Filament Assembly from Profilin-Actin*

Profilin plays a major role in the assembly of actin filament at the barbed ends. The thermodynamic and kinetic parameters for barbed end assembly from profilin-actin have been measured turbidimetrically. Filament growth from profilin-actin requires MgATP to be bound to actin. No assembly is observed from profilin-CaATP-actin. The rate constant for association of profilin-actin to barbed ends is 30% lower than that of actin, and the critical concentration for F-actin assembly from profilin-actin units is 0.3 μm under physiological ionic conditions. Barbed ends grow from profilin-actin with an ADP-Pi cap. Profilin does not cap the barbed ends and is not detectably incorporated into filaments. The EDC-cross-linked profilin-actin complex (PAcov) both copolymerizes with F-actin and undergoes spontaneous self-assembly, following a nucleation-growth process characterized by a critical concentration of 0.2 μm under physiological conditions. The PAcovpolymer is a helical filament that displays the same diffraction pattern as F-actin, with layer lines at 6 and 36 nm. The PAcov filaments bound phalloidin with the same kinetics as F-actin, bound myosin subfragment-1, and supported actin-activated ATPase of myosin subfragment-1, but they did not translocate in vitro along myosin-coated glass surfaces. These results are discussed in light of the current models of actin structure.

Profilin is a remarkable actin-binding protein. This small , ubiquitous, essential protein binds monomeric (G) actin in a 1:1 molar ratio (1). Although the profilin-actin complex is unable to spontaneously nucleate actin filaments, it can productively associate with the barbed ends specifically (2,3). This unique property confers a dual function to profilin, depending on the capping of barbed ends. When barbed ends are blocked by capping proteins, profilin sequesters G-actin. In contrast, in motile regions of the cell where uncapped barbed ends are actively elongating, the profilin-actin complex can actively participate in filament growth. The profilin-MgATPactin complex (PA) 1 can therefore be considered as an endspecific quasipolymerizable actin monomer. Recent results (4) show that this property of profilin is used to enhance the processivity of treadmilling (i.e. steady-state barbed end as-sembly)in the presence of actin depolymerizing factor. In the mechanism that was proposed based on thermodynamic data (5,6), F-actin assembly from profilin-actin is possible only because of its coupling to ATP hydrolysis, as follows. Profilinactin associates with the barbed end; the interaction of profilin with actin is weakened once the actin-bound ATP has been hydrolyzed; profilin then dissociates from that end, thus promoting the incorporation of one actin subunit in the filament and regenerating a free barbed end available for further growth from PA. Profilin is reused at each cycle and works as a catalyzer of assembly. While polymerization of actin alone proceeds in a manner uncoupled from ATP hydrolysis (7,8), these two reactions are proposed to occur in a compulsory order when actin filaments grow from profilin-actin units. In this paper, the proposed mechanism of barbed end growth from profilinactin is challenged by kinetic experiments.
Understanding the kinetics of filament assembly from profilin-actin in depth also has implications concerning the structure of the filament. In the atomic model of the actin filament proposed by Holmes et al. (9) and refined by Lorenz et al. (10), the profilin interaction area on actin is exposed at the barbed end, accounting for the association of profilin-actin to a growing barbed end, but not to a pointed end. Hence, within Holmes' model, it is anticipated that profilin can cap the barbed ends. On the other hand, examination of the actin-actin contacts in the crystals of the profilin-actin complex (11) shows evidence for a nonhelical ribbon structure of the profilin-actin complex, in which extensive actin-actin contacts, however different from those present in the Lorenz et al. atomic model of the actin filament (10) are involved. It was proposed (12,13) that barbed end growth from profilin-actin could involve the transient extension of profilin-ATP-actin ribbons. The transition from ribbon to filament would be coupled to ATP hydrolysis and dissociation of profilin. According to this model, the profilin-actin ribbon would exist in solution only in the ATP-or ADP-P ibound form, and the orientation of the actin monomer in the filament derived from the ribbon would be different from the Holmes atomic model.
In the present work, we first determined the rate parameters for barbed end elongation from profilin-actin using a turbidimetric method. We next attempted to find conditions under which profilin-actin would undergo spontaneous self-assembly, with the goal to characterize the profilin-actin polymer. The EDC-cross-linked profilin-actin complex appears able to polymerize into filaments, which exhibit the same helical structure and same thermodynamic stability as native F-actin filaments. Profilin-actin filaments therefore provide a tool to probe the structure of the actin filament and the interface of actin with other actin-binding proteins.

