 are a mixture of CA and Arp2/3 following illumination. Lanes 3 and 5 are Arp2/3 alone. Lane 2 was probed for biotinylated proteins (as in A). Lanes 3 and 4 were probed for Arp3 (using an anti-Arp3 antibody). Lanes 5 and 6 were probed for ArpC1 (using an anti ArpC1B antibody).

It is striking that the C-terminal residues of the acidic region are within range of both ArpC3 and ArpC1. These subunits come no closer to each other than approximately 50 Å in the inhibited Arp2/3 complex structure, a distance much greater than the 17-Å range of the cross-linker. Given this geometry and the VCA₂∶Arp2/3 stoichiometry determined above, it is likely that ArpC1 and ArpC3 are recognized by distinct CA peptides.

To better delineate the two binding sites, we capitalized on the ability of the cortactin NtA to compete with VCA at only one site (Fig. 1B). Classifying label transfer according to its direction of change upon addition of 25 μM NtA (decrease vs. increase or no change), yielded two groups of subunits (Fig. 3 A and B). Upon NtA addition, labeling of Arp3 and ArpC3 decreased while labeling of Arp2 and ArpC1 increased or remained constant. The subunits showing decreased signal upon NtA addition are similar to those that are labeled by NtA: Arp3, ArpC3, and ArpC2 (the latter two are only weakly labeled) (Fig. 3C). Thus, the Arp3/ArpC3 and Arp2/ArpC1 groups likely represent the two VCA-binding sites on Arp2/3 complex.

The label transfer reagent used here has a relatively long range for donation and thus reports on both directly contacted and proximal Arp2/3 complex subunits. In order to limit modification to the former we synthesized a new N-WASP CA peptide where the tryptophan (W503) near the C terminus was substituted with a Benzoyl-phenylalanine (BPa), and a biotin was added at the N terminus for detection. BPa is a photo-activatable short range (< 5 Å) cross-linker that is similar in size and hydrophobicity to tryptophan (32). When this CA peptide is bound to Arp2/3 complex and illuminated, approximately 40% of the Arp3 and ArpC1 bands shift to higher molecular weight species (Fig. 3D), which are biotinylated and contain Arp3 or ArpC1 immunoreactivity (Fig. 3E). Thus, the BPa is within 5 Å of Arp3 and ArpC1, strongly suggesting that the immediate C terminus of the VCA directly contacts these subunits. Because the closest approach of Arp3 and ArpC1 in the inhibited conformation is > 25 Å, these two cross-links must come from distinct binding sites on Arp2/3 complex.

Distinct Functions for the Two VCA-Binding Sites.

Previous mutagenic analyses of monomeric VCA or GST VCA dimers have shown distinct roles for the V, C, and A regions in activation. However, this preceding work could not distinguish whether a region of the VCA is required at one or both sites. To distinguish this, we covalently coupled WASP VCA peptides at their N termini, allowing generation of both VCA homodimers (VCA–VCA) and heterodimers in which one VCA was wild type and the second had V or C replaced by an equal length (GGS)_n linker or lacked A (VCA–xCA, VCA–VxA, and VCA–VC, Fig. 4A and Fig. S4 A–C in SI Appendix).

Fig. 4.

Activation of Arp2/3 complex by deletions of VCA–VCA. (A) Cartoon representation of the VCA–VCA materials. VCA is from human WASP. (B) Actin polymerization rates found in Arp2/3 dependent actin polymerization assays with the indicated concentrations of the VCA materials described in A. (C) Actin polymerization rates found in Arp2/3 dependent actin polymerization assays with VCA dimers with different length N-terminal linkers. Horizontal dashed line is the activity of unstimulated Arp2/3. Activity assays were performed with 10 nM Arp2/3, 4 μM Actin (5% pyrene labeled) in KMEI150, and are the average of three to six repeats. Error bars are the 1σ standard error on the mean, and in many cases are smaller than the symbol. See also Fig. S4 in SI Appendix for VCA dimer sequence details, measurements of the affinity of the materials in A for Arp2/3 and for titrations of the N-terminal linker variants.

