The Human Arp2/3 Complex Is Composed of Evolutionarily Conserved Subunits and Is Localized to Cellular Regions of Dynamic Actin Filament Assembly

To determine the amino acid sequence of the non-Arp polypeptide subunits of the Arp2/3 complex, the proteins were first subjected to SDS-PAGE, and then blotted to polyvinylidinedifluoride (PVDF)¹ membrane (Applied Biosystems, Inc., Foster City, CA). Proteins were identified by staining with Ponceau S (Hirano and Watanabe, 1990). The PVDF-immobilized proteins were reduced, S-carboxymethylated, and digested in situ with Achromobacter protease I and endoproteinase Asp-N (Iwamatsu, 1992). Digested peptides were chromatographed by reverse-phase HPLC using a Wakosil-II AR C18 300 Å column (2.0 × 150 mm; Wako Pure Chemical Industries, Ltd., Osaka, Japan). Amino acid sequencing was performed with a gas-phase sequencer (model PPSQ-10; Shimadzu Corp., Tokyo, Japan).

To identify cDNA clones encoding the individual subunits of the human Arp2/3 complex, the peptide sequences were used to search the EMBL/GenBank/DDBJ database of expressed sequence tags (dbEST). The following cDNA clones were identified in the database and were obtained from Genome Systems, Inc. (St. Louis, MO) (clone Id numbers are indicated): Arp3, 328293; Arp2, 587272; p41-Arc, 527050; p34-Arc, 547253; p21-Arc, 326562; p20-Arc, 187446; and p16-Arc, 300395. In addition, a cDNA-encoding p34-Arc was isolated from a λ phage Uni ZAP HeLa cell cDNA expression library (Stratagene, La Jolla, CA) by screening with anti-p34 antibody using standard procedures. The DNA sequence of each clone was determined by sequencing both strands (performed by the Biomolecular Resource Center Sequencing Facility at the University of California, San Francisco, CA).

Polyclonal anti-peptide antibodies that recognize p34-Arc were generated by immunizing rabbits (Berkeley Antibody Co., Richmond, CA) with the synthetic peptide (C)KVFMQEFKEGRRASHTAPQVLFSHRE, corresponding to amino acids 179–204 of the protein. An amino-terminal cysteine was added for sulfhydryl coupling. Peptide conjugation and antibody affinity purification were carried out as previously described (Sawin et al., 1992). Antibodies were concentrated by dialysis into 10 mM MOPS, pH 6.8, 150 mM NaCl, 35% glycerol.

To generate polyclonal antibodies that recognize p21-Arc (formerly referred to as the 23-kD subunit; Welch et al., 1997), the corresponding gene was amplified by PCR and cloned into pGEX-3X (Smith and Johnson, 1988) to create a glutathione S-transferase (GST) fusion protein. Purified GST fusion protein was used to immunize rabbits (Berkeley Antibody Co.). An affinity matrix for purification of anti-p21 antibodies was created by cleaving p21-Arc from the GST-p21 fusion protein using Factor Xa (New England Biolabs Inc., Beverly, MA) and covalently coupling it to affigel-10 (Bio Rad Laboratories, Hercules, CA). Antibody affinity purification was performed using standard procedures (Harlow and Lane, 1988). Antibodies were concentrated by dialysis into 10 mM MOPS, pH 6.8, 150 mM NaCl, 35% glycerol. Immunoblotting was carried out using affinity-purified anti-Arp3 (0.4 μg/ml), anti-p34 (0.04 μg/ml), and anti-p21 (0.6 μg/ ml) antibodies using the ECL system (Amersham Corp., Arlington Heights, IL).

