Insertions within the Actin Core of Actin-related Protein 3 (Arp3) Modulate Branching Nucleation by Arp2/3 Complex*

Background: The Arp2/3 complex is a molecular machine that nucleates branched actin filaments using poorly understood structural mechanisms. Results: Deletion of inserts within the actin core of Arp3 hyperactivated the complex or rendered it inactive. Conclusion: Inserts in Arp3 are required to properly modulate the activity of Arp2/3 complex. Significance: Understanding branching nucleation is required to understand how actin is regulated in vivo. The Arp2/3 (actin-related protein 2/3) complex nucleates branched actin filaments involved in multiple cellular functions, including endocytosis and cellular motility. Two subunits (Arp2 and Arp3) in this seven-subunit assembly are closely related to actin and upon activation of the complex form a “cryptic dimer” that stably mimics an actin dimer to nucleate a new filament. Both Arps contain a shared actin core structure, and each Arp contains multiple insertions of unknown function at conserved positions within the core. Here we characterize three key insertions within the actin core of Arp3 and show that each one plays a distinct role in modulating Arp2/3 function. The β4/β5 insert mediates interactions of Arp2/3 complex with actin filaments and “dampers” the nucleation activity of the complex. The Arp3 hydrophobic plug plays an important role in maintaining the integrity of the complex but is not absolutely required for formation of the daughter filament nucleus. Deletion of the αK/β15 insert did not constitutively activate the complex, as previously hypothesized. Instead, it abolished in vitro nucleation activity and caused defects in endocytic actin patch assembly in fission yeast, indicating a role for the αK/β15 insert in the activated state of the complex. Biochemical characterization of each mutant revealed steps in the nucleation pathway influenced by each Arp3-specific insert to provide new insights into the structural basis of activation of the complex.

Precise cellular control of the polymerization of actin is critical for orchestrating diverse cellular processes ranging from cell division, to cellular motility, endocytosis, exocytosis, and host-pathogen interactions. There are Ͼ60 families of known actin-binding proteins that can influence the reactions of actin, and many of these families are conserved among eukaryotes to form a core set of actin regulatory machinery. Key to this machinery are actin filament nucleators that catalyze the de novo formation of actin filaments in response to cellular signal-ing pathways, allowing precise spatiotemporal control of the initiation of actin filament networks. To date, three classes of actin filament nucleators have been discovered: Arp2/3 complex, formins, and tandem monomer-binding nucleators (1). Through distinct mechanisms, each class is thought to function by stabilizing oligomers, thereby overcoming the thermodynamic instability of actin dimers and trimers, which causes a kinetic barrier to filament formation (2). Arp2/3 2 complex (actin-related protein 2/3 complex) is unique among nucleators in that it contains two actin-related subunits that can mimic a stable actin dimer to initiate filament formation (3). X-ray crystal structures show that in its intrinsically inactive state, the other five subunits of the complex hold Arp2 and Arp3 apart, blocking formation of the Arp2-Arp3 dimer to prevent nucleation (4,5). Three-dimensional reconstructions of negatively stained branch junctions show that upon activation, Arp2 moves ϳ25 Å to form a dimer with Arp3 that mimics the short pitch conformation of two consecutive actin subunits within a filament (6).
Activation of the complex requires binding one or more nucleation-promoting factors (NPFs). The WASp/Scar family of proteins are the best characterized NPFs and contain a characteristic C-terminal sequence called VCA, the minimal sequence sufficient to activate Arp2/3 complex. The V region binds monomeric actin (7)(8)(9) and is thought to recruit the first actin subunit(s) to the daughter filament. The CA (central and acidic) binds to two sites on Arp2/3 complex and causes conformational changes that may contribute to activation (10 -14). Activation of the complex also requires binding to the side of a preexisting actin filament (mother filament) (8,15). Finally, activation requires ATP and free actin monomers. How NPFs, actin filaments, actin monomers, and ATP cooperate to dimerize the Arps and activate the complex is still unclear and remain a key question in the field.