All three regions are required at both sites for high affinity toward, and/or maximal activation of, Arp2/3 complex. VCA–VCA produced robust activation at low concentrations and saturated at concentrations near 400 nM (Fig. 4B). VCA monomer had approximately the same saturating activity as VCA–VCA but approaches saturation with a sigmoidal shape and saturates at roughly 3,000 nM, reflecting its lower overall affinity (Fig S4H in SI Appendix; see also ref. 29). The C or A deletions (VCA–VxA and VCA–VC) have potencies between those of VCA monomer and VCA–VCA, their activities both saturate at approximately 1,500 nM (Fig. 4B). VCA–VxA has a saturating activity near that of VCA–VCA, and VCA–VC has a saturating activity somewhat higher than VCA–VCA. We do not currently understand the latter’s higher maximal activity. The effects of the V deletion (VCA–xCA) are distinct. Like VCA–VCA, VCA–xCA has activity that saturates at low concentrations (approximately 200 nM) and high affinity for Arp2/3 complex (Fig. S4 G and H in SI Appendix). However, the saturating activity of VCA–xCA is very low (Fig. 4B). Thus, Arp2/3 complex is bound with high affinity, but the lack of the second V region leads to very weak nucleation activity. This suggests that activation of Arp2/3 complex might require delivery of two actin monomers, not one as generally posited.

We also constructed VCA dimers with a series of N-terminal linkers to ask whether the linker length affects the activity toward Arp2/3 complex (Fig. S4 D and E in SI Appendix). We compared the activity as a function of concentration for three of these materials, the shortest, the longest, and one intermediate (Fig. S4F in SI Appendix). Each approached saturating activity similarly; for all three materials, 300 nM of VCA dimer was only slightly less active than 600 nM of VCA dimer. However, the three materials had different saturating activities. Because these included the shortest and longest linkers, we concluded that linker length does not strongly affect the approach to saturating activity, and at 400 nM all the constructs are near their saturating activity. We measured activity at 400 nM for seven different linker lengths (Fig. 4C). It was found that for N-terminal linkers shorter than approximately 70 Å, the activity at 400 nM was roughly that observed for VCA–xCA (Fig. 4 B and C). As linker length increased beyond this value, activity at 400 nM sharply increased. This behavior provides a length scale for the separation of the two V regions when activating Arp2/3 complex.

To resolve which of the two binding sites was responding to the missing V region, we developed a means of probing them separately. To this end, we genetically fused two VCAs into a single polypeptide through a (GGS)₄ linker (VCAggsVCA, Fig. 5A and Fig. S5A in SI Appendix). VCAggsVCA has affinity for and potency toward Arp2/3 complex that is intermediate between those of the VCA monomer and the N-terminally linked VCA—VCA (Fig. 5C and Fig. S4H in SI Appendix). At saturation its activity is below that of the other two reagents. Lower saturating activity and intermediate affinity indicate that both VCAs from VCAggsVCA engage Arp2/3 complex simultaneously. That is, if two VCAggsVCA peptides bound Arp2/3 with one VCA each, the tandem material would behave like the VCA monomer, with low affinity but full activity at saturation. That this is not the case indicates that the tandem VCA binds with both VCAs engaged.

Fig. 5.

Asymmetric engagement of Arp2/3 complex by VCAggsVCA. (A) Cartoon representation of the VCAggsVCA constructs produced. VCAggsVCA materials were produced with H472C (native WASP sequence numbers) in either the N-terminal (top line) or C-terminal (second line) VCA. Deletion of the N-terminal or C-terminal V-region results in CAggsVCA and VCAggsxCA materials, respectively. (B) Label transfer experiments from the H472C mutant VCAggsVCA materials to Arp2/3 in the absence or presence of actin. (C) Actin polymerization rates in Arp2/3 dependent actin polymerization assays stimulated with the indicated VCA materials, 10 nM Arp2/3, and 4 μM actin (5% pyrene labeled) in KMEI150, and are the average of three to six repeats. Error bars are the 1σ standard error on the mean, and in many cases are smaller than the symbol. See also Fig. S5 in SI Appendix for the activity of VCAggsVxA and VxAggsVCA and the sequence of VCAggsVCA.

