Demonstration of Three Distinct Calcium-binding Sites in Villin, a Modulator of Actin Assembly*

Villin, a Ca2+-modulated F-actin-binding protein (95,000 daltons) present in microvillus core filament bundles, has been shown to contain multiple Ca2+-bind- ing sites. 46Ca Hummel-Dreyer chromatography reveals the presence of two rapidly exchanging Ca2+-binding sites with an apparent dissociation constant, K d , equal to 4.6 X 1O“j M. Use of the two proteolytically separable domains of the molecule revealed that one site is located on the 90,000-dalton core (apparent K d = 3.5 X M) while the second site is provided by the 8,800- dalton headpiece fragment (apparent K d = 7.4 X M). In addition villin displays a further very slowly exchanging or nonexchangeable high affinity Ca2+-binding site, which is situated in the core domain. Sec- ondary structural predictions and a comparison of the amino acid sequence of headpiece with other known Ca2+-binding proteins indicates one region suggestive of a Ca2+-binding site, although headpiece seems not to exhibit a classical “EF-hand” Caz’-binding structure. The actin filament bundle supporting the membrane extension of the microvilli present on intestinal epithelial cells provides a particularly interesting system. This structure can be isolated in a homogeneous membrane-free in to actin at proteins, which three bind Ca2+ in the micromolar

The preceding paper (9) has shown t,hat addition of Ca2+ induces a iarge change in the hydrodynamic properties of villin. The conformational change is due to an increase in the frictional coefficient of villin and not to dimerization. Since this change was not detected with the isolated core (9), the results raise the question of the role of core and headpiece in e a 2 + binding.
Using Ca"-binding studies we have now identified multiple sites on the villin molecule. In addition to a nonexchangeable site, we document two rapidly exchanging sites, one on headpiece and one on core. We discuss the possible functional significance of the latter two sites and correlate the headpiece Ca'+-binding site to other sites established for various Ca')+binding proteins.

EXPERIMENTAL PROCEDURES
Isolation of Villin, Core, a n d Headpiece-Villin was isolated using the modifications on the standard procedures (3,12) given before (9). Villin was digested with V-8 protease. Core and headpiece were isolated as described (13). Protein concentrations were determined by absorbance at 280 nm using an absorptivity of 0.86 liter/g-cm for headpiece, 1.20 liters/g-cm for core (13), and 1.28 liters/g-cm for villin in EGTA (9). Circular Dichroism-CD measurements were made on a calibrated Jobin Yvon Mark V autodicrograph spectropolarimeter as before (15). Measurements were made using a 0.20-cm or 0.40-cm pathlength cell with protein solutions ranging from 150 pg/ml to 1.0 mg/ml at 20.0 k 0.1 "C over a range from 350 to 200 nm. Stock solutions of core and headpiece were prepared by passing them over a pre-equilibrated Sephadex G-25 column in 10 mM Tris, 10 mM NaCI, pH 8.5, and either 50 ~L M Ca2+ or 50 p~ EGTA. All solutions were centrifuged and Millipore-filtered prior to use. Spectra were repeated using different preparations of core and headpiece.
Ca"binding Studies-Nalgene beakers, tubes, and bottles were used throughout to avoid contamination with Ca'+ from laboratory glassware. Absolute concentration of Ca2+ in the various buffers and solutions was monitored using a UNICAM SPSOB series 2 atomic absorption spectralphotometer with Chelex-treated distilled deionized H20 as the reference. 45Ca2+ binding to villin, core, and headpiece was studied using the chromatographic method of Hummel and Dreyer (11). 41Ca'Li was purchased from New England Nuclear. A Sephadex G-25 fine column (0.6 X 20 cm) at room temperature, 23 & 2 "C, was equilibrated with the desired buffer. The flow rate was controlled with a Pharmacia P-3 peristaltic pump.
In order to obtain villin and core free of exchangeable Ca", approximately 1 ml protein solution, stored in an EGTA-containing solution, was dialyzed for 48 h against 1 liter of Ca'+-free H20 (3 changes). The protein solution was centrifuged and the integrit.y of the proteins monitored with sodium dodecyl sulfate-polyacrylamide gel electrophoresis prior t,o use. The protein was equilibrated in the column buffer before loading. Using this technique, the protein was eluted from the column in 20-30 min. 0.5-ml fractions were collected, mixed with liquid scintillation fluid obtained from Baker Chemicals, and counted using a Beckman L5-230 liquid scintillation system. The ratio of Ca2+ to '"Ca for each buffer could he calculated. Ca'+ binding was determined using the area of the radioactive peak corresponding to the *%a bound to the protein according to the method of Levi et al. (17). The data were then plotted with Scatchard plots, curves fit with least squares analysis (R2 2 0.95), and the stoichiometry and binding constants calculated. An average error of +IO% in the binding constants is estimated due to experimental error and temperature variations.
