Different Molecular Mechanisms for Rho Family GTPase-dependent, Ca2+-independent Contraction of Smooth Muscle*

Abnormal smooth muscle contraction may contribute to diseases such as asthma and hypertension. Alterations to myosin light chain kinase or phosphatase change the phosphorylation level of the 20-kDa myosin regulatory light chain (MRLC), increasing Ca2+ sensitivity and basal tone. One Rho family GTPase-dependent kinase, Rho-associated kinase (ROK or p160ROCK) can induce Ca2+-independent contraction of Triton-skinned smooth muscle by phosphorylating MRLC and/or myosin light chain phosphatase. We show that another Rho family GTPase-dependent kinase, p21-activated protein kinase (PAK), induces Triton-skinned smooth muscle contracts independently of calcium to 62 ± 12% (n = 10) of the value observed in presence of calcium. Remarkably, PAK and ROK use different molecular mechanisms to achieve the Ca2+-independent contraction. Like ROK and myosin light chain kinase, PAK phosphorylates MRLC at serine 19 in vitro. However, PAK-induced contraction correlates with enhanced phosphorylation of caldesmon and desmin but not MRLC. The level of MRLC phosphorylation remains similar to that in relaxed muscle fibers (absence of GST-mPAK3 and calcium) even as the force induced by GST-mPAK3 increases from 26 to 70%. Thus, PAK uncouples force generation from MRLC phosphorylation. These data support a model of PAK-induced contraction in which myosin phosphorylation is at least complemented through regulation of thin filament proteins. Because ROK and PAK homologues are present in smooth muscle, they may work in parallel to regulate smooth muscle contraction.

In smooth muscle cells, elevation of [Ca 2ϩ ] i in response to electrical or chemical stimulation causes calmodulin to activate myosin light chain kinase (MLCK). 1 This highly specific serine/ threonine kinase phosphorylates the 20-kDa regulatory light chain of myosin (MRLC) at serine 19, thereby increasing the actin-activated ATPase activity of myosin and inducing contraction. MRLC is dephosphorylated by myosin light chain phosphatases (MLCP) resulting in muscle relaxation. The properties and activities of MLCK or MLCP can be modified by phosphorylation, providing a means to link smooth muscle contraction to other signaling pathways (1,2). Of particular interest is the recent finding that a Rho-associated kinase (ROK or p160 ROCK ) can induce Ca 2ϩ -independent contraction of smooth muscle (3) by directly phosphorylating MRLC on serine 19 (4) and by phosphorylating and inhibiting MLCP (5). The serine/threonine protein kinases PAK and ROK are activated through interactions with the Rho superfamily of Rasrelated low molecular weight GTPases (M r ϭ 21,000) (for reviews see Refs. 6 and 7). ROK specifically binds RhoA, whereas PAK (M r ϭ 62-68,000) associates with both Cdc42 and Rac but not RhoA.
Five closely related PAK isozymes have been identified in rat brain (8), human placenta and platelets (9), mouse fibroblast (10), and skeletal and vascular smooth muscle (9). In lower eukaryotes, PAK homologues include Ste20 in yeast (12) and the single-headed myosin I heavy chain kinases in Dictyostelium (13). PAK consists of two domains: an N-terminal regulatory domain that contains a Cdc42/Rac binding-domain and a C-terminal catalytic domain. The three mammalian PAK isoforms, PAK1, PAK2, and PAK3, share ϳ70% identity in overall amino acid sequence and over 90% identity within the kinase catalytic domain. Binding of GTP-Cdc42/Rac leads to autophosphorylation of PAK and activation toward exogenous substrates such as myelin basic protein (10,14,15). PAK and ROK have been implicated in alterations to the actin cytoskeleton during cell motility (for reviews see Ref. 6 and 7), suggesting that these kinases may have similar or overlapping modulatory roles.
