ATP hydrolysis on actin-related protein 2/3 complex causes debranching of dendritic actin arrays

Abstract

Extension of lamellipodia, an important dissipative process in cell motility, is driven by the turnover of a polarized dendritic array of actin filaments. Motility is driven by catalytic cycles of filament attachment to Wiskott–Aldrich syndrome protein (WASP)-activated actin-related protein (Arp)2/3 complex at the leading edge, branch formation, and detachment, allowing subsequent growth of branched filaments. The morphology, mechanical strength, and lifetime of the array are determined by the processes of filament branching, debranching, and treadmilling. All three processes are controlled by ATP hydrolysis. ATP hydrolysis on F-actin is known to be at the origin of treadmilling. Here, by using radiolabeled ATP covalently bound to Arp2/3, we show that ATP is hydrolyzed on Arp2, not on Arp3, after a delay following filament branching. Hydrolysis of ATP on Arp2 promotes debranching of filaments and acts as a clock that controls the stability of dendritic actin arrays in lamellipodia. Finally, we propose that hydrolysis of ATP on G-actin in the ternary G-actin–WASP–Arp2/3 complex on branch formation destabilizes the WASP–actin interface and energetically facilitates the detachment step in the branching reaction.

The extension of cell protrusions, generated by site-directed polymerization of actin into cohesive 3D networks, is the first elementary process in cell migration (1). The mechanical properties of these extensions, lamellipodia and filopodia, are controlled by several signal-responsive proteins, including actin-related protein (Arp)2/3 complex, which induces the formation of branched filaments (2–4); cortactin, which stabilizes the branches (5); and ABP 280/filamin A, which cross-links filaments orthogonally in gels (6). Elucidating how the regulation of the structure, stability, and plasticity of the branched, cross-linked actin array is coupled to efficient locomotion is a crucial issue. The protrusive activity at the leading edge is mimicked by the propulsive movement of Listeria or Shigella pathogens (7), endocytic vesicles (8), or functionalized microspheres (4, 9), which has been analyzed in vitro in reconstituted motility assays.

Recent data have indicated that Arp2/3 complex, activated by Wiskott–Aldrich syndrome protein (WASP) or by the Listeria protein ActA (10–14), incorporates into filaments after barbed end branching (4, 9, 15–17) and appears located at the branch junction in lamellipodia (18). In addition, the balance between filament branching and capping of barbed ends by capping proteins controls the branch spacing, which is one of the factors that determines the mechanical characteristics of the branched actin array (9, 19). Another important parameter is the rate of debranching. In vitro, branched filaments assembled in the presence of WASP and Arp2/3 complex spontaneously debranch after a delay of a few minutes after branching (2, 15, 16, 20). In motile structures like Listeria actin tails or lamellipodia, debranching of filaments is expected to create destabilization zones in the array and to generate filament pointed ends available for depolymerization. Therefore, the rate of debranching is a potentially important kinetic parameter in the control of the mechanical strength, morphology, and lifetime of the dendritic array in vivo. Yet the key elementary reactions that are involved and determine the recycling of Arp2/3 are not known.

Arp2/3 complex contains Arp2 and Arp3 (21), which bind ATP. Binding of ATP to Arp2 is enhanced after activation of Arp2/3 by WASP and is required for filament branching (22). In contrast, binding of ATP to Arp3 is not affected by WASP. These results suggest (22) that ATP binding and hydrolysis on Arp2/3 must play a pivotal role in Arp2/3 function. This issue is addressed here.

Methods

Proteins. Actin was purified from rabbit muscle, Arp2/3 from bovine brain, and other proteins used here as described (13). Actin was fluorescently labeled on Cys-374 by using pyrenyliodoacetamide and on Lys-373 by using 7-nitrobenz-2-oxa-1,3-diazol-4-yl (NBD)-Cl.

