Reversal of Caldesmon Binding to Myosin with Calcium-Calmodulin or by Phosphorylating Caldesmon*

Caldesmon, an actin-binding protein from smooth muscle and non-muscle cells, has previously been shown to bind stoichiometrically to smooth muscle my- osin in an ATP-dependent manner. We now show quan-titatively the effects of Caz’-calmodulin and phos- phorylation on the binding of caldesmon to myosin. Caz+-calmodulin reduces the binding of caldesmon to myosin with the same effectiveness as it does the binding of caldesmon to actin. However, Ca2’-calmodulin is ineffective in antagonizing the binding of the purified myosin-binding region of caldesmon to myosin. These and other results suggest that Ca2’-calmodulin binding to the COOH-terminal region of caldesmon is responsible for reversal of binding to myosin. Phos- phorylation of the NHz-terminal region of caldesmon by the co-purifying kinase, calmodulin-dependent pro- tein kinase 11, weakens but does not eliminate the binding of caldesmon to smooth muscle myosin. Fi-nally, phosphorylation of smooth muscle myosin by smooth muscle myosin light chain kinase has no effect on the binding of caldesmon to myosin. Since Ca2’- calmodulin and phosphorylation of caldesmon weaken the binding of caldesmon to both actin and myosin, these events may be coordinately regulated.

The NH2-terminal region of caldesmon is also interesting because it binds to myosin (Hemric and Chalovich, 1988; * This work was supported by Grant AR35216 from the National Institutes of Health (to J. M. C.) and Grant 1988-89-Al6 from American Heart Association, North Carolina Affiliate (to M. E. H.). A preliminary report was presented at the Biophysical Society Meeting, February 24, 1991, San Francisco, CA (Hemric et al., 1991. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Current address: Dept. of Physiology and Biophysics, University of Vermont, School of Medicine, Burlington, VT 05405.

2973.
ll To whom correspondence should be addressed.
Earlier studies have shown that both Ca2+-calmodulin (Ikebe and Reardon, 1988) and caldesmon phosphorylation (Sutherland and Walsh, 1989) alter the caldesmon-myosin interaction. However, these studies did not quantitate the effect of these modifications, nor did they localize the site of the effects. We now show that Ca2+-calmodulin binding to the COOH region of caldesmon reverses the binding of myosin to the amino-terminal region of caldesmon. Phosphorylation of sites within the amino-terminal region of caldesmon weakens the binding of caldesmon to myosin, and the effect is graded with the number of phosphates incorporated into that region of caldesmon.

MATERIALS AND METHODS
zards as described by Persechini and Hartshorne (1983) except that Proteins-Smooth muscle myosin was prepared from chicken giz-0.1% Triton X-100 was included in the first wash and myosin was extracted with 40 mM imidazole-HCI (pH 7.2), 8 mM ATP, 60 mM KCI, 1 mM EDTA, 100 mg/liter streptomycin sulfate, and 0.5 mM dithiothreitol. Caldesmon was isolated from turkey gizzards by a modification of the method of Bretscher (1984) described previously (Velaz et al., 1989). Calmodulin was purified from porcine brains by the method of Yazawa et al. (1980). Smooth muscle myosin light chain kinase was purified by the method of Walsh et al. (1983). Phosphorylation of smooth muscle myosin was performed as described by Adelstein and Klee (1982), and quantitation of light chain phosphorylation was done by scanning isoelectric focusing gels (Wells and Bagshaw, 1984). The [14C]iodoacetamide-labeled caldesmon was prepared as previously described (Velaz et al., 1989) but without cellulose phosphate chromatography. The myosin-binding fragment of caldesmon was produced as previously described (Velaz et al., 1990) with some modifications. Caldesmon was cleaved with a 1:600 weight ' The abbreviations used are: S-1, myosin suhfragment-1; HPLC, high performance liquid chromatography; Mes, 4-morpholineethanesulfonic acid. 15305 ratio of chymotrypsin for 5 min at 25 "C in a buffer composed of 0.1 M KCl, 20 mM Tris-HC1 (pH 8), 1 mM EDTA, and 1 mM dithiothreitol. The digest was applied to a 1.5 X 95-cm Spectra/Gel AcA 54 gel filtration column equilibrated with digestion buffer brought to 1 M KCl. The myosin-and several actin-binding fragments in the 20-30-kDa molecular mass range were contained in the second peak. These myosin-and actin-binding fragments were subsequently dialyzed against 60 mM NaCI, 10 mM Tris-Mes (pH 7), 1 mM dithiothreitol and applied to a Waters Protein-Pak SP R (1 X 10 cm) cation exchange HPLC column equilibrated with dialysis buffer . The fragments were eluted with a gradient of NaCl to 500 mM. A larger amino-terminal caldesmon fragment was prepared by digestion of caldesmon with submaxillary Arg protease (1:500, w/ W) for 90 min (Mornet et al., 1988). This fragment, which migrates at 90 kDa on SDS gels, was purified by gel filtration on Spectra/Gel AcA 54 followed by HPLC ion exchange chromatography on a Pharmacia Mono Q column.