There is substantial evidence linking activation and/or translocation of RhoA to Ca 2ϩ sensitivity of smooth muscle in some (16 -19) but not all situations (18,19). In addition, inactivation of RhoA (20) or ROK by a selective inhibitor (21) has been shown to correct hypertension in the spontaneous hypertensive rat model. However, the activation of RhoA and ROK cannot explain the increase in basal tone or Ca 2ϩ sensitization under all agonist-stimulated conditions (18,19), indicating that alternate signaling pathways are likely to be involved. Because the PAK kinases have been implicated in the control of motile events in nonmuscle cells (6,7), PAK is a potential modulator of smooth muscle contraction. In this manuscript it is shown that like ROK (GST-ROK), PAK (GST-mPAK3) causes Ca 2ϩindependent contraction of Triton-skinned smooth muscle fibers. However, PAK acts via a different molecular mechanism than ROK to induce Ca 2ϩ -independent force in smooth muscle fibers.

MATERIALS AND METHODS
Protein Preparations-Intact smooth muscle myosin, MRLC, MLCK, and caldesmon were purified from chicken gizzard (22)(23)(24), whereas * This work was funded by the Ontario Heart and Stroke Foundation and the Medical Research Council of Canada. 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.
¶ A Canadian Heart and Stroke Scholar. To whom correspondence should be addressed. Tel.: 613-545-6535; Fax: 613-545-6880; E-mail: JVE1@post.queensu.ca. 1 The abbreviations used are: MLCK, myosin light chain kinase; MRLC, 20-kDa myosin regulatory light chain; PAK, p21-activated protein kinase; MLCP, myosin light chain phosphatase; ROK, Rho-associated kinase; GST, glutathione S-transferase; PAGE, polyacrylamide gel electrophoresis; MES, 4-morpholineethanesulfonic acid; GTP␥S, guanosine 5Ј-O-(thiotriphosphate). PAK3 was isolated from rat brain (13). Recombinant caldesmon fragments (CaD39 and CaD40) were prepared and purified as described in Ref. 25. Recombinant Cdc42 and Rac1 were expressed and purified as described in Ref. 8. Plasmid pGST-mPAK3 carrying mouse fibroblast mPAK3 fused to GST in pGEX-KG was expressed and purified according to Ref. 8. Plasmid pDK-mPAK3 K297R carrying an inactive mPAK3 mutant with lysine amino acid residue 297 mutated to arginine (mPAK3 K297R cDNA) was fused to GST by subcloning a BamHI fragment of pDK-mPAK3 K297R into pGEX-4T3. 2 The mutation of lysine 297 in the ATP binding site of PAK is sufficient to render GST-mPAK3 K297R unable to phosphorylate isolated MRLC or myelin basic protein in vitro. The GST-mPAK3 fusion proteins were expressed in Escherichia coli and when tested together were simultaneously expressed, purified, and dialyzed against the same buffers. The catalytic subunit of recombinant ROK (Rho kinase) was expressed as a GST fusion protein in baculovirus (GST-ROK; Ref. 5). All recombinant kinases were purified on a glutathione-Sepharose affinity column (Amersham Pharmacia Biotech), concentrated in a Centriprep 30 (Amicon), and dialyzed against 10 mM imidazole, pH 7.0. As reported previously (10), GST-mPAK3 is susceptible to degradation, leading to different activities for each preparation. To ensure consistency, activities of the various GST-mPAK3 preparations were standardized against myelin basic protein phosphorylation. GST-mPAK was used in the skinned fiber assay within 4 days but could maintain sufficient activity for in vitro phosphorylation analysis if stored frozen in 50% glycerol. Various preparations of GST-mPAK3 (ϳ0.5 to 5 g/ml of active GST-mPAK3) were able to induce Ca 2ϩindependent contractions in skinned smooth muscle fibers ranging from 26.1 to 80.9% of Ca 2ϩ -dependent contraction.
