Control of Actin Dynamics by Proteins Made of β-Thymosin Repeats

Actobindin is an actin-binding protein from amoeba, which consists of two β-thymosin repeats and has been shown to inhibit actin polymerization by sequestering G-actin and by stabilizing actin dimers. Here we show that actobindin has the same biochemical properties as the Drosophila orCaenorhabditis elegans homologous protein that consists of three β-thymosin repeats. These proteins define a new family of actin-binding proteins. They bind G-actin in a 1:1 complex with thermodynamic and kinetic parameters similar to β-thymosins. Like β-thymosins, they slow down nucleotide exchange on G-actin and make a ternary complex with G-actin and Latrunculin A. On the other hand, they behave as functional homologs of profilin because their complex with MgATP-G-actin, unlike β-thymosin-actin, participates in filament barbed end growth, like profilin-actin complex. Therefore these proteins play an active role in actin-based motility processes. In addition, proteins of the actobindin family interact with the pointed end of actin filaments and inhibit pointed end growth, maybe via the interaction of the β-thymosin repeats with two terminal subunits.

tration of ATP-G-actin. Therefore, upon addition of profilin to a solution of F-actin in the presence of T␤4, the amount of T␤4actin complex decreases, i.e. F-actin increases (15). Conversely, addition of actin depolymerizing factor (ADF), which increases the concentration of ATP-G-actin at steady state, leads to increased sequestration of G-actin by T␤4, i.e. causes F-actin disassembly (16).
The structure of the T␤4-actin complex is not known at atomic resolution; however, NMR and biochemical studies consistently show that T␤4 binds actin in an extended configuration, the N-terminal segment interacting with the barbed end of the actin monomer, while the C-terminal region binds subdomain 2 at the pointed end of the actin monomer (13,14). This view satisfactorily accounts for the fact that T␤4 inhibits Gactin association with both the barbed and the pointed ends of actin filaments.
␤-Thymosins have been found in all vertebrates and in echinoderms and mollusks (17,18). Recently, a cDNA clone encoding for a 41-amino acid ␤-thymosin has been identified in the calcareus sponge Scyon raphanus (19), indicating the ancient character of ␤-thymosin among metazoa. On the other hand, ␤-thymosins are absent in yeast, amoeba, Drosophila, and plants. In Acanthamoeba castellanii, a protein called actobindin consists of two ␤-thymosin repeats and has been identified as a G-actin-binding protein that may also sequester actin dimers (20 -24). A BLAST search on the complete genomic sequences of Drosophila melanogaster and Caenorhabditis elegans identified a triplicate ␤-thymosin sequence (19). Independently, this 3-␤-thymosin repeat protein, called ciboulot (Cib), has been identified in Drosophila as being involved in the control of brain development during metamorphosis and characterized as a G-actin-binding protein (25). Amazingly, although ciboulot shares sequence homology with T␤4, the Cib-actin complex participates in filament barbed end assembly like profilin-actin (25). The profilin-like property of Cib accounts for its function in actin-based motility and axonal growth. These observations raise an issue of structural and functional relevance about the evolution of the actin binding domain of ␤-thymosins. Are some biochemical properties of ␤-thymosins found unaltered in actobindin and Cib? What are the structural features at the origin of the functional difference between T␤4 and Cib? Is actobindin, the amoeba ␤-thymosin two-repeat protein, functionally similar to T␤4 or to Cib?
Here we compare Cib and actobindin regarding their binding to G-actin, their effects on nucleotide exchange on G-actin, and their effects on actin assembly at the two ends of filaments, with either MgATP-actin or CaATP-actin. We show that actobindin is functionally similar to Cib or profilin, and differs from ␤-thymosins, regarding the control of filament dynamics. On the other hand, actobindin and Cib both slow down nucleotide exchange on G-actin like T␤4. We conclude that proteins that consist of ␤-thymosin repeats like actobindin in Acanthamoeba or ciboulot in Drosophila and probably the C. elegans homolog must play a positive role in motile properties of living organisms. The pure G-actin sequestering function of regular ␤-thymosins may result from divergent evolution.