Swiss 3T3 fibroblasts were grown in DME-H16 medium with 10% FBS, penicillin, and streptomycin. To create a confluent layer of cells for the wounding experiment, cells were plated at a density of 3 × 10⁴ cells per ml onto 22-mm diam coverslips and allowed to grow to confluence. A wound was created using a plastic pipette tip. HeLa cells were grown in DME-H16 medium with 10% FBS, penicillin, and streptomycin. Cells were plated onto glass coverslips 1–2 d before immunofluorescence experiments. Infection of HeLa cells with L. monocytogenes strain 10403S was carried out as previously described (Theriot et al., 1994). Locomoting fibroblasts were prepared from heart explants from 7-d-old chick embryos as described in Cramer et al., 1997. Cells were plated onto glass coverslips treated with 1 mg/ml poly-d-lysine (Sigma Chemical Co., St. Louis, MO), and 125 μg/ml matrigel (Collaborative Research, Inc., Waltham, MA) and were grown in DME-H16 medium with 10% FBS and 10% heat-inactivated chicken serum, penicillin, and streptomycin. Immunofluorescence experiments were carried out within 20 h of plating, which was shown by Cramer et al. (1997) to be optimal for observing rapidly locomoting cells.

For immunofluorescence experiments, cells were extracted for 30 s in cytoskeleton buffer with 0.32 M sucrose (CBS) (Small, 1981; Symons and Mitchison, 1991) also containing 0.1% Triton X-100 and 1 μg/ml phalloidin. Cells were fixed in CBS plus 4% formaldehyde for 20 min followed by −20°C methanol for 3 min (methanol fixation was necessary to observe staining with the anti-p34 and anti-p21 antibodies, indicating that they recognize only denatured protein). Indirect immunofluorescence was performed using affinity-purified rabbit anti-Arp3 (4 μg/ml), anti-p34 (0.4 μg/ ml), anti-p21 (6 μg/ml) and mouse anti-actin (5 μg/ml; Boehringer Mannheim Biochemicals, Indianapolis, IN) primary antibodies, and Texas red– conjugated goat anti–rabbit and FITC-conjugated goat anti–mouse secondary antibodies (Jackson Immunoresearch Laboratories, Inc., West Grove, PA). Fluorescent phalloidin staining of filamentous actin was carried out with 1 μg/ml FITC-conjugated phalloidin (Sigma Chemical Co.). Photomicrographs were taken using a cooled CCD camera (Princeton Scientific Instruments, Inc., Princeton, NJ) on a Nikon Optiphot-2 microscope. Images were handled digitally using Adobe Photoshop (Adobe Systems Inc., Mountain View, CA).

The gene encoding Arp3 was isolated by PCR and was cloned into pEGFP-N1 (Clontech, Palo Alto, CA) to create a fusion of Arp3 with green fluorescent protein (EGFP) attached to its carboxy terminus (pArp3-EGFP). (The DNA sequence of the Arp3 gene isolated by PCR was identical to the sequence of the Arp3 cDNA clone.) pArp3-EGFP was introduced into HeLa cells using calcium phosphate–mediated transfection as described previously (Heald et al., 1993). To visualize Arp3-EGPF in L. monocytogenes–infected cells, HeLa cells were infected 1 d after transfection. For fluorescence experiments, cells were extracted in CBS with 0.1% Triton X-100 and 1 μg/ml phalloidin for 30 s, and then fixed in CBS with 3.2% formaldehyde for 10 min. To visualize polymerized actin, cells were stained with 1 μg/ml 5-carboxytetramethyl rhodamine phalloidin (Sigma Chemical Co.) for 2 min. Cells were photographed as described above except that a blocking filter was used to eliminate bleed through from the rhodamine to the fluorescein channel.

