Phosphorylation of Caldesmon by p21-activated Kinase

We have previously shown that p21-activated kinase, PAK, induces Ca2+-independent contraction of Triton-skinned smooth muscle with concomitant increase in phosphorylation of caldesmon and desmin but not myosin-regulatory light chain (Van Eyk, J. E., Arrell, D. K., Foster, D. B., Strauss, J. D., Heinonen, T. Y., Furmaniak-Kazmierczak, E., Cote, G. P., and Mak, A. S. (1998) J. Biol. Chem. 273, 23433–23439). In this study, we provide biochemical evidence implicating a role for PAK in Ca2+-independent contraction of smooth muscle via phosphorylation of caldesmon. Mass spectroscopy data show that stoichiometric phosphorylation occurs at Ser657 and Ser687 abutting the calmodulin-binding sites A and B of chicken gizzard caldesmon, respectively. Phosphorylation of Ser657 and Ser687 has an important functional impact on caldesmon. PAK-phosphorylation reduces binding of caldesmon to calmodulin by about 10-fold whereas binding of calmodulin to caldesmon partially inhibits PAK phosphorylation. Phosphorylated caldesmon displays a modest reduction in affinity for actin-tropomyosin but is significantly less effective in inhibiting actin-activated S1 ATPase activity in the presence of tropomyosin. We conclude that PAK-phosphorylation of caldesmon at the calmodulin-binding sites modulates caldesmon inhibition of actin-myosin ATPase activity and may, in concert with the actions of Rho-kinase, contribute to the regulation of Ca2+sensitivity of smooth muscle contraction.

modeling processes that accompany lamellapodia and filopodia formation in many types of non-muscle cells (6). One of the key downstream effectors of Rac1 and cdc42 is the Ser/Thr kinase, PAK (7). We have shown previously that infusion of Tritonskinned guinea pig Taenia coli smooth muscle fibers with constitutively active PAK3 induces Ca 2ϩ -independent contraction to about 60% of the force obtained in the presence of Ca 2ϩ / calmodulin (8). PAK-induced contraction is accompanied by an increase in the phosphorylation of caldesmon and desmin but not MLC even though PAK is able to phosphorylate MLC at Ser 19 in vitro (8,9). These results suggest that Rac and cdc42, in contrast to Rho, induce smooth muscle contraction by altering the properties of actin and/or intermediate filaments.
Smooth muscle caldesmon is speculated to be a thin filament regulatory protein by virtue of its ability to inhibit actin-tropomyosin activated ATPase activity of myosin (10 -12). It is an 89-kDa protein that binds in an extended conformation along filaments of actin-tropomyosin. It houses binding sites for myosin (13), tropomyosin (14), calmodulin (15,41), and actin (16). In vitro, caldesmon-mediated inhibition of actomyosin ATPase activity can be regulated by calcium binding proteins, including calmodulin (17,18) or caltropin (19).
Our finding that PAK induces Ca 2ϩ -independent contraction in skinned smooth muscle fibers (8) and the recent demonstration that an unknown kinase besides MAPK phosphorylates gizzard caldesmon in vivo (20) suggest that PAK phosphorylation of caldesmon may be involved in the regulation of Ca 2ϩ sensitivity of smooth muscle contraction. Here, we report biochemical evidence supporting a role of PAK in the Ca 2ϩ sensitivity of smooth muscle contraction via phosphorylation of caldesmon. Specifically, we have identified the sites of phosphorylation and have studied phosphorylated caldesmon with respect to its affinity for actin-tropomyosin and calmodulin and its ability to inhibit actomyosin ATPase activity.

MATERIALS AND METHODS
Protein Preparation-h-Caldesmon and ␣ ␤ tropomyosin were purified from chicken gizzards essentially as described by Bretscher (21). Skeletal muscle actin was purified from rabbit muscle as outlined in (22). Smooth muscle myosin S1 was prepared by papain cleavage of gizzard myosin (23). Rabbit skeletal myosin S1(A1) was prepared by cleavage with chymotrypsin as described in (24). Recombinant murine PAK3, was expressed from the plasmid pGST-mPAK3 in Escherichia coli JM101 and/or JM110 cells as described before (8).