MATERIALS AND METHODS
Proteins-Actin was purified from rabbit muscle, isolated as CaATP-G-actin by Sephadex G 200 chromatography in G buffer (5 mM Tris-Cl, pH 7.8, 0.2 mM ATP, 1 mM dithiothreitol, 0.01% NaN 3 ). Actin was pyrenyl labeled on cysteine 374 (26) and NBD 1 -labeled on lysine 373 (27). Gelsolin was a kind gift from Dr. Yukio Doi (University of Kyoto, Kyoto, Japan) and actobindin was purified from Acanthamoeba castellanii (20). Thymosin ␤4 and profilin were prepared as described previously (15). The fusion protein GST-Cib cloned in the expression vector p-GEX2T* (Amersham Bioscience) was induced in Escherichia coli strain BL21 and purified (15). The Cib protein was then cleaved off the GST moiety with thrombin, dialyzed against 20 mM Tris-Cl, 1 mM dithiothreitol, pH 7.5, and stored at Ϫ80°C. The concentration of Cib was derived from amino acid analysis, from which a standardized bicinchoninic acid assay was developed.
Actin Polymerization Measurements-Steady-state measurements of F-actin were derived from fluorescence measurements of pyrenyl-labeled actin. Actin (10% pyrenyl-labeled) was polymerized under physiological ionic conditions (0.1 M KCl, 1 mM MgCl 2 ) in the absence or presence of gelsolin (at 1:300 molar ratio to actin) and in the presence or absence of Cib or actobindin at the indicated concentrations. The value of the equilibrium dissociation constant K C for the Cib-actin (or actobindin-actin) complex [CA] was derived from measurements of the amount of assembled actin at steady state (after 18-h incubation) (see the following equations), where [A U ] and [A] are the concentrations of unassembled actin and free G-actin at steady state and [C 0 ] is the total concentration of actobindin or Cib.
Initial rates of filament growth from the barbed and the pointed ends were measured spectrofluorometrically using either spectrin-actin seeds (28) or gelsolin-actin seeds, respectively. Gelsolin-actin seeds were prepared by mixing gelsolin and a 2.5 molar equivalent CaATP-G-actin in G buffer. The adequate amount of seeds was added at time 0 together with salt to a solution of MgATP-or CaATP-G-actin (10% pyrenyl-labeled) and Cib or actobindin. The initial rate of pyrene fluorescence increase was converted into molar amount of assembled Factin using a critical concentration plot derived from the same actin solution. Data were analyzed and simulated as described in Yarmola et al. (29) Depolymerization rates at the pointed ends were measured by 20-fold dilution of a solution of 2.5 M gelsolin-capped filaments (50% pyrenyllabeled) in F buffer containing known amounts of Cib or actobindin.
Nucleotide Exchange on G-actin-Kinetics of nucleotide exchange on monomeric actin were monitored using the change in fluorescence of ⑀ATP upon binding to G-actin (30). ATP-G-actin 1:1 complex was obtained by Dowex-1 treatment and diluted to 1 M in 5 mM Tris-Cl, pH 7.8, 0.1 mM CaCl 2 , 1 mM dithiothreitol, in the presence of Cib or actobindin. The increase in fluorescence ( exc ϭ 350 nm, em ϭ 410 nm) upon addition of 5 M ⑀ATP was monitored using a Safas flx spectrofluorometer. The fluorescence time courses were satisfactorily analyzed in terms of a monoexponential process. The observed first order rate constant displayed hyperbolic saturation behavior with the concentration of Cib or actobindin, consistent with a model in which Cib or actobindin shuttle from one molecule of G-actin to the other at a faster rate than nucleotide dissociates from G-actin or from Cib-actin or actobindin-actin complexes. The equilibrium dissociation constant for the Cib-actin or actobindin-actin complex was derived from the dependence of the apparent exchange rate constant on the concentration of Cib or actobindin, as follows, where 0 ] are the free and total concentrations of G-actin and Cib or actobindin and [CA] is the concentration of the complex; k 1 and k 2 are the rate constants for nucleotide dissociation from G-actin and from the CA complex, respectively; K C is the equilibrium dissociation constant of the CA complex. The values of k 1 and k 2 were determined experimentally in the absence and in the presence of saturating amounts of Cib. The value of K C was derived from the adjustment of the calculated curves k obs ([C 0 ]) to the data.