MATERIALS AND METHODS
Proteins-Actin was purified from rabbit skeletal muscle acetone powder (14) and isolated as CaATP-G-actin through Sephadex G-200 chromatography (15) in G buffer (5 mM Tris/Cl Ϫ , 0.1 mM CaCl 2 , 0.2 mM ATP, 1 mM dithiothreitol, 0.01% NaN 3 , pH 7.8). MgATP-G-actin (Ͻ20 * This work was supported in part by the Association Francaise contre les Myopathies, the Association pour la Recherche contre le Cancer, the Ligue Nationale contre le Cancer, and a grant from the European Economic Community (Contract CHRX-CT94-0652). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed. Fax: 33 01 69 82 31 29; E-mail: carlier@lebs.cnrs-gif.fr. 1 The abbreviations used are: PA, profilin-MgATP-actin complex; M) was prepared by incubation of CaATP-G-actin with 0.2 mM EGTA and 1 mol eq plus 10 M excess MgCl 2 for 3 min and used immediately afterward. MgATP-G-actin was polymerized by the addition of 2 mM MgCl 2 and 0.1 M KCl. CaATP-G-actin was polymerized by the addition of 0.1 M KCl. Actin was NBD-labeled (16). Spectrin-actin seeds were prepared from human erythrocytes, and their molar concentration was determined as described (17) and by titration by gelsolin. Gelsolin-actin seeds were prepared at a 2 M concentration by incubation of 2 M human plasma gelsolin (a generous gift from Dr. Yukio Doi) with 4.2 M CaATP-G-actin in G buffer. Profilin was isolated from bovine spleen by poly(L-proline) affinity chromatography (18). S 1 (A 1 ) and S 1 (A 2 ) isoforms of chymotryptic myosin subfragment-1 were resolved by SP-trisacryl chromatography (19).
Fluorescence and Light Scattering Measurements-Fluorescence measurements were carried out at 20°C using a Spex Fluorolog 2 spectrofluorimeter at the following wavelengths: NBD-actin, exc 475 nm and em 530 nm; rhodamine-phalloidin, exc 530 nm and em 575 nm; tryptophan fluorescence, exc 295 nm and em 330 nm.
Turbidity Measurements-Turbidity measurements were performed at 20°C at 310 nm using a Cary 1 Varian spectrophotometer with 1-cm path cuvettes. All solutions were thoroughly filtered and degassed before the experiment.
Measurement of the Initial Rate of Filament Elongation-Initial rates of filament growth were measured using turbidity or NBD fluorescence. Spectrin-actin seeds were used to initiate elongation at the barbed end (20), while gelsolin-actin seeds initiated elongation at the pointed end. At time 0, MgATP-G-actin (or CaATP-G-actin), in the absence or presence of profilin or profilin-actin covalent complex, was supplemented simultaneously with seeds and salts, and the time course of assembly was recorded. The actin concentrations were chosen so that the spontaneous nucleation could be neglected.
The rate of barbed end growth J at a given total concentration of G-actin ([A o ]) and in the presence of different total amounts [P o ] of profilin was monitored turbidimetrically. Data were analyzed as follows, In Equation 1, [S] is the concentration of spectrin-actin seeds; [A] and [PA] are the concentrations of free and profilin-bound G-actin; and k ϩ A and k ϩ PA are the corresponding barbed end association rate constants. The contribution of the off rates was neglected in equation 1 because measurements were carried out at actin concentrations well above the critical concentration. The change in J as the concentration [P o [ of profilin was increased reflected the increase in the contribution of PA to barbed end assembly as G-actin was gradually saturated by profilin. The concentration of PA was calculated as follows, where K PA represents the equilibrium dissociation constant for profilin-actin. Measurements of J in absorbance units/s were converted in M assembled F-actin/s using a critical concentration calibration curve as described previously (21), from which the polymerization of 1 M Factin led to an increase of 0.0017 absorbance units at 310 nm (1-cm optical path).
Combining Equations 1 and 2 led to the expression of J as a function of [P o ], [A o ], and K PA . Analysis of the data within this expression led to the determinations of k ϩ A (from the measurement of J in the absence of profilin), k ϩ PA (from the measurement of J in the presence of saturating amounts of profilin), and K PA , by adjustment of the theoretical curve to the data at different concentrations of profilin.
The rate of barbed end growth was also measured at different concentrations of G-actin and either in the absence or in the presence of saturating amounts of profilin (i.e. 1 mol eq of G-actin plus an excess of 5 M bovine profilin or of 30 M Arabidopsis thaliana profilin 3 (6)). The resulting J(c) plots for actin and profilin-actin were used to derive the values of k ϩ B for actin or profilin-actin and of the critical concentrations C c for polymerization, as follows.
All experiments were performed using freshly prepared G-actin in G buffer containing 1 mM dithiothreitol to be sure that Cys 374 was thoroughly reduced, so that all of the actin in the preparation was able to bind profilin with high affinity. We reasoned that if a very small proportion of G-actin was oxidized, it would give rise to spontaneous polymerization at high actin concentrations even in the presence of an excess of profilin. Therefore, where appropriate, the spontaneous polymerization measured in the absence of seeds was subtracted from the data.
Chemical Cross-linking of Actin and Profilin-The cross-linking of Glu 364 of actin to Lys 125 of bovine profilin (which corresponds to Lys 115 of Acanthamoeba profilin) was performed using EDC as described (22,23). Briefly, actin was dialyzed at 4°C versus 2 mM HEPES, 0.2 mM CaCl 2 buffer (pH 7.8); profilin was prepared in 5 mM HEPES, 1 mM dithiothreitol, 0.2 mM CaCl 2 buffer (pH 7.8). Samples were brought to 20°C. Actin (15 M, 60 ml) was activated with 2 mM EDC and 2 mM sulfo-N-hydroxysuccinimide for 20 min at pH 6.5. 20 M profilin and 0.2 mM dithiothreitol were then added, the pH was adjusted to 7.8, and the reaction was allowed to proceed for 30 min before it was quenched by 10 mM glycine. The yield of the cross-link was about 15% (Fig. 1, inset,  lane a).