Placing a cysteine in either the N- or C-terminal C region separately (Fig. 5A), we repeated the label transfer with VCAggsVCA as the donor. In the absence of actin these reactions showed little asymmetry, with each C region labeling Arp3 similarly (Fig. 5B). However, when the reactions were repeated in the presence of excess LatrunculinB:Actin, the two tandem VCA constructs behaved differently (Fig. 5B). The N-terminal VCA preferentially labeled Arp2 and ArpC1, while the C-terminal VCA preferred Arp3. The preference is not complete, but the result nevertheless suggests that VCAggsVCA binds Arp2/3 complex in a preferred orientation, with the N-terminal VCA occupying the Arp2 site and the C-terminal VCA occupying the Arp3 site. This asymmetry allows the two sites to be probed independently.

To determine where actin needs to be delivered, we generated one tandem lacking the N-terminal V region (CAggsVCA) and a second where the C-terminal V region was replaced with an equal length GGS sequence (VCAggsxCA) (Fig. 5A). VCAggsxCA, which is unable to deliver actin to Arp3, saturates at appreciably lower activity than the wild-type tandem (Fig. 5C). CAggsVCA, which is unable to deliver actin to Arp2, approaches saturation more slowly than the wild-type construct and may have lower saturating activity as well. The combined cross-linking and mutagenesis data on the tandem VCAggsVCA dimer suggest that delivery of actin to Arp3 is critical to activation of Arp2/3 complex, and that delivery to Arp2 also plays an important but secondary role in the process.

Discussion

The data presented here, in conjunction with existing data in the literature, are sufficient to propose a model for the two VCA-binding sites on Arp2/3 complex. Other studies have used chemical cross-linking (21–25), NMR line broadening (21, 24, 33), direct binding of VCA fragments (24, 26), and SAXS (9) to create models of how VCA binds Arp2/3 complex (see Table S2 in SI Appendix). Additionally, conserved Arp2/3 complex surfaces (28), structures of the V region bound to actin (10, 15), and the sequence similarity of the C and V regions (9, 10) inform our model building. Finally, binding of a CA peptide is sufficient to induce a conformational change in Arp2/3 complex (9, 12, 14), and, in the branch reconstruction, Arp2 and Arp3 take on a short-pitch actin dimer-like conformation (6). Below we assemble these data into a model of the VCA-bound Arp2/3 complex nucleation intermediate.

The Two VCA-Binding Sites on Arp2/3 Complex.

Our BPa crosslinking and biotin label transfer data show that there are two distinct VCA-binding sites on Arp2/3 complex. The response of the label transfer experiments to NtA competition shows that one site is near Arp2 and ArpC1 and the other is near Arp3 and ArpC3 (Fig. 3B). We refer to the two sites as the Arp2 and Arp3 sites, respectively, to emphasize the subunit to which actin monomer is likely delivered by the corresponding V regions (see below).

To create a model of the VCA–Arp3 site interaction (Fig. 6A) we began with the observation that replacing the C-terminal W503 with BPa enabled efficient cross-linking to Arp3. In addition, biotin can be transferred from a cysteine near the C terminus of N-WASP, E498C (Fig. 3), or the WAVE1 C terminus (Fig. S3 in SI Appendix), to both Arp3 and ArpC3. Thus, we locate the C-terminal Acidic/Trp element on Arp3 but within 17 Å of ArpC3. This location is consistent with the previous observation that a spin label on N-WASP CA C506 broadened NMR signals in ArpC3 (range < 25 Å) (21). A conserved basic patch (identified in ref. 28 and termed A-1) serves as a likely binding site for this highly acidic element. Consistent with this placement, W503 binds at this site in a recent crystal structure of the A-region bound to Arp2/3 complex (34).

Fig. 6.