The Arp3 subunit plays a central role in the activity of Arp2/3 complex. Isolated complexes lacking the Arp3 subunit are inactive in actin polymerization assays (16). Residues at the barbed end of Arp3 contact the actin subunit at the end of the daughter * This work was supported, in whole or in part, by National Institutes of Health Grant GM092917 (to B. J. N). □ S This article contains supplemental Figs. S1-S7, Movies S1-S4, and additional references. 1 To whom correspondence should be addressed: Institute of Molecular Biology, 1229 University of Oregon, Eugene, OR 97403-1229. Tel.: 541-346-7412; E-mail: bnolen@uoregon.edu. filament in the branch junction structure, consistent with a role for Arp3 in templating new filaments (6). In addition, Arp3 is critical for interaction with NPFs because cross-linking and an x-ray crystal structure of a tripeptide fragment of A bound to Arp2/3 complex indicate one of the NPF binding sites is on the Arp3 subunit (11,(17)(18)(19)(20)(21). In molecular dynamics simulations a loop in Arp3 blocks movement of Arp2 into the short pitch conformation, so Arp3 may also be important for maintaining the constitutive inactivity of the complex (22). Finally, analysis of the three-dimensional reconstruction of Arp3 at a branch junction places one end of Arp3 within 5 Å of the mother filament of actin (6), so Arp3 may also play a role in linking the complex to the side of filaments. Despite close sequence and structural similarities with actin, Arp3 shows significant differences that may be critical for the function of the complex. Several large insertions to the actin core are present in all Arp3 sequences but not in actin (23), and numerous solvent-exposed regions within the core are conserved in Arp3 but different from actin (24). We hypothesized that these differences may be critical for roles of Arp3 in Arp2/3 complex function, specifically in functions distinct from daughter filament templating, such as mother filament binding, NPF binding, or maintenance of constitutive inactivity. Here, we used mutational analysis of the complex to study three key regions in Arp3 that differ from actin. We asked how these regions influenced the activity of the complex in vitro and in vivo. We found that individual mutations in the Arp3-specific sequences caused differential effects on the complex. Two of the mutants completely inactivated the complex in vitro, whereas the third mutation made the complex more responsive to activation by NPFs. Using biochemical and biophysical methods, we determined which aspects of complex function were influenced by each mutation. Importantly, we found that the ␤4/␤5 insert mediates interactions with actin filaments, and the hydrophobic plug contributes to the integrity of the complex but is not required for stabilization of the nucleus. Our data also indicate that the ARPC3 subunit is critical for activity of the complex and NPF binding. These observations, along with observations of the influence of the mutants on actin assembly in vivo, provide new insights into the precise steps of Arp2/3 complex activation and suggest a complex pathway coordinated by multiple critical structural features.

EXPERIMENTAL PROCEDURES
Protein Purification and Labeling-Schizosaccharomyces pombe Arp2/3 complex was isolated as described previously, with some modifications (25). After recovering the complex from an ammonium sulfate precipitation step, we purified it further using a GST-VCA affinity column, an anion exchange column, and a gel filtration column. With the exception of the Arp3⌬HPlug mutation, all of the mutants purified with all seven subunits intact, indicating that none of the mutations caused unfolding or disassembly of the complex. The Arp3⌬HPlug complex eluted from the affinity column at lower salt concentrations than the other complexes and had contaminating bands not present in the other preparations. The Arp3⌬HPlug complex affinity elution lacked the ARPC3 subunit, but the rest of the complex was intact and eluted as a single peak in the gel filtration step. GST-Wsp1-VCA and Wsp1-VCA were purified, and Wsp1-VCA was labeled with tetramethylrhodamine as described previously (25). Gelsolin (G4 -G6 fragment) and pyrene or Oregon Green-labeled rabbit skeletal muscle actin were purified as described previously (26 -29).
Pyrene Actin Polymerization Assays-Polymerization assays were assembled as described previously (20), and fluorescence measurements were made on a Tecan Safire2 plate reader using an excitation wavelength of 365 nm and an emission wavelength of 407 nm. The maximum rate of polymer formation was determined by plotting the slope of each polymerization curve at each time point and converting relative fluorescence units/s to nM actin/s assuming that the total amount of polymer at equilibrium is equal to the total concentration of actin minus 0.1 M, the critical concentration.
Actin Filament Binding Assays-Gelsolin-actin seeds for pointed end binding assays were prepared as described previously (30). Pointed end polymerization assays contained 25 nmol of gelsolin actin seeds, 2 M of 15% pyrene-labeled actin, and 0 -2 M of wild type or mutant S. pombe Arp2/3 complex. Copelleting assays were carried out as described previously (20).