Purification of the Profilin-Actin Covalent Complex-The mixture obtained at the end of the cross-linking reaction was first loaded on a DEAE-cellulose column (DE52; Whatman; 2.5-cm diameter, 10-cm length) equilibrated in DEAE buffer (5 mM HEPES, 0.2 mM CaCl 2 , 0.2 mM ATP, pH 7.8) (Fig. 1). Free profilin was recovered in the flowthrough. Following a wash step with one column volume of DEAE buffer supplemented with 50 mM NaCl ( Fig. 1, inset, lane b), the mixture of free actin and covalent profilin-actin complex was eluted with DEAE buffer containing 250 mM NaCl ( Fig. 1, inset, lane c). The eluate was applied to a poly(L-proline)-agarose column (1.5-cm diameter, 12-cm length) equilibrated in PLP buffer (10 mM Tris/Cl Ϫ , 0.2 mM CaCl 2 , 0.2 mM ATP, 0.1 M glycine, 0.1 M KCl, pH 7.8). Free actin was recovered in the flow-through ( Fig. 1, inset, lane d), and profilin-actin covalent complex was then eluted with PLP buffer containing 30% Me 2 SO, according to the procedure used for isolation of profilactin complex (24). The eluted complex was rapidly concentrated to about 30 M by ultrafiltration over a Diaflo PM30 membrane in a 100-ml Amicon cell, dialyzed against G buffer, and centrifuged at 400,000 ϫ g for 45 min.
The covalent complex obtained at this stage was at least 90% pure as juged by SDS gel electrophoresis and appeared able to self-assemble reversibly upon the addition of salt. The capacity of covalent profilinactin to undergo self-assembly declined with time following isolation of the complex. About 70% of the covalent profilin-actin complex that eluted from poly(L-proline)-agarose was polymerizable on day 1. This proportion declined by 2-fold within 2 weeks. A cycle of polymerization was added as a final step of the purification of covalent profilin-actin, as follows. The solution of covalent PA was incubated with 0.2 mM EGTA and 40 M MgCl 2 for 3 min and then overnight with 2 mM MgCl 2 and 0.1 M KCl at room temperature. The solution was centrifuged at 400,000 ϫ g for 45 min. The pellet, which typically contained about 70% of initial PA cov , was resuspended and dialyzed against G buffer for 24 h, with a brief sonication after 16 h. The solution was again centrifuged at 400,000 ϫ g for 45 min. No visible pellet was found. The supernatant consisted of pure covalent profilin-actin ( Fig. 1, inset, lane e). About 2.5 mg of covalent profilin-actin complex were obtained, corresponding to 7% of the initially reacted actin. This material was used within a week. Small amounts of oligomers of high molecular mass, occasionally present after the cross-link, were eliminated in the purification steps (compare lanes a and e). Several independent preparations of covalent profilin-actin complex yielded a material that behaved in a reproducible fashion in the different experiments. The concentration of the covalent profilin-actin complex was determined by the Bradford assay (Bio-Rad reagent) using G-actin as a standard.
Electron Microscopy-Profilin-actin covalent complex (1.5 M) in G buffer was first converted into MgATP-bound PA cov by incubation with EGTA and MgCl 2 as described above for G-actin and was induced to polymerize by the addition of 2 mM MgCl 2 and 0.1 M KCl at 20°C. At different times of the polymerization process, 4-l aliquots of the sample were deposited on air glow-discharged carbon coated grids. Following adsorption to the grid for 10 s, the specimens were negatively stained with 2% uranyl acetate and observed in a CM12 (Philips) electron microscope. Electron micrographs taken at a 35,000-fold magnification on electron image plates (Eastman Kodak Co.) were developed for 5 min in full-strength D19 developer (Kodak). Images were selected according to the quality of their optical diffraction pattern (no astigmatism, defocus value in the range 0.5-1 m, presence of the first and sixth layer lines). Images were digitized with a rotating drum microdensitometer (Optronics P-1000) using a scan raster of 25 m. The filaments were straightened and cut to a length corresponding to 4 -7 axial repeats and a width slightly larger than the diameter. The axial repeat is typically equal to 36.5 nm. The mean value of the densities surrounding the filaments was subtracted from the images, which were floated in 512 ϫ 512 arrays. Fourier transforms were calculated. They were characterized by the presence of three layer lines (ll0, ll1, and ll6), which can be indexed by the selection rule l ϭ Ϫ6n ϩ 13m.
Sedimentation Assay for Binding of S 1 to F-PA cov -Actin or PA cov complex in G buffer was freed from ATP by Dowex-1 treatment (25), polymerized at 5 M, and split into samples supplemented with 0, 4, 6, 8, 10, or 20 M S 1 (A 1 ). Samples were centrifuged 15 min later at 400,000 ϫ g for 10 min at 20°C in the TL 100 Beckman ultracentrifuge. Pellets were resuspended in the original volume of G buffer. Supernatants and resuspended pellets were submitted to SDS-polyacrylamide gel electrophoresis and Coomassie Blue-stained.
