Demonstration of at Least Two Different Actin-binding Sites in Villin, a Calcium-regulated Modulator of F-actin Organization*

Villin, one of the calcium regulated modulator pro- teins of F-actin organization, restricts F-actin to short filaments in the presence of calcium and bundles F- actin in the absence of calcium. Limited in vitro proteolysis of villin generates, in addition to a large core fragment (apparent M, = 90,000) previously described, a small headpiece (Mr = 8,500). The finding that the F- actin nucleation and severing activity of villin, but not its bundling activity, is retained by the core suggested that the headpiece may be directly involved in bun- dling. Headpiece has now been purified and characterized. It shows strong F-actin binding both in the pres- ence and absence of calcium, leading to a final stoichiometry of 1 headpiece to 1 F-actin monomer. Headpiece also inhibits villin-induced F-actin bundling. Thus villin expresses at least two distinct actin-binding sites lo-calized on separate functional domains. Protein se- quence analysis documents that the core comprises the NH2-terminal portion of intact villin, whereas the head- piece covers the COOH-terminal 76 amino acids. We provide the amino acid sequence of the headpiece, which is currently the smallest F-actin binding peptide. control, regulated at least in part by a group of modulator proteins, called proteins (1-4). These proteins reveal strong Caz+

trations of villin lead to the formation of tightly packed Factin bundles similar in morphology to microvillus cytoskeletons and derivatives of these structures still containing villin (1,5,12,13). The transition between these two opposed functions seems to be governed by the saturation of the Ca2+ binding site, which is characterized by a micromolar dissociation constant (5). We have recently reported that mild in vitro proteolysis separates the bundling function from the severing function. Treatment of villin (M, = 95,000) with Staphylococcus V-8 protease gave rise to a large core fragment (M, = 90,000) which could be isolated by actin affinity chromatography (4,9). The core retains the Ca2+-dependent nucleation and severing activity of villin, but lacks the bundling activity observed with villin in the absence of Ca2+ (4). Here we characterize for the fist time the second proteolytic fragment, the headpiece, which is separated from core by actin affinity chromatography. We show that the headpiece covers the carboxyterminal 76 residues of the villin polypeptide chain, whereas the core is restricted to the NHZ-terminal part. The core and headpiece account for different structural and functional domains of t h e v a i n molecule. The headpiece binds to F-actin independently of Ca" and acts as an inhibitor of villin-induced F-actin bundling. The combined results allow us to discuss the function of at least two distinct actin binding sites in the intact vain molecule.

MATERIALS AND METHODS
Isolation of Villin, Core, and Headpiece-Villin was isolated from chicken brush borders by our standard procedure (1,9). Villin was digested with V-8 protease by two methods, each a slight modification of that described in a preceding paper (9). Procedure "A" was used to prepare digests which could be used directly in F-actin bundling assays without the need for dialysis or lyophilization. In this procedure, villin (6-12 mg/ml) in storage buffer consisting of 10 mM NaCI, 10 mM PIPES,' 0.1 m~ dithiothreitol, 3 mM NaNs, 0.1 mM CaCI2, pH 7.0, was digested with Staphylococcus V-8 protease (5 pg/ml) for 30 min at room temperature. V-8 protease was removed from the digest by affinity chromatography on monospecific rabbit anti-V-8 protease IgGs covalently bound to Sepharose 4B (1 mg of monospecific IgG/ ml of settled gel) equilibrated in the storage buffer. The V-8 proteasedepleted digest was then used directly to assay for actin-bundling activity. Alternatively, the solution was applied to a DNase-Sepharose column saturated with actin (9) and equilibrated with storage buffer. The flow-through from this column has been previously shown to contain only the villin headpiece without significant contamination by actin, villin, or vain core. Procedure "B" was employed for large scale preparation of villin headpiece to be used in actin binding assays and protein-chemical studies. Villin (7-10 mg/ml) was dialyzed against 50 nm ammonium bicarbonate containing 0.1 nm CaCI2 and digested in this solution with 5 pg/ml of V-8 protease for 45 min at room temperature. V-8 protease was removed as before, and the solution was applied to a DNase-I column saturated with actin and equilibrated in 50 mM ammonium bicarbonate, 0.1 mM CaC12. Column fractions were monitored by absorbance a t 280 nm and the flowthrough fractions containing the headpiece were pooled and lyophilized. The lyophilized peptide was redissolved in distilled water, clarified by low speed centrifugation, and applied to a Sephadex (3-50 column (77 X 1 cm) (Pharmacia) equilibrated in water. Peak fractions were pooled and aliquots were stored lyophilized, or used immediately. Using this procedure, 1.5-2 mg of homogeneous villin headpiece was routinely isolated from 30-35 mg of intact villin.