Photo-Cross-Linking of 8-Azido-ATP to Arp2/3. ATP was covalently bound to either Arp3 alone or Arp2 and Arp3 as follows. Arp2/3 (2 μM) in buffer A (5 mM Tris·HCl, pH 7.5/100 mM KCl/1 mM MgCl₂/0.1 mM CaCl₂) was incubated with 5 μM 8-azido-[γ-³²P]ATP (ALT, Lexington, KY) with or without 20 μM WA, the C-terminal domain of N-WASP, and photoirradiated for 45 s. The reaction was stopped by 1 mM DTT. Autoradiography of the electrophoresed photoirradiated material showed that 18% of Arp3 and 5% of Arp2 had [γ-³²P]ATP covalently bound in the absence of WA, whereas 12% Arp3 and 23% Arp2 had [γ-³²P]ATP covalently bound in the presence of WA. The active covalent adduct was isolated as follows. Free 8-azido-[γ-³²P]ATP was eliminated by treatment by hexokinase (10 units/ml) and 1 mM glucose. Covalent [γ-³²P]ATP-Arp2/3 complex was separated from free 8-azido-ADP, glucose, and glucose 6-[³²P]phosphate by gel filtration (Sephadex G-25) in buffer A + 1 mM DTT. After the hexokinase/gel filtration steps, Arp2 carried either exchangeable ADP or covalently bound [γ-³²P]ATP. Only the latter species was able to branch filaments. To evaluate the filament-branching activity of Arp2/3 photo-cross-linked to ATP, Arp2/3 photoirradiated in the presence of 8-azido-ATP and WA and treated by hexokinase/glucose was chromatographed on ATP-agarose to remove the fraction of unreacted Arp2/3 that still possessed exchangeable ATP-binding sites. It was verified that the ATP-agarose-treated Arp2/3 contained 2 mol of [³²P]ATP per mol of Arp2/3 and that the rates of branched actin polymerization were identical within 10% in the absence or presence of added free ATP. The filament-branching activity of the photo-cross-linked azido-ATP-Arp2/3 complex prepared as described previously was measured in a polymerization test by using 1:1 ATP–G-actin complex (5 μM), 1 μM WA, and comparison with standards of ATP-Arp2/3. Of the branching activity of Arp2/3, 75% was recovered when 8-azido-ATP had been photo-cross-linked to both Arp2 and Arp3.

ATP Hydrolysis on Arp2/3 During Polymerization into Branched Filaments. Cleavage of ATP (acid-labile P_i) was monitored by extraction of ³²P-labeled phosphomolybdate (23) during polymerization of 5 μM ATP–G-actin 1:1 complex (4% pyrenyl-labeled) and 1 μM WA with 100 nM Arp2/3 complex carrying covalently bound [γ-³²P]ATP on both Arp2 and Arp3 in buffer A containing 1 mM DTT, 50 μM BeCl₂, and 10 mM NaF to prevent actin depolymerization after exhaustion of ATP. The specific radioactivity of [γ-³²P]ATP-labeled Arp2/3 was in the range 3–30 Ci/mmol (1 Ci = 37 GBq). The concentrations of actin and WA used were selected for producing the maximum number of branched filaments from 100 nM Arp2/3. The kinetics of ATP hydrolysis on Arp2/3 were unchanged when free ATP was present in the assay (compare Fig. 1 B and C). For P_i release measurement, samples were centrifuged at 300,000 × g for 300 s at room temperature and the supernatant was submitted to ³²P_i measurement as described in the previous section. When only Arp3 had been photo-cross-linked to azido-[γ-³²P]ATP (in absence of WA), the hydrolysis of ATP on Arp3 during formation of branched filaments was monitored in the presence of free ATP so that unlabeled ATP was bound to Arp2, allowing branching to occur. All experiments shown have been reproduced at least three times.

Fig. 1.