Proposed binding sites for VCA on Arp2/3 complex. Structural models of the VCA-binding sites on Arp2/3. Models are constructed as described in the text using the information summarized in Table S2 in SI Appendix. (A–C) Cartoon depiction of CA binding orientation shown in thick black lines, on top of the inhibited Arp2/3 complex shown in surface representation. Subunits are colored as: Arp3-orange, Arp2-red, ArpC1-green, ArpC2-cyan, ArpC3-magenta, ArpC4-blue, and ArpC5-yellow. Conserved patches (28) are indicated by color changes to the surface. C-2 (on Arp2) and A-2 (on Arp3) are indicated by black surface. Conserved basic residues of M-3 (on ArpC1) and A-1 (on Arp3) are shown in blue, highly conserved, nonbasic residues of M-3 are shown in white surface. (A) The proposed Arp3 binding site. (B) A reference orientation of Arp2/3, with labeled subunits. (C) The proposed Arp2/ArpC1 binding site. (D) Delivery of actin to Arp2/3 in a long-pitch dimer orientation relative to Arp3 results in a large steric clash with Arp2. (E) An Arp2/3 complex model with Arp2 moved to a short-pitch actin dimer position with respect to Arp3. (F) A model for Arp2/3 bound to two VCA peptides and two actins. Actin monomers have been added in a long-pitch positions relative to Arp2 and Arp3. Arrows indicate the rotation of molecules relative to the reference view of the Arp2/3 complex in panel B. Panels E and F are in the same orientation.

Label transfer experiments suggest that the entire length of the CA binds to or near Arp3 (Fig. 3B), consistent with previous cross-linking of position 462 (the N terminus of the C region) to this subunit (21). Because of sequence similarities between the C and V regions, it was previously suggested that the C region could contact subunits of Arp2/3 complex in a manner similar to the V region:actin complex (9, 10, 15). By this analogy, residues 466–476 of the C region would form a helix that doubles back along the back side of Arp3, contacting a conserved patch (identified in ref. 28 and termed A-2). This placement is consistent with NMR spin label experiments showing that residue 481 (the C terminus of the C region) is within 25 Å of ArpC3, but N-WASP residue 462 is outside of this range (21). Moreover, it would position the V region at the barbed end of Arp3, allowing it to deliver actin in a long-pitch orientation.

We model VCA at the Arp2 site using a similar logic (Fig. 6C). First, our W503BPa cross-linking data show that the immediate C terminus of the VCA is bound to ArpC1 (Fig. 3E). This location is consistent with biotin transfer from the E498C donor (Fig. 3B), and the previous demonstration that the VCA can bind isolated ArpC1 in a manner dependent on the conserved C-terminal tryptophan residue (26). Given the highly acidic nature of the A region, the most likely binding site is the conserved basic patch near the ArpC1/Arp2 interface (M-3 in ref. 28). Interaction with this site would explain biotin transfer to ArpC1 throughout the A region (Fig. 3B). The proximity of this basic patch to ArpC5 could also explain the reported cross-linking of N-WASP VCA to this subunit of acanthamoeba Arp2/3 complex (25). Further, disruption of the acidic region binding surface may also explain the lethal phenotype of the Arc40-143 allele in budding yeast when this surface is mutated (35).

As at the Arp3 site, we place the C region on Arp2 as a helix mimicking the interaction of the V region with actin (9, 10, 15, 36). This places the C region in contact with a conserved patch (designated C-2 in ref. 28). This location also positions the C terminus of the helix in contact with the M-3 patch (28) on ArpC1, possibly explaining a report that an isolated C-helix peptide can bind this subunit (24). In this orientation, E473 and A481 are close enough to ArpC1 to allow label transfer, but the A462 is out of range, as we observe (Fig. 3B). This organization of the C-helix would position the V region to deliver actin monomer to Arp2 (9, 10). Others have also suggested this binding site for the C-helix but have then placed the acidic region projecting toward Arp3, rather than ArpC1, to satisfy the Arp3 crosslinking data using a single VCA (9, 10). Our model satisfies these data with two distinct VCA peptides.

A Model for an Arp2/3∶VCA₂∶Actin₂ Assembly.