Headpiece as purified by the method of Glenney et al. (13) is essentially Can+ free ( t 5 X lo-' M ) and was used without further purification. The buffer used to determine Caz+ binding at pH 8.5 was 50 m M Tris, 50 mM NaCI, 0.1 m M dithiothreitol, 1.0 m M MgC12, plus the desired level of Ca2+ and 45Ca'+. At pH 7.0 and buffer system remained the same. Villin Renaturation-Villin was denatured and renatured using essentially the method of Hager and Burgess (18). 100 p1 of villin (22 mg/ml) was dialyzed into 6 M guanidine-HCI, 0.5 mM EGTA, 50 mM Tris, pH 8.5, at room temperature for 24 h to remove free Ca". The villin was then dialyzed into Caz+-free ( t 5 X 10" M Ca") 6 M guanidine-HC1, 50 mM Tris, pH 8.5, as before, in order to remove all EGTA. 45Ca2' to -5 X IO-* M was added to the protein solution and allowed to equilibrate for 1 h, at which time the solution was diluted 50-fold with Ca2+-free 50 mM Tris, 50 m M NaCl, pH 8.5. The protein was permitted to renature for 18 h at room temperature. The protein-'%a2+ solution was dialyzed against a 2000-fold volume of pH 8.5 buffer containing 50 mM Tris, 50 mM NaCI, 0.1 mM dithiothreitol, 1.0 mM MgC12, 10 @M CaCli for 24 h. Aliquots of dialysate and buffer were removed and the 4"Ca counted with a liquid scintillation system (Beckman), and the total concentration of Ca" was determined using atomic absorption spectroscopy.

Villin Contains Two Rapidly Exchanging Calcium Sites
Located on the Headpiece a n d the Core-Ca'+-binding experiments on villin, core, and headpiece were undertaken to define the stoichiometry and affinity of the Ca2+ interactions. The results obtained by the Hummel-Dreyer chromatography at pH 8.5 are shown in Fig. 1. Fig. 1 shows a typical elution profile. The peak of 45Ca coincides with the protein elution, in this case headpiece, and is separated from the resultant trough. This profile is indicative of a system in equilibrium and not suffering from protein overloading (22, 23). 45Ca binding was observed over a range from 1 X M Ca2+ to 20 X M ca'+ in 50 mM Tris, 50 mM NaC1,O.l M dithiothreitol, 1.0 mM MgC12, pH 8.5, using a series of protein-loading concentrations. No effect of protein concentration dependence was observed on the binding parameters nor did the elution profiles show trailing shoulders of unbound ligand. Thus it is unlikely that this system exhibits ligand-mediated association as discussed by Cann and Hinman (23). The Scatchard plots are shown in Fig. 2 and are linear within the range of Ca'+ concentration studied. The respective stoichiometries and binding constants are summarized in Table I     Villin Contains a Slowly Exchanging Ca2+ Site-Although villin reveals two rapidly exchangeable Ca2+ sites in solution, this result cannot provide information on the possible existence of any slowly exchanging or nonexchangeable Ca2+ sites. The stoichiometry of slowly exchanging Ca2+ was observed in native villin using a different approach. Duplicate samples of villin at 3.6, 2.1, and 1.0 mg/ml were dialyzed 24 h against 10 mM Tris and 1 mM EGTA, pH 8.0, to remove any exchangeable Ca". The protein was then exhaustively dialyzed against a 2000-fold volume of Ca'+-free 10 mM Tris, 50 mM NaC1, 0.1 mM dithiothreitol, pH 8.0, buffer a t room temperature with two changes of 24 h each. The total Ca2+ concentrations of both the dialysis buffer and the villin solutions were monitored using atomic absorption spectroscopy. The Ca"' concentration of the dialysis buffer was 50.5 X 1O"j M Ca, the limit of the atomic absorption spectralmeter's resolution, while the Ca2' concentration present in the dialyzed villin solution was 38 X M for the highest protein concentration used. The concentration of Ca2+ measured represented a molar ratio of 1.2 Ca2+ atoms per villin monomer when averaged for the three villin concentrations tested. This result indicates that one very slowly exchanging, z.e. tightly bound, Ca2+ remains associated with native villin.
