The Structure and Function of the Actin-binding Domain of Myosin Light Chain Kinase of Smooth Muscle*

In addition to its kinase activity, the myosin light chain kinase (MLCK) of smooth muscle has an actin binding activity through which it can regulate the actin-myosin interaction of smooth muscle (Kohama, K., Okagaki, T., Hayakawa, K., Lin, Y., Ishikawa, R., Shimmen, T., and Inoue, A. (1992) Biochem. Biophys. Res. Commun. 184, 1204–1211). In this study, we have analyzed the actin binding activity of MLCK and related it to its amino acid sequence by producing native and recombinant fragments of MLCK. Parent MLCK exhibited both calcium ion (Ca2+) and calmodulin (Ca2+/CaM)-sensitive and Ca2+/CaM-insensitive binding to actin filaments. The native fragment, which consists of the Met1–Lys114 sequence (Kanoh, S., Ito, M., Niwa, E., Kawano, Y., and Hartshorne, D. J. (1993)Biochemistry 32, 8902–8907), and the recombinant NN fragment, which contains this 1–114 sequence, showed only Ca2+/CaM-sensitive binding. An inhibitory effect of the NN fragment on the actin-myosin interaction was observed by assayingin vitro motility and by measuring the actin-activated ATPase activity of myosin. The recombinant NN/41 fragment, which is constructed without the Met1–Pro41 sequence of the NN fragment, lost both the actin binding activity and the inhibitory effect. We confirmed the importance of the 1–41 sequence by using a few synthetic peptides to compete against the NN fragment in binding to actin filaments. The experiments using recombinant fragments and synthetic peptides also revealed that the site for CaM-binding is the Pro26–Pro41 sequence. The site for the Ca2+/CaM-insensitive binding, which is shown to be localized between the Ca2+/CaM-sensitive site and the central kinase domain of MLCK, exerted no regulatory effects on the actin-myosin interaction.

Myosin light chain kinase (MLCK) 1 has an important regu-latory role in smooth muscle contraction (see Ref. 1 for a review). MLCK phosphorylates the 20-kDa light chain of smooth muscle myosin together with calmodulin in the presence of Ca 2ϩ ions (Ca 2ϩ /CaM), thereby activating the myosin, which can then interact with actin filaments to induce contraction.
In addition to this kinase activity, MLCK can act as an actin-binding protein; MLCK is present in association with the sarcomeric I-band (2), and it binds to actin filaments with a high affinity (3)(4)(5). We have shown that MLCK can inhibit the ATP-dependent interaction between actin and myosin by binding to actin filaments. This inhibition can be relieved by Ca 2ϩ /CaM, as was demonstrated by avoiding the complication derived from its kinase activity (6 -8), although the demonstration was limited to in vitro only.
Such an inhibition, however, is not peculiar to MLCK. Caldesmon (see Ref. 9 for a review) and calponin (10) are known to inhibit the actin-myosin interaction by binding to actin, an inhibition that is relieved by Ca 2ϩ /CaM. Twenty years ago, Hartshorne et al. (11) obtained an inhibitory fraction from chicken gizzard in a preliminary form. The activity of the 130-kDa component of their fraction relates to the actin-binding property of MLCK. By measuring the content of MLCK, caldesmon, and calponin in the myofibril of smooth muscle (12), we have shown previously how the actin-binding property of MLCK participates in regulating smooth muscle contraction.
The relationship between the structure of MLCK and the function of the kinase activity in phosphorylating the myosin light chain and how Ca 2ϩ /CaM modulates the activity has been well established (see Ref. 13 for a review). However, there has been little study of the structure-function relationship of the actin binding activity except for the purification of an actinbinding fragment from MLCK (14).
In this study, we have investigated which sequence of MLCK is responsible for the actin binding activity, which sequence inhibits the actin-myosin interaction, and which sequence binds CaM to relieve the inhibition. Our approach has been to cleave MLCK to prepare native fragments containing the actin binding activity (14), to design MLCK cDNA to express the recombinant fragments of actin binding activity in Escherichia coli, and then to analyze them biochemically. A few peptides have been synthesized to confirm these analyses. We have shown that: (i) MLCK has two actin-binding sites on the Nterminal side away from the central kinase domain, (ii) the binding site responsible for the inhibitory effect is at the Met 1 -Pro 41 sequence, and (iii) CaM interacts with MLCK at the Pro 26 -Pro 41 sequence.