Approximately 200 g of caldesmon was dissolved in 200 l of 100 mM NH 4 HCO 3 , pH 7.9, containing 10 g of endoproteinase Glu-C. The digestion was carried out overnight at room temperature. The digest solutions were evaporated to dryness and redissolved in 5% acetic acid  Phosphopeptides were detected by precursor ion scanning (precursors of m/z 79) in negative ion mode on a API 3000 triple quadrupole mass spectrometer (Perkin Elmer/SCIEX Concord, ON, Canada). Precursor ion spectra were acquired in multiple channel acquisition mode, typically over a period of 3 min (m/z 400 -2000, 0.5 mass units step size, 5 msec dwell time). Argon was used as the collision gas, and the collision offset voltage was 80 V. Phosphopeptide sequencing was achieved by MS/MS using a prototype quadruple time-of-flight mass spectrometer (QqTOFMS, Perkin Elmer/SCIEX) equipped with a nanoelectrospray ionization source. Product ion spectra were carried out in positive ion mode using argon as the collision gas and a collision energy of 60 eV (laboratory frame of reference). MS/MS spectra were typically acquired every 2 s over a period of 3 min.
Calmodulin-Caldesmon Interaction-Interaction between calmodulin and phosphorylated and nonphosphorylated caldesmon was studied using intrinsic Trp fluorescence as described previously (15). The binding buffer was 20 mM Tris-HCl, pH 7.2, 0.5 mM CaCl 2 , 100 mM NaCl, 1 mM DTT. The excitation wavelength was 295 nm with a slit width of 10 nm. Intensity measurement was made with a 290-nm filter at 330 nm and a slit width of 10 nm. Binding curves were fitted to a binding equation to obtain dissociation constants as described before (15) except that binding stoichiometry was set to 1 mol of caldesmon per mol of calmodulin as previously determined (15).
Actin-binding Assays-Two nmol of actin was mixed with 0.4 nmol of tropomyosin and 0 to 1.5 nmol of phosphorylated or nonphosphorylated caldesmon in 200 l of binding buffer (40 mM Tris, pH 7.5, 100 mM NaCl, 5 mM MgCl 2 , and 1 mM DTT). Protein mixtures were allowed to equilibrate for 30 min prior to centrifugation at 100,000 ϫ g in a Beckman TL-100 ultracentrifuge. Pellets were rinsed with actin-binding buffer once and dissolved in 100 l of 0.05% (v/v) TFA in water. Relative amounts of caldesmon, tropomyosin, and actin were determined by high performance liquid chromatography using a Zorbax SB300-C8 HPLC column (5).
Actin-activated S1-ATPase Assays-Actin-activated myosin S1 ATPase assays in the presence and absence of tropomyosin were conducted in a 96-well ELISA plate (100 l assay volume). Inorganic phosphate was determined colorimetrically as described in (38). ATPase buffer contained 40 mM Tris, pH 7.5, 50 mM NaCl, 5 mM MgCl 2, and 1 mM DTT. Myosin S1 ATPase activity was determined for reaction mixtures containing 0 -4 M caldesmon, 10 M actin, and 0.5 M S1 with or without 2 M tropomyosin for 10 min at 37°C. ATPase reaction was initiated with 4 mM ATP and terminated by adding 100 l of a solution containing 3% ascorbic acid, 0.5 M HCl, 4% SDS, and 0.5% (w/v) ammonium molybdate. Color was allowed to develop for 6 min prior to the addition of 2% sodium citrate and 2% sodium m-arsenite followed by 10-min incubation before absorbance at 650 nm was measured in a Molecular Dynamics E-max plate reader. Phosphate content was determined by comparison to a potassium phosphate standard curve. Less than 10% of the ATP was hydrolyzed over the course of the reaction. Phosphate release was linear within the 10-min reaction.

RESULTS
Chicken gizzard h-caldesmon was phosphorylated in vitro using a constitutively active murine GST-PAK3 to a maximum of 2 mol of phosphate per mol of caldesmon as shown in Fig. 1. Only phosphorylated Ser was recovered from a hydrolysate of 32 P-labeled caledesmon (Fig. 1, inset).