ATP hydrolysis occurs on Arp2 after actin filament branching. (A) Photo-cross-linking of 8-azido-[γ-³²P]ATP to the Arp2/3 complex (2 μM) in the absence or presence of 20 μM WA. Samples were submitted to SDS/PAGE. Lane 1, Coomassie blue staining of Arp2/3. γ-³²P-labeled subunits were revealed by using a PhosphorImager. Lanes 2 and 3, nonphotoirradiated Arp2/3 in the absence and presence of WA; lanes 4 and 5, photo-cross-linked 8-azido[γ-³²P]ATP-Arp2/3 in the absence and presence of WA. (B) Quantitation of the correlation between branch formation and ATP hydrolysis by Arp2/3. Branched filaments were assembled with 100 nM photo-cross-linked [γ-³²P]ATP-Arp2/3, 1 μM WA, and 5 μM pyrenyl-labeled actin-ATP 1:1 complex (no free ATP was present, ensuring that all of the branching events were due to photo-cross-linked [γ-³²P]ATP-Arp2/3). Aliquots were removed at intervals to measure acid labile ³²P_i (filled red circles) or centrifuged before ³²P_i extraction to measure released P_i (open red circles). The time course of production of branches (blue line) was derived from the polymerization curve (15). (Inset) Enlarged view of the early stage of the reaction (same symbols). Spontaneous ATP hydrolysis on Arp2/3 in the absence of actin is shown in green circles. (C) ATP hydrolysis occurs on Arp2, not on Arp3, after filament branching. 8-Azido[γ-³²P]ATP was photo-cross-linked to Arp2/3 in the absence (open symbols) or presence (filled symbols) of WA as in A. Corresponding kinetics of actin assembly into branched filaments (open and filled blue squares) and associated hydrolysis of [γ-³²P]ATP bound to either Arp3 alone (open red circles) or to both Arp2 and Arp3 (filled red circles) were monitored simultaneously, as in B, except for the presence of 100 μM ATP, allowing activation of Arp2 when [γ-³²P]ATP is photo-cross-linked to Arp3 only. A control sample was tested (green circles) in which Arp2/3 was prepared with WA and 8-azido-[γ³²P]ATP as above but not photoirradiated. No hydrolysis of azido-ATP was recorded, testifying that the measured hydrolysis in the curve with red circles occurs strictly on covalently cross-linked ATP-Arp2 and is not a parasite reaction occurring on actin.

Binding of CrATP to Arp2/3. The displacement of 1,N⁶-etheno-ATP (ε-ATP) bound to Arp2/3 by CrATP was performed in the presence or absence of WA, using the enhancement of fluorescence of ε-ATP after binding Arp2/3 as described (22). CrATP was as efficient as MgATP in displacing ε-ATP from either Arp3 in the absence of WA or both Arp2 and Arp3 in the presence of WA.

Dependence of Filament Branching on the Nucleotide Bound to Actin and Arp2/3. MgADP–G-actin was prepared by incubating MgATP–G-actin with hexokinase (10 units/ml) and 1 mM glucose, followed by Sephadex G-25 filtration in 5 mM Tris·HCl (pH 7.5)/100 μM MgCl₂/10 μM ADP/0.2 mM DTT. CrATP-actin was prepared as described (23) by addition of 500 μM CrATP to Mg[³H]ADP-actin (prepared as described previously by using ³H-labeled ATP).

To assemble branched filaments from MgADP-actin or CrATP-actin and ATP-Arp2/3, Arp2/3 was photo-cross-linked to azido-ATP, gel-filtered, and treated by ATP-agarose as described in the previous section. A sedimentation assay of the polymerized actin was performed to check that CrATP had displaced 85% of actin-bound [³H]ADP.

To polymerize filaments with CrATP bound to actin and Arp2/3, CrATP was added to MgADP-actin and to Arp2/3 preequilibrated with 10 μM ADP.

Kinetics of Debranching. Filament debranching was monitored by light microscopy as described (16). Actin (5 μM) was rapidly polymerized in the presence of WA and Arp2/3, under indicated conditions of bound nucleotide, in 20 mM imidazole, pH 7.0 (which ensures stability of CrATP over the time course of the experiments). As soon as 90% polymerization was reached, aliquots were removed from the solution at time intervals, supplemented with rhodamine-phalloidin, diluted 500-fold, and processed for observation in an Olympus (Melville, NY) AX-70 microscope with a ×100 (numerical aperture 1:35) objective and a Lhesa (Cergy-Pontoise, France) LH4046 video camera. The number of branches per micrometer of actin filaments was calculated by using METAMORPH software. At each time point, a total filament length of 5,000–30,000 μm and about 500 branches at the earliest time were counted.

Double-Label Seeded Growth Assay. Branched filaments were assembled from unlabeled G-actin (2.5 μM), 0.5 μM WA, and 50 nM Arp2/3 complex; let sit for 90 min until debranched; and stabilized by addition of 2.5 μM rhodamine-phalloidin. These filaments were used as seeds by 20-fold dilution into a solution of 2 μM G-actin for 1–2 min. After addition of 2 μM Alexa 488-labeled phalloidin, filaments were diluted 500-fold and processed for rhodamine/Alexa 488 fluorescence microscopy observation as described in the previous section.