ATPase Measurements-Actomyosin MgATPase activity of myosin subfragment-1 was measured in the presence of 17 M F-actin or F-PA cov and 1 M S 1 (A 1 ) at 20°C in a buffer containing 5 mM Tris/Cl Ϫ , pH 7.8, 0.1 M KCl, 2 mM MgCl 2 , 1 mM ␥-32 P-labeled ATP. The reaction was started by the addition of S 1 (A 1 ). Aliquots of the reaction mixture were removed from the solution at 30-s intervals; brought into 1 N HCl, 10 mM ammonium molybdate; and processed for 32 P i extraction as described (8).
Actomyosin Motility Assay-Actin or PA cov was polymerized overnight at 1 M in the presence of 0.05 M KCl, 2 mM MgCl 2 , and 1 M tetramethyl rhodamine-phalloidin. In vitro motility assays were conducted using whole myosin tethered to monoclonal antibodies (anti LMM 5C3-2) immobilized onto a nitrocellulose-coated glass coverslip, as described (26). The rhodamine-phalloidin filaments were diluted to 3-5 nM in motility buffer (25 mM imidazole, 25 mM KCl, 4 mM MgCl 2 , 5 mM 2-mercaptoethanol, 0.2 mM CaCl 2 , 7.5 mM ATP, 0.1% methylcellulose, pH 7.6) supplemented with oxygen scavengers. Observations were made on a Zeiss microscope using a 100ϫ Plan Apochromat objective equipped with epifluorescence optics. Images of moving filaments were recorded with a SVHS video recorder.

Filament Assembly at the Barbed Ends from Profilin-Actin Requires Mg 2ϩ and Not Ca 2ϩ as Divalent Metal Ion Tightly
Bound to G-actin-Since profilin binds derivatized actin very poorly (27), the conventional method using the increase in fluorescence of N-pyrenyl-carboxyamidomethyl-or NBD-labeled actin as a probe of polymerization is inadequate in its presence. Turbidimetry, which has proven useful to monitor actin polymerization in the case of ADF/cofilin (21), was used. The initial rate of barbed end assembly from spectrin-actin seeds was measured at a given G-actin concentration and in the presence of increasing amounts of profilin. Fig. 2a shows that when filament barbed ends elongated from CaATP-G-actin subunits, profilin inhibited the growth in a concentration-dependent fashion. Total inhibition of growth was observed at saturation by profilin. Data, analyzed as described under "Materials and Methods," indicate that profilin binds CaATP-Gactin with an equilibrium dissociation constant of 1.2 Ϯ 0.2 M and that the profilin-CaATP-actin complex does not participate in barbed end assembly. Preassembled F-actin (5 M) was also used to measure barbed end growth in the presence of 5 M CaATP-G-actin, either in the absence or presence of 20 M profilin. Turbidity increased in the absence of profilin, demonstrating active barbed end growth, but decreased when profilin was added together with G-actin. 70% depolymerization of the F-actin seeds was observed within 5 min. In conclusion, in the presence of CaATP-actin, profilin acts as a pure G-actin-sequestering protein. In particular, it does not cap the barbed ends.
In contrast, when barbed ends elongated from MgATP-Gactin subunits, profilin caused only a partial (40%) inhibition of growth. The data (Fig. 2a) are quantitatively consistent with the view that profilin binds MgATP-G-actin with an equilibrium dissociation constant of 0.1 M, and the profilin-MgATPactin complex associates productively with the barbed ends of actin filaments. At saturation by profilin, filaments grew from profilin-actin units exclusively, at a rate 40% lower than from MgG-actin alone. No further inhibition was observed at concentration of profilin as high as 100 M, again indicating no detectable capping of the barbed ends by profilin.
In conclusion, in the presence of CaATP-actin, profilin acts as a pure G-actin-sequestering protein. In contrast, the complex of profilin with MgATP-actin participates in barbed end assembly. The different behaviors of profilin-Ca-actin and profilin-Mg-actin are in agreement with the conclusion of previous works (5, 6) that the effective participation of profilin-actin in barbed end growth required ATP hydrolysis to be coupled to polymerization, a condition that was fulfilled for profilin-Mgactin but not for profilin-Ca-actin. The hydrolysis of ATP during polymerization of profilin-Mg-actin was measured at a high concentration (46 M Mg-[␥-32 P]ATP-G-actin 1:1 complex plus 60 M profilin). A short sonication was applied immediately after the addition of salt to enhance the rate of polymerization by fragmentation. Filament assembly showed a short lag followed by a rapid increase in light scattering (complete in less than 1 min). ATP was hydrolyzed in a manner that showed no detectable uncoupling from polymerization. However, a definite proof that elongation is coupled to hydrolysis requires the demonstration that the rate of elongation be kinetically limited at high concentration by the rate of ATP hydrolysis.