To locate the phosphorylation sites, caldesmon phosphorylated by GST-PAK to 2 mol of phosphate/mol of protein was subjected to digestion by endoproteinase Glu-C, which cleaves peptide bonds on the COOH-side of Glu residues. The resulting digest was analyzed for phosphorylated peptides by precursor ion scanning as described in "Materials and Methods." Two major doubly deprotonated ions were detected ( Fig. 2A, peaks a  and b) together with a number of minor peaks representing minor sites of phosphorylation. MS/MS sequence analysis determined that peak a (m/z 648.0) and peak b (m/z 754.0) correspond to two singly phosphorylated peptides, Gly 651 -Val-Arg-Asn-Ile-Lys-p-Ser-Met-Trp-Glu 660 and Thr 678 -Ala-Gly-Leu-Lys-Val-Gly-Val-Ser-p-Ser-Arg-Ile-Asn-Lys-Glu 691 , A and B, respectively (data not shown). Ser 657 , being the only Ser in peptide A, can be unambiguously assigned as the site of phosphorylation in this peptide. There are two adjacent Ser residues in peptide B however. MS/MS sequence analysis indicated that Ser 687 is most likely the site of phosphorylation (Fig. 2C) because a b-type fragment ion at m/z 813.5 corresponding to the unphosphorylated peptide, Thr-Ala-Gly-Leu-Lys-Val-Gly-Val-Ser 686 , was detected (data not shown). The nonphosphorylated counterparts of peptides A and B were not detected, indicating that Ser 657 and Ser 687 were fully phosphorylated, largely accounting for the observed stoichiometry of 2 mol of phosphate/ mol of protein (Fig. 1). The precursor ion scan of the endoproteinase Glu-C digest of unphosphorylated caldesmon shows no trace of peak a (m/z 648.0) or peak b (m/z 754.0), indicating that Ser 657 and Ser 687 are genuine PAK-target sites (Fig. 2B). Background peaks at m/z 529, 543, and 558, which amount to less than 12% of peak b, were detected in the unphosphorylated  (15,16) in subdomain 4. We therefore examined whether Ser 657 and Ser 687 were accessible to PAK when caldesmon formed a complex with Ca 2ϩ /calmodulin. GST-PAK3-phosphorylated caldesmon-calmodulin complex at a similar initial rate but reach a stoichiometry of 1.2 mol of phosphate/mol of protein (Fig. 1). Calmodulin was not phosphorylated by PAK, and Ca 2ϩ did not affect PAK activity under the same conditions (data not shown). This result suggests that binding of calmodulin to sites A and B of caldesmon renders Ser 657 and/or Ser 687 less accessible to PAK.
To determine whether introduction of phosphate groups to Ser 657 and Ser 687 at the calmodulin-binding sites A and B can affect calmodulin-binding, we compared binding of phosphorylated and nonphosphorylated caldesmon to calmodulin using intrinsic Trp fluorescence measurements (Fig. 3). Phosphorylated and nonphosphorylated caldesmon have similar fluorescence spectra, each exhibiting a similar emission maximum at 350 nm, which suggests that the phosphate groups do not cause significant changes in the environments surrounding the Trp residues which are major determinants for calmodulin-binding (15). Binding of Ca 2ϩ -calmodulin, which contains no Trp, increased the intrinsic Trp fluorescence of caldesmon by a maximum of about 70% (Fig. 3B) and caused a blue shift of the emission maximum from 350 to 340 nm (Fig. 3A). On the other hand, calmodulin increases the fluorescence intensity by less than 40% at saturation accompanied by a smaller shift in emission maximum from 350 to 345 nm (Fig. 3B). As shown in  We compared the ability of caldesmon and its phosphorylated counterpart to interact with actin-tropomyosin and to inhibit actin-activated myosin ATPase activity because the calmodulin-binding site A has been shown to bind actin (27), and site B is in the middle of the tropomyosin-linked actin-binding and inhibitory region in subdomain 4 (28). As shown in Fig. 4, caldesmon phosphorylated to 2 mol of phosphate/mol of protein has a modest reduction in affinity for actin-tropomyosin; K d was increased by less than 2-fold from 1.0 to 1.7 M. However, phosphorylation of caldesmon induces a significant release of inhibition of actin-S1 ATPase (Fig. 5) in the presence or absence of tropomyosin. At 0.2 mol/mol of caldesmon/actin, nonphosphorylated caldesmon inhibits actin-activated skeletal myosin S1 ATPase activity by 80% in the presence of tropomyosin, whereas about 40% inhibition was observed by the same amount of phosphorylated caldesmon (Fig. 5B). Similar results were obtained using smooth muscle S1 (data not shown). In the absence of tropomyosin (Fig. 5A), caldesmon is much less effective in inhibition as reported by others (28,29); 0.4 -0.5 mol/ mol of nonphosphorylated caldesmon/actin is required to cause a 40% inhibition whereas similar amounts of phosphorylated caldesmon inhibit by 20%. DISCUSSION This study provides biochemical evidence to support the hypothesis that phosphorylation of caldesmon by PAK may play a role in inducing Ca 2ϩ -independent contraction in smooth muscle. The strategic location of Ser 657 and Ser 687 in the calmodulin-binding sites A and B provides a crucial clue to our under- phosphorylate caldesmon in vitro. The sequences surrounding Ser 657 (Arg-Asn-Ile-Lys-Ser 657 -Met-Trp-Glu) and Ser 687 (Lys-Val-Gly-Val-Ser-Ser 687 -Arg-Ile-Asn) in caldesmon, and Ser 19 (Gln-Arg-Ala-Thr-Ser 19 -Asn-Val-Phe) in MLC have a hydrophobic residue in the ϩ2 position which agrees with Brzeska et al. (39) who showed that a Tyr at the ϩ2 position is strongly preferred by PAK1. As well, seven of the eight autophosphorylation sites in PAK1 have a hydrophobic residue in position ϩ2 (40). However, Tuazon et al. (34), using a series of synthetic peptide substrates, identified the signature determinants for PAK1 phosphorylation as KRES, which bears little resemblance to the caldesmon and MLC phosphorylation sites except for the presence of a basic residue between positions Ϫ1 and Ϫ5. It appears, therefore, secondary structures and a hydrophobic amino acid at the ϩ2 position are equally important determinants for PAK recognition.