Results

Monitoring ATP hydrolysis on Arp2/3 complex during polymerization of actin into branched filaments is a challenging task because actin, which is at concentrations about three orders of magnitude higher than Arp2/3, hydrolyzes ATP after polymerization. The potential contribution of Arp2/3 represents about 0.1% of the overall hydrolysis of ATP during assembly of branched filaments. This difficulty was overcome by measuring hydrolysis of [γ-³²P]ATP covalently bound to Arp2 and/or Arp3 exclusively while actin was polymerizing with bound unlabeled ATP. We have shown that the K_d for binding ATP to Arp2 is 40 μM in the absence of WA and 0.1 μM in the presence of WA (22). Hence, at low concentration of ATP (<10 μM), ATP binds and is photo-cross-linked to Arp3 only in the absence of WA and to both Arp2 and Arp3 in the presence of WA. The yield of photoincorporation of ATP was greatly enhanced when 8-azido-[γ-³²P]ATP was used instead of ATP, whereas the effect of WA on the binding of ATP to Arp2 remained unaltered after replacing ATP by 8-azido-ATP (Fig. 1 A). The autocatalytic polymerization curves recorded with Arp2/3 can be analyzed and modeled assuming that Arp2/3 generates new filaments by barbed end branching exclusively and is not a conventional nucleating agent (15). That the polymerization curves remained unchanged with photoirradiated covalent ATP-Arp2/3 instead of Arp2/3 that freely exchanges ATP demonstrates that the filament-branching activity of Arp2/3 is unaltered by the covalent cross-linking of ATP (Fig. 2B).

Fig. 2.

ATP hydrolysis on Arp2 is required for actin filament debranching. (A) ATP hydrolysis on Arp2 is not a consequence of debranching. ATP hydrolysis on Arp2/3 was monitored during polymerization as in Fig. 1, in the absence of phalloidin (filled green circles) or after addition of 10 μM phalloidin (red circles) at the end of the polymerization at 200 s (arrow). Filled red circles, cleavage of ATP; open red circles, P_i release in the presence of phalloidin. Actin assembly in the presence or in the absence of phalloidin was recorded simultaneously (blue line). (B) Filament branching does not depend on the actin-bound nucleotide. The assembly reaction contained 5 μM actin (continuous lines) or 7.5 μM actin (dashed lines), 1 μM WA, 100 nM Arp2/3 at pH 7.0, and the indicated combinations of nucleotide bound to Arp2/3 and actin. X-linked ATP = cross-linked ATP. When indicated, only photo-cross-linked ATP-Arp2/3 was present due to pretreatment by ATP-agarose (see Methods). (C) P_i release from Arp2 is required for filament debranching. (Upper) Filaments branched with the indicated combinations of nucleotides were stabilized at 180 and 3,500 s by rhodamine-phalloidin. (Lower) The time course of filament debranching was monitored after polymerization was performed by using the same conditions and combinations of bound nucleotides as in B. Each experiment was repeated at least three times.

ATP-actin was assembled in branched filaments in the presence of Arp2/3 complex carrying photo-cross-linked [γ-³²P]ATP on both Arp2 and Arp3 (see Methods). Polymerization was complete in 400 s, whereas ATP was hydrolyzed on Arp2/3 complex in an exponential process strongly uncoupled from the branching reaction, with a rate constant of (8 ± 1)·10^—4 s^—1 (half-time of 800 ± 100 s; Fig. 1B). A plateau was reached at which 4.8 nM ATP had been hydrolyzed. Kinetic analysis of the polymerization curve (15) showed that up to 4.5 nM filaments were created by branching. Hence, one molecule of ATP was hydrolyzed on Arp2/3 complex after formation of a branch. The time courses of release of P_i and cleavage of ATP superimposed, indicating that P_i release is faster than chemical cleavage of ATP on Arp2/3.

When branched filaments were assembled with Arp2/3 complex carrying photo-cross-linked [γ-³²P]ATP on Arp3 only, while ATP was noncovalently bound to Arp2, no hydrolysis of radioactively labeled ATP was detected even 2 h after formation of branched filaments (Fig. 1C). In the parallel sample containing photo-cross-linked [γ-³²P]ATP on both Arp2 and Arp3 and branching at the same rate, ATP was hydrolyzed after branch formation. Altogether, these results demonstrate that ATP is hydrolyzed on Arp2, not on Arp3, with a delay of several minutes after filament branching.