MATERIALS AND METHODS
Preparation of Proteins-All procedures were carried out at 0 -4°C. The purity of proteins was routinely monitored by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (see below) so * This work was supported in part by grants from the Fujisawa Foundation and the Mitsubishi Foundation and by grants-in-aid for scientific research from the Ministry of Education, Science and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed. Tel.: 81-27-220-7960; Fax: 81-27-235-1401 or 81-27-220-7962. 1 The abbreviations used are: MLCK, myosin light chain kinase; CaM, calmodulin; PAGE, polyacrylamide gel electrophoresis; NTCB fragment, chicken gizzard MLCK fragment produced by the cleavage with 2-nitro-5-thiocyanatobenzoic acid; NN fragment, expressed fragment (amino acids 1-337 of bovine MLCK); NC fragment, expressed fragment (amino acids 319 -721 of bovine stomach MLCK); NN/25 fragment, expressed fragment (amino acid 26 -337 of bovine stomach MLCK); NN/41 fragment, expressed fragment (amino acid 42-337 bovine stomach MLCK). that they were Ͼ95% pure. The purified proteins, except actin and myosin (see below), were divided into aliquots and stored at Ϫ80°C until they were used. All columns (see below) were incorporated into a high performance liquid chromatography system (model L-6200; Hitachi, Tokyo, Japan). MLCK was prepared by the method of Adelstein and Klee (15) with slight modification (16). In short, MLCK was extracted from the smooth muscle of chicken gizzard and was subjected to ammonium sulfate fractionation. After fractionation, MLCK was purified using column chromatography in DEAE-Toyopearl 650M columns and subsequently in SP-Toyopearl 650M (Tosoh, Tokyo, Japan) columns; this MLCK was used in all experiments unless otherwise specified. We also purified MLCK from bovine stomach smooth muscle by a modification (6) of the method of Kuwayama et al. (17) and used this for a few experiments, in which it is specified as bovine stomach MLCK.
MLCK was cleaved by 2-nitro-5-thiocyanatobenzoic acid (NTCB) complex at a Cys site, and an actin-binding fragment was purified as described by Kanoh et al. (14). The fragment consisted of the Met 1 -Lys 114 sequence (14,18) and is referred to as the NTCB fragment.
Skeletal muscle actin was purified from an acetone powder of chicken breast muscle (19) and used as actin filaments after polymerization. The concentrations of actin filaments were expressed in terms of monomericactin. Myosin was purified from chicken gizzard as described by Nakamura and Nonomura (20), who modified the method of Ebashi (21) to remove phosphatase activities. The myosin was phosphorylated by incubation with MLCK in the presence of Ca 2ϩ /CaM for 10 min at 25°C as described by Okagaki et al. (22) and used as myosin in our experiments. CaM from bovine brain, glucose oxidase, and catalase were purchased from Sigma.
Expression and Purification of Recombinant MLCK Fragments-For the expression of recombinant MLCK fragments, the pET system (Novagen, Madison, WI) was used according to the manufacturer's instructions. Series of cDNA fragments (see Fig. 6 for their topology) encoding various domains of bovine stomach MLCK (23) were amplified by the polymerase chain reaction. The amplified fragments were subcloned into a pET21 vector and verified by DNA sequencing. The recombinant MLCK fragment was overproduced in E. coli BL21(DE3). The cells were grown to an absorbance at 600 nm of 0.5-0.6 in 2 liters of Luria broth medium containing ampicillin (50 g ml Ϫ1 ) and then were induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside. All of the fragments were soluble in low salt. Therefore, cells expressing the fragments were collected by centrifugation, and then were suspended in 40 ml of a solution of 14 mM 2-mercaptoethanol and 20 mM Tris-HCl (pH 7.5). The resulting supernatant was fractionated with ammonium sulfate. The fragments were purified from the fraction by column chromatography using a combination of DEAE-Toyopearl 650M and CM-Toyopearl 650M columns (Tosoh, Tokyo, Japan). When any minor contaminants were found in SDS-PAGE, they were removed by gel filtration with Superose 12HR (Pharmacia Biotech Inc.). The MLCK fragments were: NN-, NC-NN/25, and NN/41 fragments, which were composed of Met 1 -Lys 337 , Pro 319 -Val 721 , Pro 26 -Lys 337 , and Lys 42 -Lys 337 of bovine stomach MLCK (23), respectively. The amino acid sequences of these fragments were confirmed by sequencing about 7 residues from their N termini using an Applied Biosystems model 477A analyzer.