Caldesmon Phosphorylation-Caldesmon and the co-purifying calmodulin-dependent protein kinase I1 were purified from turkey gizzards as previously described (Velaz et al., 1989) except, prior to heat treatment, extracted caldesmon was applied to a Cibricon Blue agarose (Bio-Rad) equilibrated with 0.5 M KCl, 25 mM Tris-HC1 (pH 8), 1 mM EGTA, 1 mM EDTA, and 1 mM dithiothreitol. The caldesmonkinase complex was eluted when the equilibration buffer was raised to 1 M KCI. Caldesmon was phosphorylated in 0.1 M NaCI, 20 mM Tris-HC1 (pH 7.5), 5 mM MgClZ, 1 mM dithiothreitol, 0.125 mM CaCI2, 0.6 PM calmodulin, and 1 mM [32P]ATP (1 mCi) at 25 "C for 90 min and then boiled for 5 min to terminate phosphorylation and inactivate any phosphatases. The phosphorylation was dependent on the presence of Ca2+-calmodulin. Under conditions giving 1.2 mol of Pi/mol of caldesmon, in the presence of Caz+-calmodulin, only 0.18 mol of Pi/mol of caldesmon was observed in the presence of EGTA and absence of added calmodulin. Phosphorylated caldesmon was dialyzed and applied to Sepharose G-50 gel filtration chromatography for final purification.
Affinity Chromatography-Smooth muscle myosin and calmodulin were covalently linked to CNBr-activated Sepharose 4B (10 mg of protein/ml of resin) according to the manufacturer's directions (Pharmacia LKB Biotechnology Inc.). Unreacted groups on the column were blocked with ethanolamine to minimize the introduction of charged groups.
Electrophoresis and Autoradiography-SDS-polyacrylamide gels were run by the procedure of Laemmli (1970) and stained with Coomassie Blue. Dried gels were exposed to Kodak BB Blue film at -70 "C using Lightning Plus intensifying screens (Du Pont-New England Nuclear).
Binding Studies-The binding of radioactively labeled caldesmon to smooth muscle myosin was determined by a low speed sedimentation assay described previously (Hemric and Chalovich, 1990). During modification, a fraction of caldesmon lost its ability to bind myosin and was determined to be 26% for ['4C]iodoacetamide-labeled caldesmon and 22.5% for [32P]caldesmon. The fraction of protein that sedimented in the absence of smooth muscle myosin was 3% for ["C] iodoacetamide-labeled caldesmon and 4.5% for [32P]caldesmon. Finally, the fraction of smooth muscle myosin that did not sediment was 1%.
Mathematical Modeling-Two models of the binding of caldesmon with myosin and calmodulin were considered. In the first model, calmodulin binds to a single site on caldesmon (Kz), which is located in the COOH-terminal region of caldesmon and which precludes myosin subfragment binding (Kl). In the second model, calmodulin binds to a site at the NHz-terminal region of caldesmon (Kz), which precludes myosin binding at K,, and to a second site at the COOHterminal region (K3), which does not prevent the binding of myosin (see Table I). The observable reactants in the system are the bound forms of myosin (i.e. myosin with bound caldesmon with or without a maximum of 2 molecules of calmodulin). Theoretical values of the bound myosin with respect to changes in concentration of calmodulin (or caldesmon and calmodulin) were calculated using an iterative  Table I). The conditions were: 1 mM ATP, 30 mM NaCl, 5 mM MgClZ, 10 mM imidazole-HC1 (pH 7), 1 mM CaClZ, and 1 mM dithiothreitol, at 25 "C. Theoretical curves were drawn assuming one (solid line) or two (broken line) calmodulin binding sites on caldesmon.
numerical procedure. Data fitting was done using the MATLAB mathematical modeling program (The Mathworks, Inc.).'