To further understand the role of the Arp2/3 complex in the control of actin assembly in cells, we determined the predicted amino acid sequences of all of the subunits of the human protein complex. Using peptide sequence information (Welch et al., 1997) to search the EMBL/GenBank/DDBJ dbEST we obtained and sequenced full-length human cDNA clones encoding Arp2 and Arp3 (these sequence data are available from EMBL/GenBank/DDBJ under accession numbers AF006082 for Arp2 and AF006083 for Arp3). The Arp2 cDNA encodes a predicted protein of 394 amino acids with a molecular mass of 45 kD (43 kD as determined by SDS-PAGE; Welch et al., 1997). Human Arp2 is 99% identical to chicken Arp2 (these data are available from EMBL/GenBank/DDBJ under accession number X73971), and is also highly similar to Arp2 proteins from Drosophila melanogaster (81% identical; accession number X78486), Caenorhabditis elegans (76% identical; accession number Z71181), A. castellanii (72% identical; accession number U29609), and S. cerevisiae (67% identical; accession number X61502). Human Arp3 is 418 amino acids in length and has a predicted molecular mass of 47 kD (50 kD as determined by SDS-PAGE; Welch et al., 1997). It is identical in sequence to bovine Arp3 (accession number D12816), and shows extensive similarity to Arp3 proteins from D. melanogaster (80% identical; accession number X71789), A. castellanii (70% identical; accession number U29610), Dictyostelium discoideum (69% identical; accession number Z46418), Neurospora crassa (59% identical; accession number U79737), S. pombe (59% identical; accession number M81068), and S. cerevisiae (58% identical; accession number Z49565).

We also searched the dbEST database using peptide sequences derived from the five non-Arp subunits of the Arp2/3 complex. Full-length cDNA clones encoding each subunit were identified, and their DNA sequences were determined. The predicted polypeptides encoded by these clones are similar in sequence to proteins from a wide variety of eukaryotes.

Based on its high degree of sequence similarity, the p41-Arc subunit of the platelet Arp2/3 complex is a new member of the Sop2 family of proteins. As is shown in Fig. 1,A, p41-Arc is 67% identical to the previously identified human Sop2 protein Sop2Hs (Balasubramanian et al., 1996), making them similar but distinct members of the same family. p41-Arc is also similar to Sop2 proteins from D. melanogaster (49% identical; accession number L39625), S. pombe (37% identical; accession number Y08998), and S. cerevisiae (34% identical; accession number Z36103). Previous evidence has suggested that Sop2 family proteins encode the 41-kD subunit of the Arp2/3 complex. The canonical member of this family, S. pombe Sop2p, associates with Arp3 in a protein complex, and mutations in the sop2 and arp3 genes interact genetically. Moreover, Sop2 proteins contain sequences that are similar to peptide sequences from the 40-kD subunit of the A. castellanii Arp2/3 complex (Balasubramanian et al., 1996). Like previously identified Sop2 family proteins, human p41-Arc contains WD repeat sequences (also known as β-transducin repeats). We identified seven potential WD repeats in p41-Arc (Fig. 1 A) that correspond to the consensus sequence described by Neer et al. (1994) with one to five mismatches. These repeats are highly conserved in both sequence and position within p41-Arc and Sop2Hs.

The 41- and 40-kD subunits of the platelet Arp2/3 complex (Fig. 2, far left lane), which differ slightly in their mobility by SDS-PAGE, are different versions of the p41-Arc polypeptide. Digestion of p41 and p40 with Achromobacter protease I and endoproteinase Asp-N produced identical peptide maps by reverse phase HPLC (not shown). This suggests that p41 and p40 may differ in their state of posttranslational modification. Furthermore, this indicates that the platelet Arp2/3 complex contains seven polypeptides, identical to the number found in the A. castellanii complex. Both p41 and p40 are the only subunits of the platelet Arp2/3 complex that are present in substoichiometric quantities, while the other six subunits are represented in slightly larger and roughly equal amounts (Welch et al., 1997). However, when summed together, the quantity of the p41 and p40 is stoichiometric with that of the other subunits. This indicates that the seven polypeptide subunits of the complex are present in a 1:1:1:1:1:1:1 stoichiometry, similar to that found in the A. castellanii Arp2/3 complex (Machesky et al., 1994; Mullins et al., 1997).