Centrifugation Assay for the Actin Binding Activity of MLCK and Its Fragments-Actin filaments at a concentration of 12 M were mixed with specified amounts of MLCK, its fragments, and/or CaM in 50 mM KCl, 20 mM Tris-HCl (pH 7.5), 5 mM MgCl 2 , and 0.1 mM EGTA or 0.1 mM Ca 2ϩ unless otherwise specified in the figure legends. The mixture was incubated for 0.5 h at 25°C. The mixture after incubation was centrifuged at 100,000 ϫ g for 45 min in a Beckman Airfuge at 25°C. The amount of MLCK both in the supernatants and the pellets was determined by SDS-PAGE followed by densitometry as described above, and then the amount of MLCK bound to actin filaments was calculated (24). The assay for the binding activity of the fragments of MLCK was done in the same way. The results were analyzed on Scatchard plots.
Interaction between Calmodulin, MLCK, and Its Fragments-The binding interaction between CaM and MLCK or its fragments was also measured by surface plasmon resonance with the IAsys Cubette System (Fissons, Cambridge, United Kingdom). CaM was immobilized to the cubette of carboxymethylated dextran matrix via N-hydroxysuccinimide and 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide according to the manufacturer's protocol. MLCK and its recombinant fragments, after dissolving in 50 mM KCl, 5 mM MgCl 2 , 20 mM Tris-HCl (pH 7.5), and 0.1 mM Ca 2ϩ or 0.1 mM EGTA, were added to the cubette. Then the resultant resonance at 25°C was recorded. In some experiments, the binding of MLCK or its fragments was monitored in the presence of various concentrations of synthetic peptides.
The Actin-Myosin Interaction-The effect of MLCK and its fragments on the ATP-dependent interaction between actin and myosin was analyzed by means of an in vitro motility assay on a myosin-coated surface (see Ref. 25 for the method), as modified for smooth muscle myosin by Okagaki et al. (22). In short, 3 nM actin filaments labeled with rhodamine-phalloidin (Molecular Probes, Eugene, OR) were mounted on a coverslip coated with myosin together with various concentrations of MLCK in 50 mM KCl, 30 mM Tris-HCl (pH 7.5), 4 mM MgCl 2 , 1 mM ATP, 25 mM dithiothreitol, and 0.1 mM EGTA. To see the effect of Ca 2ϩ /CaM, EGTA was replaced by 0.1 mM Ca 2ϩ and CaM at the same molar concentration as MLCK. Glucose oxidase (0.2 mg ml Ϫ1 , Sigma), catalase (0.04 mg ml Ϫ1 , Sigma), and glucose (4.5 mg ml Ϫ1 ) were added to the above solution to prevent photobleaching of rhodamine. The movement of actin filaments was recorded under a fluorescence microscope equipped with a videocamera and a recorder. The average velocity was measured using an image analyzer (Image Sigma III, Nippon Avionics, Tokyo, Japan).
The effect of recombinant MLCK fragments on the actin-myosin interaction was also examined by measuring ATP hydrolyzed at 25°C for 10 min with 0.05 M phosphorylated myosin in the presence of various concentrations of MLCK and its fragments under the conditions of 2 M actin filaments, 0.5 mM ATP, 60 mM KCl, 5 mM MgCl 2 , 0.1 mM EGTA, and 20 mM Tris-HCl (pH 7.5) unless otherwise specified. The liberated phosphate was quantified in duplicate or triplicate by the malachite green method of Kodama et al. (26). We ensured that the rate of ATP hydrolysis was constant and that the specific activity of the ATPase did not differ from our previously reported values. Therefore, we expressed the activity as a normalized ATPase activity (27).
Other Methods-SDS-PAGE was carried out using the method of Laemmli (28) with a slight modification (24) so that we could check the purity (Ͼ95%) of MLCK and its fragments. Protein concentrations were determined using the methods of Bradford (29) and Lowry et al. (30) with bovine serum albumin as the standard.
The molecular masses used for calculating concentrations (M) of proteins and peptides were obtained from their amino acid sequences as follows: myosin, 440 kDa; actin, 41.8 kDa; CaM, 16 Homology of the specified amino acid sequences of MLCK with the sequences of other proteins was examined using MacDNASIS Pro 020 -00 (Hitachi Software Engineering, San Bruno, CA) to search the GenBank/NBRF-PIR/SWISS-PROT data bases. Isoelectric points of the synthetic peptides were calculated according to Skoog and Wichman (31).

Actin-binding Properties of Recombinant Fragments of MLCK-
We designed a fragment so that it consisted of Met 1 -Lys 337 of bovine stomach MLCK (see Fig. 6 for its topology in MLCK), and subjected this NN fragment to an actin-binding assay. As shown in Fig. 1a by open circles, the amount of the NN fragment bound to actin filaments increased with increasing concentration of the NN fragment. The K a of the binding of the NN fragment to actin filaments was 3.72 Ϯ 0.20 ϫ 10 5 M Ϫ1 (mean Ϯ S.E., n ϭ 3). The binding was abolished totally by Ca 2ϩ /CaM (Fig. 1b, open circles). Thus, we deduce that the 1-337 sequence contains a site for Ca 2ϩ /CaM-sensitive actin binding.