⑀-ATP Binding Assay-⑀-ATP binding was measured as described previously (31). Briefly, the fluorescence of a solution containing 50 M ⑀-ATP, 10 mM Hepes, pH 7.0, 50 mM KCl, 1 mM EGTA, 1 mM MgCl 2 , 1 mM DTT, 1.4% acrylamide, and 0.5 M Arp2/3 complex was measured at 420 nm (excitation wavelength ϭ 340 nm) on a Flouromax3 fluorescence spectrophotometer. The concentration of ⑀-ATP in the solution was decreased stepwise by mixing with an identical solution lacking ⑀-ATP, and the background ⑀-ATP fluorescence was subtracted assuming a linear relationship between ⑀-ATP concentration and fluorescence at saturation concentrations of ⑀-ATP. The resulting isotherm was fit with the following equation where r ϭ measured fluorescence, r f is fluorescence in the absence of ⑀-ATP, r b is the fluorescence at saturating ⑀-ATP, R is the concentration of ⑀-ATP, L is the concentration of Arp2/3 complex, and K d is the dissociation constant.
Fluorescence Anisotropy Binding Assays-The anisotropy of a solution containing 75 nM SpWsp1-Rh-VCA, 2 or 3 M Arp2/3 complex, 10 mM imidazole, pH 7.0, 50 mM KCl, 1 mM MgCl 2 , 1 mM EGTA, 1 mM DTT, and 0.2 mM ATP was measured using excitation and emission wavelengths of 557 nm and 574 nm, respectively. The solution was diluted stepwise with an identical solution lacking Arp2/3 complex and the anisotropy measured for each concentration of Arp2/3 complex. Data were fit as described for ⑀-ATP binding assay. For competition assays, 0.5 M Arp2/3 complex plus 75 nM SpWsp1-Rh-VCA were titrated with GST-Wsp1-VCA and the anisotropy measured. Data were fit with a previously described procedure (32).
Total Internal Reflection Fluorescence (TIRF) Microscopy-TIRF chambers were constructed, and the reaction setup was carried out as described previously (29). TIRF images were collected on a Nikon TE2000-U microscope outfitted with a Nikon 100ϫ TIRF objective and a charge-coupled device (CCD) (ORCA-AG, Hamamatsu), or an EM-CCD camera (iXon3, Andor). Images were collected using 100-ms exposures at 10-s intervals.
Yeast Genetics-The arp3⌬␣K/␤15 mutant fission yeast strain was constructed by knocking out one copy of ARP3 and 280 upstream nucleotides in a diploid strain with a kanamycin resistance cassette using a homologous recombination procedure (33). The kanamycin resistance cassette was then replaced with mutant or wild type ARP3 containing 280 upstream nucleotides and the URA4 marker before sporulating the diploid and selecting for URA negative strains. The arp3⌬␤4/␤5, arp3⌬hplug, and wild type control strains were constructed by cloning ARP3 (with 280 nucleotides upstream and 186 nucleotides downstream) into pJK148, creating the desired mutations by PCR, and integrating the linearized plasmid into the leu1-32 locus of a haploid strain as described previously (34). For the S. pombe arp3⌬␣K/␤15, arp3⌬␤4/␤5, and arp3⌬hplug mutants, the residues deleted were as follows: 357-371, 44 -61, 295-306. Wild type ARP3 was knocked out with a KanMX6 marker, and FIM1 was tagged with GFP using a NatMX6 marker. For each mutant strain, mutated arp3 was sequenced to verify that the mutation was present, and junctions for knockouts and insertions were verified by PCR.
Microscopy-Movies of live S. pombe cells were made using a Nikon Eclipse TE2000-U microscope, a Yokogawa CSU10 spinning disk head (Wallac), and a Hamamatsu ORCA-AG or Andor EM-CCD iXon3 camera. Fim1-GFP fluorescence data were collected using an argon laser (Dynamic Laser). FM4-64 fluorescence data were collected using a Sapphire 561-nm laser (Coherent). Movies used for patch counting and timeaveraged images were taken at 2 frames/s for 2 min (240 frames total) at 30°C and can be viewed in the supplemental Movies S1-S4. Movies for kymographs were collected at 5 frames/s for 40 s (200 frames total) at 25°C. Typical patches at the cell tips were identified for each strain, and a kymograph spanning the length of the movie was made along the direction of motion for the patch. FM4-64 uptake assays were carried out as described previously (35). The open source software package Micromanager was used for collection of all microscopy data (36).