The four other subunits of the human Arp2/3 complex— p34-Arc, p21-Arc, p20-Arc, and p16-Arc (formerly the 34-, 23-, 22-, and 19-kD subunits as determined by SDS-PAGE; Welch et al., 1997)—are similar to proteins from diverse eukaryotic organisms. However, they do not share similarity with any proteins of known function. These therefore define four novel protein families, which are shown in Fig. 1,B. The p34-Arc family is comprised of the human and S. cerevisiae proteins (34% identical). Furthermore, a region in p34-Arc is similar in sequence to a peptide from 35-kD subunit of the A. castellanii complex (Fig. 1,C). A partial sequence of human p34-Arc had previously been determined and its gene maps to chromosome 13 in the BRCA2 region (Couch et al., 1996). Like p34-Arc, p21-Arc has a single homologue in S. cerevisiae (45% identical; Fig. 1,B) and contains a region similar to a peptide in the 18-kD subunit of the A. castellanii complex (Fig. 1,C). The p20-Arc family has members in multiple organisms (Fig. 1,B) including C. elegans (73% identical in its amino-terminal 168 amino acids), S. pombe (68% identical), and S. cerevisiae (67% identical). In addition, it contains a sequence that is similar to the 19-kD subunit of the A. castellanii complex (Fig. 1 C). Finally, p16-Arc resembles proteins in C. elegans (34% identical), S. pombe (26% identical), and S. cerevisiae (21% identical). The S. cerevisiae proteins that share sequence similarity with human p34-Arc, p21-Arc, p20-Arc, and p16-Arc have recently been shown to be components of the yeast Arp2/3 complex (Winter et al., 1997).

Although the sequences of these subunits have not provided clues to their potential function in the Arp2/3 complex, we have used the sequence information as well as the cloned genes to generate reagents to probe the function of the complex in cells. We have raised polyclonal antibodies to a GST-p21 fusion protein and to a peptide from p34-Arc. Affinity-purified antibodies recognize p21-Arc and p34-Arc in the pure Arp2/3 complex (not shown), and specifically recognize 21- and 34-kD proteins in whole cell extracts made from a variety of mammalian and avian cell types (Fig. 2). In addition, we have taken advantage of the previously described anti-Arp3 antibody (Welch et al., 1997), which specifically recognizes Arp3 in the same cell extracts (Fig. 2). Together, these reagents provide a means of examining the function of the Arp2/3 complex by determining its intracellular distribution.

The Arp2/3 complex promotes the assembly of actin filaments at the surface of L. monocytogenes. Moreover, the Arp3 subunit of the complex is localized to the actin tails assembled by moving intracellular bacteria and to actin clouds surrounding stationary bacteria (Welch et al., 1997). To confirm that other subunits of the Arp2/3 complex exhibit the same localization pattern, we carried out immunofluorescence experiments on L. monocytogenes– infected HeLa cells using anti-p34 and anti-p21 antibodies. As is shown in Fig. 3, A and C, both p34-Arc and p21-Arc are localized to L. monocytogenes assembled actin clouds, and throughout tails with a distribution that mirrors that of Arp3 (Welch et al., 1997; and see below) and polymerized actin (Fig. 3, B and D). Punctate cytoplasmic staining also was evident, and was reduced by briefly pre-extracting cells with detergent in the presence of phalloidin to stabilize actin structures (see Materials and Methods).

To further confirm the observed localization pattern, we created a tagged Arp3 with green fluorescent protein (GFP) at its carboxy terminus. We chose to create this particular fusion because S. cerevisiae Arp3 tagged at its carboxy terminus can complement a null mutation in the ARP3 gene (Winter et al., 1997). HeLa cells were transfected with pArp3-EGFP plasmid DNA, and 1 d later were infected with L. monocytogenes. Arp3-GFP fluorescence was observed in actin clouds and tails with a distribution indistinguishable from that seen by immunofluorescence (Fig. 3 E; actin localization in F). GFP alone did not localize to actin clouds or tails (not shown). As was seen in the immunofluorescence experiments, cytoplasmic Arp3-GFP fluorescence was always apparent, even in cells expressing lower levels of the fusion protein. A brief pre-extraction of the cells before fixation significantly reduced this cytoplasmic background.