Equilibrium and Kinetic Measurements of the Interaction of Cib with G-actin-The change in fluorescence of NBD-labeled G-actin was used as a probe for the formation of the complexes of G-actin with T␤4, actobindin, or Cib. Static fluorescence measurements were carried out in a Spex Fluorolog 2 spectrofluorimeter in G buffer for CaATP-G-actin and in G buffer supplemented with 10 M MgCl 2 and 0.2 mM EGTA for MgATP-G-actin. Samples contained 1.5 M NBD-G-actin and different amounts of Cib. Excitation and emission wavelengths were 475 nm and 525 nm, respectively. The equilibrium dissociation constant for the complex was derived from the dependence of the fluorescence change on the total concentration of T␤4, actobindin, or Cib, analyzed using Equation 4.
The kinetics of interaction of NBD-G-actin with Cib were monitored by fluorescence ( exc ϭ 470 nm, slit 0.25 mm) using a stopped-flow apparatus (DX-18 MV, Applied Photophysics), with a 270-s noise filter. A solution of K 2 Cr 2 O 7 was placed on the emission beam to eliminate light scattered at the excitation wavelength. Four to six superimposable time courses were averaged for each concentration of Cib.

Direct Binding and Kinetics of Interaction of Cib and Actobindin with G-actin-
To evaluate the thermodynamic and rate parameters of the interaction of Cib and of actobindin with G-actin, we sought suitable spectroscopic probes. While binding of T␤4 to G-actin is accompanied by a 20% increase (11), and binding of profilin by a 25% decrease (30) in tryptophane fluorescence of actin, no change was observed with Cib. T␤4 also causes a large change in the fluorescence of AEDANS-labeled actin (14). Binding of Cib to AEDANS-G-actin caused a 7-nm blue shift in the excitation and emission spectra and a 5% increase in fluorescence ( exc ϭ 340 nm, em ϭ 480 nm), suggesting a less polar environment of the AEDANS fluorophore in the Cib-actin complex than in G-actin (an effect conspicuously opposite to the one observed (14) with T␤4), but the signal was too small to be useful in kinetic experiments. Like T␤4, neither actobindin nor Cib affected the fluorescence of pyrenyl-labeled G-actin, in agreement with previous reports (22,25). On the other hand, the fluorescence of actin in which lysine 373 was NBD-labeled was increased by 25% upon binding of Cib, actobindin, or T␤4 ( Fig. 1). A similar change has been reported when actobindin bound to actin in which cysteine 374 was labeled by IANBD (22). The equilibrium dissociation constants K C of the 1:1 complexes of G-actin with T␤4, Cib, and actobindin were derived from the analysis of the dependence of the fluorescence change on protein concentration according to Equation 4 (Fig. 1). Similar values of K C were obtained for all proteins, typically 2-4 M for CaATP-G-actin and 0.7-2 M for MgATP-G-actin (Table I).
Latrunculin A is a drug that interacts with G-actin with high affinity (5 M Ϫ1 ) and prevents polymerization (31). Latrunculin A has been shown to noncompetitively inhibit T␤4 binding to G-actin, decreasing the affinity of actin for T␤4 by approximately an order of magnitude in the ternary complex (32). The present data show that in the presence of saturating (20 M) concentrations of latrunculin A, the change in fluorescence of NBD-actin (1.5 M) upon binding Cib or actobindin or T␤4 was lower. The value of K C for binding T␤4 or Cib or actobindin to NBD-G-actin was increased about 2-fold by latrunculin A (Table I). The data are consistent with the formation of a ternary complex between G-actin, latrunculin A, and either Cib or actobindin or T␤4. No binding of Cib or actobindin to ADP-actin was detected using the change in NBD-fluorescence (data not shown). It is known that the fluorescence of NBD-actin, in contrast to pyrenyl-actin, is not affected by the bound nucleotide (33). In conclusion, like T␤4 and profilin, Cib has a high specificity for binding ATP-G-actin.
The change in NBD fluorescence was used to monitor the kinetics of Cib-actin complex formation. Cib bound to NBD-Gactin within a single exponential process. The apparent first order rate constant k obs increased practically linearly with Cib concentration in the range 0 -40 M Cib (Fig. 2). The apparent association rate constant (k ϩ ϭ 1.6 M Ϫ1 ⅐s Ϫ1 ) was derived from the slope, and the apparent dissociation rate constant (k Ϫ ϭ 14 s Ϫ1 ) was derived from the lower limit of k obs at low Cib concentration. The value of the ratio k Ϫ /k ϩ (3 M) was in good agreement with the equilibrium dissociation constant derived from the dependence of the fluorescence change on Cib concentration. The value of 14 s Ϫ1 for k Ϫ was then confirmed by a competition experiment in which Cib was displaced from the Cib-NBD-actin complex by a 10-fold excess of unlabeled actin.
