The C-terminal Tail of UNC-60B (Actin Depolymerizing Factor/Cofilin) Is Critical for Maintaining Its Stable Association with F-actin and Is Implicated in the Second Actin-binding Site* h S

Actin depolymerizing factor (ADF)/cofilin changes the twist of actin filaments by binding two longitudinally associated actin subunits. In the absence of an atomic model of the ADF/cofilin-F-actin complex, we have identified residues in ADF/cofilin that are essential for filament binding. Here, we have characterized the C-termi-nal tail of UNC-60B (a nematode ADF/cofilin isoform) as a novel determinant for its association with F-actin. Removal of the C-terminal isoleucine (Ile 152 ) by carboxypeptidase A or truncation by mutagenesis elimi-nated F-actin binding activity but strongly enhanced actin depolymerizing activity. Replacement of Ile 152 by Ala had a similar but less marked effect; F-actin binding was weakened and depolymerizing activity slightly enhanced. Sali and Blundel (39). Evaluation of the initial model, superimposed with the known struc- tures, suggested changes in the alignment in the region of the insertion at UNC-60B Lys 58 –Lys 65 . Multiple cycles of realignment and rebuilding of the models were completed. Side chain conformations, along the length of the UNC-60B sequence, were optimized manually to satisfy individual residue structural functions ( i.e. van der Waals packing, ionic interactions, H bonds between side chain and main chain, etc.) as seen in the known structures. Additional refinement and minimization of the model are underway. UNC-60B Actin-binding Assays— Copelleting assays of UNC-60B with rabbit muscle F-actin were performed as described previously (40) in a buffer containing 0.1 M KCl, 2 m M MgCl 2 , 20 m M HEPES-NaOH, 1 m M dithi- othreitol, pH 7.5. All protein solutions except F-actin were clarified by ultracentrifugation prior to the pelleting assays. Ultracentrifugation was performed with a Beckmen Airfuge at 28 pounds/square inch for 20 min. Assay for nucleation of actin polymerization was performed as described previously (30) except that rabbit muscle actin was used as seeds in this study. Nondenaturing Polyacrylamide Gel Electrophoresis— Nondenaturing polyacrylamide gel electrophoresis was performed as described by Safer (43). CaATP-G-actin and UNC-60B were incubated in G-buffer (2 m M Tris-HCl, 0.2 m M CaCl 2 , 0.2 m M dithiothreitol, 0.2 m M ATP, pH 7.5) for 30 min at room temperature. The samples were then supplemented with 0.25 volume of a loading buffer (50% glycerol, 0.05% bromphenol blue) and electrophoresed using a Bicine/triethanolamine buffer sys-tem. The proteins were visualized by staining with Coomassie Brilliant Blue R-250 (Sigma). ; work station (Mountain View, CA). The goal was to produce a model that fit inside the envelope of the electron microscopy reconstruction well and also taking into account available genetic and biochemical data concerning the cofilin residues involved in actin contacts (27, 28, 30, 31). Ribbon diagrams were generated using Ribbons 2.65 (49), saved as Inventor format files, and displayed in IRIS Explorer.

Actin depolymerizing factor (ADF) 1 /cofilins are a family of actin-regulatory proteins ubiquitous among eukaryotes that are essential for rapid turnover of the actin cytoskeleton in vivo (1)(2)(3). ADF/cofilin enhances the turnover of actin filaments by both increasing the rate of depolymerization from their pointed ends (4,5) and by severing them thereby increasing the number of ends (5)(6)(7)(8)(9)(10)(11)(12)(13)(14). These two activities can be uncoupled by point mutations (15,16). Binding of ADF/cofilin to F-actin changes the twist of the filaments (17) and leads to destabilization of the lateral contacts of actin monomers within the filaments (18). This ADF/cofilin-mediated structural change in F-actin is important for cooperative binding of ADF/cofilin to F-actin (17,19), resulting in the coexistence of bare and ADF/ cofilin-decorated actin filaments, as observed in the lamellipodia of cultured cells (20).