Aliquots of the reaction mixture were analyzed for protein phosphorylation using Whatman P81 paper and SDS-PAGE/autoradiography as described previously (29). Phosphorylated amino acids were identified by thin-layer electrophoresis after hydrolysis of the phosphorylated proteins in 6 N hydrochloric acid as described previously (29 -31). Twodimensional tryptic peptide maps were produced as described previously (29 -31).
GST-mPAK3 and GST-ROK Phosphorylation in Skinned Muscle Fibers-Skinned fibers mounted on a U-shaped pin were incubated at 25°C under various conditions using the same conditions as in the skinned fiber assays (relaxing, contracting, and relaxing plus GST-mPAK or relaxing plus GST-ROK); except when required, assay buffers contained 1 mM [␥-32 P]ATP (0.25 mCi/ml) instead of 7.2 mM ATP. The GST-mPAK3 and GST-ROK concentrations induced ϳ70% of maximum Ca 2ϩ -dependent contraction. After 90 min of incubation, fibers were submerged in ice-cold 15% trichloroacetic acid and 2 mM inorganic phosphate followed by acetone to inactivate the kinases/phosphatases. This ensures preservation of the phosphorylation levels. Fibers were stored at Ϫ20°C until analysis.
Samples of radioactively labeled proteins from skinned fibers were resolved by 12.5% (see Fig. 4, A and B) and 10% (see Fig. 4, D and E) SDS-PAGE. Duplicate protein samples resolved simultaneously by 10% SDS-PAGE were transferred to two nitrocellulose blots, and one blot was probed with anti-caldesmon antibody (see Fig. 4, D and E), and the other blot was probed with anti-desmin antibody (see Fig. 4D; ROK data not shown) as described above. Alignment was accomplished by pin-holing the nitrocellulose and autoradiography film with [␥-32 P]ATP. Identification of proteins phosphorylated in the presence of either PAK (see Fig. 4D) or ROK (see Fig. 4E) was then carried out by exposure of the nitrocellulose blots to a 32 P PhosphorImager screen (Screen GP, Kodak). This allowed not only for the accurate alignment between Western blots and the subsequent autoradiograph results, but also for detection of radioactive emissions only, because the PhosphorImager screen does not detect chemiluminescent emissions. Phosphorimages of the blots were developed using the Storm PhosphorImager 820 (Molecular Dynamics, Sunnyvale, CA). This series of experiments has been performed using four different guinea pig t. coli skinned fiber preparations and two GST-mPAK3 preparations each yielding the identical pattern and alignments of autoradiography and Western blots.
Proteins from skinned fibers that were radioactively labeled in the presence of GST-mPAK3 as described above were also resolved by two-dimensional gel electrophoresis (see Fig. 4E) according to a standard protocol (Bulletin 1144) from Bio-Rad. Proteins were resolved in the first dimension by isoelectric focusing on a mini-Protean II isoelectric focusing gel electrophoresis apparatus (Bio-Rad) using an ampholyte mixture of 10% pH 3.5 to 10.0 and 90% pH 4.0 to 6.5. Protein resolution in the second dimension was carried out by 12.5% SDS-PAGE. The two-dimensional gels were stained with Coomassie Blue and dried, and autoradiography was performed directly on film (X-Omat Blue XB-1, Kodak). To determine the pH gradient, blank gels resolved in the first dimension were cut into evenly sized slices, each slice immersed in 1 ml of 10 mM potassium chloride, vortexed, and incubated at room temperature for 1 h, and then the pH of the solution was determined. Linear regression of the pH and gel distance was carried out (y ϭ 0.3386x ϩ 4.375, where y ϭ pH and x ϭ cm from acidic end of gel) to determine the actual pI of the various phospho-proteins. Migration in the second dimension was compared with broad range protein markers (New England Biolabs, Beverly, MA).