RESULTS
The effect of calmodulin on the binding of caldesmon to myosin was determined by two experimental protocols. In the first, shown in Fig. lA, the amount of caldesmon bound to myosin was measured as a function of the total added calmodulin concentration at saturating Ca2+. In this type of experiment, the amount of caldesmon bound to myosin decreased to less than 4% of the initial value a t high Ca2+calmodulin concentrations. In the second protocol, a series of curves were generated where the caldesmon bound was measured as a function of the caldesmon concentration at constant ratios of calmodulin to caldesmon. Thus, the calmodulin and caldesmon were increased simultaneously. An example of this type of experiment is shown in Fig. 1B. Both types of experiments were analyzed to two models of calmodulin binding (see "Materials and Methods" and Table  I). The solid line shows the best fit to the simple competitive model in which a single Ca2+-calmodulin site is assumed. The broken line is the fit of the model with two calmodulin binding sites, one com- The program used to solve these equations, in the MATLAB language, may be obtained by writing to R. I. Shrager.

TABLE I
Estimated association constants for binding of caldesmon to both Ca2+-calmoduZin and myosin Several data sets were analyzed by two models. Sets 1-4 were measured with varied caldesmon and calmodulin concentrations at a constant ratio of calmodulin to caldesmon. The molar ratios of calmodulin to caldesmon were 0,0.25,0.75, and 1.3 for sets 1 through 4. Set 6 was measured as the concentration of calmodulin was increased at constant caldesmon and myosin concentrations. Association constants have units of M" and are defined by the following scheme where C is caldesmon, M is myosin, and X is Ca2+-calmodulin, Kz and K3 are calmodulin binding to the competitive and noncompetitive sites, respectively, as shown below. 2.0 X 1 0 5 2.6 x lo5 3, 4 9.6 X 1 0 4 4.9 X 105 Two calmodulin binding site model 2, 3 , 4 , 6 1.5 X 105 2.8 X lo5 3.8 x 10' 2, 6 1.9 X lo5 3.9 X io6 7.3 x 10' 3, 4 9.6 X 1 0 4 4.9 X 105 1.1 X 100 a Conditions were the same as those described in Fig. 1. petitive and the other noncompetitive for myosin binding. Both models can be fitted well to the data and give almost indistinguishable curves. More complex models, which incorporated a cooperative interaction between the presumptive noncompetitive calmodulin binding site and myosin, did not improve the fit (not shown). Both models allow estimations to be made for the association constants of myosin to caldesmon ( K J , calmodulin to the competitive site of caldesmon ( K 2 ) , and calmodulin to the noncompetitive site of caldesmon ( K3). Values obtained from fitting the different data sets are listed in Table I. In the single calmodulin binding site model, the affinity of caldesmon for both calmodulin and myosin is about the same (1-5 x 10' M-'). In this model caldesmon can bind to either myosin or calmodulin but not to both simultaneously. If, on the other hand, we assume that there are two calmodulin binding sites then the affinity of calmodulin to the site which is competitive for myosin binding (the NHz-site) is about 6-fold greater than for the site which is noncompetitive for myosin (the COOHsite) binding if we exclude data sets 3 and 4, which are not consistent with model 2.
It is not likely that binding of Cap+-calmodulin is nearly as strong at the NHp-site as at the COOH-site. We observed that only small amounts of the NHp-terminal, chymotryptic, myosin-binding fragment of caldesmon bound to calmodulin-Sepharose affinity columns in the presence of calcium. In contrast, all of the actin-binding fragments in the 20-kDa molecular mass range bound to the column (data not shown). We confirmed this result using the larger NHz-terminal fragments prepared from a submaxillary arginine-C protease digestion of caldesmon. This fragment begins at the NH2 terminus and ends around residue 482, where the 35-kDa CNBr fragment begins (Mornet et al., 1988). We observed, as did Mornet et al. (1988), that this fragment binds very weakly, if at all, to a calmodulin affinity column (data not shown). Therefore, if the NHz-region binds to calmodulin, it does so with an affinity much less than that of the COOH-terminal fragments. In contrast to the prediction of model 2, it is likely that Ca2+-calmodulin binding to the COOH-region of caldesmon controls binding of myosin to the NHz-terminal region of caldesmon.