Villin core, recovered from the DNase column by elution with EGTA-containing buffer, was redigested with 5 pg of V-8 protease/ml for 45 min, passed through the anti-V-8 column, precipitated with 50% ammonium sulfate, and applied to a Sephadex G-150 column (1.5 X 55 cm) equilibrated in 0.8 M KCI, 10 m~ PIPES, pH 7.0, in order to remove a small amount of 40,000-50,000-dalton fragments as described previously. The final preparation revealed only a single polypeptide with an apparent molecular weight of 90,000 when analyzed by SDS-polyacrylamide gel electrophoresis (4).
Low Speed Centrifugation Assay for F-actin Bundles-In order to study the bundling activity of villin by a biochemical assay, low speed centrifugation was used. Using this procedure, villin-actin bundles could be harvested by low speed centrifugation under conditions where normal F-actin remained in the supernatant. G-Actin at 0.5 mg/ml was incubated with varying amounts of villin, or equivalent amounts of the V-8 protease-depleted digest of villin (determined spectrophotometrically). The sample was adjusted to 50 mM KCI, 10 m~ PIPES, 1 mM MgCI2, 0.1 m~ ATP, 0.1 mM dithiothreitol, 0.1 mM CaC12, 5 mM EGTA, pH 7.0, in a total volume of 200 pl. After incubation for 1 h a t room temperature, the samples were centrifuged for 15 min a t 12,000 X g (Eppendorf centrifuge, model 5412). Supernatants were carefully separated from the pellets, and both fractions were adjusted to equal volumes with S D S sample buffer and boiled for 3 min. Samples were run on 7.5-20% polyacrylamide gradient SDS gels, stained with Coomassie blue, and quantitated by densitometry as described (14). Results were also monitored by electron microscopy. Pellets were fixed directly with 1% glutaraldehyde containing 0.2% tannic acid, embedded, and processed for sectional analysis as described (1).
F-actin Binding Assays-Villin, headpiece, or both were incubated a t various ratios with actin in the buffer described above for the bundling assay, except that EGTA was excluded where indicated. After 1 h at room temperature, the samples were centrifuged for 20 min a t 28 p.s.i. (150,000 X g) in a Beckman Airfuge. Supernatants were separated from pellets, adjusted to equivalent volumes with SDS sample buffer, run on 7.5-20% polyacrylamide gradient S D S gels, stained, and quantitated by densitometry.
Binding to Monomeric Actin-G-actin was centrifuged at 28 p.s.i. for 30 min in a Beckman Airfuge. The supernatant was mixed with headpiece (molar ratio, 2 actins to 1 headpiece), incubated for 1 h at room temperature in G-buffer (2 mM Tris, 0.2 mM dithiothreitol, 0.2 mM ATP, 0.2 mM CaC12, pH 7.6). and applied at 4 "C to a Sephadex G-50 column (77 X 1 cm) equilibrated in the same buffer.