RESULTS
The Arp3 ␤4/␤5 Insert Mediates Interactions of Arp2/3 Complex with Actin Filaments-The loop between ␤4 and ␤5 is a hotspot for insertions within the actin core (23), and in actin this loop, called the DNaseI loop, makes intersubunit contacts within filaments (37)(38)(39). In Arp3, the sequence of the ␤4/␤5 loop diverges completely from actin and is longer by 7-47 amino acids relative to actin (supplemental Fig. S1). In crystal structures of Arp2/3 complex, the ␤4/␤5 loop is disordered (4,5), but the EM model of a branch junction places it near the mother filament of actin (6), and molecular dynamics simulations indicate that it forms favorable contacts with the filament (40) (Fig. 1, A and B). Therefore, we hypothesized that this insert influences interactions of Arp2/3 complex with the mother filament. We created a haploid fission yeast strain with 18 residues deleted from the Arp3 ␤4/␤5 loop to test its role in activity of the complex (supplemental Fig. S1). For this and each of the mutants described herein, the mutant arp3 was under control of its native promoter to ensure it was expressed at endogenous levels, and wild type ARP3 was not expressed.
We purified the mutant Arp3⌬␤4/␤5 complex using previously published procedures and tested its activity in pyrene actin polymerization assays (25) (supplemental Fig. S2). The Arp3⌬␤4/␤5 complex had activity similar to the wild type complex when activated by Wsp1-VCA (Fig. 1, C and D). Recent experiments demonstrated that VCA binds to two sites on Arp2/3 complex and that dimeric NPFs are more potent activators of the complex because they simultaneously or sequentially engage both sites (11,12). Therefore, we also tested the ability of GST-Wsp1-VCA to activate the complexes. We found that the Arp3⌬␤4/␤5 complex was more strongly activated by GST-VCA than wild type complex (Fig. 1E). Titration with a range of GST-VCA concentrations showed a characteristic biphasic response in which the maximum polymerization rate increased as the complex became saturated and then decreased due to inhibition of spontaneous nucleation by GST-VCA ( Fig.  1F) (8). At optimal GST-VCA concentrations, the Arp3⌬␤4/␤5 complex gave a maximum elongation rate of 0.009 m/s compared with 0.0065 m/s for the wild type complex. To determine whether the Arp3⌬␤4/␤5 insert plays a general role in "damping" the activity of the complex, we made the equivalent mutation in Saccharomyces cerevisiae Arp3 (⌬50 -80) and found that not only was the budding yeast ⌬␤4/␤5 insert complex more active in the presence of GST-VCA than wild type complex, it also showed greater constitutive and Wsp1-VCAstimulated activity than wild type complex (Fig. 1G). To determine whether the mutation caused defects in branch formation or stability, we used TIRF microscopy to directly visualize polymerization of Oregon Green actin in the presence of wild type or mutant complex. The S. pombe ⌬␤4/␤5 insert complex formed branches indistinguishable from wild type complex, and we observed very few instances of dissociation of branches nucleated by either wild type or mutant complex (Fig. 1H). The Arp3⌬␤4/␤5 complex activated by GST-VCA produced greater branch densities than wild type complex, verifying our observations from the bulk assays (Fig. 1I).
We next asked whether deletion of the Arp3 ␤4/␤5 insert influences activator binding. We first used a fluorescence anisotropy assay that measures binding predominantly to the high affinity NPF binding site (11). The Arp3⌬␤4/␤5 deletion had a small but significant effect on binding of VCA to this site. Rhodamine-labeled Wsp1-VCA (Rh-VCA) bound the Arp3⌬␤4/␤5 insert complex with an affinity of 0.15 M, compared with 0.24 M for the wild type complex ( Fig. 2A and Table 1). Competing Rh-VCA off the complex with GST-VCA yielded a K d value of 7 nM for both mutant and wild type complexes, suggesting that the increased activator affinity cannot explain the enhanced response to GST-VCA by the Arp3⌬␤4/␤5 mutant (Fig. 2B).