We also examined this question more directly by measuring the effect of calmodulin on the binding of the purified myosinbinding fragment of caldesmon to myosin. While the purified myosin-binding fragment bound to smooth muscle myosin, as shown in Fig. 2 A , the binding was rather insensitive to calmodulin. Interestingly, however, if the NHz-terminal fragment was not separated from the other caldesmon fragments in the digest, its binding was reversed upon the addition of Caz+-calmodulin as shown in Fig. 2B. A component of the caldesmon molecule distinct from the NH2-region is required for reversal by Cap+-calmodulin, and it need not be covalently linked to the NH2-region. Adding the purified 35-kDa actinand calmodulin-binding, COOH-terminal fragment to the purified myosin-binding fragment was not sufficient for restoring Cap+-calmodulin sensitivity (data not shown).
Phosphorylation of caldesmon by its co-purifying calmodulin-dependent protein kinase I1 has been shown to eliminate its binding to a smooth muscle myosin-Sepharose affinity column (Sutherland and Walsh, 1989). Fig. 3 shows a similar affinity column experiment using caldesmon phosphorylated with [32P]ATP. Both the absorbance and radioactivity of the effluent are shown. The first peak of phosphorylated caldesmon did not interact with the column even under low ionic strength conditions. This caldesmon contained an average of 3 mol of phosphate/mol of caldesmon. A second population of phosphorylated caldesmon was retained by the column and eluted at 0.4 M NaCl (unphosphorylated caldesmon elutes at 0.5 M NaC1). The caldesmon retained by the column contained an average of 2 mol of phosphate/mol of caldesmon. Therefore, the binding became very weak only with the incorporation of 3 mol of Pi/mol of caldesmon.
Direct binding of the phosphorylated caldesmon to smooth muscle myosin was measured in the absence of ATP and the results shown in Fig. 4. Under these conditions, the affinity of phosphorylated caldesmon for myosin is quite low and saturation of binding could not be achieved. However, a reasonable fit was obtained assuming that the stoichiometry of binding was unchanged by caldesmon phosphorylation (2 mol of caldesmon/mol of myosin). This is reasonable since the stoichiometry is apparently dependent only on the myosin conformation (Lu and Chalovich, 1993). The association constant was reduced to 34% (2 mol of Pi/mol of caldesmon) and 17% (3 mol of Pi/mol of caldesmon) compared to unphosphorylated caldesmon.
Higher levels of saturation of binding could be obtained in the presence of ATP where the binding of caldesmon to myosin is enhanced. Fig. 5 shows that incorporation of 1.4 mol of Pi reduces the association constant to 24%, whereas the presence of 3.1 mol of Pi reduces the association constant to 4% of the initial value. In both cases, the best fit to the data occurs with a stoichiometry within experimental error of that measured with zero phosphorylation (1:l). Thus, the association constant of caldesmon for myosin decreases with increasing levels of caldesmon phosphorylation.