RESULTS AND DISCUSSION
Overlay assays with Cdc42 and Rac1 provide a sensitive means to detect PAK kinases in smooth muscle. [ 35 S]GTP␥S-Cdc42 bound to bands of 62 and 65 kDa in extracts of guinea pig taenia coli smooth muscle, whereas [ 35 S]GTP␥S-Rac1 detected a single band of 62 kDa (Fig. 1, A and B). An antibody raised against the N-terminal 13 amino acid residues of mouse fibroblast mPAK3 reacted with a protein of 62 kDa in guinea pig smooth muscle and a protein of the same molecular mass in rat aorta (Fig. 1C). These results indicate that smooth muscle contains one, and possibly two, PAK isoforms (Fig. 1A). PAK was absent from Triton-skinned smooth muscle fibers (Fig. 1, A  and B), suggesting that, like ROK (3), PAK is either a cytoplasmic or membrane-bound enzyme.
Triton-skinned guinea pig taenia coli smooth muscle fibers were induced to contract in a Ca 2ϩ -independent manner when 2 R. Cerione, personal communication.
incubated in the presence of recombinant, constitutively active GST-mPAK3 ( Fig. 2A). The force induced by GST-mPAK3 (ϳ5 g/ml; 55 nM) in relaxing buffer (pCa Ͻ 8.0) reached a maximum level equivalent to 62 Ϯ 12% (n ϭ 10) of that achieved by addition of a calcium-containing activation solution (pCa 4.3). Under the same conditions, the inactive PAK mutant, GST-mPAK K297R , was unable to induce force in the absence of Ca 2ϩ (Fig. 2B). In previous studies, Ca 2ϩ -independent smooth muscle contraction has been induced through the use of unregulated forms of MLCK (1, 2), by addition of phosphatase inhibitors (26,33,34), or most recently by another Rho family GTPase-dependent kinase, ROK (3). In all cases the degree of smooth muscle contraction correlates with an increase in the level of MRLC phosphorylation. In the case of ROK, contraction is promoted by the direct phosphorylation of MRLC on serine 19 (4) in addition to the phosphorylation and inhibition of MLCP (5).
This dual effect of ROK was demonstrated by the use of wortmannin, which is a potent inhibitor of MLCK but does not affect the activity of either ROK (3) or PAK (Fig. 2C). The addition of a constitutively active GST-ROK catalytic domain to Triton-skinned smooth muscle fibers produces a wortmannin-sensitive contraction at pCa 6.5 (Ca 2ϩ -dependent contraction) as well as a wortmannin-insensitive contraction at pCa Ͻ 8.0 (Ca 2ϩ -independent contraction) (3). On the other hand, wortmannin at a concentration of 1 mM had little effect on the contraction of smooth muscle induced by GST-mPAK3 at low calcium (Fig. 2D), even though this concentration is sufficient to completely inhibit MLCK-dependent contraction at elevated Ca 2ϩ (data not shown). These results demonstrate that the Ca 2ϩ -independent contraction promoted by PAK occurs without a requirement for MLCK activity. Furthermore, it seems unlikely that PAK promotes contraction by inhibiting MLCP because the Ca 2ϩ -independent contractions achieved with phosphatase inhibitors are invariably dependent on MLCK activity and are abolished by MLCK inhibitors (e.g. 26,33,34). Thus, PAK most likely works by direct phosphorylation of a contractile protein rather than altering either MLCK or MLCP.
These results prompted an investigation into whether PAK directly phosphorylates MRLC, thus achieving contraction in a "traditional" manner. In vitro analysis shows that GST-mPAK3 phosphorylates intact chicken gizzard smooth muscle myosin to 2 mol of phosphate/mol (Fig. 3A). Phosphate is incorporated into a single serine residue of MRLC (Fig. 3A). Furthermore, MLCK was unable to phosphorylate MRLC following GST-mPAK3 treatment (Fig. 3B), indicating that PAK and MLCK both phosphorylate serine 19. Indeed, identical two-dimensional tryptic phosphopeptide maps were obtained from MRLC phosphorylated by either MLCK or GST-mPAK3 (Fig. 3C). These results predict that PAK, like MLCK and ROK, promotes smooth muscle force generation by increasing MRLC phosphorylation levels. However, under conditions where GST-mPAK3 induced Triton-skinned smooth muscle fibers to contract with ϳ70% of the maximal force obtained in the presence of Ca 2ϩ , no significant increase in the level of MRLC phosphorylation is observed (Fig. 3D). In fact, the level of MRLC phosphorylation remains similar to the level of relaxed fibers (absence of GST-mPAK3 and calcium, Fig. 3D, lane 4) even as force induced by GST-mPAK3 increases from 26 to 70% (Fig. 3D, lanes 1-3). The uncoupling between MRLC phosphorylation and force genera- tion implies that PAK does not directly or indirectly activate myosin but must employ an alternative and novel mechanism to contract the skinned muscle fibers.