The actin-based motility of L. monocytogenes may be mechanistically similar to actin-driven cellular movements such as the protrusion of lamellipodia. To evaluate whether the Arp2/3 complex may be involved in actin polymerization–driven membrane protrusion, we determined its intracellular distribution. We observed the localization of Arp2/3 complex subunits in a variety of tissue culture lines by immunofluorescence. Mouse Swiss 3T3 fibroblast cells gave the most consistent results, although a similar staining pattern was evident in all cell lines tested (including human HeLa epithelial-like cells and mouse MC7 3T3 fibroblasts). As is shown in Fig. 4, Arp3 (A), p34-Arc (C), and p21-Arc (E) were distributed throughout the thin actin-rich lamellipodia at the edges of Swiss 3T3 cells (actin staining in the same cells is shown in Fig. 3,B, D, and F). Staining was not observed along the entire margin of the cell, but was confined to lamellipodia (Figs. 4 and 5). Strikingly, other actin-containing structures such as stress fibers and focal adhesions were not stained. The punctate cytoplasmic staining pattern was especially prominent in thicker regions of the cell in the vicinity of the nucleus (Fig. 5).

Stationary Swiss 3T3 fibroblasts produce actin-rich lamellipodia that actively protrude and retract but are not involved in directional motility. These were randomly dispersed over the cell surface as is illustrated by the distribution of the p21 subunit of the Arp2/3 complex and actin in these cells (Fig. 5, A and B). To ascertain whether the Arp2/3 complex may be involved in membrane protrusion during cell migration, we observed the localization of the complex in motile fibroblasts. We generated a population of moving Swiss 3T3 fibroblasts by wounding a confluent layer of cells. Cells at the margin of the wound responded immediately by generating membrane ruffles at their edges facing the wound, indicating that they had begun to reorient their actin cytoskeleton. Within 30 min, they began to move outwards towards the center of the wound. p21-Arc was localized to the broad actin-rich lamellipodia at the leading edges of these moving cells (Fig. 5, C and E, and actin localization in D and F).

We also generated a population of motile primary cultured fibroblasts and looked at the distribution of Arp2/3 complex subunits in these cells by immunofluorescence. To prepare moving fibroblasts, tissue explants from the hearts of 7-d-old embryonic chicks were plated onto coverslips. Cells migrate outward from explants within 12 h, and continue to move rapidly for up to 20 h (Cramer et al., 1997). These cells assumed a morphology typical of locomoting cells, with a cone- or fan-shaped cell body and broad lamellipodia at their leading edges (Fig. 6). Both Arp3 (Fig. 6,A) and p34-Arc (Fig. 6 C) were localized throughout lamellipodia. Neither protein was found in actin bundles, fully consistent with their distribution in Swiss 3T3 fibroblasts.

The seven subunits of the human Arp2/3 complex define seven protein families that are present in diverse eukaryotic organisms. Six share sequence similarity with subunits of the A. castellanii complex (sequence information for A. castellanii p14 is not yet available). Moreover, every subunit has a homologue S. cerevisiae, and the yeast equivalents of human p34-Arc, p21-Arc, p20-Arc, and p16-Arc have recently been shown to be components of the Arp2/3 complex (Winter et al., 1997). This indicates that the structure of the Arp2/3 complex has been conserved through evolution and suggests a corresponding conservation of its activity in all eukaryotic cells. The complex appears to be essential for cell viability and actin cytoskeletal function, as deletion of the genes encoding Arp2, Arp3, and Sop2p is lethal in yeast, and less severe mutations cause defects in actin organization and disrupt actin-dependent processes (Lees-Miller et al., 1992; Schwob and Martin, 1992; Moreau et al., 1996; McCollum et al., 1996; Balasubramanian et al., 1996; Winter et al., 1997).

The Arp2/3 complex induces actin polymerization at the cell surface of L. monocytogenes (Welch et al., 1997), suggesting that it performs a similar function in cells. It is likely to promote actin assembly by filament nucleation. Structural modeling of Arp2 and Arp3 proteins (Kelleher et al., 1995) has suggested that these subunits may serve as a template for the nucleation of polymerization, similar to the function performed by γ-tubulin in the initiation of microtubule assembly (Moritz et al., 1995; Zheng et al., 1995). Although structural considerations led Kelleher et al. (1995) to propose that the Arp subunits interact with the pointed end of the filament, interaction with the barbed end remains a possibility. Of the Arp2/3 complex subunits, the sequences of Arp2 and Arp3 proteins are among the most highly conserved between species, consistent with their proposed role in nucleating actin polymerization.