To verify that the affinity of Cib or actobindin is not affected by NBD labeling, the following experiment was carried out. Increasing amounts of actobindin or Cib were added to two parallel samples of actin, containing either 10 or 50% NBDactin, polymerized at 1.6 M in the presence of 4 nM gelsolin. The linear decrease in the amount of F-actin was consistent with the same affinity of Cib or actobindin for G-actin independently of the proportion of labeled actin (supplementary data, Fig. 1).
Actobindin and Cib Slow Down Nucleotide Exchange on Gactin-Thymosin ␤4 is known to slow down nucleotide dissociation from G-actin, while profilin accelerates it. It has often been proposed that the effect of profilin on nucleotide exchange supports its effect on motility in vivo. Both actobindin and Cib, which enhance actin-based motility like profilin in an in vitro reconstituted motility assay (25), slow down nucleotide dissociation from G-actin, like T␤4 (Fig. 3). The exchange rate was decreased 12-and 9-fold by actobindin and by Cib, respectively, while it was decreased 20-fold by T␤4 under the same conditions. Analysis of the data using Equation 2 yielded values of 2.0 M and 1.7 M for the equilibrium disssociation constants for binding of actobindin and Cib, respectively, to CaATP-Gactin in G buffer.
This result indicates that Cib and actobindin share some of the binding features of T␤4. Their association with G-actin is linked to the slower dissociation of ATP. However, the effects of these two proteins on the dynamics of actin filaments are independent of their effects on nucleotide exchange on G-actin.
The Actobindin-Actin Complex Participates in Barbed End Assembly of MgATP-actin, Like Cib-Actin and Profilin-Actin-The effect of actobindin on the steady-state amount of 10% pyrenyl-labeled F-actin was measured under physiological conditions (0.1 M KCl, 1 mM MgCl 2 ) in the absence and in the presence of gelsolin (Fig. 4). When the barbed ends of filaments  were capped, actobindin caused depolymerization of F-actin. The amount of depolymerized actin increased linearly with the concentration of actobindin, consistent with sequestration of MgATP-G-actin by actobindin in a 1:1 complex, with an equilibrium dissociation constant K C of 6 M. This value is in good agreement with previous measurements (22)(23)(24), as well as with the value derived from nucleotide exchange kinetics (Fig.   3). A similar value (K C ϭ 2.5 M) had been found for the Cib-actin complex using the same assay (25). In contrast, actobindin did not depolymerize F-actin when barbed ends were free. The very slow decrease in F-actin versus actobindin concentration is consistent with the lowering of the steady-state concentration of G-actin upon addition of actobindin. As an example, at 20 M actobindin, the measured concentration of unassembled actin was 0.15 M (Fig. 4). This result implies that actin-actobindin complex, like profilin-actin complex, participates in barbed end assembly as well as actin itself, hence it lowers the energetic contribution of G-actin to monomer-polymer exchanges at steady state. In conclusion, actobindin-actin complex, like Cib-actin and profilin-actin, can stabilize the barbed ends, via monomer-polymer exchange reactions, as efficiently as G-actin. The validity of the above conclusions relies on the assumption that the affinity of actobindin for actin is not affected by pyrenyl labeling. This was first established by Lambooy and Korn (20). We have confirmed this conclusion both for Cib and actobindin by checking that the value of K C derived from assays shown in Fig. 4 was independent of the proportion of pyrenyl-actin.