The structural fold of the ADF/cofilin family proteins (21)(22)(23) is similar to that of the individual segments of the gelsolin family (24). Although the structure of gelsolin segment-1-actin complex has been solved (25), predicted models of ADF/cofilinactin complex that were based on the topology of this complex (21,26) differed significantly and can not be used to predict F-actin binding. When one of these predicted structures was extended to develop models of ADF/cofilin-F-actin complex (17), the C terminus of ADF/cofilin was placed away from the actin surface. However, alanine-scanning mutagenesis of yeast cofilin (27) revealed that helix ␣4 near the C terminus (Fig. 1b) is a part of the actin-binding surface that confers filament binding.
Mutagenesis of yeast cofilin (27) revealed two actin-binding surfaces as follows: one that is essential for G-actin binding and a second that is required for F-actin. The essential G-actinbinding surface includes the N terminus, a portion of helix ␣3, and the turn connecting strand ␤6 and helix ␣4 (27) (Fig. 1b). Helix ␣3 includes two highly conserved basic residues that have been implicated in monomer binding in several ADF/ cofilin proteins by chemical cross-linking (28), mutagenesis (16,29,30), and peptide competition (28,31). The N terminus is also involved in actin binding (32), and this region includes the phosphorylation site Ser-3 (Ser-6 in plant ADF) that is responsible for phosphorylation-dependent inactivation of actin binding by ADF/cofilin (33)(34)(35)(36). These regions of the molecule are close together in the three-dimensional structures and probably form a binding surface that interacts with subdomains 1 and 3 of G-actin (21)(22)(23).
There is little detailed information regarding the site on ADF/cofilin that is required for F-actin binding. Two regions have been implicated in filament binding as follows: (i) a cluster of charged residues (15,27) in a loop connecting ␤2 and ␤3 in destrin and at a similar position at the beginning of ␤5 in yeast cofilin, and (ii) a group of charged residues in helix ␣4 in yeast cofilin (27). In addition, mutations of conserved tyrosine residues (Tyr 67 and Tyr 70 in maize ADF3) in the apolar core of the molecule results in uncoupling of monomer and filament binding activities (37).
We previously identified a novel function for the C-terminal tail of ADF/cofilin (30). The tail is predicted to lie outside helix ␣4 (Fig. 1c) and is not included in the crystal structures of either yeast cofilin (23) or Acanthamoeba actophorin (38), suggesting structural flexibility in this region. A mutant form of UNC-60B (a Caenorhabditis elegans ADF/cofilin isoform) that lacks three C-terminal amino acids (Fig. 1c) shows loss of both F-actin binding and severing activity but still depolymerizes F-actin, possibly through its monomer sequestering activity (30). Importantly, a C. elegans strain that expresses this mutant is defective in proper assembly of actin into myofibrils (30), indicating that the C-terminal tail of UNC-60B is functionally significant in vivo. Therefore, in this study, we characterized the role of the C-terminal tail of UNC-60B in its actin regulating activity.

EXPERIMENTAL PROCEDURES
Homology Modeling of the Structure of UNC-60B-Homology modeling of UNC-60B was based on the known structures of Saccharomyces cerevisiae cofilin (Protein Data Bank code 1cfy) (23), Acanthamoeba castellanii actophorin (Protein Data Bank code 1ahq) (22), and Sus scrofa destrin (Protein Data Bank code 1ak7) (21). Coordinate files were obtained from Protein Data Bank. The initial alignment, generated by pattern-induced (local) multiple alignment, showed 10% identical and 57% conserved residues between UNC-60B and cofilin, actophorin and destrin (Fig. 1a). A model was built based on this initial alignment using the program Modeller by Sali and Blundel (39).