The other five subunits of the Arp2/3 complex may be important for stabilizing its structure, regulating its biochemical activity, and/or determining its intracellular localization. The sequence and biochemical properties of one of these subunits, p41-Arc, suggest that it plays an important regulatory role. This subunit is a member of the Sop2 family of proteins and is most similar to another human protein, Sop2Hs. The similarity between these proteins suggests that they may both function as the p41 subunit of the human Arp2/3 complex. Multiple versions of the p41 subunit may adapt the functions of the complex to different cell types or developmental stages. Platelets appear to use only p41-Arc since the peptides derived from the 41- and 40-kD polypeptides were all identical to it in sequence, and clearly different from the corresponding regions in Sop2Hs.

p41-Arc, like Sop2Hs, contains seven putative WD repeats that are evenly dispersed over the entire sequence of the protein. The presence of WD repeats may provide some insight into the potential functions of this subunit. These repeats, which were first identified in the G_β subunit of heterotrimeric GTP-binding proteins, have since been found in between four to ten copies in proteins with a wide variety of functions (for review see Neer et al., 1994). The repeat unit is involved in the formation of a characteristic three-dimensional structure known as a β-propeller, each repeat constituting one blade of the propeller (Wall et al., 1995; Lambright et al., 1996; for review see Neer and Smith, 1996). Because many WD repeat-containing proteins participate in the formation of multiprotein complexes, one possibility is that p41-Arc is involved in assembling and maintaining the structure of the Arp2/3 complex.

Another intriguing possibility is that p41-Arc mediates the interaction of the complex with other molecules that regulate its localization or activity. One known ligand of WD repeats is the pleckstrin homology (PH) domain (for reviews see Lemmon et al., 1996; Shaw, 1996), which is found in a variety of cell signaling and cytoskeletal proteins that interact with cell membranes. In addition to binding WD repeat sequences, PH domains can simultaneously bind phosphatidyl inositol 4,5-bisphosphate, and thus may serve as adapters that link WD repeat–containing proteins to the cell membrane. We have found that the Arp2/3 complex is localized to the cell edges in lamellipodia. An interaction between p41-Arc in the Arp2/3 complex and a PH domain–containing protein may mediate the recruitment of the complex to the cell membrane and may also serve to activate the actin polymerization promoting capacity of the complex. A protein that recruits the Arp2/3 complex to cell membranes would be functionally similar to the L. monocytogenes ActA protein, which is required for actin polymerization by bacteria in vivo (Domann et al., 1992; Kocks et al., 1992) and for the activity of the Arp2/3 complex at the bacterial surface (Welch et al., 1997).

Posttranslational modification of p41-Arc may also regulate the activity and intracellular distribution of the complex. We have determined that the 41- and 40-kD subunits of the platelet Arp2/3 complex are the same polypeptide, suggesting that they may differ in their state of posttranslational modification. However, the nature of this putative modification and its effect on the Arp2/3 complex remain to be determined.

The four other subunits of the Arp2/3 complex (p34-Arc, p21-Arc, p20-Arc, and p16-Arc) define four novel protein families. The extent of sequence conservation between family members from different species indicates that the structure of these subunits has been conserved through evolution. However, the sequences of these proteins do not indicate what their role in the structure or regulation of the complex may be.

Two of these subunits of the A. castellanii complex—the 35-, and either the 19-, 18-, or 14-kD subunits—were found by chemical cross-linking to be intimately associated with actin (Mullins et al., 1997). The Arp3 subunit and actin could be also cross-linked. It was suggested that these subunits mediate the ability of the A. castellanii complex to bind the sides of actin filaments and to bundle actin filaments into parallel arrays in vitro. However, these subunits do not exhibit any similarity to known actin-binding proteins, implying that they may possess novel actin-binding motifs.