In agreement with the above data, actobindin failed to com- pletely inhibit barbed end growth in a seeded polymerization assay (Fig. 4, inset). When all MgATP-G-actin was in complex with actobindin, the rate of barbed end growth was 70% of the rate observed with G-actin alone, indicating that the actobindin-actin complex associates with barbed ends with a rate constant only 30% lower than G-actin. This value is identical to the one found first for profilin-actin (30) and Cib-actin (25) association with barbed ends. Fig. 1 show that actobindin and Cib form a 1:1 complex with MgATP-G-actin that does not participate in pointed end filament assembly. Formation of this complex is expected to cause inhibition of filament pointed end elongation from G-actin subunits. Seeded growth assays using gelsolin-actin seeds were carried out at two different concentrations of G-actin. The decrease in the initial rate of elongation off gelsolin-actin seeds upon addition of actobindin should reflect saturation of G-actin by actobindin. Hence more actobindin should be needed to saturate a higher amount of G-actin. Typically, since the pointed end critical concentration is 0.5 M, the rate of growth is expected to reach zero when 1.5 M complex is formed at 2 M actin and when 4.5 M complex is formed at 5 M actin. 50% inhibition of elongation should therefore be reached at an actobindin concentration A 50% ϭ 3.75 M at 2 M actin and A 50% ϭ 6.34 M at 5 M actin (using the law of mass action with actin-actobindin complex ϭ 1/2(total actin Ϫ 0.5 M) and K C ϭ 5 M). In the case of Cib, 50% inhibition of filament growth should be achieved by addition of 2.25 M Cib at 2 M actin and 4.3 M Cib at 5 M actin. The experimental data differed from the expected behavior. Elongation of filaments off gelsolin-actin seeds was inhibited by actobindin and Cib, but the rate of growth at pointed ends displayed superimposable concentration dependencies (within experimental error) at 2 M and at 5 M G-actin and could not be accounted for by the calculated curves within a sequestration model (Fig. 5). Half-inhibition was observed at a lower concentration of actobindin or Cib than expected within the sequestration activity. The superimposable curves at two different actin concentrations suggest that actobindin or Cib bind to pointed ends with high affinity, preventing their growth. Analysis within pointed end capping with a K P value of 0.6 M and a sequestration constant K C of 5 M for actobindin and 2.5 M for Cib satisfactorily accounted for the data (Fig. 5). The fact that a higher affinity is observed for binding to the pointed end than for binding to monomeric actin suggests, in agreement with previous results (23), that actobindin and Cib may interact with two actins at the pointed end, via their T␤4 repeats. Similar data were obtained using CaATP-actin instead of MgATP-actin and a polymerization buffer that contained 0.1 M KCl.

Association of Actobindin or Cib with Filament Pointed Ends Blocks Pointed End Growth but Not Depolymerization-The steady-state measurements displayed in
To determine whether actobindin and Cib prevent pointed end disassembly, dilution-induced depolymerization of gelsolin-capped filaments was performed. The initial rate of depolymerization from the pointed ends was not affected by either actobindin or Cib up to 10 M (see supplemetary data Fig. 2). In conclusion, association of actobindin or Cib with pointed ends prevents growth but not depolymerization. The behavior of Cib and actobindin has some similarity here with DNase I, which also prevents pointed end growth but does not prevent depolymerization (34). The failure of Cib and actobindin to prevent pointed end depolymerization may be in relation with the poor binding of these proteins to ADP-actin, which is exposed at depolymerizing pointed ends, while ATP-actin is at the end of growing filaments. Therefore at the steady state of assembly of gelsolin-capped filaments, the relevant reaction is essentially the sequestration of monomeric actin by actobindin or Cib.
Actobindin and Cib Act as Purely G-actin Sequestering Proteins When CaATP Is Bound to Actin-Polymerization of CaATP-actin is quasi-reversible due to the slow hydrolysis of ATP on CaATP-F-actin. Consistently, the critical concentrations are practically identical (0.6 M) at the two ends. Profilin-CaATP-actin failed to polymerize at the barbed ends (35). Both actobindin and Cib displayed the same behavior as profilin regarding interaction with CaATP-actin and caused depolymerization of F-actin (i.e. sequestration of G-actin) at the barbed and at the pointed end (Fig. 6). Values of 8 Ϯ 1 M were derived from the data for the equilibrium dissociation constants of the complexes of CaATP-G-actin with either acto- bindin or Cib. In conclusion, the general features of the different interactions of profilin with CaATP-actin and MgATP-actin are displayed by actobindin and Cib. DISCUSSION We have shown that regarding its biochemical properties and biological function, the amoeba protein actobindin is a member of a new family of actin-binding proteins. These proteins consist of two or three ␤-thymosin repeats and share some biochemical properties with ␤-thymosins. For instance they bind ATP-G-actin specifically with thermodynamic and rate parameters very similar to T␤4 and slow down nucleotide dissociation from G-actin-like ␤-thymosins, indicating that their binding sites appreciably overlap the T␤4 binding site on Gactin. This conclusion is in agreement with results from EDCcross-linking experiments, which indicated that lysine 18 of T␤4 (13) and lysine 16 of actobindin (36) make contacts with the 4 N-terminal acidic residues of actin. The rate constant for association to G-actin is lower than the expected diffusionlimited rate constant, suggesting, as proposed for T␤4 (14), that the formation of a low affinity rapid equilibrium collision complex is followed by a reversible isomerization step leading to a tighter interaction; alternatively, the protein might be in rapid equilibrium between several conformational states, one of which only binds G-actin.