Miscellaneous Procedures-Actin was purified from bovine skeletal muscle by the method of Spudich and Watt (15), using the modification of MacLean-Fletscher and Pollard (16). Protein concentrations were determined spectrophotometrically using extinction coefficients of A%' = 13 for villin, A: $ = 8.64 for the headpiece, and A%" = 6.6 for actin. In the case of headpiece, the value was monitored by amino acid analysis. Immunodiffusion was performed at room temperature using 1% agarose with 0.9% NaCI, 10 m~ imidazole, 0.02% NaNs, and rabbit antisera raised against native villin. Negative stain analysis was done as described (14).
Protein-Chemical Techniques-NH2-terminal sequence determination of villin and core was performed using the SDS-dansyl technique of Weiner et al. (17). Preparative peptide maps of tryptic or thermolytic digests of the headpiece were obtained on Whatman 3MM paper as described (18). Peptides were detected and recovered by standard procedures and characterized by amino acid analysis and extensive dansyl-Edman degradation. Amide assignments were based on the Offord plot and the results obtained by leucine aminopeptidase digestion (18). Digestion of villin, core, and headpiece by carboxypeptidase Y (Sigma Chemical Co.) was performed in 0.1 M sodium acetate, pH 5.5, containing 4 M urea. All other methods used are standard procedures and have been given elsewhere (18).

RESULTS
Isolation of Headpiece-In a previous report, we demonstrated that mild proteolysis of villin by V-8 protease yielded a core peptide of M , = 90, 000. This cleavage is very specific as judged by one-and two-dimensional electrophoretic analysis (4). Subsequently, we were able to identify a further proteolytic fragment (M, = 8,000) which we call the headpiece. The initial products of V-8 digestion are the core and the headpiece (Fig. 1). Upon longer incubation times, further partial cleavage of the core into intermediate molecular weight peptides (M, = 40,000-50,000) is observed. Under these conditions, the headpiece becomes heterogeneous, and a closely spaced doublet is observed on SDS-polyacrylamide gels (see Ref. 9). In order to explore the biochemical properties of the headpiece, we used digestion conditions yielding the headpiece as a single band on SDS gels, even though villin was only incompletely converted to core (-90%) (see under "Materials and Methods").
Since headpiece does not bind to actin immobilized on DNase-Sepharose, a simple purification scheme was devised. The flow-through from the DNase column was further subjected to gel fitration on Sephadex G-50. This preparation results in a good yield of homogeneous headpiece, which is salt-free and easily stored in lyophilized form ( Fig. 1). Villin core can be recovered from the DNase column by elution with EGTA (4,9).
The headpiece is electrophoretically homogeneous and migrates on 20% acrylamide gels in the presence of SDS as a single band with an apparent molecular weight of approximately 8,000 (not shown). That headpiece provides major antigenic sites in the intact villin molecule is shown by immunodiffusion experiments using rabbit antisera elicited with native intact villin (Fig. 2). No cross-reactivity with villin core is observed (Fig. 2), confming the assumption that headpiece is not a secondary digestion product of villin core, but provides a separate domain of the villin molecule. Wells of the double diffusion plate were filled with: A, rabbit antiserum directed against native villin; B and E , intact villin; C, villin core; and D, villin headpiece. Note the lack of identity between the core and headpiece, but the partial identity between the core (or headpiece) and intact villin. F-actin Binding of Headpiece-We have previously shown that isolated villin core does not bundle F-actin (4). It was therefore of interest to determine whether removal of headpiece is responsible for this loss of bundling. Use of anti-V-8 antibodies coupled to Sepharose enabled us to effectively remove the protease in these initial digests and to assay immediately for bundling activity. Since we found that villininduced F-actin bundles could be harvested by low centrifugal forces under conditions where normal F-actin is not sedimented (Fig. 3), we used this semiquantitative assay. When villin was mixed with actin under bundling conditions, ie. in the presence of EGTA, most of the actin became sedimentable when the molar ratio of villin to actin became higher than 1: 10 ( Fig. 4). When the same amount of V-8 protease-depleted digest of villin containing both core and headpiece was substituted, actin was not sedimented, confirming the loss of bundling activity by the cleavage itself.