Previously, we investigated the importance of the sequence between Lys 42 and Ala 80 to the inhibitory effect of MLCK by searching for a part of MLCK that is homologous to the actinbinding domain of caldesmon (6). We synthesized a peptide consisting of Lys 42 -Ala 80 of MLCK and used it to compete against the actin binding activity of the NN fragment. How-ever, as shown in Fig. 1c by the open circles, the binding was not at all affected.
We obtained another peptide, this time of the Met 1 -Pro 41 sequence, and found its inhibitory effect on the actin binding of the NN fragment as shown by the filled circles in Fig. 1c. Thus, the sequence responsible for the binding of the NN fragment to actin filaments, i.e. the Ca 2ϩ /CaM-sensitive site of MLCK, must be located at the N terminus of the MLCK molecule. We divided peptide Met 1 -Pro 41 into peptide Met 1 -Gly 25 and pep-tide Phe 26 -Pro 41 . Both of them failed to inhibit the actin binding (open and filled triangles in Fig. 1c), suggesting that there is a critical region for actin binding within the 1-41 residues.
To confirm this idea by another method, we produced the NN/41 fragment, which is devoid of the Met 1 -Pro 41 sequence of MLCK, and the NN/25 fragment which is devoid of the Met 1 -Gly 25 sequence of MLCK (see Fig. 6 for their topology in MLCK), and then assayed for their actin binding activity. As shown by filled triangles in Fig. 1a, the binding of the NN/41 fragment to actin filaments remained at a basal level. Although the NN/25 fragment definitely bound to actin filament with K a ϭ 1.01 Ϯ 0.53 ϫ 10 5 M Ϫ1 (mean Ϯ S.E.,n ϭ 3), the affinity was much lower than that of the NN fragment. The length of the Pro 26 -Pro 41 sequence when added to the NN/41 fragment (i.e. the NN/25 fragment) is probably too short to endow it with an actin-binding ability as high as that of the NN fragment, confirming the importance of the 1-41 sequence as an actinbinding site.
We also produced the NC fragment, which comprises the Pro 319 -Val 721 sequence of bovine stomach MLCK (see Fig. 6 for its topology in MLCK). We assayed the NC fragment for its actin-binding ability. As shown in Fig. 1a by filled circles, it shows a definite actin binding activity with a K a ϭ 1.83 Ϯ 0.47 ϫ 10 5 M Ϫ1 (mean Ϯ S.E., n ϭ 3). We used CaM in the presence of Ca 2ϩ to compete against the actin binding of the NC fragment, but failed to observe any effect of Ca 2ϩ /CaM (see Fig.  1b, filled circles). Considering the Ca 2ϩ /CaM-sensitive, actin binding activity of the NN fragment, parent MLCK would be expected to bind to actin filaments in both Ca 2ϩ /CaM-sensitive and insensitive ways, which will be shown later and discussed (see Fig. 8a).
Direct Detection of Interaction of Recombinant Proteins with CaM-We immobilized CaM on an IAsys cubette surface, and allowed MLCK to interact with the surface. In the presence of Ca 2ϩ , the surface plasmon resonance of IAsys increased with the increase in the concentration of MLCK. However, in the presence of EGTA, the increase is only slight (data not shown). Such a difference indicates that the IAsys system could be reliably used to detect binding activity of MLCK and its fragments to CaM.
As shown in Fig. 2a by open (MLCK) and filled (bovine stomach MLCK) circles, the difference in the resonance between Ca 2ϩ and EGTA at the specified concentrations was readily increased at low concentrations of both MLCKs and approached saturation at their higher concentrations. This Ca 2ϩdependent change in the resonance of the NN fragment shown in Fig. 2a by filled squares is quite different from that of parent MLCK; it increased gradually. However, it was never saturated, indicating the low affinity of the fragments for CaM.
We assayed the NN/41 fragment with IAsys using the same surface of the cubette. As shown in Fig. 2a, the Ca 2ϩ -dependent changes in the resonance were not detected with this fragment (filled triangles), indicating that the CaM-binding site is not located in the C-terminal portion away from Lys 42 . The failure to detect these changes suggests that the site is probably within the sequence of Met 1 -Pro 41 , and so we allowed the Met 1 -Pro 41 peptide to compete against the CaM binding activity of the NN fragment. As shown by the triangles in Fig. 2b, the Ca 2ϩ -dependent changes in the resonance of the NN fragment reduced with the increase in the concentration of the peptide. The molecular mass of the peptide itself was too low to produce a resonance (data not shown), so the reduction in the resonance is an indication of the binding of the 1-41 sequence of MLCK to CaM.