On the other hand, unlike ␤-thymosins, these proteins are not pure G-actin sequesterers but actually regulate the dynamics of filament assembly using the same mechanism as profilin, i.e. their complex with MgATP-G-actin participates in barbed end growth exclusively. This property is observed in kinetic assays of seeded filament growth. Its consequence at steady state is the lowering of the steady-state concentration of free G-actin, consistent with copolymerization of actin and actobindin-actin or Cib-actin complexes. The failure of actobindin to depolymerize actin at steady state when barbed ends are free or inhibit barbed end growth had been noticed previously but had been explained differently (24). Due to the ability of their complex with actin to participate in barbed end growth, proteins of the actobindin family play an active role, like profilin, in actin-based motility processes. We have shown that the Drosophila homolog, ciboulot, is required for axonal growth during central brain development in metamorphosis of the fly and that either Cib or actobindin can replace profilin in a reconstituted motility assay (25). The present work indicates that actobindin likewise must be required for some motile processes of the amoeba. Like for profilin, the participation of actobindin-actin or Cib-actin to barbed end growth is observed with MgATP-actin only, and a pure G-actin sequestering function is observed with CaATP-actin. These results point to a possible role of ATP hydrolysis associated with actin polymerization in this function. Actobindin consists of two imperfect ␤-thymosin repeats, while the Drosophila and C. elegans proteins harbor three ␤-thymosin repeats, suggesting that their functional difference with T␤4 might correlate with the fact that the ␤-thymosin motif is repeated. This view is not supported, however, by the following observations. Similar G-actin binding motifs showing similarity with T␤4, called "verprolin homology region" or WH2 domain, are found in WASp family proteins, and in the ActA protein of Listeria, where they have been demonstrated to share the same functional homology with profilin (Refs. 37-39 and see Ref. 40 for a recent review). In these proteins, the G-actin binding module is generally not repeated, except in the case of N-WASp (the neural form of WASp), which contains a tandem of two verprolin homology regions. Hence it seems likely that a subtle difference in sequence in the G-actin binding motif, rather than the repeat of the motif, is responsible for the functional difference between ␤-thymosins and these proteins. The repeated sequences would then simply help to enhance the affinity of these proteins for G-actin. Further studies of the biochemical properties of the isolated repeats are required to challenge this view. A mutagenetic analysis of these proteins should also help to elucidate the structural basis for the change in function of the ␤-thymosin motif. In this respect, previous studies have shown that changing the sequence 17 LKKTET 22 in the actin binding motif of T␤4 into LKETET caused T␤4-induced actin aggregation in low ionic strength buffer (9). In addition to this segment, T␤4 interacts with G-actin via helix 1 (residues 5-16) and helix 2 (residues 31-39) (42). According to Safer and co-workers (13,14), C-terminal helix 2 interacts with subdomain 2 at the pointed end of G-actin. This contact may prevent association of the T␤4-actin complex to the barbed end of an actin filament. In binding to G-actin, Cib and actobindin do not sterically interfere with subdomain 2 association to a barbed end.
Finally, ␤-thymosin repeat proteins display the original property, not shown by either profilin or ␤-thymosins, to prevent pointed end growth by a capping effect. This result has no physiological significance since in vivo pointed ends only depolymerize, but it is interesting from a structural point of view. The affinity for pointed ends is about 5-fold higher than for monomeric actin. This result is surprising and paradoxical. Actually, the view that actobindin-actin and Cib-actin participate in barbed end growth intuitively suggests that these proteins are transiently bound to the terminal subunit at the barbed end of the filament, but cannot be bound to the pointed end. To accommodate the unexpected pointed end capping we propose that two ␤-thymosin repeats of a single protein may interact with two terminal subunits at the pointed end. The structure of actobindin or Cib bound to the pointed end may be similar to the structure of the reported high affinity complex of actobindin with covalently cross-linked actin dimers obtained by reacting F-actin with para-phenylene-bis-maleimide (43). The covalent bond connects lysine 191 to cysteine 374 of a laterally adjacent subunit along the genetic helix (41), i.e. may reconstitute the pointed end structure of a filament. Resolution of the three-dimensional structure of the complex of actobindin or Cib with G-actin is required to challenge this hypothesis.