Evaluation of the initial model, superimposed with the known structures, suggested changes in the alignment in the region of the insertion at UNC-60B Lys 58 -Lys 65 . Multiple cycles of realignment and rebuilding of the models were completed. Side chain conformations, along the length of the UNC-60B sequence, were optimized manually to satisfy individual residue structural functions (i.e. van der Waals packing, ionic interactions, H bonds between side chain and main chain, etc.) as seen in the known structures. Additional refinement and minimization of the model are underway.
Proteins-Recombinant wild-type UNC-60B was purified as described (40). Rabbit muscle actin was purified as described (41), and G-actin was further purified by gel filtration with Sephacryl S-300 or purchased from Cytoskeleton Inc. (Denver, CO). We obtained indistinguishable results using either actin preparation. C. elegans actin was purified from wild-type N2 strain (obtained from the Caenorhabditis Genetics Center, St Paul, MN) as described (42). Bovine pancreas carboxypeptidase A (CPA) (EC 3.4.17.1) was purchased from Roche Molecular Biochemicals.
Electrospray Ionization Mass Spectrometry-The protein solution was acidified to pH ϳ2 with 10% trifluoroacetic acid, and the protein was chromatographed on a Jupiter-C4 column (1 ϫ 150 mm) equilibrated in 0.1% aqueous trifluoroacetic acid and eluted using a linear gradient of acetonitrile in 0.08% aqueous trifluoroacetic acid. The column eluate was monitored at 210 nm, and the fractions containing the protein (eluting as a major peak at ϳ30% acetonitrile) were collected for further analysis. The fractions were mixed 1:1 with isopropyl alcohol/ water/acetic acid (50:50:1) and directly infused at a flow rate of 15 l/min into a MicroIon source of model API3000 triple quadrupole mass spectrometer. Following spectra deconvolution (using the BioSpec BioReconstruct routine), the mass of the protein was determined.
N-terminal Sequence Analysis-The proteins were subjected to automated Edman degradation in a Procise-cLC sequencer (PE Biosystems) using a standard manufacturer's protocol.
Mutant UNC-60B Proteins-Mutations were introduced in synthetic oligonucleotides that complement the 3Ј-end of the coding sequence of the cDNA for UNC-60B. The cDNA fragments that carry mutations were amplified by polymerase chain reaction using these mutant primers and a common forward primer (5Ј-GATCCCATGGCTTCCGGAGT-CAAAGTTG). The amplified DNA fragments were digested by NcoI and BamHI at the sites introduced by the primers and cloned into pET-3d (Novagen, Madison, WI). The sequences of the inserts were verified by DNA sequencing not to contain any polymerase chain reaction-induced errors. The mutant primers used are 5Ј-CTAGGGATCCTTATCTTTG-GTTGGACATCAGGTC for ⌬I, 5Ј-CTAGGGATCCTTATTGGTTGGAC-ATCAGGTCGC for ⌬RI, 5Ј-CTAGGGATCCTTAGATTCTTTAGTTGG-ACATC for ⌬QRI, 5Ј-CTAGGGATCCTTAGGCTCTTTGGTTGGACATC for I152A, and 5Ј-CTAGGGATCCTTAGATTGCTTGGTTGGACATC for R151A. The mutant UNC-60B proteins were expressed and purified as described previously for wild-type UNC-60B (40).
Actin-binding Assays-Copelleting assays of UNC-60B with rabbit muscle F-actin were performed as described previously (40) in a buffer containing 0.1 M KCl, 2 mM MgCl 2 , 20 mM HEPES-NaOH, 1 mM dithiothreitol, pH 7.5. All protein solutions except F-actin were clarified by ultracentrifugation prior to the pelleting assays. Ultracentrifugation was performed with a Beckmen Airfuge at 28 pounds/square inch for 20 min. Assay for nucleation of actin polymerization was performed as described previously (30) except that rabbit muscle actin was used as seeds in this study.