Even though calmodulin-dependent protein kinase I1 phosphorylates the NH2 terminus of caldesmon preferentially, some COOH-terminal sites are also phosphorylated. Ikebe and Reardon (1990) showed that with a total phosphorylation of 2.8, the distribution of 32Pi was 28.6% in Ser-26, 36.4% in Ser-73, 12.5% in Ser-726, and the rest in other sites. Thus, even with 3 total phosphates in caldesmon, the two NH2- terminal sites are not totally phosphorylated. To demonstrate that phosphorylation of the NHz-terminal sites of caldesmon is responsible for weakening of its binding to myosin, the binding of cleaved caldesmon to myosin was studied. Fig. 6 shows overloaded SDS-polyacrylamide gels of the supernatants and pellets formed after sedimenting cleaved :'2Pi-labeled caldesmon with myosin. By comparing the Coomassie Bluestained gels with the autoradiographs, it can be seen that the actin-binding fragments (ABF) were not significantly phosphorylated in this experiment. With only 2 mol of "P incorporated, only a small amount of phosphorylation in the actin binding regions is expected. In addition, phosphorylation a t Ser-726 may be underestimated due to the production of small polypeptides during chymotrypsin digestion. In contrast, the upper band of the myosin-binding fragments (MBF) was heavily phosphorylated (an unidentified component of intact caldesmon is highly phosphorylated and its digestion products are visible in lane b'). It is noteworthy that, even in this heavily loaded gel, none of the phosphorylated myosin-binding fragment is observed to co-sediment with myosin. Quantitation of this result was done by measuring the radioactivity remaining in the supernatant following sedimentation with myosin. While 14% of the nonphosphorylated myosin-binding fragment, in caldesmon digests, bound to myosin, only 2% of the phosphorylated myosin-binding fragment bound to myosin. Therefore, phosphorylation at the NHz-terminal region by calmodulin-dependent protein kinase I1 is responsible for the reversal of caldesmon binding to myosin.
Phosphorylation of smooth muscle myosin regulatory light chains does not have an effect on the affinity of myosin for  FIG. 4. Effect of caldesmon phosphorylation of the binding of caldesmon to smooth muscle myosin. Binding was measured in the presence of 10 p~ smooth muscle myosin and varied concentrations of caldesmon phosphorylated with ["PIATP. The conditions were the same as in Fig. 1, except that ATP and CaCI2 were excluded. The theoretical curves are for a single class of binding sites with a stoichiometry of 2 caldesmon molecules/myosin molecule and association constants of 5.6 X lo' actin (Sellers et al., 1982). We investigated whether light chain phosphorylation affected myosin binding to caldesmon. The binding of caldesmon to smooth muscle myosin phosphorylated to about 0.66 mol of phosphate/mol of subfragment-1 is shown in Fig. 7. The binding to phosphorylated myosin is fit reasonably well by the curve obtained from our earlier experiments of caldesmon binding to unphosphorylated smooth muscle myosin (Hemric and Chalovich, 1990). Thus, phosphorylation of smooth muscle myosin has no effect on its binding to caldesmon. tryptically cleaved, phosphorylated caldesmon to smooth muscle myosin-ATP. Caldesmon was phosphorylated to 2 mol of phosphate/mol of caldesmon and digested with chymotrypsin. 4 pM cleaved caldesmon was co-sedimented with 3.1 p~ smooth muscle myosin under the same conditions as in Fig. 1 except for the omission of CaCI2. Samples from the supernatant and pellet and standards of caldesmon and cleaved caldesmon were applied to an SDS 7-20% polyacrylamide gradient gel as follows: intact caldesmon ( a and a'), digested caldesmon (6 and b'), supernatant (c and c'), and pellet ( d and d ' ) . The gel was stained with Coomassie Blue ( a d ) and exposed to x- ray film (a'-d'). The positions of intact caldesmon (CaD), myosin-binding fragment ( M B F ) , and the actin-binding fragments (ABF) are shown. Note the presence of a low molecular weight polypeptide, which is highly labeled (lone a') but not visible in the Coomassie Blue-stained gel (lane a ) . Digestion of this polypeptide may contribute to the radioactivity seen in lanes b' and c' below the myosin-binding fragment.

DISCUSSION
Calmodulin inhibits the direct binding of caldesmon to myosin in a concentration-dependent manner. It is likely that the reversal of binding of caldesmon to both myosin and actin are controlled by the binding of Ca2+-calmodulin to the same site on caldesmon. Thus, the value of the binding constant for binding of Ca2+-calmodulin to the inhibitory site, K2, is 3-5 X 10% M-' from myosin competition data and 3.6 X lo6 M" from actin competition data (estimated from Velaz et al.