Actin filament binding is required for activation of Arp2/3 complex (8, 15), so we hypothesized that deletion of the Arp3 ␤4/␤5 insert may increase NPF-induced activation by influencing interaction of the complex with F-actin. Therefore, we used actin filament copelleting assays to measure filament binding. Neither wild type nor mutant S. pombe complexes reproducibly copelleted with actin filaments, so we used the S. cerevisiae Arp2/3 complex for our assays. Titration with actin filaments increased copelleting of wild type ScArp2/3 complex, but binding appeared to saturate at 50% with an apparent K d of 1.1 M (Fig. 2C). In contrast, the Arp3⌬␤4/␤5 complex saturated at near 100% bound, with a K d of 0.5 M. These data indicate that deletion of the Arp3⌬␤4/␤5 insert increases the affinity of Arp2/3 complex for actin filaments, providing an explanation for the increased responsiveness of the Arp3⌬␤4/␤5 insert complex to NPF activation. The incomplete saturation of filaments by wild type complex could be due to a portion of the complex being inactive. However, both complexes were expressed and purified identically, and the wild type complex shows activity identical to the Arp3⌬␤4/␤5 complex when activated by Wsp1-VCA, so we argue that dissociation during the pelleting assay is a more likely explanation.

Increased Activity of the Arp3⌬␤4/␤5 Deletion Does Not
Cause Defects in Actin Patch Dynamics-To determine whether in vitro nucleation activity was correlated with actin dynamics in vivo, we used confocal microscopy to image actin dynamics in live yeast cells. Genetic and chemical inhibition studies showed that the assembly of endocytic actin patches requires Arp2/3 complex (41)(42)(43). Previously reported mutations in Arp2/3 complex that increase constitutive activity or response to NPFs in vitro cause temperature sensitivity or actin defects in budding yeast (44). Therefore, we asked whether the increased activity caused by the Arp3⌬␤4/␤5 deletion affected actin patch dynamics in S. pombe. To monitor the patches, we added a GFP tag to S. pombe Fim1, an actin-cross-linking protein recruited to the patches with the same timing as actin, Arp2/3 complex, and other actin-binding proteins (45,46). Cortical assembly of this actin module initiates ϳ7 s before reaching a peak concentration, then the patch moves away from the cortex into the cytoplasm, disassembling over 9 -15 s (45). Therefore, this system provides a well characterized in vivo readout for the Arp3 mutants.
The distribution of actin patches was indistinguishable comparing the wild type and Arp3⌬␤4/␤5 strains (Fig. 3A). Wild type and the Arp3⌬␤4/␤5 nucleated patches also had similar lifetimes and internalization rates (Fig. 3, B and C, and supplemental Movies S1 and S2). To determine whether the increased activity of the Arp3⌬␤4/␤5 complex affects endocytosis, we pulsed cells with the lipophilic dye FM4-64 and imaged its movement from the plasma membrane into endosomes and vacuoles (47). The dye was taken up by wild type and Arp3⌬␤4/␤5 cells almost immediately under our assay condi-tions, whereas a control strain with Wsp1 deleted showed slow uptake, with significant amounts of dye still at the membrane 20 min after the pulse (Fig. 3, D and E). Together, these data demonstrate that the increased activity of the Arp3⌬␤4/␤5 insert deletion does not cause defects in vivo.
The Arp3 Hydrophobic Plug (HPlug) Is Required for Activity and Integrity of the Complex-Actin contains a 15-amino acid loop at the interface of subdomain 3 and 4 referred to as the hydrophobic plug. The HPlug is conserved in Arp3 and shows 40% similarity to actin (supplemental Fig. S3). Because the HPlug stabilizes short pitch interactions in actin filaments (37,39,48,49), we hypothesized that in Arp3 it may stabilize the Arp3-Arp2 interface in the activated conformation. The electron tomography branch junction model shows the HPlug of Arp3 positioned near but not contacting Arp2 (Fig. 4A) (6). We replaced the entire HPlug with a GSG linker and purified the mutant complex from S. pombe (Fig. 4B). Surprisingly, ARPC3 did not coelute with the Arp3⌬HPlug complex from a GST-VCA affinity column, despite the fact that the remaining six subunits eluted as a single peak (Fig. 4C).
With 0.8 M Wsp1-VCA present, up to 200 nM Arp3⌬HPlug complex showed no detectible nucleation activity in pyrene actin polymerization assays, whereas 200 nM wild type SpArp2/3 created ϳ1.6 nM barbed ends and had a maximum polymerization rate of 15 nM/s (Fig. 4D). The activity of the Arp3⌬HPlug mutant could not be rescued with high concentrations of Wsp1-VCA, because it showed no activity at concentrations of Wsp1-VCA up to 2.5 M or GST-VCA up to 1 M (Fig. 4E, data not shown). Together, these data indicate that deletion of the hydrophobic plug inactivates the complex in vitro. Although we cannot completely rule out the possibility that contaminating proteins in the final Arp3⌬HPlug preparation inhibit nucleation activity, fractions containing the contaminant proteins did not inhibit wild type Arp2/3 complex in a separate set of experiments (supplemental Fig. S4).