Cryoelectron Microscopy and Modeling of the Structure of UNC-60B-F-actin Complex-Decorated filaments were prepared in one of two ways. In most cases, pre-formed actin filaments (3 M) were incubated at room temperature for ϳ2 h with 10 M UNC-60B in 20 mM MOPS, pH 7, 0.1 M KCl, 2 mM MgCl 2 , 1 mM dithiothreitol, 0.5 mM EGTA. Although ATP was present in the polymerized F-actin solution, no additional ATP was added upon dilution into the UNC-60B-containing buffer, resulting in a final concentration of Ͻ0.04 mM ATP. For partial decoration of filaments, 10 M F-actin was decorated on ice overnight in the presence of 10 M UNC-60B using the same buffer conditions and diluted to a final actin concentration of 2.5 M just prior to plunging. Filament decoration was confirmed by pelleting assays in a Beckman Airfuge at 20 pounds/square inch for 15 min followed by SDS-polyacrylamide gel electrophoresis. Decorated filaments were rapidly frozen on holey carbon films in ethane slush cooled with liquid nitrogen and then imaged using a JEOL 1200EX transmission electron cryomicroscope at the National Center for Macromolecular Imaging at Baylor College of Medicine. Micrographs were recorded at an accelerating voltage of 100 kV and a nominal magnification of ϫ 30,000.
Suitable micrographs were digitized using a Zeiss SCAI film scanner at a final resolution of 4.7 Å per pixel. Tobacco mosaic virus particles (0.05 mg/ml; kindly provided by Dr. Ruben Diaz, Florida State University) were included in the samples to provide an internal magnification standard. The defocus of each micrograph was determined by incoherent averaging of regions of protein embedded in ice (44). The micrographs used in this study were recorded at Ϫ2.1and Ϫ2.6-m defocus. Filament images were analyzed using the helical image processing package PHOELIX (45,46). Mean crossover lengths were measured from positions of the first layer line in computed diffraction patterns of computationally straightened filaments.
For image reconstructions, layer line data were collected to a radial resolution of 0.032 Å Ϫ1 (within the first node of the contrast transfer function) and separated into near and far side data sets. Filaments that best conformed to the 20:9 selection rule and that showed good symmetry between the near and far sides of the Fourier transform were selected for further processing. A total of 22 data sets were averaged, and a three-dimensional structure was calculated by Fourier-Bessel inversion (47) to a resolution limit of 0.032 Å Ϫ1 radially using the following layer lines (l, n): 0, 0; 2, 2; 4, 4; 5, Ϫ5; 7, Ϫ3; 9, -1; 11, 1; 13, 3; 15, 5. The mean phase residuals during the last round of alignment of the 22 data sets were 31.8°and 47.6°for the correct and incorrect polarity, respectively. The resulting reconstruction was aligned to a previously calculated F-actin reconstruction (17) in Fourier space to aid in model building.
Pseudo-atomic models of UNC-60B-F-actin were derived using the actin subunits from the Lorenz filament model (twisted to match the UNC-60B-induced actin structure) and the homology-based model of UNC-60B. The UNC-60B model structure was positioned interactively using the programs O (48) and Iris Explorer (Numerical Algorithms Group, Downers Grove, IL) running on a Silicon Graphics Indigo R4000 work station (Mountain View, CA). The goal was to produce a model that fit inside the envelope of the electron microscopy reconstruction well and also taking into account available genetic and biochemical data concerning the cofilin residues involved in actin contacts (27,28,30,31). Ribbon diagrams were generated using Ribbons 2.65 (49), saved as Inventor format files, and displayed in IRIS Explorer.