Villin core was studied in a similar manner to observe if the slowly exchanging site is located in the core domain or if it lost upon digestion. Samples of core a t 2.7 mg/ml after having been dialyzed in the identical manner as villin were monitored using atomic absorption spectroscopy. The dialysis buffer contained approximately 0.6 X 1O"j M Ca, while the Ca'+ concentration present in the dialyzed core solution was 33 PM. This represents a molar ratio of 1.2 Ca'+ atoms/core monomer and demonstrates that the slowly exchanging Cas+ is retained in the core domain.
To further test this hypothesis, villin was denatured in 6 M guanidine-hydrochloride as outlined under "Experimental Procedures" and then renatured in the presence of 5 X lo-' M 45Ca. The 45Ca-exposed villin was then dialyzed against two changes, 12 h each, of 1000-fold volume of a buffer containing 1.0 IJM Ca" and 1.0 mM MgC12. Aliquots were removed and tested for 45Ca and for villin activity. The results showed that approximately 90% of the protein was renatured and active, as determined by F-actin binding, Ca'+-induced severing, and actin-DNase binding (for methods see Refs. 3,12,and 13). The 45Ca in the dialysis buffer revealed an average of 64 cpm/ ml while that within the dialysis tubing contained an average of 2550 cpm/ml. The concentration of Ca in the protein solution was determined using atomic absorption spectroscopy. The mol/mol ratio of Ca bound to villin was 0.85. This level can be accounted for by 45Ca. This calculation does not take into account the small loss of protein during renaturation or possible competition upon renaturation by other divalent cations. The results with denatured-renatured villin and native villin are in complete agreement and indicate that villin could contain a very high affinity Ca" site similar to those reported elsewhere (25,26).
Circular Dichroism Spectra of Core a n d Headpiece-Spectra were observed in the presence of either 50 IJM EGTA or 50 PM CaC12. As shown in Fig. 3A, villin core exhibits a change in spectra in the presence of Ca2+ similar to that previously observed with villin (9).
The increases from 10,600 -t 300 to 11,000 +-300 upon the transition from EGTA to Ca2+ buffers. The fraction of polypeptide existing in the a-helical conformation was calculated using the approximation of Chen et al. (16). Upon the binding of Ca2+, the extent of a-helix in core increases from 27 to 29%. The CD spectrum of headpiece, Fig. 3B, although different from that of core or villin, indicates some Ca2+induced structural changes. However, headpiece exhibits no change in the (e), , , , within experimental error, which indicates there are no large Ca2+-induced changes in the net amount of polypeptide in a-helix.
Therefore, the observed spectral changes are attributed to net changes in the amount of protein present in unordered a n d P structures. These spectral changes, although by no means definitive, are nevertheless indicative of Ca2+ interactions with both core and headpiece.
Secondary Structure of Headpiece a n d the Possible Location of a Calcium Site-Villin headpiece represents the smallest protein known to bind both actin and Ca". Since the i primary sequence of villin headpiece is known (13), a comparison of headpiece and some of the known Ca"-binding proteins could provide some clues to the possible location of the Ca2+ site within the 76 amino acid residues comprising the headpiece.