The Arp3 HPlug is not predicted to be involved in VCA interactions based on the previously proposed models for binding (9, 12, 50) (supplemental Fig. S5). However, the affinity of the Arp3⌬HPlug mutant for Wsp1-Rh-VCA measured by fluorescence anisotropy was decreased 2-fold compared with wild type  Arp2/3 complex (0.50 M versus 0.24 M) (Fig. 4G, Table 1). The HPlug deletion had an even greater influence on GST-VCA binding, showing a 10-fold reduction compared with the wild type complex (Fig. 4G and Table 1). Previous cross-linking and NMR studies have shown that VCA contacts ARPC3 (17), so dissociation of ARPC3 caused by deletion of the HPlug may contribute to the reduced affinity. However, reduced activator binding alone cannot explain the inactivity of the Arp3⌬HPlug complex because activity could not be rescued with high activator concentrations. ATP binding causes the nucleotide cleft of Arp3 to close, and closure of the cleft is thought to be required for activation. Therefore, we used an ⑀-ATP (1,N6-ethenoadenosine-5Ј-O-triphosphate) binding assay to determine whether deletion of the HPlug influences nucleotide binding. Binding of ⑀-ATP to the wild type complex caused an increase in fluorescence that we used to measure the K d (2.2 M). In contrast, we did not observe an increase in ⑀-ATP fluorescence when titrating into a solution of the Arp3⌬HPlug complex (Fig. 4F). This indicates that ⑀-ATP does not bind or that there is no increase in fluorescence upon binding. Either possibility demonstrates an unanticipated structural connection between the HPlug and the nucleotide cleft.
The Arp3 Hydrophobic Plug Is Not Required for Nucleation Activity or Endocytosis in Fission Yeast-To determine whether the deletion of the Arp3 HPlug caused defects in vivo, we first compared the growth rate of the arp3⌬hplug strain with the wild type. Surprisingly, the mutant grew as well as wild type at 30°C in rich medium (Fig. 5A). We next used the strategy described above to monitor actin patch dynamics and found that although the arp3⌬hplug strain had fewer patches on aver-age per cell, patch dynamics and FM4-64 uptake were normal in the mutant (Fig. 5, B-G, and supplemental Movie S3). These data demonstrate that the Arp3 HPlug deletion does not cause defects in nucleation activity in vivo and suggest that it does not play an essential role in stabilizing the short pitch dimer. Instead, we propose that it stabilizes the interaction between ARPC3 and Arp3 and that the loss of the ARPC3 subunit we observed during purification results in the in vitro inactivity of the Arp3⌬HPlug complex. This is consistent with previous data that showing that ARPC3 is required for full nucleation activity and crystal structures showing direct contacts between ARPC3 and the Arp3 HPlug (4,16).
Deletion of the Arp3 ␣K/␤15 Insert Inactivates Arp2/3 Complex-Subdomain 3 in Arp3 contains an insertion of 12-16 amino acids relative to actin that extends helix ␣K by three turns and lengthens the loop between ␣K and ␤15 (Fig. 6, A-C, and supplemental Fig. S6). Molecular dynamics simulations suggested that this insert forms a steric bumper that blocks movement of Arp2 into the short pitch conformation (22). Therefore, we hypothesized that deletion of this insert would cause constitutive activity in the complex. However, we found that the Arp3⌬␣K/␤15 complex was inactive in pyrene actin polymerization assays and could not be activated by VCA or GST-VCA at any concentration we tested (Fig. 6, D and E). This demonstrates that the insert plays an essential role in the activated complex. The EM structure of a branch junction shows the ␣K/␤15 insert projects from the Arp3 barbed end toward the pointed end of the adjacent actin subunit in the daughter filament, suggesting that it may stabilize the interaction of pointed ends of actin filaments and/or VCA-recruited monomers with Arp3 (Fig. 6B) (6). We tested the ability of the Arp3⌬␣K/␤15 complex to bind to the pointed ends of actin filaments by measuring the elongation rate of actin filament seeds capped with gelsolin at their barbed ends as a function of Arp2/3 concentration. The Arp3⌬␣K/␤15 complex capped pointed ends as well as wild type Arp2/3 complex (Fig. 6F). This indicates that the Arp3⌬␣K/␤15 insert does not contribute to interaction of the complex with the pointed ends of preformed actin filaments, but does not rule out a potential role for the ␣K/␤15 insert in contacting the pointed end of actin monomers (see below). Because helix ␣K contributes residues to the binding pocket for A identified in a recent crystal structure (11), we next asked whether the ␣K/␤15 insert influences activator binding. The K d values of Arp3⌬␣K/␤15 complex for VCA and GST-VCA were 0.13 M and 9.0 nM, close to the values for the wild type complex (Fig. 6G). Therefore, inactivity of this mutant was not due to defects in activator binding.