The NC fragment was also constructed to include a Ca 2ϩ / CaM-insensitive, actin-binding site, which was confirmed by the actin-binding assay (see Fig. 1). Accordingly, the NC fragment did not cause any Ca 2ϩ -dependent change in the resonance (Fig. 2a, open triangles).
To identify the position of the CaM-binding site in the 1-41 sequence, we synthesized peptides of Met 1 -Gly 25 and Pro 26 -Pro 41 . Because of their low molecular mass, their interaction with CaM could not be detected directly by resonance (data not shown). Therefore, their interaction with CaM was detected by observing the interaction of the NN fragment with the CaMcoated surface of an IAsys cubette in the presence of various concentrations of the peptides (Fig. 2b). The peptide that competed with the NN fragment was Pro 26 -Pro 41 .
We also tested the NN/25 fragment for CaM-binding ability. As shown by asterisks in Fig. 2a, its binding ability is comparable to that of the NN fragment. Taken together with the competing abilities of the peptides, we conclude that the boundary of the CaM-binding sequence of MLCK is at Pro 26 .
Effect of Recombinant Proteins on the Actin-Myosin Interaction-As shown in Fig. 3a by filled circles, the motility of actin filaments on a myosin-coated glass surface was inhibited with the increase in the concentration of the NN fragment. Halfmaximal inhibition was observed in the presence of 41.1 Ϯ 5.5 nM (mean Ϯ S.E., n ϭ 3) of the NN fragment. The inhibition was relieved by Ca 2ϩ /CaM (Fig. 3a, open circles). Actin-activated ATPase activities of myosin (Fig. 3b, filled circles) were reduced with the increase in the concentration of the NN fragment. Half-maximal inhibition was observed in the presence of 0.276 Ϯ 0.028 M (mean Ϯ S.E., n ϭ 3). Ca 2ϩ /CaM effectively relieved the inhibition; the inhibition caused by 1 M NN fragment was abolished by an 8 -10-fold molar excess of CaM over the NN fragment (Fig. 3c, filled circles).
It must be noted that the concentration required for inhibiting the activity half-maximally was about 6.7-fold higher than the concentration required for inhibiting the motility. Such a discrepancy has been noted by Sato et al. (8) by using aortic MLCK. The difference in the concentrations of actin filaments used for the motility and ATPase assays may be one possible explanation.
The regulatory activity of the NN/41 fragment was also tested both with motility and ATPase assays. As expected from the absence of actin binding activity (Fig. 1a), we failed to find any effect, irrespective of the presence or absence of Ca 2ϩ /CaM (Fig. 3, a-c, squares). Taking the absence of the actin binding activity of the NN/41 fragment (Fig. 1a) into consideration, the absence of any effect indicates that Met 1 -Pro 41 is the sequence of MLCK that is responsible for the inhibition of myosin through actin binding activity. The NN/25 fragment, which showed a weak actin binding activity (Fig. 1a), failed to inhibit the effect in the motility and ATPase assays (Fig. 3, a and b,  asterisks), suggesting that the residues on both N-and Cterminal sides of Gly 25 are required for inhibiting the actinmyosin interaction.
The NC fragment, on the other hand, bound to actin filaments in a Ca 2ϩ /CaM-insensitive manner (Fig. 1b). As shown in Fig. 3 (a and b) by triangles, the NC fragment did not exert any regulatory activity as monitored both by the motility and ATPase assays. The absence of activity was confirmed in the presence of various concentrations of Ca 2ϩ /CaM (Fig. 3c, triangles). Thus, we conclude that the Ca 2ϩ /CaM-insensitive binding site of MLCK has no regulatory activity.
Confirmation with Parent MLCK and Its Native Fragment-We allowed parent MLCK to bind to actin filaments in a similar way to its recombinant fragments. In the presence of 1 mM EGTA, MLCK bound to actin filaments at its high affinity site at 0.15 Ϯ 0.02 mol/mol actin (mean Ϯ S.E., n ϭ 3) with an affinity constant K a ϭ 2.16 Ϯ 0.65 ϫ 10 6 M Ϫ1 (mean Ϯ S.E., n ϭ 3). At its low affinity site, MLCK bound to actin filaments with a K a ϭ 4.67 Ϯ 0.17 ϫ 10 5 M Ϫ1 (mean Ϯ S.E., n ϭ 3). In the presence of 1 mM Ca 2ϩ and CaM at an 8-fold molar excess over MLCK, binding of MLCK to the actin filaments had only a single K a of 3.1 ϫ 10 5 M Ϫ1 . Because this value is closer to the K a of the low affinity site, we conclude that the actin binding at the high affinity site is Ca 2ϩ /CaM-sensitive.