Since villin core has been shown to retain the Ca"-dependent G-actin binding site(s) of intact villin (4, 9), we explored the possibility that the proteolytically derived headpiece may retain some biological activity to account for the loss of bundling. This is directly confirmed by the finding that headpiece binds to F-actin ( Fig. 5; Table I) but not to G-actin. The headpiece can saturate F-actin close to a molar ratio of 1 headpiece/actin monomer. This binding is Ca2'-independent and occurs equally well in the presence or absence of EGTA (Table I). As expected, high levels of headpiece inhibit the Factin binding of intact villin, and high levels of villin (but not core) inhibit the binding of the headpiece (Table I).
When G-actin was mixed with headpiece and applied to a Sephadex G-50 column, all actin eluted at the void volume, and headpiece was observed as a single peak eluting at the Morphology of Headpiece-F-actin Complexes-When actin was mixed with subsaturating amounts of headpiece, the morphology of the resulting actin filaments appeared to be unchanged in negative stain analysis (Fig. SA). As the headpiece level approaches saturating amounts (1 bound headpiece/actin), many filaments can be found which appear to be straighter or stiffer than normal F-actin (Fig. 6B). Although not all filaments in the latter. preparations appear to be straight, and normal F-actin can still be found as well, the long straight filaments are consistently found in the headpiece-treated samples and are absent in the controls. Cur-   ( B and D). Note also the loss of din-induced bundling (C) when the headpiece is included (D).
rently, the reason for these stiffer filaments is unknown. Actinheadpiece mixtures (Fig. 6, A and B ) do not show the F-actin bundles induced by villin (Fig. 6C). When headpiece is added to actin-villin mixtures, bundling is strongly reduced, and bundles which are observed appear to be only loosely organized (Fig. 6D).

Protein-Chemical Characterization of Villin Headpiece-
The amino acid composition (Table 11) indicates the purity of the preparation. Assuming the presence of 1 residue each of histidine and methionine predicts the presence of approximately 75 amino acid residues. The approximate minimal molecular weight of 8,500 is in excellent agreement with the value of 8,000 estimated from SDS-polyacrylamide gels and Sephadex G-50 gel filtration data in the presence of 0.1 M KCl.
For this molecular weight value, the amount of cysteine, isoleucine, and tyrosine is below the limit of detection, i.e. 0.05 residues/mol in our experiments. Thus, headpiece has a unique amino acid composition, because total villin contains tyrosine, isoleucine (1) and cysteine.2 Dansylation detected only valine as NH2-terminal residue. The COOH-terminal phenylalanine residue was identified by carboxypeptidase Y digestion (yield, 0.7 mol/mol). This experiment also predicted leucine as penultimate residue (see also below).
Two 1-mg preparations of headpiece were subjected to J. R. Glenney Val -Phe -Thr -A l a -Thr -T h r -Thr -Leu -V a l -P r o -T h r -L y s -Leu -G l u -Thr -Phe -P r o -Leu -Asp -Val -2 1 Leu -V a l -Asn -Thr -A l a -A l a -G l u -Asp -Leu -P r o -Arg -G l y -Val -Asp -P r o -Ser -Arg -Lys -Glu -Asn -

TABLE I1
Amino acid composition of isolated villin headpiece and the composition calculated from the amino acid sequence (see Fig. 7) The amino acid composition given in column A was calculated from the experimental data, adjusting for hydrolysis loss of serine and threonine and correcting for increased amounts of isoleucine, leucine, and valine upon prolonged hydrolysis. Column B gives the amino acid analysis calculated from the amino acid sequence analysis of the headpiece given in Fig. 7. Note the excellent agreement between columns A and B. Cysteine, isoleucine, and tyrosine are present in isolated headpiece at levels below our detection limit (ND), i.e. at levels below 0.05 mol/mol of headDiece.