Deletion of the Arp3⌬␣K/␤15 Insert Causes Defects in Actin Patch Dynamics and Endocytosis-Compared with wild type, the Arp3⌬␣K/␤15 strain showed a marked reduction in the number of actin patches (Fig. 7, A and B). Actin patches in the Arp3⌬␣K/␤15 strain had longer average lifetimes than wild type patches and frequently failed to internalize (Fig. 7, C-E, and supplemental Movie S4). These data are consistent with observation that the Arp3⌬␣K/␤15 mutant showed severely compromised nucleation activity in vitro. The Arp3⌬␣K/␤15 strain showed slowed FM4-64 uptake, indicating that the defects in nucleation activity correlate with defects in endocytosis (Fig. 7, F and G).

DISCUSSION
The Role of the ⌬␤4/␤5 Insert in Arp2/3 Complex Function-Whereas the EM structure of Arp2/3 complex at a branch junction shows that all seven subunits contact the mother filament, only the ARPC2 and ARPC4 subunits have been mutationally mapped and directly tested for their contributions to F-actin binding (51,52). Here we show that deleting the Arp3⌬␤4/␤5 insert had a significant influence on F-actin binding, consistent with the observation that the pointed end of Arp3 is near the mother filament in the EM model. How might improved F-actin affinity increase the activity of the complex? Kinetic modeling indicates that a slow activation step occurs after the complex binds to filaments and that slow filament binding is partially rate-limiting at actin concentrations used in vitro (53). Therefore, deletion of the ␤4/␤5 may increase the on rate for Arp2/3 binding to filaments to increase the nucleation rate.
Our observations on the Arp3⌬␤4/␤5 complex raise additional important questions about the mechanism of activation and in vivo function of the complex. First, the S. pombe Arp3⌬␤4/␤5 mutant shows increased responsiveness to GST-VCA but not VCA. This points to fundamental mechanistic differences between monomeric versus dimeric VCAs and highlights the need for measurements of kinetic and thermodynamic constants for interactions of Arp2/3 complex, dimeric-VCAs, and actin. Second, we note that the ␤4/␤5 insertion, although present in all Arp3 sequences we examined and in some cases as long as 47 amino acids, is neither conserved nor structured (supplemental Fig. S1) (4,5), drawing into question how it influences interactions with actin filaments. One possibility is that this insert serves as a nonspecific steric bumper that evolved to block homo-oligomerization in Arp3. This function would not result in pressure to retain specific sequences, but the insertion itself would be retained. This is the case for at least one other Arp protein, Arp4, which contains an insertion with weak or no sequence conservation at an invariant position in subdomain 4 (54). The Arp4 insertion forms a flexible loop that appears to block the pointed end to prevent homopolymerization. Such a steric bumper mechanism may also fine tune the affinity of Arp2/3 complex for actin filaments. Consistent with this, overlaying snapshots of a molecular dynamics simulation of Bos taurus Arp3 onto the branch junction model shows that most of the observed conformations of the ␤4/␤5 loop in unbound Arp3 clash with the subdomain 1 in the M2 actin subunit (supplemental Fig. S7) (55). Finally, the Arp3⌬␤4/␤5 insert deletion increased actin assembly rates in vitro but not in vivo, suggesting slow actin filament binding by the complex (or fast release) does not limit actin assembly in endocytic patches. This may be due to high local concentrations of filaments or accessory factors that influence interactions of the complex with filaments (45,56).