Strikingly, ATP hydrolysis on Arp2/3 complex kinetically correlated with filament debranching, which occurred with the same half-time of 800 ± 100 s when filaments were branched with either covalently or noncovalently bound ATP (Fig. 2C). To determine whether ATP hydrolysis on Arp2 is the cause or the consequence of the dissociation of the daughter filament from the mother filament, phalloidin was added at the end of the polymerization process to slow down debranching (2). The kinetics of ATP hydrolysis and P_i release on Arp2 were not affected by phalloidin, indicating that ATP hydrolysis is not consecutive to and likely precedes debranching (Fig. 2 A). Cortactin, which synergizes with WA to activate Arp2/3 complex (5), also slowed down debranching but, like phalloidin, did not affect ATP hydrolysis on Arp2 (data not shown).

To examine whether the nature of the nucleotide bound to actin affected branching or debranching, Arp2/3-stimulated polymerization of actin in branched filaments was carried out by using different combinations of nucleotides bound to actin/Arp2/3, respectively, as follows: MgATP/MgATP, MgADP/MgATP, CrATP/MgATP, and CrATP/CrATP. CrATP is a metal ion exchange-inert analog of MgATP in which the stability of the bonds between the Cr(III) ion and the β- and γ-phosphates of ATP prevents the release of the cleaved γ-phosphate of ATP on F-actin (23). Branching of filaments occurred irrespective of the actin-bound nucleotide (Fig. 2B), in full agreement with a recent report (17). The rate of debranching was the same whether ATP was covalently or noncovalently bound to Arp2/3 and was not affected by the nucleotide bound to actin. Specifically, branched filaments assembled from either MgATP-actin or CrATP-actin using MgATP-bound Arp2/3 debranched at the same rate, but filaments branched with CrATP-bound Arp2/3 debranched 12-fold more slowly (Fig. 2C), indicating that if P_i is not released from Arp2, the daughter filament does not dissociate from the mother filament. Altogether, these results demonstrate that ATP hydrolysis/P_i release on Arp2 at the branch drives dissociation of the daughter filament but they do not support the expressed view (20) that ATP hydrolysis and P_i release from F-actin, which occurs 10-fold faster (k = 6·10^—3s^—1; ref. 24), cause filament debranching. That branched filaments made from ADP-actin and ATP-Arp2/3 branch as well as those assembled from ATP-actin does not support the proposal that Arp2/3 complex branches filaments by preferential side-binding to the terminal ADP–P_i–F-actin subunits, rather than by binding to the barbed ends (25, 26).

In the present solution studies, the time at which WA putatively dissociates from the branch is not known, whereas in vivo, dissociation of the branch from membrane-bound WASP precedes filament growth. To examine whether dissociation of the activator affects ATP hydrolysis on Arp2/3 or debranching of filaments, assembly of branched filaments was performed under conditions mimicking the in vivo context, using N-WASP, the neural homolog of WASP, immobilized at the surface of polystyrene beads (9). Beads generated autocatalytic polymerization curves identical to the ones observed with soluble WA. As reported (9, 27), branched filaments released from bead-immobilized N-WASP were observed in the light microscope. Hydrolysis of [γ-³²P]ATP covalently cross-linked to Arp2/3 complex followed the same slow kinetics as when Arp2/3 was activated by soluble WA (data not shown). Hence, ATP hydrolysis on Arp2/3 at the branch and debranching are not rate-limited by the dissociation of WA but may occur at a later stage after the faster dissociation of WA.