RESULTS
Homology Modeling of the Structure of UNC-60B-The structure of UNC-60B was modeled based on the known structures of three ADF/cofilin proteins. The model presented in Fig. 1c incorporated the tertiary structures of the known ADF/cofilin proteins, a three-layer ␣-␤-␣ sandwich, with an internal fivestranded ␤-sheet. Significant differences in the model resulted from insertions in the UNC-60B sequence between helix ␣2 and strand ␤3 (UNC-60B Lys 58 -Lys 65 inserted between Pro 58 and Asn 60 of cofilin) and between strands ␤3 and ␤4 (UNC-60B Arg 80 -Thr 86 inserted between Gly 75 and Gly 78 of cofilin). There is also an addition of 4 residues at the C terminus (UNC-60B Asn 149 -Ile 152 ), the structure of which is modeled after the destrin C-terminal structure (residues Leu 156 -Ala 159 ).
Ile 152 of UNC-60B Is Required for Maintaining Stable Association with F-actin-The C-terminal isoleucine at position 152 of UNC-60B was removed by carboxypeptidase, and its effect on the activity of UNC-60B was examined. We found that CPA removed only Ile 152 because Arg 151 is not a substrate of CPA. Removal of Ile 152 was confirmed by electrospray mass spectrometry. Molecular mass of control UNC-60B was determined as 16,919 atomic mass units that corresponded to residues 2-152 of UNC-60B (the predicted mass, 16,915 atomic mass units). However, the spectrum of CPA-treated UNC-60B yielded a major peak of 16,806 atomic mass units and a minor one of 16,921 atomic mass units. These correspond to residues 2-151 (the predicted mass, 16,802 atomic mass units) and 2-152, respectively. The mass of the major peak was consistent with that of UNC-60B-⌬I (the experimentally determined mass, 16,803 atomic mass units), a mutant lacking Ile 152 . Area calculations of the peaks from the mass spectra suggested that ϳ17% of the total UNC-60B was not processed with CPA assuming that both forms were ionized to the same extent. Nterminal sequencing of CPA-treated UNC-60B yielded the identical sequence to control UNC-60B starting from Ala 2 , indicating that there was no detectable amino-or endopeptidase activity in the CPA preparation. In addition, there was no evidence of major alteration in secondary structure as examined by circular dichroism (plots not shown), providing evidence that removal of Ile 152 does not significantly disrupt the structure of the C-terminal helix ␣4.
Interaction of CPA-treated UNC-60B with F-actin was surprisingly different from that of control UNC-60B. Whereas control UNC-60B cosedimented with rabbit muscle F-actin without significantly depolymerizing it (Fig. 2a, lane 2, compare with lane 1) (40,42), CPA-treated UNC-60B increased the amount of unpolymerized actin in the supernatant and did not cosediment with F-actin (Fig. 2a, lane 3). Thus, CPA-treated UNC-60B depolymerizes F-actin and prevents re-polymerization. CPA itself did not cause actin depolymerization (Fig. 2a,  lane 4). Furthermore, preincubation of CPA with its inhibitor, 1,10-phenanthroline, abolished the ability of CPA to change the activity of UNC-60B (Fig. 2a, lane 5). These results indicate that removal of the C-terminal Ile from UNC-60B converts it from a filament-binding protein to a depolymerizing factor. Nearly complete depolymerization of F-actin was achieved at molar ratios at greater than 1:1 of CPA-treated UNC-60B to actin monomer (Fig. 2b). CPA-treated UNC-60B also showed enhanced actin depolymerizing activity using C. elegans actin as compared with control UNC-60B (data not shown). Since rabbit muscle actin consists of a single isoform that has been very well characterized, we used rabbit muscle actin in the biochemical experiments to correlate better the biochemical data with the structural interpretation as described below.