The simplest explanation for the higher potency of VCA dimers than monomers in activating Arp2/3 complex is that two VCAs bind the Arp2 and Arp3 sites simultaneously at some point during the nucleation process. In addition, our ultracentrifugation data show that two VCA monomers can recruit two actin monomers to Arp2/3 complex. To construct a model of this Arp2/3 complex∶VCA₂∶actin₂ assembly, we began by noting a significant steric clash between Arp2 and an actin delivered to the barbed end of Arp3 in the inhibited structure of Arp2/3 complex (Fig. 6D). This clash would be relieved by rotation of Arp2 to create a short-pitch pseudoactin dimer with Arp3 (Fig. 6E), consistent with electron microscopic reconstructions of the filament branch junction (6). Adding actin monomers in a long-pitch orientation to the barbed ends of Arp2 and Arp3, V regions based on their known complexes with actin, and the two CA interactions described above leads to a model for an Arp2/3∶VCA₂∶actin₂ assembly (Fig. 6F).

In this model the N termini of the two V regions are approximately 65 Å apart, allowing them to be bridged by the VCA—VCA N-terminal linker (which could span approximately 150 Å in an extended conformation). Decreasing the linker to 80 Å or less resulted in a loss of activity (Fig. 4C). This represents a good match to the separation of V regions in our model given that both actin monomers partially block the direct path and that there are entropic penalties for completely extending any disordered linker (37). Presumably, when the N-terminal linker is too short, one V region reverts to a linker function, converting the molecule into an analog of VCA–xCA. This would explain the low activity seen at shorter linker lengths.

In considering how VCAggsVCA could bind Arp2/3 complex, there are two potential connections from the C terminus of one VCA to the N terminus of the other. In our model (Fig. 6F), the C terminus of the Arp2-bound VCA is approximately 60 Å from the N terminus of the Arp3-bound VCA. For the opposite connection, the C terminus of the Arp3-bound VCA is approximatley 110 Å from the N terminus of the Arp2-bound VCA and must diverge substantially from a straight-line path to avoid both actins and Arp3. The 22-residue linker in VCAggsVCA could only span the first distance (Fig. S5A in SI Appendix). Thus, the N- and C-terminal VCAs should bind to the Arp2 and the Arp3 sites, respectively, consistent with our label transfer data (Fig. 5B). The fact that VCAggsVCA binds and activates, and does so with preferences to the predicted binding sites serves as an important test of our model.

Implications for Arp2/3 Complex Activation.

What role do two VCAs (Fig. 1) and two actins (Fig. 2) serve in activation of Arp2/3 complex? The increased potency of VCA dimers suggests that both VCAs contribute to the process. The tandem VCA experiments show that the V regions (and thus the actins) play distinct roles. The C-terminal, Arp3 targeted, V region is more important for activation, as its deletion results in approximately 75% decrease in maximal activity (Fig. 5C). Given that the asymmetry of these materials is on this order, loss of the Arp3 targeted V region may reflect a complete inactivation of the material when actin cannot be delivered to Arp3. In contrast, deletion of the N-terminal, Arp2 targeted V region results in only a slight loss of activation (approximately 30%). The greater importance of actin delivery to Arp3 is consistent with the observation that an Arp2/3 complex containing an Arp2–V region fusion can be activated by GST-CA (12), because inspection of the structure suggests that the introduced V region could deliver actin to Arp3 about as well as to Arp2.

The very low activity of VCA—xCA (approximately 25% of VCA–VCA) seems inconsistent with the primary importance of actin delivery at only one site. If VCA–xCA bound Arp2/3 complex equally in the two possible orientations, its activity should be roughly 50% that of VCA–VCA. That the activity is ≪ 50% suggests that the less active orientation, with VCA:actin at the Arp2 site, is favored. This can be explained if V:actin contributes to affinity more significantly at the Arp2 site than at the Arp3 site. Phrased differently, VCA:actin binds more tightly than CA at the Arp2 site but not at the Arp3 site. This difference is consistent with a clash in inactive Arp2/3 complex between Arp2 and actin bound to Arp3 (Fig. 6D), which suggests that in the absence of a rearrangement of Arp2, VCA:actin should bind with an affinity similar to CA alone. VCA–xCA binding would then be biased toward the orientation that delivers actin to Arp2, resulting in inactive Arp2/3 complex. An analogous argument can be made to explain the approximately 90% activity of VCA–VxA (see Fig. 4B and Fig. S5B in SI Appendix). If CA contributes similarly at both sites, this argument would imply that VCA:actin binds more tightly at the Arp2 site than at the Arp3 site. This would be consistent with reported SAXS data showing preferential association of one VCA:actin to the Arp2 site when mixed at a 1∶1 ratio of VCA:actin to Arp2/3 complex (9) (as distinct from our study, where we used an excess of VCA:actin, Fig. 2).