As we describe below (see Fig. 8a), the Ca 2ϩ /CaM-insensitive binding is confirmed by the persistence of actin binding activity in MLCK when the concentration of CaM was increased. Sellers and Pato (4) showed Ca 2ϩ /CaM-sensitive binding of MLCK to actin filaments. However, we conclude that parent MLCK binds to actin filaments in both Ca 2ϩ / CaM-sensitive and -insensitive ways, confirming the Ca 2ϩ / CaM-sensitive and -insensitive binding of the NN-and NC fragments, respectively.
We produced the NTCB fragment from MLCK by chemical cleavage. In the presence of 1 mM EGTA, the NTCB fragment showed actin binding activity, confirming the results of Kanoh et al. (14). Scatchard plots of the data showed that the NTCB fragment bound to actin filaments maximally at 0.2 mol/mol actin with a single K a of 4.8 ϫ 10 5 M Ϫ1 . In the presence of 1 mM Ca 2ϩ and CaM at a 9-fold molar excess over MLCK, we detected no binding activity of the NTCB fragment to actin filaments. We consider that the actin binding of the fragment is very similar to that of the NN fragment, i.e. Ca 2ϩ /CaM-sensitive, although we are aware that the K a of the NTCB fragment is lower than that of the Ca 2ϩ /CaM-sensitive site of parent MLCK (K a ϭ 2.16 ϫ 10 6 M Ϫ1 ). The difference is discussed in more detail under "Discussion." Fig. 4 shows the competition experiment for synthetic peptides against the actin binding of the NTCB fragment. Resembling the result of the NN fragment (Fig. 1c), neither peptide Met 1 -Gly 25 nor peptide Pro 26 -Pro 41 was able to abolish the actin binding of the NTCB fragment. The abolition was brought about only by peptide Met 1 -Pro 41 , confirming the idea that the boundary of the Ca 2ϩ /CaM-sensitive actin-binding site is probably around Pro 41 .
The inhibitory effects of MLCK and the NTCB fragment on the actin-myosin interaction are shown in Fig. 5. MLCK effectively inhibited the motility of actin filaments in the presence of EGTA. However, there was no inhibition in the presence of Ca 2ϩ /CaM. Similar inhibition and its relief by Ca 2ϩ /CaM were observed for the NTCB fragment, confirming the results ob-tained with the NN fragment (Fig. 3). The higher concentration of the NTCB fragment compared with that of parent MLCK (Fig. 5, compare a and b) is attributable to their different affinities to actin filaments.
Taken together with the CaM binding activity of the NTCB fragment (see Fig. 3, asterisks), the above data for the NTCB fragment confirm those for the NN fragment. We consider that the NN fragment is a recombinant form of the NTCB fragment and that it contains the whole sequence (Met 1 -Lys 114 ) of the NTCB fragment (see Fig. 6).

DISCUSSION
This report analyzes two classes of actin-binding site of MLCK, i.e. Ca 2ϩ /CaM-sensitive and Ca 2ϩ /CaM-insensitive sites as shown in Table I. The former is involved in the inhibitory effect of MLCK on the actin-myosin interaction. The latter site has no regulatory activity.
The inhibitory effect of MLCK on the actin-myosin interaction is brought about by its actin binding activity. This actinlinked nature was first demonstrated with a Nitella-based motility assay (see Ref. 25 for the method), where we allowed MLCK to bind to the actin cables that run on the inner surface of the plasma membrane of Nitella internodal cells (6). With the myosin-coated motility assay used in the present study, Sato et al. (8) suggested the actin-linked nature of the effect after observing that the inhibitory effect was obscured with the increase in the concentrations of actin filaments. The present study (Table I) demonstrates the actin-linked nature of the effect by using actin-binding fragments of MLCK which were produced (i) by purifying the NTCB fragment with conventional protein chemistry and (ii) by producing recombinant fragments with the pET expression system. MLCK, with respect to its actin-linked regulatory role as follows. The Ca 2ϩ /CaM-sensitive, actin-binding site is localized at the extreme N terminus of MLCK, consisting of Met 1 -Pro 41 (Fig. 6, a and b). The localization was determined (i) by comparing actin binding activity of the recombinant NN fragment containing only the Ca 2ϩ /CaM-sensitive site with that of the NN/41 fragment, which is devoid of the 1-41 sequence of the NN fragment (Fig. 1a); and (ii) by using a synthetic peptide of Met 1 -Pro 41 to compete against the actin binding of the native NTCB fragment (Fig. 4) and its recombinant form, the NN fragment (Fig. 1c).