FIG. 7. Effect of myosin light
chain phosphorylation on the binding of caldesmon to smooth muscle myosin*ATP. Binding was measured in the presence of 2.5 ~I M smooth muscle myosin and varied concentrations of [14C]iodoacetamide-labeled caldesmon. The conditions were the same as in Fig.  1 except for the omission of CaCI2. The theoretical curve is the same as that calculated for the binding of caldesmon to unphosphorylated smooth muscle myosin (Hemric and Chalovich, 1990). The curve is calculated for a stoichiometry of 0.92 mol of caldesmon/mol of myosin and a binding constant of 9.9 X 10' "I. (1989)). In contrast, NHn-terminal myosin-binding fragments of caldesmon bind weakly, if at all, to calmodulin affinity columns (present study;Mornet et al., 1988). Actin-binding fragments, from the COOH-terminal region, bind tightly to such column under the same conditions. Furthermore, deletion mutants that encode for the myosin binding region (residues 1-239, see Hayashi et al. (1991)) or the myosin binding region plus the helical region (residues 1-458, see Hayashi et al. (1991); residues 1-578, see Redwood et al. (1990)) do not bind to calmodulin affinity columns. Although our modeling cannot exclude two binding sites for Can+calmodulin as suggested by others (Wang et al., 1989), it appears that the COOH-terminal Ca2+-calmodulin site regulates binding of caldesmon both to actin and to myosin.
If Ca2+-calmodulin binding to the COOH-terminal region of caldesmon affects myosin binding at the NHz-region of caldesmon, then it is likely that caldesmon is folded under some conditions. This is true because there is a large extended helical region separating these two domains (Wang et al., 1991) and direct communication between them is unlikely. Folding of caldesmon is known to occur since the thiol groups at the two ends of caldesmon can form a disulfide bond (Lynch et al., 1987).
Can+-calmodulin-dependent protein kinase 11, which copurifies with caldesmon (Abougou et al., 1989;Scott-Woo et al., 1990), was used in the present study. This kinase apparently binds very tightly to caldesmon, since it is separated from caldesmon only with great difficulty. It is also interesting that phosphorylation by this kinase decreases the binding of caldesmon to both actin (Ngai and Walsh, 1987) and to myosin (Sutherland and Walsh, 1989). Interestingly, serines 73 and 26, within the myosin binding region of caldesmon, are phosphorylated preferentially to Ser-726 and Ser-587 in the actin binding region (Ikebe and Reardon, 1990). Likewise, we find that the NH2-region of caldesmon is predominantly phosphorylated (Fig. 6). Phosphorylation of the NHZ-region of caldesmon is sufficient to reverse binding to myosin, since the phosphorylated myosin-binding fragment of caldesmon binds only weakly to myosin.
Phosphorylation of caldesmon with Ca''-calmodulin-dependent protein kinase I1 does not totally eliminate binding to myosin but produces a graded inhibition depending on the level of phosphorylation. The incorporation of 3 phosphates per caldesmon reduces the binding to constant to 4% of its initial value. Such a reduction in the affinity of caldesmon to myosin could alter the cross-linking of actin to myosin, which presumably is responsible for the increased interaction between actin and myosin under some conditions (Hemric and Haeberle, 1992;Walker et al., 1989;Hegmann et al., 1991;Lash et al., 1986;Hemric and Chalovich, 1988). Marston et al. (1992) have recently reported their results of the interaction of caldesmon with myosin. We are in agreement with their data in terms of the binding constant of caldesmon to smooth myosin and with the lack of effect of myosin phosphorylation on this interaction. However, differences exist between their observations with sheep aorta caldesmon and ours with turkey gizzard caldesmon. For example, they found that the stoichiometry of binding is 3 caldesmon molecules per myosin and that neither the stoichiometry nor the affinity are altered in the presence of ATP. Our present report (Figs. 4 and 5) supports our previous finding (Hemric and Chalovich, 1990) that the affinity of caldesmon for myosin increases in the presence of ATP. Furthermore, we observed the stoichiometry to be 1:1, in the presence of ATP, and 2 or 3 caldesmon molecules per myosin in the absence of ATP.
The different binding observed could be due to different sources of caldesmon. Marston et al., (1992) also reported that caldesmon does not bind to skeletal myosin or to the S-1 region of smooth myosin. In contrast, we have observed binding to skeletal heavy meromyosin Chalovich, 1988, 1990) and a weak interaction with the S-1 region of smooth and skeletal S-1 Chalovich, 1988 andFig. 6 of Hemrich and. At present, we do not know the reason for these different observations.