While the V region is clearly needed for delivery of the initial actins to Arp2/3 complex, binding of CA (or VCA in the absence of actin) has been found to induce a conformational change in the assembly, as shown by FRET (12) and electron microscopy (13, 14). These studies have used a mixture of monomeric and dimeric CA regions, such that the importance of CA binding at either site is difficult to distinguish. However, the SAXS model of VCA:actin bound at the Arp2 site suggests that engagement of the Arp2 site alone may be sufficient to induce a conformational change. It is most likely that the activating conformational change is thermodynamically favored by cooperative delivery of actin to Arp3 and the effects of CA binding. It remains unclear how these events and filament binding are kinetically ordered during the nucleation process.

Conclusion

Previous Arp2/3 nucleation mechanisms have invoked only a single VCA binding and a single actin delivered to create the nucleus. Our model holds that delivery of actin to Arp2 occurs with high affinity but activates Arp2/3 complex only weakly, while actin delivery to Arp3 occurs with low affinity but is critical for activity. This suggests that nucleation fundamentally proceeds through two VCAs and two actins, even when presented as monomers. As argued below, in the biological context, WASP proteins likely function as dimers and higher order oligomers. In such systems, engagement of the Arp2 site with high affinity should drive engagement of the low affinity Arp3 site.

We propose that at some point in the nucleation process, even when induced by VCA monomers, the VCA₂∶actin₂∶Arp2/3 complex assembly described above is likely to exist. Where does this fit into the nucleation pathway? The most likely possibility is that this assembly, with Arp2 and Arp3 aligned, forms in solution prior to filament binding. Such a complex has two actins in addition to the Arp2 and Arp3 subunits and, thus, could function as a nucleus. An interesting possibility is that this Arp2-Arp3-actin₂ nucleus may somehow be prevented from growth by association with the VCAs, perhaps similarly to how VCA prevents spontaneous actin nucleation (20). This inhibition could be relieved either by fluctuations of the VCA on the nascent barbed end or in response to a specific signal—e.g., mother filament binding or ATP hydrolysis in Arp2 or actin. An alternative possibility is that Arp2/3 complex binds mother filament first, and the 2∶2∶1 stoichiometry is assembled on filament directly. In this case, mother filament could also contribute to the alignment of Arp2 and Arp3. Either possibility would allow coordination of the various binding events and nucleation such that branch formation occurs with high fidelity.

In cells, WASP proteins integrate a wide variety of signals to control Arp2/3-mediated actin dynamics. These signals include membrane phosphoinositides, multivalent SH3 proteins, and Rho GTPase gradients, which collectively cluster WASP proteins at membrane (7). This observation has several interesting implications. First, VCA dimers are likely the functional form of WASP proteins in cells, reinforcing a biochemical mechanism in which two VCAs act together to promote nucleation. Second, it provides a mechanism of organizing the branched filament network, because Arp2/3 engaged at two points may bias the orientation of the daughter filament toward the membrane (where WASP protein N termini are engaged with activating factors). Finally, the cellular mechanisms of WASP signal integration appear to have coevolved with the molecular mechanism of Arp2/3 activation, illustrating the ability of different organizational scales to influence one another. We hope that this mechanistic insight into Arp2/3 function will drive both future biochemical and biological studies of actin dynamics.

Materials and Methods

Protein Preparation.