Calponin (10) and caldesmon (9) are actin-binding proteins in smooth muscle that inhibit the actin-myosin interaction in a similar way to MLCK. Recent biochemical studies have narrowed the length of amino acid sequence responsible for actin binding activity down to 37 amino acids for calponin (33), and 32 and 46 amino acids for caldesmon (34). In agreement with these values, a length of 41 amino acids was shown to be a Ca 2ϩ /CaM-sensitive, actin-binding site. This was demonstrated as follows. (i) When the peptide of Met 1 -Pro 41 was split into Met 1 -Gly 25 and Pro 26 -Pro 41 , both peptides lost the antagonism for the actin binding of the NTCB (Fig. 4) and NN fragments (Fig. 1c). (ii) The extension of fragment length from the NN/41 fragment to the NN/25 fragment failed to restore full actin binding activity (Fig. 1a). This weak actin-binding ability of the NN/25 fragment suggests the possibility that some residues within the 1-41 sequence may not be involved in the actin binding.
It must be noted that the calculated pI of the sequence of Met 1 -Pro 41 is 10.67. Because actin is a highly acidic protein, we wonder whether the alkaline property of the sequence could cause MLCK to bind to actin nonspecifically. As shown in Fig  4, the peptide of the Lys 42 -Ala 80 sequence did not affect the ability of the NTCB fragment to bind to actin filaments, although its pI was similarly alkaline (pI ϭ 10.51). Therefore, the interaction between MLCK and actin filaments at the site of Met 1 -Pro 41 is not attributable to the nonspecific binding between the acidic and alkaline sequences. We searched for homology of Met 1 -Pro 41 with other actin-binding proteins and found sequences that show a 35% identity within the ␣-actinin sequence, a 32.3% identity within the dystrophin sequence, a 30.0% identity within the villin sequence, and a 28.1% identity within the coronin sequence. Unlike the Ca 2ϩ /CaM-sensitive, actin-binding site, the position of the Ca 2ϩ /CaM-insensitive, actin-binding site was not determined precisely (asterisk in Fig. 6a). What we can conclude is that the Ca 2ϩ /CaM-insensitive site is included in the NC fragment as shown in Fig. 6. The sequence of Pro 319 -Lys 337 in the NC fragment overlaps with the NN fragment and is obviously devoid of the Ca 2ϩ /CaM-insensitive site (compare the NC fragment with the NN fragment in Fig. 6b). Therefore, the Ca 2ϩ /CaM-insensitive site must be present somewhere in the Gly 338 -Val 721 sequence of bovine stomach MLCK.
Is there any other sequence in MLCK that binds to actin filaments in a Ca 2ϩ /CaM-sensitive manner? The NN/41 and NC fragments cover the N-terminal portion of MLCK except for the 1-41 sequence. Neither the NN/41 fragment nor the NC fragment showed Ca 2ϩ /CaM-sensitive binding to actin filaments (Table I, Fig. 1). We obtained the central kinase domain by proteolyzing MLCK, 2 and the C-terminal myosin-binding domain by purifying telokin (32). Because these domains failed to show actin binding activity, 2 the 1-41 sequence is the sole sequence responsible for the Ca 2ϩ /CaM-sensitive binding of MLCK that allows it to exert its regulatory role in the actinmyosin interaction.
Although parent MLCK shares the 1-41 sequence with the native NTCB fragment and the recombinant NN fragment, there is a difference in K a values. When the Ca 2ϩ /CaM-dependent site was separated from parent MLCK (K a ϭ 2.16 ϫ 10 6 M Ϫ1 ) as the NTCB fragment, its K a was reduced to 4.8 ϫ 10 5 M Ϫ1 . The K a of the NN fragment, a recombinant form of the NTCB fragment, is also low, i.e. 3.72 ϫ 10 5 M Ϫ1 . Such a reduction in the affinity upon cleavage from a parent molecule has also been noticed for the actin binding activity of the ␣-actinin family (35). On the other hand, affinity to actin filaments of the Ca 2ϩ /CaM-insensitive site is affected only slightly upon its separation from the parent molecule. The NC fragment, which carries only this site, binds to actin filaments with a K a ϭ 1.83 ϫ 10 5 M Ϫ1 (Fig. 1a), an affinity that is comparable with the K a values of the Ca 2ϩ /CaM-insensitive sites of MLCK (K a ϭ 4.67 ϫ 10 5 M Ϫ1 as measured in the absence of the Ca 2ϩ /CaM, and K a ϭ 3.1 ϫ 10 5 M Ϫ1 as measured in the presence of Ca 2ϩ /CaM).