Kinetic Parameters for Barbed End Assembly from Profilin-Actin Complex-The rate constants for profilin-MgATP-actin association to and dissociation from the barbed ends can be derived from the analysis of the dependence of the rate of  (Fig. 2b) were derived from growth rate measurements using spectrinactin seeds. The values found for the association rate constants were 4.9 Ϯ 0.5 M Ϫ1 s 1 for Mg-actin, in agreement with Pollard and Mooseker (28), and 3.5 Ϯ 0.4 M Ϫ1 s Ϫ1 for profilin-Mg-actin. The rate constants for MgATP-actin and profilin-MgATP-actin dissociation from the barbed ends were derived from the ordinate intercepts of the plots. Values of 0.6 Ϯ 0.2 s Ϫ1 and 1 Ϯ 0.3 s Ϫ1 were found for actin and profilin-actin, respectively. In other words, when barbed ends are actively elongating from profilin-actin as well as from G-actin subunits, terminal subunits dissociate at a rate 1 order of magnitude lower than the rate of dissociation of ADP-F-actin (8 s Ϫ1 ). In conclusion, the present kinetic data rule out a simple model for barbed end growth from profilin-actin, according to which actin incorporation and profilin release from the barbed end would be tightly coupled to P i release, as described in Fig. 7 (Scheme I). In this scheme, the dissociation rate constant measured in a regime of growth is the dissociation rate constant of ADP-actin, k D . The resulting J(c) plot for profilin-actin would then be a straight line with an ordinate intercept equal to k D , and the critical concentration (abscissa intercept) would be 1 order of magnitude higher than the experimentally observed value of 0.3 M.
In a range of high concentrations of profilin-actin, the J(c) plot deviated from linearity and curved downward (Fig. 2b,  inset), indicating that the association of profilin-actin units to barbed ends was rate-limited by another reaction. We propose this reaction to be ATP hydrolysis, since this feature appears to be characteristic of the barbed end growth from profilin-actin and was not observed when the elongation from G-actin was measured. Quantitatively identical data were obtained with bovine profilin and A. thaliana profilin 3 (compare open circles and closed diamonds in Fig. 2b, inset).
Profilin Is Not Detectably Incorporated in Rapidly Growing Filaments-When observed in electron microscopy, filaments assembled from profilin-actin units displayed a structure strictly identical to those assembled from unliganded actin. No profilin was detected in the pellets of sedimented filaments. Efforts were made to detect nonhelical ribbon-like extensions of filaments that might be transiently formed at the barbed ends in regimes of rapid growth. At very high G-actin concentration, ATP hydrolysis is known to be more largely uncoupled from F-actin assembly (7,29), hence profilin might remain transiently bound to F-ATP-actin stretches, possibly in a ribbonlike structure (12). Specimens undergoing rapid assembly at high concentrations of profilin-actin were rapidly observed in the initial stages of the reaction. No structural difference was seen between the core and the end-proximal regions of the filaments assembled from profilin-actin. The possibility that the ATP-bound profilin-actin ribbon at the tip of the growing filament was too small to be detectable by EM cannot be discarded. Finally, [BeF 3 Ϫ ,H 2 O], a structural analog of P i that binds to F-ADP-actin filaments and reconstitutes the intermediate state in ATP hydrolysis on F-actin (30), was used to examine whether profilin would be able to bind to F-ADP-P*actin. F-ADP-BeF 3 -actin was incubated overnight in the presence of different concentrations of profilin. SDS-polyacrylamide gel electrophoresis of the pellets of the sedimented samples showed no profilin bound to F-ADP-BeF 3 -actin.
In conclusion, all assays failed to detect the incorporation of profilin into F-actin. To increase the probability of formation of a pure profilin-actin polymer and to identify the nature of the actin-actin contacts that can be formed when profilin remains bound to actin, a purification of the covalently crosslinked profilin-actin complex was elaborated (see "Materials and Methods").
Self-assembly of PA cov -The covalent profilin-actin complex was assayed for its nucleotide binding properties. ATP was found bound to the complex following gel filtration on Sephadex G-25 (PD10; Amersham Pharmacia Biotech) in G buffer containing no ATP. To assess whether profilin retained the property to enhance the rate of nucleotide exchange in the PA cov complex, ⑀-ATP was mixed with either noncovalent profilin-ATP-actin or PA cov -ATP 1:1 complex (Ca-ATP-actin, in G buffer, with 20 M free Ca 2ϩ ) in the stopped flow, and the increase in fluorescence of ⑀-ATP associated with the exchange of ⑀-ATP for bound ATP was monitored. Comparison of the nucleotide exchange kinetics on the noncovalent and covalent complexes, shown in Fig. 3a, demonstrates that, in the presence of 20 M free Ca 2ϩ ions, the rate constants of nucleotide exchange were 0.085 s Ϫ1 and 0.025 s Ϫ1 for the noncovalent and covalent complex, respectively, while a rate constant of 0.0006 s Ϫ1 was measured for nucleotide exchange on unliganded actin under the same conditions. Nucleotide exchange therefore was Samples were kept at room temperature for 18 h before being centrifuged for 45 min at 400,000 ϫ g. Closed circles represent the concentration of PA cov in the supernatant measured by the Bio-Rad assay. Open circles are the calculated amounts of polymerized PA cov derived from the difference between the concentrations of total PA cov and of PA cov in the supernatant. accelerated 150-and 45-fold on noncovalent and covalent profilin-actin, respectively. The fact that the covalent profilinactin complex is purified by poly-L-proline chromatography also indicates that the poly-L-proline binding property of profilin (and of profilin-actin) is not altered in the covalent complex. In conclusion, the covalent complex is biochemically very similar to the noncovalent complex. PA cov was able to self-assemble upon the addition of salt. The time courses of spontaneous polymerization of actin alone, PA cov alone, and mixed actin and PA cov were compared (Fig.  3b). The turbidity curve observed when actin and PA cov are mixed together could not be described as the sum of the time courses recorded for actin alone and PA cov alone, which suggests that PA cov can copolymerize with F-actin. The effect of PA cov on the rates of filament growth at the barbed and pointed ends confirmed this conclusion. PA cov had no effect on the rate of growth at the pointed ends and very slightly slowed down the rate of G-actin assembly onto spectrin-actin seeds monitored by the change in fluorescence of NBD-actin, in agreement with the data shown in Fig. 3b. Notably, no high affinity capping of barbed ends by PA cov was observed. Altogether, these results indicate that PA cov can copolymerize with F-actin.