In contrast to the effects on F-actin, CPA treatment of UNC-60B did not affect its interaction with G-actin. Both control UNC-60B and CPA-treated UNC-60B were resolved as a single band on nondenaturing polyacrylamide gels (Fig. 3, lanes 2-5), whereas actin gave a doublet and smeared bands (Fig. 3, lane  1). When actin and either control or CPA-treated UNC-60B were mixed, a third band appeared (Fig. 3, lanes 6 -9, arrows). The intensity of this band increased as the concentration of UNC-60B was increased (Fig. 3, lanes 6 -9). This increase was accompanied by a decrease in the intensities of both actin and UNC-60B bands. These results indicate that the third band FIG. 1. Homology modeling of the structure of UNC-60B. a, sequence alignment of UNC-60B, yeast cofilin (Y-cof), actophorin (Actoph), and porcine destrin. Locations of helices ␣1-4 are indicated by bold and underlined letters. Colored residues correspond to the ones in b and c and are explained below. b, crystal structure of yeast cofilin (Protein Data Bank code 1cfy). Essential residues for actin-binding are shown in red, and residues that confer F-actin binding are shown in yellow (27). Note that residues 1-5 and 140 -143 are not visible in the crystal structure. Therefore, Val 6 is colored red instead of the essential residues 1-5. c, a predicted structure of UNC-60B. Missense mutations that have been reported by Ono et al. (30) are shown in blue. Residues in the C-terminal tail (C-tail) that are truncated in the unc-60(r398) mutation are shown in green. The coordinates of this model (unc60b_model) are available as Supplemental Material.
represents the actin-UNC-60B complex. Results were similar with control or CPA-treated UNC-60B (compare Fig. 3, upper and lower panels), suggesting that CPA treatment of UNC-60B did not affect its G-actin binding activity.
The role of Ile 152 of UNC-60B was further investigated by using recombinant mutant UNC-60B proteins. Mutant UNC-60B-⌬I, which lacks Ile 152 , showed substantial actin-depolymerizing activity and cosedimented poorly with F-actin (Fig.  4b). This is in marked contrast to wild-type which binds F-actin but does not significantly depolymerize it (Fig. 4a). These results are consistent with those using CPA treatment, except that CPA-treated UNC-60B exhibited stronger actin-depolymerizing activity than UNC-60B-⌬I. UNC-60B-⌬I was indistinguishable from CPA-treated UNC-60B in its interaction with G-actin, molecular mass, and N-terminal sequence (data not shown). Treatment of UNC-60B-⌬I with CPA did not change its behavior (data not shown), suggesting that the quantitative difference may reflect slightly different folding of the bacterially expressed ⌬I mutant compared with CPAtreated UNC-60B. Incomplete processing of UNC-60B with CPA (17% of UNC-60B remained intact) is unlikely to cause this difference because a mixture of control UNC-60B and UNC-60B-⌬I at the equivalent ratio showed no difference in actin depolymerizing activity compared with UNC-60B-⌬I alone (data not shown).
To investigate further the significance of the isoleucine residue at position 152, we produced the mutant UNC-60B-I152A in which Ile 152 was replaced with Ala. UNC-60B-I152A showed weaker binding to F-actin than wild-type UNC-60B but had increased depolymerizing activity compared with the control (compare Fig. 4, a and c). In the presence of 30 M UNC-60B-I152A, 6.6 Ϯ 0.24 M actin was precipitated, and with 30 M wild-type, 8.0 Ϯ 0.07 M actin was precipitated (Fig. 4, c and a,  respectively). UNC-60B-I152A was indistinguishable from wild type in its interactions with G-actin (data not shown). These results suggest that Ala can partially substitute for Ile at position 152 but is not as efficient as Ile in maintaining stable association with F-actin. Taken together, these data demonstrate the importance of Ile 152 for UNC-60B to make a stable complex with F-actin and inhibit filament disassembly.
Arg 151 of UNC-60B Is Required for F-actin Binding and Efficient Actin Depolymerization-We investigated the roles of adjacent C-terminal amino acids of UNC-60B using bacterially expressed mutant UNC-60B proteins with additional truncations at the C terminus. Mutant UNC-60B-⌬RI cosedimented poorly with F-actin but showed considerable depolymerizing activity. This mutant was not as effective as UNC-60B-⌬I (compare Fig. 4, b and d), suggesting that Arg 151 is needed for efficient depolymerizing activity. Additional truncation of Gln-150 (UNC-60B-⌬QRI, formerly designated as the r398 mutant (30)) did not cause any detectable difference in the depolymer- izing activity compared with UNC-60B-⌬RI (Fig. 4e). UNC-60B-⌬RI and UNC-60B-⌬QRI were indistinguishable from wild type in their interaction with G-actin (data not shown).