Actin, pyrene actin, and Arp2/3 complex were prepared as previously described (20, 38, 39). Actin was stored in G-buffer (2 mM Tris HCl, 200 μM ATP, 1 mM sodium azide, 0.5 mM DTT, 100 μM calcium chloride) for up to two months; Arp2/3 complex was stored in KMEI (10 mM imidazole pH 7.0, 50 mM KCl, 1 mM MgCl₂, 1 mM EGTA) + 0.5 mM DTT and used within three weeks of preparation or used from stocks flash frozen in KMEI + 0.5 mM DTT, 100 μM ATP and 20% weight to volume sucrose and stored at -80 °C. Residual sucrose and ATP from these frozen Arp2/3 stocks do not appreciably affect the assays and are consistent within any compared data. Prior to use in experiments requiring nonreducing conditions, Arp2/3 complex was concentrated and buffer exchanged (Superdex200, GE) into KMEI without DTT, and actin was dialyzed against G-buffer without DTT.

N-WASP VCA, N-WASP CA, WASP VCA, and Cortactin NtA derived materials were prepared as previously described or with minor adaptations (29). WASP VCAggsVCA preparation was adapted from the method for WASP VCA. Covalent VCA heterodimers were prepared similarly to N-WASP N-terminally linked VCA, but isolation of pure heterodimers required ion exchange methods developed for each material (see SI Appendix). Cysteine labeling with Alexa Fluor 488 C₅ maleimide (Invitrogen) or label transfer reagent (Profound Mts-ATF-Biotin, Pierce) was performed for two hours at room temperature, following manufacturer guidelines.

Actin Assembly Assays.

Actin assembly assays were performed as previously described (38) but with three modifications. First, assembly assays were performed in KMEI with the KCl concentration adjusted to 150 mM. Second, with the exception of Fig. S5B in SI Appendix (which was collected using the previously described equipment), 8 or 16 assays were performed simultaneously using a Varioskan Flash plate reader (Thermo Scientific). Third, quantitation was performed by finding the actin polymerization rate at t₅₀.

Analytical Ultracentrifugation.

All analytical ultracentrifugation experiments were carried out in an Optima XL-I centrifuge using an An50-Ti rotor (Beckman–Coulter). Samples (approximately 390 μl) were placed in charcoal-filled, dual-sector Epon centerpieces. Absorbance and/or interference optical systems were used to monitor the sedimentation of the proteins at a rotor speed of 50,000 rpm, except for the mass-measurements experiments, which were conducted at 30,000 rpm. The stoichiometry of the Arp2/3∶actin∶VCA complex was calculated using the “leading” and “lagging” methodology previously outlined (40). Data were analyzed using SEDFIT and SEDPHAT (available at http://www.analyticalultracentrifugation.com) (30, 41). Additional details are available in Supplementary Materials and Methods in SI Appendix and in a separate methods paper (31).

Cross-Linking and Label Transfer.

VCA, CA, and NtA label transfer reactions followed established methods (42), but MTS-ATF-Biotin (Pierce) was used as the label transfer reagent, and blots were probed with Neutravidin-HRP Pierce. Peptides labeled with MTS-ATF-Biotin (5 μM) were mixed with 1 μM Arp2/3 complex in KMEI lacking reducing agent. Asymmetric crosslinking reactions were performed with 2 μM Arp2/3 complex, 1.5 μM labeled VCAggsVCA protein, 5 μM actin and 19 μM Latrunculin B (Calbiochem, #428020), in KMEI with 100 μM ATP added.

N-terminally biotinylated N-WASP CA peptides with W503 changed to BPa, where BPa is benzoyl-phenylalanine (32), were synthesized using FMOC methods. Cross-linking to Arp2/3 complex used filtered light from a 200 W mercury lamp. Arp3 and ArpC1 were detected by Western blotting.

Molecular Modeling

Molecular modeling performed in Pymol (Delano Scientific) using the following published coordinates: Actin:WASP VCA (15), inhibited bovine Arp2/3 complex with Arp2 completely built (27, 28), filamentous actin (43). Possible C-region binding sites on Arp2 or Arp3 were found by superimposing the actin:WASP VCA structure (15) upon Arp2 or Arp3, respectively. The Arp2/3∶VCA₂∶actin₂ model was created by rotating Arp2/3 complex to align Arp3 to the first monomer in an actin tetramer derived from the filamentous actin coordinates. Arp2 was repositioned to align it with the second actin, and the remaining two actins appear in the model.