Amino Acid
A B proteolytic cleavage using either trypsin or thermolysin, respectively. The resulting peptides were separated by the twodimensional preparative fingerprint system on Whatman 3 " paper (18). Peptides were detected by fluorescamine staining, recovered by elution, and characterized by amino acid analysis. Peptides were subjected to manual dansyl-Edman degradation to obtain maximal sequence information. In the case of a few peptides, secondary enzymatic digestion with chymotrypsin was employed. The resulting fragments were isolated and characterized as above. Amide assignments were based on the Offord plot and the results provided by leucine amino peptidase digestion. The combined results provided the amino acid sequence shown in Fig. 7. The amino acid composition calculated from the sequence is in excellent agreement with that found experimentally (Table 11), and provides a calculated molecular weight of 8,536 for the headpiece.

Aspartic acid
Headpiece Provides the COOH-terminal Domain of Villin-Manual dansyl-Edman degradation in the presence of SDS (17) showed that villin and villin core have the same NHzterminal sequence. Valine, glutamic acid or glutamine, and leucine are found in positions 1-3. Since headpiece shows the NHp-terminal sequence Val-Phe-Thr, the core fragment covers the NHp-terminal part of the villin polypeptide chain. Direct proof for the location of the headpiece at the COOH terminus of viUin was obtained by the characterization of the COOH-terminal residue of villin and headpiece. Carboxypeptidase Y digestion indicated a leucine-phenylalanine sequence (final yield of phenylalanine, 0.8 mol/mol) for both villin and headpiece. In agreement with the placement of the core at the NH2 terminus and the glutamic acid specificity of V-8 protease, hydrazinolysis identified glutamic acid as the COOHterminal end group in the core (yield, 0.3 mol/mol).

DISCUSSION
Villin displays two structurally and functionally distinct domains separable by mild in vitro proteolysis with V-8 protease. Dansyl-Edman degradation and the determination of the COOH-terminal residues of villin, core, and headpiece clearly show that the headpiece covers the COOH-terminal 76 amino acid residues of the villin polypeptide chain, whereas the core seems to span the remainder of the molecule, including the NH2 terminus. Currently, we don't know whether core and headpiece are directly continuous, i.e. whether V-8 protease gives rise to a single proteolytic cut or if a further fragment(s) is generated, which is lost in the final preparation of core and headpiece. If the latter case were true, the putative further fragment(s) must be very small, considering the molecular weights of villin, core, and headpiece. Our proteinchemical studies document a unique amino acid sequence of the headpiece, curiously lacking cysteine, isoleucine, and tyrosine. The calculated molecular weight of headpiece is 8,536, making this fragment the smallest actin-binding protein.
The limited action of proteolytic enzymes on native proteins can be envisioned as a process in which certain flexible hinge regions connecting tightly packed functional domains are preferentially and specifically cleaved (see, for instance, Refs. 19 and 20). That the headpiece consists of a tightly packed domain is directly suggested by the fact that there are 5 internal glutamic acid residues not available to V-8 cleavage. The headpiece must be prominently displayed on the surface of the native villin molecule, since it provides major antigenic sites for antibodies raised against native villin. Several such sera show strong reactivity with isolated headpiece.
Headpiece does not seem to bind to G-actin under our conditions, as indicated already by its isolation procedure. It does, however, bind to F-actin. The binding is independent of the presence or absence of Cap' and shows saturation at a molar ratio close to 1 headpiece for each actin monomer present in sedimentable F-actin form. This stoichiometry is obtained when actin and headpiece are mixed at a relative ratio of 1:4, and preliminary results indicate a dissociation constant in the micromolar range. Thus it is not surprising that headpiece inhibits villin-induced F-actin bundling observable in the absence of Cap+ and that villin inhibits headpiece binding to F-actin under similar experimental conditions. The combined results indicate that headpiece reveals one of the F-actin-binding sites responsible for the F-actinbundling activity of villin observed in the absence of Ca2'.