The above results then raise the issue of the nature of the elementary process that causes dissociation of the filament from surface-bound activator after branch formation, thus enabling growth of the mother and daughter filaments. Filament branching involves the association of the ternary G-actin–WA–Arp2/3 complex with a barbed end. The G-actin–WA moiety of the ternary complex, like profilin–actin, associates productively with the barbed end and participates in barbed end growth (13). Accordingly, the covalently cross-linked WA-actin complex copolymerizes with actin (M.-F.C., unpublished results). We proposed (4, 15) that the association of the G-actin–WA moiety of the ternary complex with the barbed end of the mother filament would initiate the formation of a branch. Within this hypothesis, ATP hydrolysis on WA-bound G-actin, after barbed end association, should destabilize the interaction between the two proteins and facilitate the detachment of the product of the branching reaction. This hypothesis was tested by monitoring the binding of WA to MgATP–G-actin and MgADP–G-actin comparatively by using the quenching of fluorescence of NBD-G-actin linked to WA binding as a probe. Fig. 3 shows that, like profilin (28), WA binds MgADP–G-actin with a 7-fold-lower affinity than MgATP–G-actin, in support of the described hypothesis.

Fig. 3.

ATP hydrolysis destabilizes the interaction between WA and G-actin. Fluorescence of NBD-labeled MgATP–G-actin (red circles) or MgADP–G-actin (blue circles) in low ionic strength buffer was quenched by addition of WA as indicated. The changes in fluorescence (λ_ex = 475 nm, λ_em = 530 nm) are normalized and expressed in terms of concentration of WA–actin complex. Dashed lines represent the stoichiometric titration curve (infinite affinity). Solid curves are calculated with values of K_d of 0.1 and 0.7 μM for the ATP-bound and ADP-bound forms of WA–actin complex, respectively. The same result was obtained at low ionic strength and at high ionic strength (two reproducible experiments were carried out at each ionic strength).

Visualization and quantitation of rhodamine-actin and Alexa 488-Arp2/3 complex in actin tails by using fluorescence microscopy have revealed that as the fluorescence intensities of actin and Arp2/3 decline, the actin:Arp2/3 ratio is maintained constant along the tail (9). This result indicates that Arp2/3 complex is lost at the same rate as actin by depolymerization. Because ADP-bound Arp2/3 is unable to branch filaments (22, 29), ATP exchange for bound ADP has to occur to recycle Arp2/3 subsequent to ATP hydrolysis. To determine whether a second branching reaction can take place from mother filament-bound Arp2/3 after dissociation of the daughter branch and exchange of ATP for bound ADP, branched filaments were left aside until debranched, stained with rhodamine-phalloidin, and used as seeds by diluting them in a solution containing G-actin. Filament stretches formed in 2 min were stained with Alexa 488-labeled phalloidin. Alexa 488-labeled stretches grew endwise only from rhodamine-labeled seeds. No side growth of filaments from “old” Arp2/3 branching sites was observed, indicating that Arp2/3 complex branches only once in the polymerization process (data not shown). In conclusion, recycling of Arp2/3 complex is linked to filament depolymerization. Nucleotide exchange on Arp2 is rapid and probably is not rate-limiting in the Arp2/3 ATPase cycle in vitro (rate constant on the order of 1 s^—1; ref. 29) but it might be affected by regulators in vivo.

Discussion

We have shown that ATP hydrolysis regulates Arp2/3 function. Although no hydrolysis of ATP can be detected on isolated or on activated Arp2/3 complex (22, 29), Arp2 is committed to hydrolyze ATP after filament branching. In contrast, no hydrolysis of ATP occurs on Arp3. A comprehensive scheme (Fig. 4) featuring consecutive elementary steps in Arp2/3-induced filament branching summarizes the results and emphasizes the kinship among actin, tubulin, small G proteins, and Arp2/3 complex regarding the role of nucleotide exchange and hydrolysis in biological activity. Like actin, Arp2/3 complex hydrolyzes a bound nucleotide once committed in a macromolecular assembly. As on actin, ATP hydrolysis on Arp2 is not required for the building of this assembly but it promotes the destabilization of protein–protein interactions and leads to disassembly of the scaffold. However, differences exist among actin, Arp2, and Arp3 despite the quasi-identity of residues in the ATP-binding sites of the three proteins (21). ATP is not hydrolyzed on Arp3 after branching and ATP is hydrolyzed more slowly on Arp2 than on Mg-F-actin. These differences may be related to differences in the protein–protein contacts that trigger hydrolysis of ATP on actin in the filament and on Arp2/3 at the branch. The nature of the partner protein that dissociates from Arp2 in the debranching process is not known. Because the Arp2/3 complex has seven subunits, several scenarios are possible. One possibility, consistent with similarities between Arp2 and actin, is that when ATP is hydrolyzed on Arp2, an Arp2–actin bond in the branch, homolog of the ATP-sensitive actin–actin interface in the filament, is weakened. Phalloidin then might slow down debranching by stabilizing this actin–Arp2 bond.