The localization of subunits of the Arp2/3 complex in yeast and A. castellanii is consistent with its proposed role in promoting actin polymerization in cells. The S. cerevisiae Arp2 and Arp3 proteins are found in the motile cortical actin patches, but not in cytoplasmic cables that consist of bundled actin filaments (Moreau et al., 1996; Winter et al., 1997). S. pombe Arp3 shares a similar distribution, and in addition is absent from the contractile ring (McCollum et al., 1996). In ameba, subunits of the complex colocalize with actin at the cell cortex (Machesky et al., 1994; Kelleher et al., 1995). Staining is enriched in actin fibrils that may represent bundles of actin filaments, in agreement with the bundling activity exhibited by the A. castellanii complex in vitro (Mullins et al., 1997).

We have now determined the localization of multiple subunits of the Arp2/3 complex in vertebrate cells. The Arp3, p34-Arc, and p21-Arc subunits are found in the lamellipodia of motile and stationary fibroblasts. They are also localized to L. monocytogenes assembled actin clouds and tails in infected cells. In contrast, Arp3, p34-Arc, and p21-Arc were not detected in cellular bundles of actin, including stress fibers in tissue culture fibroblasts and graded polarity bundles (Cramer et al., 1997) in primary cultures of chick embryo fibroblasts. Moreover, they have not been observed in filopodia, which also contain bundled actin (DePace, A., M. Welch, and T. Mitchison, unpublished results).

Actin filaments in lamellipodia and those in actin bundles differ in their spatial organization and dynamic properties. Filaments in lamellipodia are arranged as an orthogonal meshwork and are oriented at an angle of ∼45°, relative to the direction of protrusion (Small et al., 1995). They exhibit rapid polymerization dynamics, turning over with a t_1/2 of <1 min in some cell types (Theriot and Mitchison, 1991). This rapid turnover is similar to that seen for filaments in L. monocytogenes tails (Theriot et al., 1992). Conversely, filaments in stress fibers and other cellular bundles are arranged in closely spaced parallel arrays. They exhibit much slower polymerization dynamics than filaments in lamellipodia (Theriot and Mitchison, 1991). We do not detect the Arp2/3 complex in actin bundles, suggesting that the filament side-binding and bundling activity exhibited by the complex in vitro (Mullins et al., 1997) may not be a primary determinant of its function in vertebrate cells. Instead, we find that the Arp2/3 complex is present with the rapidly polymerizing and depolymerizing populations of actin filaments in lamellipodia and L. monocytogenes tails, further strengthening the analogy between bacterial motility and lamellipodial protrusion. Together with the biochemical activity of the Arp2/3 complex, these results suggest a model in which the complex promotes actin polymerization in the lamellipodium and participates in lamellipodial protrusion. The ability of the complex to bind filaments might serve to retain the complex in the lamellipodium, making it readily available for additional rounds of polymerization.

Before we can put together a comprehensive model of actin polymerization dynamics in lamellipodia, several outstanding questions need to be answered. For example, it is not clear whether actin polymerization occurs throughout the lamellipodium, matching the distribution of Arp2/3 complex, or only at the leading edge. In addition, it is not known whether the mechanism of new actin filament polymerization involves the elongation of existing filaments or continuous nucleation of new filaments. Finally, the mechanisms that regulate the spatial and temporal control of actin polymerization have yet to be determined. Further characterization of the interactions of the Arp2/3 complex with actin filaments and with factors that regulate its activity and localization should help shed light on the answers to these outstanding problems.

This work was supported by grant from the National Institutes of Health (GM48027) to T.J. Mitchison, and a postdoctoral fellowship from the National Institutes of Health to M.D. Welch.

CBS

cytoskeleton buffer with sucrose

dbEST

database of expressed sequence tags

EGFP and GFP

green fluorescent protein

GST

glutathione S-transferase

PH

pleckstrin homology