The results complement our previous characterization of the biological properties of the villin core (4,9). This large fragment retains the Ca2+-dependent properties previously documented for villin. The core shows severing of preformed F-actin and restricts, in a concentration-dependent manner, the length of F-actin filaments, when added to G-actin before salt-induced polymerization. This phenomenon is probably explained by the potent nucleation activity of villin or its core on actin assembly in the presence of Ca2+, since villin core caps the barbed end leading to a unidirectional elongation toward the pointed end. Consequently, treadmilling of actin monomers through the resulting filaments is inhibited (9). The final result, directly governed by the molar ratio of actin to villin, is the expression of more and necessarily shorter filaments, as observed independently by electron microscopical studies (1, 9-11). The formation of a distinct isolatable villin or villin core complex with three monomeric actins only emphasizes the degree of actin fiament length regulation and is observed upon the appropriate increase of the molar ratio of villin to actin (4, 5, 9). The loss of F-actin-bundling activity upon transition from villin to its core (4) is now explainable by the lack of the headpiece, which itself reveals an F-actin-binding site. Any simple model of F-actin bundling predicts the presence of at least two F-actin-binding sites in order to allow neighboring filaments to be bridged. We have previously identified a strong but Ca2+-dependent site in the core (4, 9), and here we show the existence of a Ca2+-independent site on the headpiece. Since F-actin bundling occurs in the absence of Ca2+, only the headpiece site qualifies as a site used in bundling. Although we have no direct evidence for the participation of the core part in bundling activity, this appears likely in view of the following observations. First, the headpiece does not induce bundling by itself (Fig. 6). Second, the core contains at least one or possibly two actin-binding sites active in the presence of Ca2+ (4,9). Third, Ca2+-dependent F-actin severing by villin core, a mechanism distinct from core-induced nucleation of actin assembly (9), seems to point to an endofiament mechanism (e.g. acting within the filament rather than at the end). Thus, although we cannot rule out the possibility of a yetunidentified third actin-binding site, expressed in the presence of EGTA, we prefer the conclusion that upon depletion of Ca2+, a conformational change allows the transition of the core site from a capping to a bundling state. It does not seem unreasonable to assume that although the binding of core to F-actin may be very weak in the presence of EGTA (4), the situation may change when the headpiece containing villin is used. A cooperative effect may be achieved by both actinbinding domains along the length of adjacent filaments resulting in the observed bundles. In support of this hypothesis, we found that binding of isolated headpiece to F-actin in EGTA (bundling conditions) is consistently weaker than the binding of intact vain (Table I).
The Ca2+-dependent functions of the core can be explained by either a single actin-binding site or the presence of two sites (9). The existence of an isolatable stoichiometric Ca"dependent complex with three monomeric actins or two actins when DNase is present (4, 9) favors the second possibility. This is further supported by the finding that another F-actinsevering protein, i.e. fragmin from P. polycephalum, shows in line with its lower molecular weight of 43,000 only a complex with one monomeric actin (2,8). Whatever the final number and properties of actin-binding sites of villin will be, our current results have clearly established the presence of at least two structurally and functionally distinct sites.
All other currently described Ca2+-dependent F-actin-severing proteins are similar to villin core in that they lack the EGTA-dependent F-actin-bundling activity of villin. They also show a lower polypeptide molecular weight than viUin, ranging in size between M, = 43,000 and 91,000 (discussed in Ref. 4). If these severing proteins were not only functionally but also structurally related to villin, they could arise by the expression of a related gene product omitting the "headpiece" domain at the COOH-terminal end. Alternatively, in vivo proteolytic processing could result in a shortened peptide lacking the headpiece region. In addition, proteolysis resulting during prolonged purification schemes must be considered as well. Further protein-chemical studies may be able to test these possibilities.
The presence of two isoelectric variants in purified villin (1, 5) has so far remained unexplained. The quantitative conversion of villin into core and headpiece together with the unique amino acid sequence of the headpiece clearly show that the two isoforms are closely related. The difference in isoelectric points resides in the core (4) and may either be due to local sequence heterogeneity or, more likely, to a partial posttranslational modification, such as, for instance, a phosphorylation. The possible physiological importance of this heterogeneity remains to be elucidated.