The characteristic secondary structure observed for calmodulin-like Ca2+-binding proteins consists of the "EF-hand" architecture (helix-loop-helix) (28). Does villin headpiece also exhibit such a structure? The secondary structure of headpiece was predicted using the method of Chou and Fasman (19,20). The results shown in Table I1 indicate a predominantly P-sheet structure on the NH2 terminus interrupted by a turn and a short segment of random structure. The region exhibiting the putative sequence homology would be predicted to contain two extremely strong p-turns extending from amino acids 29-32 and 34-37, followed by a stretch of random structure. The region between residues 44 and 52, inclusive, is particularly interesting, as the average predicted values for helix and P-sheet for this region are 1.18 and 1.20, respectively. As discussed by Chou and Fasman (21), such variable regions are capable of undergoing conformational changes depending on environmental conditions. The secondary structure of headpiece, however, is not characteristic of the "EF-hand" architecture, although the loop structure at the possible Ca2+ site seems similar to those observed in other Ca"-binding proteins (15). Tufty and Kretsinger (28) have proposed a test to ;ecognize "EF-hand" (helix-loop-helix) regions based on the amino acid sequence of a protein. If a minimum of 10 out of 16 critical structure forming and liganding residues can be aligned to a 29-residue test sequence, the region is considered a calcium binding "EFhand". When this test is applied to villin headpiece no sequence can be found which contains at least 10 of 16 of the test residues. The sequence which scores the highest (6 of 16) is shown in Table 111. Since headpiece is a calcium-binding protein, it may contain a variant binding site. Table I11 also contains the sequences for the crystallographically determined calcium-binding sites of carp parvalbumin, from which the "10 of 16" test was derived, and vitamin D-dependent calciumbinding protein (ICaBP). The structure of ICaBP, however, includes a variant Ca'+-binding site, residues 11-32, which does not meet the "10 of 16" test requirements (25). Another

TABLE I1
The predicted secondary structures of villin headpiece The amino acid sequence of villin headpiece is shown with the predicted secondary structures (PS). €3, P-sheet. T, P-turn, R, random, H, helix. *, variable region (see text); (Ala, A; Arg, R; Asn, N; Asp, D; Cys, C; Gln, Q; Glu, E; Gly, G; His, H; Ile, I; Leu, L; Lys, K; Met, M; Phe, F Pro, P; Ser, S; Thr, T; Trp, W; Tyr, Y; Val, V). The method Ca2+ binding site shown in Table 111. of Chou and Fasman (19,20) was used for structural predictions. The underlined region represents the proposed (PS) B B B B B B (9). Amino acid sequences of some calcium-binding proteins with a

I V F T A T T T L V I~P T K L E T F P L D~~V L V N T A A E D L : I O P
The two structural domains of villin, obtained through limited comparison to villin headpiece proteolysis, are a Ca"4ndependent F-actin binding headpiece The sequences are aligned to show the homologous calcium-binding (&Ir = 8,800) and a Ca"+-sensitive core (Iw, = 87,000) which regions. The residues underlined are either assumed to be involved in both nucleates actin and restricts filament length (13, 14).
analysis (for details see text). The standard amino acid abbreviations calcium coordination or known to do this by x-ray crystallographical core, however, is not capable of exhibiting the Cat+-induced are provided in Table 11. hydrodynamic change observed in villin (9). The questions of how villin is able to perform both bundling and length restriction activities and how these are related to the observed which attempts to localize and describe the Ca"-binding sites MCBP"." (CD) 4:FAI I Q Q D K s G F I E E D E L K L F L Q , ; " (EF) changes have been approached by this study protein containing variant Ca"-binding sites is the S-100 protein, also shown in Table 111. The variant Ca"-binding sites of S-100 and ICaBP contain both amino acid substitutions and additions within the proposed sequences for CaZ+ coordination when compared to MCBP. Headpiece may represent a more distantly related Ca"+-binding site containing yet other amino acid changes. As pointed out by Szebenyi et al. (25), variant EF-hands are not always detectable through primary sequence comparisons to parvalbumin and, therefore, the recognition of such variant sites will require the crystallographic determination of their three-dimensional structures. These conclusions would indicate that not all Ca"-binding proteins require parvalbumin-like EF-handedness, and that in the general case, only a particular secondary and tertiary structure is required to provide for the Ca'+ coordination.
Since the headpiece does bind Ca"', the predictions presented here represent the alternative methods available to attempt to localize the Cat+'-binding site. We recognize that these results are by nature speculative and must be further tested by crystallographic methods.
The role of the very slowly exchanging, tightly bound, Ca2+ in villin is unknown, but may represent a structural site unaccessible t.o the solvent. As shown, the location of this structural site is on the 87,000-dalton core fragment. A similarity may exist between villin and certain Ca"-binding proteins whose structures have been crystallographically determined. Szebenyi et al. (25) reported the existence of a structural Ca" site in vitamin D-dependent calcium-binding protein from bovine intestine, and Moews and Kretsinger (26) have identified the presence of a Ca'+ site, inaccessible to solvent, in carp muscle calcium-binding parvalbumin. Thus, although this category of site is not without precedent in other Ca"+-binding proteins it provides the first report in a protein known to exert Ca"-dependent modulation of actin filaments.