The kinetics of spontaneous polymerization of pure PA cov at different concentrations was examined by turbidimetry. The polymerization curves shown in Fig. 3c consisted in a lag phase followed by an exponential increase, suggestive of a nucleationgrowth process similar to the polymerization of actin itself. The lag phase was much more pronounced for CaATP-PA cov than for MgATP-PA cov . This feature again is strongly reminiscent of the slower nucleation of filaments from CaATP-actin than from MgATP-actin (31). Shearing of the polymers by pipetting accelerated the polymerization process of PA cov , consistent with the view that fragmentation of the PA cov polymers increases the number of elongation sites, as observed for F-actin filaments. The extent of turbidity change at the end of the polymerization process increased linearly with the concentration of PA cov in the polymerizing sample. Samples of PA cov polymerized at different concentrations in the range 1.5-10 M were sedimented at 400,000 ϫ g for 45 min when the turbidity plateau was reached. The concentration of unassembled PA cov present in the supernatants was 0.2 M for all samples. From these data, the specific increase in turbidity per M assembled PA cov was found to be 0.048 cm Ϫ1 at 310 nm, a value 28-fold higher than the one (0.0017 Ϯ 0.0002) measured for F-actin (21), indicating that the size of the PA cov polymer that scattered light was much larger than the size of the actin filament.
A conventional critical concentration plot was derived from measurements of the amounts of unassembled PA cov present in the supernatants of sedimented samples prepared by serial dilutions of a preassembled solution of 8 M PA cov followed by 18-h incubation at room temperature. Data shown in Fig. 3d demonstrate that PA cov polymerized with a critical concentration of 0.2 M.
The ATP bound to actin in the PA cov complex was hydrolyzed during the polymerization of PA cov . When 7.7 M PA cov was polymerized in the presence of ␥-32 P-labeled ATP, no 32 P was found in the pellet of the sedimented material at the end of the polymerization process. 75 M P i were found in the supernatant after 16 h, indicating that PA cov filaments turn over.
In conclusion, the polymerization of PA cov shares mechanistic similarities with the polymerization of F-actin.
PA cov Polymerizes into Helical Filaments-The structure of the PA cov polymer was examined by electron microscopy. Observation of negatively stained specimens of assembled PA cov at steady state showed a homogeneous population of bundled filaments. The ends of the bundles often appeared blunt, sug-gesting that the filaments that composed the bundle grew together in a synchronous fashion. The kinetics of formation of these bundles was examined in electron microscopy and turbidity simultaneously, at a concentration of PA cov low enough (1.5 M) for the different steps of nucleation and growth to be clearly time-resolved. Electron micrographs of the polymerizing sample at different times of the polymerization process shown in Fig. 4 indicate that during the lag time only short individual filaments were formed, which gradually interacted with each other in bundles. Essentially, bundles of ϳ4 -6 filaments were visible on the grid at the end of the lag phase (t ϭ 6 min) when the change in turbidity was less than 1% of the total change. Bundles became thicker with time. The Fourier transforms of the PA cov filaments displayed clear layer lines at spacings of 36 and 6 nm, corresponding to the first and sixth layer lines of F-actin. These results indicate that PA cov polymerizes into helical filaments that have the same helical periodicities as F-actin. Fourier transforms of the bundles observed at later stages of assembly show an identical pattern. Bundling most likely results from the change in charge of filaments due to profilin, which is a basic protein. Native filaments contain repulsive charges that maintain a distance between each polymer in solution. Binding of ligands can abolish the repulsion between filaments and favor their mutual interactions (32). The poly-L-proline binding site exposed at the surface of profilin may also mediate hydrophobic contacts between the PA cov filaments and favor the formation of bundles. Accordingly, bundling was less extensive at low ionic strength.
To understand whether the bundling resulted from the EDC/ NHS and/or dimethyl sulfoxide treatment of actin, in a control experiment the uncross-linked actin recovered from the poly-L-proline affinity column at the end of the PA cov preparation (see "Materials and Methods") was supplemented with 30% Me 2 SO and processed like the covalent complex. Turbidity measurements showed that this G-actin material polymerized at the same rate and to the same extent as standard actin and did not form bundles (data not shown).