According to Dabrowska et al. (3), the concentrations of MLCK and actin filaments in chicken gizzard smooth muscle cells are estimated to be 3 M and 830 M, respectively, and MLCK in the cells is adequately absorbed by the actin filaments. Our estimation of affinity for actin of the Ca 2ϩ /CaMsensitive site of MLCK explains their data, indicating that 0.36% of actin filaments in the cells are in the MLCK-bound form. The concentration of MLCK that gives half-maximal inhibition of the motility of actin filaments is 2.1 nM (Fig. 5a). Using a 3 nM concentration of actin filaments in the motility assay, we calculate that 0.42% of actin filaments would be in the MLCK-bound form. A comparable figure has also been calculated from the concentration of MLCK that gives halfmaximal inhibition of the actin-activated ATPase (see Fig. 2 Fig. 1, a and b. Panel b,  Ref. 6). To establish the physiological role of the actin binding activity of MLCK (36), it remains to be demonstrated whether or not such a small percentage of actin filaments in smooth muscle cells could participate in regulating their contraction. Alternatively, the in vivo binding of MLCK to actin filaments may serve not to regulate the actin-myosin interaction per se but simply to localize the kinase activity to the vicinity of its substrate.
Actin binding of calponin (10) and caldesmon (9) is regulated by Ca 2ϩ /CaM in a similar way to MLCK (see Fig. 1). The sites that bind CaM in calponin (10) and caldesmon (9) are located close to the sites for their actin binding. Therefore, the CaMbinding site that regulates the actin binding of MLCK would be expected to be close to the actin-binding site. Accordingly, we demonstrated that the CaM-binding site is included in the actin-binding site, i.e. Met 1 -Pro 41 sequence, by showing (i) that the NN/41 fragment that is devoid of the Met 1 -Pro 41 sequence fails to bind CaM (Fig. 2a) and (ii) that the Met 1 -Pro 41 peptide competes against the NN fragment for CaM binding (Fig. 2b). Further, when we divided the 1-41 peptide into two peptides of Met 1 -Gly 25 and Pro 26 -Pro 41 , the sequence that competes for CaM binding is the latter (Fig. 2b). The calculated pI values of the former (pI ϭ 10.25) and the latter (pI ϭ 10.82) are similar. Therefore, the failure of the former peptide to bind to CaM suggests that the interaction between CaM and the latter peptide was not caused by the nonspecific binding of an alkaline peptide to an acidic protein such as CaM. As shown in Fig. 7, the IQ motif, a consensus sequence in CaM-binding proteins (37), can be assigned to the latter peptide, although the matching was not complete.
MLCK has another CaM-binding site for regulating its kinase activity; the site is the Ala 796 -Ser 815 sequence for chicken gizzard MLCK (18) and the Ala 1002 -Ser 1021 sequence for bovine stomach MLCK (23) (see Fig. 6a for their topology). Does peptide Pro 26 -Pro 41 interact with the 796 -815/1002-1021 sequence? We confirmed that this sequence is a CaM-binding site as follows. Peptide Ser 787 -Ser 815 and peptide Ala 796 -Ser 834 , both of which contain the 796 -815/1002-1021 sequence, effectively abolished phosphorylation of myosin by MLCK in the presence of Ca 2ϩ /CaM (Fig. 8b, open circles and triangles). However, peptide Pro 26 -Pro 41 did not affect this kinase activity of MLCK (Fig. 8b, filled circles), demonstrating that the amino acid sequence of MLCK that binds to CaM regulate its kinase activity is totally different from the one that regulates its actin binding activity.
Alternatively, we allowed Ca 2ϩ /CaM to antagonize the actin binding of parent MLCK in the absence and presence of peptide Pro 26 -Pro 41 (Fig. 8a). In its absence, the actin binding of parent MLCK was antagonized by Ca 2ϩ /CaM as shown by filled circles in Fig. 8a. When the peptide was present, such an antagonism by Ca 2ϩ /CaM was much weakened (Fig. 8a, open circles), confirming that the 26 -41 sequence works as a CaM-binding site for regulating actin binding activity of MLCK. On the other hand, peptide Ser 787 -Ser 815 (Fig. 8a, filled triangles) exerted no effect. Thus, we concluded that parent MLCK has at least two classes of CaM-binding sites, which rules out the possibility that the CaM binding activity of the 26 -41 sequence was produced upon fragmentation of MLCK.