Fig. 4.

Model for the control of actin filament branching–debranching and turnover by ATP binding and hydrolysis on actin and Arp2/3 complex. The following elementary steps are involved in filament branching and debranching. Step 1, activation of Arp2/3 complex by formation of the ternary complex G-actin–WA–Arp2/3. Step 2, association of a filament to the immobilized activated complex. Step 3, ATP hydrolysis on actin after incorporation at the branch destabilizes its interaction with WA, leading to detachment of the branched filament. Step 4, growth of the mother and daughter branches. Step 5, ATP hydrolysis on Arp2 at the branch causes debranching. Step 6, depolymerization of filaments. Step 7, nucleotide exchange on G-actin and Arp2/3 allows recycling of active actin and Arp2/3.

Overall, our results and conclusions regarding the role of ATP binding and hydrolysis on Arp2/3 complex differ from the views expressed by others (20, 29). First, ATP hydrolysis on F-actin was thought to cause debranching (20), because phalloidin, which slows down P_i release from F-actin, slows down debranching. However, those experiments had been done with rhodamine-phalloidin, which actually accelerates P_i release (30), hence the release of P_i from F-actin could not be responsible for debranching. The present results in addition show that debranching is not slowed down when CrADP-P_i is bound to F-actin. Second, the finding that adenosine 5′-[β,γ-imido]triphosphate (AMPPNP)-Arp2/3 did not branch filaments suggested that ATP hydrolysis was required at some step of the branching process (29) or that, consistent with the low affinity of AMPPNP for Arp2, AMPPNP-Arp2/3 did not have the active structure of ATP-Arp2/3 (22). The present results show that ATP hydrolysis on Arp2/3 is not required for filament branching, in support of the latter conclusion.

The fate of Arp2/3 complex after debranching is an open issue. That in a motility assay the actin:Arp2/3 ratio remains constant along the actin tail of propelling microspheres (9) indicates that Arp2/3 is not released from the branch junction after debranching but remains bound to either the mother or the daughter filament. It is known (15) that Arp2/3 does not cap the pointed end, which supports the view (illustrated in Fig. 4) that it remains bound to the mother filament after debranching.

It is remarkable that, in addition to ATP hydrolysis-driven treadmilling, two other dissipative reactions accompany the formation of branched actin arrays and regulate their stability and mechanical strength as well as motile behavior. The first dissipative reaction is associated with the cycle of filament attachment–branching–detachment (9, 31, 32). After association of the mother filament with surface-immobilized G-actin–WASP–Arp2/3 complex, ATP hydrolysis (P_i release) occurs on the actin subunit that is incorporated in the branched structure (33). We have shown here that hydrolysis of ATP greatly weakens the interaction of actin with the C-terminal domain of N-WASP. We propose that ATP hydrolysis on this actin subunit is likely to contribute in the detachment of the product of the branching reaction from immobilized N-WASP. The second ATP hydrolysis event occurs on Arp2 incorporated in the branch and destabilizes the interaction of Arp2/3 with the daughter branch, causing debranching. Hydrolysis of ATP on Arp2 is slow (rate constant 8·10^—4 s^—1), imposing a delay between the time courses of formation of a branched filament array and of its disappearance by debranching.

In conclusion, as illustrated in Fig. 4, ATP hydrolysis acts as a double clock controlling the proportion of attached and detached filaments during stationary movement on the one hand and the lifetime and morphology of branched filament arrays on the other hand. In vitro, the half-time of debranching is 10 min. Assuming that filaments debranch at the same rate in vivo, in a cell crawling at 1 μm/min, 26% of the filaments must have debranched at a distance of 3 μm from the leading edge, thus bringing an appreciable destabilization of the actin array at the rear of the lamellipodium. It is also possible that some cellular factors accelerate debranching in vivo. Potential regulation of ATP hydrolysis on the actin protomer incorporated at the branch and on Arp2 is expected to have a bearing on the protrusive force/velocity and stability of branched actin arrays. How these reactions further combine with the activities of other proteins such as fimbrin, filamin, and cortactin that control the overall architecture of the actin array is an open issue.