We further characterized the role of Arg 151 using UNC-60B-R151A in which Arg 151 was replaced with Ala. UNC-60B-R151A cosedimented poorly with F-actin and showed significant actin-depolymerizing activity despite the presence of Ile 152 (Fig. 4f). This suggests that Arg 151 is required for F-actin binding by UNC-60B, whereas Ile 152 has an accessory function. Interestingly, there were quantitative differences in the depolymerizing activities of UNC-60B-R151A and the truncated mutants. At molar ratios Ͻ1:1, UNC-60B-R151A depolymerized actin to lesser extents than UNC-60B-⌬I, -⌬RI, or -⌬QRI. However, at higher molar ratios (Ͼ2:1), the extent of depolymerization by UNC-60B-R151A was greater than that of UNC-60B-⌬RI, or -⌬QRI, but slightly weaker than UNC-60B-⌬I. Thus, it is possible that these mutants affect actin depolymerization in different ways.
Both Arg 151 and Ile 152 of UNC-60B Are Required for F-actin Severing-We explored the role of the C-terminal tail of UNC-60B in severing actin filaments, and we found that both Arg 151 and Ile 152 are required for this activity. Wild-type UNC-60B increased the ability of F-actin seeds to nucleate actin polymerization by 2.5-fold (Fig. 5), which is strong evidence that UNC-60B increased the number of ends by severing filaments. However, this activity was greatly reduced using UNC-60B-⌬I or UNC-60B-I152A (Fig. 5). The severing activity of these mutants correlates with their weaker F-actin binding activity (Fig.  4c). CPA-treated UNC-60B has no effect on the nucleation rate (Fig. 5), which again suggests quantitative differences between CPA-treated UNC-60B and recombinant UNC-60B-⌬I. Thus, Ile 152 of UNC-60B is required for both filament binding and efficient severing.
Arg 151 is also essential for the severing activity of UNC-60B. Either truncation (UNC-60B-⌬RI and -⌬QRI) or replacement with Ala (UNC-60B-R151A) abolished this activity (Fig. 5). These results suggest that both Arg 151 and Ile 152 are essential for actin severing activity by UNC-60B.
Structure of UNC-60B-F-actin Complex-We first confirmed that UNC-60B altered the F-actin twist by computing the Fourier transforms of decorated filaments and measuring the height of the first layer line that arises from the twist of the two-start actin helix. UNC-60B produced a mean filament twist (expressed as rotation per subunit) of 161.6°for C. elegans F-actin. Filaments prepared using rabbit muscle actin were twisted to a mean twist of 161.7°per subunit. Thus, as was found with human cofilin and human ADF (17), UNC-60B reduces the filament crossover length to about 270 Å irrespective of the source of the actin. Electron cryomicroscopy of filaments decorated with sub-saturating amounts of UNC-60B confirmed that the C. elegans protein shares the ability of the human protein to bind filaments cooperatively.
A three-dimensional reconstruction was calculated from electron micrographs of frozen-hydrated UNC-60B-F-actin filaments essentially as described previously (17) (Fig. 6a). This reconstruction was used to provide a molecular envelope into which the UNC-60B and actin subunit could be positioned (Fig.  6b). In addition to using the reconstruction as a constraint, available biochemical and genetic data implicating the long helix ␣3, the N terminus, and the C-terminal residues (27,28,30,31) were used to help orient the UNC-60B molecule on the filament.