Binding of phalloidin to the PA cov filaments was monitored by the increase in fluorescence of tetramethylrhodamine-phalloidin (33,34). The time courses for binding 1 mol eq of tetramethylrhodamine-phalloidin to 1.6 M either F-actin or F-PA cov were superimposable (Fig. 5), indicating that the binding site of phalloidin is very similar in the two structures.
Interaction of the Myosin Head with the Profilin-Actin Filament-Several assays were carried out to understand how the cross-link of profilin to actin affects the actin-myosin interface and its functional properties. The binding of myosin subfragment-1 to standard filaments and PA cov filaments in the absence of ATP was examined in parallel in a sedimentation assay. Data shown in Fig. 6 show that S 1 (A 1 ) binds tightly to PA cov filaments; however, the stoichiometry was lower than the 1:1 value confirmed here for S 1 binding to F-actin. A maximum of 0.4 -0.5 S 1 was bound per PA cov subunit.
In the electron microscope, filaments assembled from PA cov and partially decorated (as above) by S 1 failed to display the conventional arrowhead decoration by S 1 . The myosin heads were attached irregularly to the PA cov filament with randomly distributed orientations. No diffraction pattern could be derived from the images (data not shown).
In an ATP-free low ionic strength buffer, S 1 (A 1 ) is known to induce the rapid formation of G-actin-S 1 oligomers, followed by their slower condensation into arrowhead-decorated F-actin-S 1 filaments. These two reactions can be monitored by light scattering (19,35). Under those ionic conditions, no oligomer nor any assembly process could be detected within 1 h when S 1 (A 1 ) at concentrations up to 20 M was added to PA cov -ATP 1:1 complex (8 M) in ATP-free G buffer. In the control experiment carried out with 8 M ATP-G-actin 1:1 complex and 12 M S 1 (A 1 ), the polymerization was complete in 5 min.
On the other hand, the MgATPase of S 1 (A 1 ) was clearly enhanced by PA cov filaments. The rate of ATP hydrolysis, under different ionic conditions (presence versus absence of 0.1 M KCl), was 2.6-fold lower than the ATPase rate measured in the presence of the same amount of F-actin (17 M) in a parallel sample.
The sliding movements of actin and PA cov filaments over myosin-coated glass surfaces were assayed in parallel. While actin filaments moved at a speed of 6.6 Ϯ 1.2 m/min, the profilin-actin filaments remained immobile and firmly attached to the myosin-coated surface. Filaments containing 50% actin and 50% PA cov moved almost as well as standard actin filaments with an average speed of 3.9 Ϯ 1.0 m/min and displayed a discontinuous, somewhat stick-slip motion. Filaments containing 10% PA cov moved at the same speed and displayed the same regular motile behavior as standard filaments.
Overall, these data suggest, but do not prove, that the actin-S 1 interface remains functional in the profilin-actin filament, but some steric hindrance due to the presence of profilin inhibits the movement of the myosin head, which is necessary for its translocation along actin filaments. Similarly, the fact that S 1 fails to induce oligomer and F-acto-S 1 assembly from PA cov indicates that S 1 cannot interact with 2 PA cov molecules as it does with G-actin. DISCUSSION Under physiological ionic conditions (MgATP-actin, 2 mM MgCl 2 , 0.1 M KCl), filaments actively elongate from profilinactin units, at a rate that is only 30 -40% lower than from G-actin subunits. These results are in agreement with an earlier report (36), which used Acanthamoeba profilins and the Limulus acrosomal processes as seeds; however, we find no evidence for a weak capping activity of profilin even at physiologically high concentrations. Profilin is not incorporated into the filaments, even in the ATP-or ADP-P i -bound state, at variance with the expectations from a ribbon-to-helix model.
Kinetic data show that filament growth from noncovalent profilin-actin, although mechanistically coupled to ATP hydrolysis, is not kinetically coupled to P i release. Evidence for this conclusion is provided by the low critical concentration for barbed end assembly from profilin-actin and extrapolation of the J(c) plot to a low value of the dissociation rate constant k Ϫ , which is inconsistent with the presence of rapidly dissociating ADP subunits at the barbed ends in a regime of growth. In the proposed alternative scheme (Fig. 7), growth from profilinactin and ATP hydrolysis are tightly coupled, and a stabilizing ADP-P i cap (37) prevents the occurrence of rapidly dissociating ADP subunits at the ends of elongating barbed ends. The downward curvature of the plot at high profilin-actin concentration (with both bovine and plant profilins) is due, within this scheme, to ATP hydrolysis, which kinetically limits profilin release and subsequent filament growth. The rate of ATP hydrolysis would be about 80 s Ϫ1 under physiological ionic conditions and would be consistent with the value measured here for the association rate constant (4.9 M Ϫ1 ⅐s Ϫ1 ). If a higher value of the association rate constant is used, e.g. 10 M Ϫ1 ⅐s Ϫ1 (38), then the corresponding ATPase rate would be 163 s Ϫ1 . J(c) measurements carried out at lower ionic strength (1 mM MgCl 2 , no KCl) yielded a limit value of J(c) of 25 Ϯ 5 s Ϫ1 (data not shown). This value is in reasonable agreement with results showing that the rate of hydrolysis accompanying filament elongation reached a limit of 13 s Ϫ1 in the presence of 1 mM MgCl 2 (29).