The Hydrophobic Plug Is Distinct in Arp3 and Actin-By analogy to actin, we hypothesized that the hydrophobic plug of Arp3 stabilizes the short pitch Arp2-Arp3 dimer in activated Arp2/3 complex. However, our data show that the HPlug plays a distinct role in Arp3 compared with actin: it helps attach ARPC3 to the complex, but is not required to lock Arp2 in the short pitch position. Recent high resolution cryo-EM structures and improved x-ray fiber diffraction models of actin filaments show short pitch actin dimers packed tightly together with the HPlug of one subunit mediating contacts with the pointed end of its short pitch neighbor (38,39). Mutational studies demonstrated the contribution of these contacts to stabilizing short pitch contacts in filaments in vitro and in vivo (48,57). That we could completely remove the HPlug from Arp3 without affecting actin filament nucleation in vivo suggests that Arp2 and Arp3 do not perfectly mimic the filament-like conformation during nucleation. The electron tomography reconstruction of a branch junction supports this idea because Arp2 is not packed tightly against Arp3 in this model, and the Arp3 HPlug cannot reach Arp2 (Fig. 4A) (6).
The ARPC3 subunit is essential for viability in S. pombe, and reduced levels of ARPC3 cause defects in actin patch assembly (58). Our data suggest that these defects are caused by failure of the ARPC3-deficient complex to nucleate filaments. Reconstitution experiments show the ARPC3 subunit is also critical for nucleation by human Arp2/3 complex, suggesting a conserved role for this subunit in branching nucleation (16). We speculate that deletion of the Arp3 HPlug in S. pombe causes dissociation  Fig. 3B for comparison). G, quantification of FM4-64 uptake. Error bars show S.D. of measurements. Uptake rate for endocytosis-defective Arp3⌬␣K/␤15 strain is shown for reference. of ARPC3 during purification, but not in vivo, explaining the lack of a phenotype for this mutation.
A New Role of the ␣K/␤15 Insert in the Activity of the Complex-Molecular dynamics simulations suggested that the Arp3 ␣K/␤15 insert plays a role in maintaining constitutive inactivity of the complex (22). However, deletion of the ␣K/␤15 insert did not constitutively activate the complex, but instead inactivated it. Although our data cannot rule out a potential role for this insert in locking the complex in the inactive state, they suggest that the insert plays a critical role in the activated state. Our structural analysis indicated that the Arp3⌬␣K/␤15 insert might interact with pointed ends of daughter filaments at branch junctions and/or with VCArecruited actin monomers during nucleation. The defects in nucleation activity of the Arp3⌬␣K/␤15 insert deletion in vitro and in vivo support this model. However, the Arp3 ␣K/␤15 insert deletion had no influence on the ability of the complex to cap pointed ends of preformed actin filaments. Whether Arp2/3 complex uses the same set of contacts to cap pointed ends, attach daughter filament ends to branch junctions, and interact with VCA-recruited actin monomers during nucleation is currently unknown. A recent cryo-EM structure shows that the conformation of the pointed end of an unbranched actin filament is distinct from conformation of the pointed end of a daughter filament at a branch junction, providing evidence that these processes are not structurally analogous (59). These observations highlight the need for developing single molecule methods to biophysi-FIGURE 6. The Arp3⌬␣K/␤15 mutation abolishes nucleation activity. A, ribbon diagram of Arp3 from Arp2/3 complex structure (Protein Data Bank code 1P9K). The Arp3⌬␣K/␤15 insert (blue) is in subdomain 3. B, model of Arp2/3 complex at a branch junction (6) showing the position of the Arp3⌬␣K/␤15 insert (blue) and the two first actin subunits in the daughter filament (D1 and D2). C, sequence alignment of Arp3 ␣K/␤15 insert from diverse species. At, Arabidopsis thalania; Dm, Drosophila melanogaster; Dr, Danio rerio; Hs, Homo sapiens; Dd, Dictyostelium discoideum; Sc, Saccharomyces cerevisiae; Sp, Schizosaccharomyces pombe. Identical residues (green boxes), similar residues (cyan boxes), and residues deleted (red) in the Arp3⌬␣K/␤15 mutant are identified. D and E, plot of maximum polymerization rate versus NPF concentration for 75 nM Arp2/3 complex titrated with Wsp1-VCA or GST-VCA. F, time course of pyrene actin polymerization from gelsolin-capped actin filament seeds incubated with a range of concentrations of wild type or Arp3⌬␣K/␤15 complex. G, fluorescence anisotropy binding assay showing titration of 75 nM Wsp1-Rh-VCA with wild type or Arp3⌬␣K/␤15 complex. Inset, competition binding assay showing titration of 0.5 M wild type or Arp3⌬␣K/␤15 complexes and 75 nM Wsp1-Rh-VCA with GST-Wsp1-VCA.
cally probe interactions of the complex with filament ends in distinct structural contexts.