The resulting orientation is similar to an earlier model of human ADF in terms of the long helix involvement and its "polarity" relative to the filament. Residues Val 104 -Ser 113 of the long helix are still in proximity to actin, but the remaining 7 residues (Val 114 -Ser 120 ) are rotated away from the filament. Interestingly, the portion of helix ␣3 that falls near actin includes Ser 112 and Ser 113 ; replacement of either of these residues by apolar groups alters F-actin binding (30). Significantly, both the N and C termini are positioned near the actin interface in the new orientation, consistent with a number of biochemical and genetic studies. DISCUSSION In this study, we show that alterations of UNC-60B at the C terminus inhibit F-actin binding and severing without changing G-actin binding. We found that F-actin binding of UNC-60B ( Fig. 4) was correlated with the extent of filament severing (Fig. 5). This suggests that severing occurs as a consequence of physical distortions in the helix that arise as a result of the change in twist induced when these proteins bind to the side of filaments (Fig. 6). This is consistent with previously suggested models (17,18).
Cryoelectron microscopy and structural modeling of UNC-60B-F-actin complex suggest that the C terminus is involved in the second actin-binding site. The alterations at the C terminus are not likely to cause major disruption of the structural fold, because we were unable to detect any changes in far UV circular dichroism or in their interactions with G-actin. However, we cannot exclude the possibility of minor conformational changes caused by these alterations. The structures of the C-terminal regions of yeast cofilin (23) and actophorin (22) are disordered in the crystals, which suggests that the C terminus is not well ordered although it may be stabilized upon binding to F-actin. The fact that Ile 152 is a substrate of CPA indicates that this residue is exposed on the surface of the molecule and readily accessible to its active site.
Our reconstructions and structural modeling suggest that both the N and C termini are positioned near the actin interface, but the predicted interactions they make with the filament are very different. The N terminus is positioned near the subdomain 1 of the upper actin subunit, whereas the C terminus falls near subdomain 2 of the lower actin. Since cofilin and profilin compete for binding to monomeric actin (8,50), and profilin binds on the "bottom" of actin between subdomains 1 and 3, this implies that the N terminus, but not the C terminus, is involved in G-actin interactions. The C terminus, on the other hand, primarily influences F-actin interactions. Mutational analysis of human cofilin suggested that F-actin binding occurs initially through this site with the lower actin, followed by interaction with the upper actin subunit via the G-actin site (15). Interestingly, subdomain 2 is generally accepted as a molecular "sensor" for the nucleotide state (51,52). The finding that F-actin binding by actophorin increases the rate of phosphate release (12) raises the possibility that the C terminus is directly involved in the nucleotide sensitivity of cofilin for actin. It will be interesting to find out whether the C terminus is involved in the activation of phosphate release. If the C-terminal tail of UNC-60B is flexible, preliminary molecular graphic simulations suggested that it could extend into the ATP-binding pocket of actin and interact with the nucleotide-binding site.
By using these mutants, we have also shown enhanced depolymerizing activity. The mechanism by which this occurs is uncertain, but our preliminary experiments using gelsolincapped filaments show that these mutants activate subunit dissociation from the pointed ends of filaments. The extent of depolymerization differed between different mutants even in the absence of any severing. This supports recent reports that depolymerization and severing activities of ADF/cofilin can be uncoupled (15,16). Detailed kinetic analysis of depolymerization by our mutant versions of UNC-60B should provide further insight into this mechanism.
Our data suggest that the C terminus of ADF/cofilin is the structural determinant of the quantitative differences in Factin binding and actin-depolymerizing activities that have been observed for different ADF/cofilin family members (5,7,19,40,53). The C-terminal region is highly divergent in both sequence and length (54). For example, human ADF is a much more potent depolymerizing agent than human cofilin (55), and the largest region of sequence divergence is located in the C-terminal 20% of the protein. Moreover, two C. elegans isoforms, UNC-60A and UNC-60B, show far greater sequence divergence and functional difference than the two human isoforms, with the biggest differences in sequence seen near the C terminus (40,56). It will be interesting to examine the structural basis of these differences and the extent to which various C-terminal sequences are involved in functional diversity.