The noncovalent profilin-actin complex can, to some extent, be considered as a polymerizing monomer with a low critical concentration of 0.3 M. This view corroborates in vivo observations of stabilization of actin filaments in cells overexpressing profilin (39). When filaments are assembled at steady state in the presence of profilin, the partial critical concentrations of free G-actin and profilin-actin can be derived from the combination of the laws of mass action, mass conservation, and statistical copolymerization (40) described by following equations, where C 1  The covalently cross-linked profilin-actin complex provides a tool to probe the structure of the actin filament. The EDCcross-link made here between a carboxylate of G-actin, preactivated by EDC, and an NH 2 group of bovine spleen profilin is thought to be the homolog of the EDC cross-link between Glu 364 of actin and Lys 115 of profilin in the Acanthamoeba profilin-actin complex. According to Schutt et al. (12), this corresponds to a cross-link between Glu 364 of actin and Lys 125 of profilin in the vertebrate profilin-actin complex. The covalent profilin-actin complex retained the divalent metal ion/nucleotide binding properties of the noncovalent profilin-actin complex. However, its assembly properties were very surprising. Considering the two current models of actin assembly of Holmes et al. (9,10) and of Schutt et al. (11), two alternative behaviors were expected. Within the Holmes/Lorenz model of the filament, the covalent profilin-actin complex was expected to be unable to self-assemble, due to steric conflict between subdomain 2 and profilin along the long pitch helix. On the other hand, PA cov was expected to efficiently cap the barbed ends of actin filaments. Within Schutt's hypothesis, the covalent profilin-actin complex was expected to possibly polymerize into a nonhelical ribbon polymer.
The experimental facts do not fully meet the behaviors expected from either model. The data show that the covalent profilin-actin does not cap the standard actin filaments and easily co-polymerizes with actin. Moreover, the EDC-crosslinked profilin-actin complex easily self-assembles into filaments that display the same helical periodicity and thermodynamic stability as standard F-actin filaments, suggesting that the actin-actin contacts are similar in F-actin and in F-PA cov and that the stabilization of the profilin-actin interface by the covalent cross-link does not inhibit F-actin assembly. This conclusion is also supported by the identical kinetics of phalloidin binding to F-actin and F-PA cov and by the functional binding of the myosin head to the profilin-actin filament.
In order to accommodate all of the data within Holmes' model, we propose that as soon as ATP is cleaved following profilin-actin association to the barbed end, the affinity of profilin for actin falls by several orders of magnitude. When the two proteins are covalently linked, the loss in affinity causes the profilin to swing out the filament around the covalent link located in the outer part of subdomain 1, allowing the association of a subsequent PA cov unit to the barbed end. In this movement, the actomyosin interface remains unhindered to some extent, allowing partial decoration of F-PA cov by S 1 in rigor and acto-S 1 ATPase activity.
The huge loss in affinity of profilin for actin following cleav- Samples were centrifuged at 400,000 ϫ g. The supernatants and resuspended pellets were submitted to SDS-polyacrylamide gel electrophoresis. In the F-actin ϩ S 1 samples (top panel), the volume of loaded pellets corresponds to a 2 times higher amount of the initial actin plus S 1 solution than the supernatants. The positions of S 1 heavy chain, actin, and PA cov are indicated. Scanning of the gel patterns (not shown) showed that in the F-actin plus S 1 sample, S 1 increases linearly in the supernatant above a total concentration of 5 M, consistent with 1:1 binding ratio of S 1 to F-actin; conversely, in the PA cov plus S 1 sample, S 1 increases linearly in the supernatant above a total concentration of 2 M S 1 , consistent with a maximum binding ratio of 0.4 S 1 per PA cov subunit. age of ATP is a very strong assumption that has to be made to interpret the present data within Holmes' model, in particular to account for the low critical concentration for PA cov assembly. It would explain why profilin does not cap the barbed ends. It implies that the simple cleavage of ATP causes a very large structural change of subdomains 1 and 3, which does not appear in the atomic model built from the structure of CaATP-Gactin. This implication needs to be challenged by further experiments. The loss in affinity of profilin for actin might also tentatively be explained by the unfolding of profilin following the covalent cross-link; this explanation, however, seems unlikely, since the native properties of profilin-actin (poly-L-proline binding, enhancement of nucleotide exchange) appear to be conserved in the PA cov complex, indicating that cross-linked profilin is not unfolded.
Another possible interpretation of our data, which would be inconsistent with Holmes' model, is that both the profilin-G-actin interface and the actin-actin bonds that build the F-actin filament are conserved upon polymerization of the covalent profilin-actin. In the absence of covalent link, the loss in affinity of profilin for F-actin would not allow the profilin to remain bound to F-actin. This interpretation also accounts for the failure of profilin to cap the barbed ends, but it implies that the orientation of actin in the filament is different from both the Holmes and Schutt models. High resolution analysis of the structure of the F-PA cov filament and image reconstruction are now under way to solve those issues.