In contrast to preliminary equilibrium dialysis experiments in which villin was observed to display a single exchangeable Ca2+-binding site with a Kc, = 2.5 X lo-" M at pH 7.3 (8), the results of the Hummel-Dreyer 4"Ca-binding studies summarized here clearly reveal two rapidly exchanging Ca" sites, one of which is located on core and the other on headpiece. The two sites, being located on different domains of the villin molecule are probably not identical, although the Kd values obtained for each separately are reasonably close. The binding data from villin represent a sum of the stoichiometry of headpiece and core binding and also appear to represent the K d as an intermediate to the values obtained for the two separate sites. No evidence has yet been observed indicating any type of cooperative effects between the two binding sites, although more extensive studies at lower concentrations of Ca2+ with varying pH values and salt conditions would be necessary before this possibility could be eliminated. The 45Ca elution profiles reveal no suggestion of Ca2+-induced association, in complete agreement with the hydrodynamic data in the accompanying paper (9). The combined results argue that Ca'+-induced polymerization is not involved in villin regulation. However, the functional role of the two Ca2+ sites is open to speculation. Previous studies (12-14) have shown that villin core retains Ca"-dependent F-actin nucleation and length restriction activities and G-actin-DNase I binding. Thus, this line of evidence suggests that the single exchangeable core Ca2+-binding site would control these activities. The headpiece fragment, by contrast, has exhibited no Ca"-dependent activities thus far as it seems to bind equally well to F-actin in the presence of EGTA or Ca2+ (13), but is required for Factin cross-linking by villin in the absence of Ca".
Hydrodynamic data presented previously have shown that v a i n core is incapable of displaying the extensive hydrodynamic change observed in intact villin upon the addition of calcium (9). Nevertheless, core retains both Ca2+ binding K,i = 3.5 X M, and Ca"-induced alterations in the secondary structure as observed in CD spectra. This evidence supports the concept that headpiece is involved in an essential role for the expression of the Ca2'-induced conformational change of intact villin as observed in hydrodynamic changes. Upon addition of Ca'+ intact villin alters its sedimentation coefficient to become a distinctly more asymmetric molecule (9), while villin core does not show this property. This suggests that the headpiece domain must be repositioned relative to the core in the intact villin molecule to give rise to the observed increase in asymmetry. The exact mechanism by which this repositioning occurs cannot be elucidated by our solution studies, but secondary structural predictions of headpiece suggest one possible site of action within this domain, the "variable" region located at the COOH-terminal side of the proposed Ca"-binding loop. Although inherently speculative by nature, since limited by calculated secondary structure prediction methods and the lack of true three-dimensional structural evidence, such a hypothesis is interesting, since the proposed site may represent a structure in which the local environment can significantly alter the structure of the polypeptide backbone (see under "Results"). It must also be stated that while no direct evidence exists at this time to conclusively link the observed Ca"-dependent conformational hydrodynamic change to the known Ca2+-dependent actin modulation activities of v a i n a similar range of Ca2' concentrations is required to induce the changes in conformation (9) and to affect the stability of the microvillus actin filament bundle which contains villin as a structural component (3,7,8,10). Since two rapidly exchanging Ca"-binding sites have been observed the possibility exists that one is responsible for the Ca"-sensitive activities of villin that are retained by the core, nucleation, and filament length restriction, while the second would be responsible for the large conformational changes observed which require the headpiece domain. This hypothesis does not exclude the hinge mechanism as outlined in the previous paper (9), but rather provides additional insight by suggesting a complementary mechanism for the Ca2' regulation of the conformational change. In such a model, the hinge mechanism would allow for cross-linking of actin filaments by villin in the absence of Ca" provided the actinbinding site located in the headpiece domain is in the proper orientation, ie. when it is in a more compact conformation. Upon the binding of Ca2+, the relative position of the headpiece domain is shifted in relation to the core to a more asymmetric conformation, resulting in an orientation of the headpiece domain's actin-binding site that is no longer suitable for the cross-linking of F-actin filaments. This mechanism would have the advantage that it does not require Ca'+ sensitivity for headpiece binding to F-actin filaments, agreeing with previous reports (13). It would also allow for the possibility that the headpiece Ca2+ site is unrelated to the previously reported Ca2+-dependent activities of core, but would still be essential for the Ca"-regulated transitions of the intact villin molecule. Although the physical-chemical data collected thus far cannot allow one to determine the exact mechanism responsible for the regulation of villin's Ca'+-sensitive bundling and severing activities, they clearly have required a more elaborate model for villin's activities. Nevertheless, the presence of three Ca"-binding sites together with the large Ca'+-induced conformational change has provided a much clearer picture of villin's structure.