Not unexpected, we found that phosphorylation of Ser 657 and Ser 687 interfered with interaction between calmodulin and caldesmon. We have shown previously that although Trp 659 and Trp 692 in sites A and B, respectively, are major determinants for caldesmon-calmodulin interaction, though amino acid residues surrounding the Trp residues also contribute to optimal binding (15). NMR data showed that sites A and B in synthetic peptides simultaneously bind to the two hydrophobic regions of calmodulin affecting all eight Met residues in the "Met puddles" (35) and become ␣-helical upon binding to calmodulin (36,37). Furthermore, the helix formed by site A is amphiphilic such that Ser 657 is located on the polar surface (36). Introduction of phosphate groups at these sites likely interferes with the contacts between the polar surface of site A and calmodulin but should have a minor impact on hydrophobic interactions, as would be suggested by fluorescence data (Fig.  3A), which indicate that phosphorylation of caldesmon alone does not affect the environment surrounding Trp residues in sites A and B. This is also consistent with our finding that binding of calmodulin to sites A and B attenuates subsequent phosphorylation of caldesmon by PAK, indicating that Ser 657 and/or Ser 687 become less accessible to PAK (Fig. 1).
The actin-binding sites span an extended region in subdomain 4 of caldesmon (28). Introduction of two phosphates at Ser 657 and Ser 687 is unlikely to induce extensive disruption in the actin-binding regions, thus abrogating interaction. This may account for the modest reduction in affinity of phosphorylated caldesmon for actin-tropomyosin (Fig. 4), which might perhaps be because of a reorientation of the caldesmon molecule along the actin-tropomyosin filament toward forming a noninhibitory state. This interpretation is consistent with our finding that phosphorylation significantly reduces the ability of caldesmon to inhibit myosin S1-ATPase. A synthetic peptide spanning from Gly 651 to Ser 667 containing site A has been shown to bind actin and enhance contraction in saponintreated single hyper-permeable ferret aorta smooth muscle cells (27). Ser 687 and site B are situated in the midst of a tropomyosin-dependent actin-binding region, residues 669 -710, which is believed to be involved in tropomyosin-linked caldesmon inhibition of actomyosin ATPase activity (28). It is conceivable that phosphorylation of Ser 687 and Ser 657 is responsible for altering the tropomyosin-dependent and tropomyosin-independent inhibition, respectively.
As shown in Fig. 5B, PAK-phosphorylation attenuates (ϳ50%), but does not abolish, caldesmon inhibition of actin-TM-activated myosin S1 ATPase activity invoking a phosphorylation mechanism by which caldesmon function can be modulated independently of Ca 2ϩ /calmodulin. It appears that caldesmon can exist in a number of states endowed with different inhibitory activities depending on its phosphorylation status and binding to Ca 2ϩ /calmodulin. One of these states, which is generated by Ca 2ϩ -independent phosphorylation of Ser 657 and Ser 687 with PAK, possesses intermediate inhibitory activity compared with the fully inhibitory nonphosphorylated caldesmon and the noninhibitory Ca 2ϩ /calmodulin-caldesmon complex. It is possible that introduction of phosphate groups to the calmodulin-binding region may engender a simulacrum of the Ca 2ϩ -calmodulin-bound state of caldesmon, thus accounting for the partial reversion of inhibition. Formation of a complex between phosphorylated caldesmon and calmodulin, however, appears unfavorable in view of results showing that the affinity of phosphorylated caldesmon for Ca 2ϩ /calmodulin is reduced 10-fold and that Ser 657 and/or Ser 687 in the caldesmoncalmodulin complex are less accessible to PAK (Fig. 3). Taken together, data from this study and others suggest that Rac1cdc42 and Rho GTPase may act in concert to target thinand thick-filament, respectively, modulating Ca 2ϩ sensitivity of smooth muscle contraction.