To begin to define the molecular basis of PAK-induced contraction, it is critical to identify the proteins phosphorylated by mPAK3 in the skinned smooth muscle fibers. Protein substrates for mPAK3 were labeled with 32 P under conditions where GST-mPAK3 produces ϳ70% of maximal Ca 2ϩ -dependent force (Fig. 4). One-and two-dimensional gel electrophoretic analyses of the proteins labeled during a PAK-induced contraction were performed. With one-dimensional SDS-PAGE, two proteins with approximate molecular masses of 58 and 145 kDa are more highly phosphorylated in the presence than the absence of GST-mPAK3 (Fig. 4, A and D, compare lanes 1 and 2). Little if any, phosphorylation of MRLC is detected in the fibers contracted with GST-mPAK3 (Figs 3D and 4A).
The 58-and 145-kDa proteins were identified by Western blot analysis as desmin and caldesmon, respectively (Fig. 4D). Furthermore, the pI of the 58-and 145-kDa proteins were determined by two-dimensional gel electrophoresis using a pH gradient from 4.0 to 6.5 followed by 12.5% SDS-PAGE (Fig. 4C).
The pI values for the 58-kDa protein are 5.59 Ϯ 0.04 and 5.37 Ϯ 0.04 for the mono-and diphosphorylated forms, respectively, and although the amino acid sequence of guinea pig desmin is not known, these pI are in the range expected for desmin (human and chick unphosphorylated desmin, Swiss P17661 and P02542, pI of 5.21 and 5.45, respectively). The 145-kDa protein, identified as caldesmon, has pI values of 5.63 Ϯ 0.03 and 5.38 Ϯ 0.04 for the mono-and di-phosphorylated forms, respectively. Again, the amino acid sequence of guinea pig caldesmon is not known, but these observed pI values are in the range of pI for human and chick unphosphorylated h-caldesmon (PIR JH0628 and A33430) of 5.62 and 5.56, respectively. Importantly, there is no known contractile protein other than caldesmon with a molecular mass between 70 and 150 kDa that is present in Triton-skinned muscle fibers with pI close to 5.60. For example, smooth muscle MLCK (Swiss P11799) and myosin light chain phosphatase PP2A 130-kDa subunit have calculated pI values of 4.72 and 5.09, respectively.
GST-ROK, the GST fusion protein of the constitutively active catalytic domain of ROK (3), causes Triton-skinned smooth muscle fibers to contract in a Ca 2ϩ -independent manner with up to 80% of maximal force (Ref. 3; data not shown). Under these conditions, the major proteins phosphorylated in the skinned fibers by GST-ROK are MRLC, desmin, and two proteins migrating at positions greater than 158 kDa (Fig. 4, B and  E). Clearly, neither of these high molecular mass proteins are caldesmon (145-kDa phospho-protein, Fig. 4E). One is most likely the catalytic domain of MLCP, which is known to be phosphorylated by ROK in vitro and has an approximate molecular mass of 158 kDa by SDS-PAGE.
Comparison of the protein substrates for ROK and PAK under conditions where GST-ROK and GST-mPAK3 induce similar amounts of Ca 2ϩ -independent force (79.5 versus 71.1%, respectively), indicated that GST-ROK incorporated more phosphate into MRLC than did GST-mPAK3 (Fig. 4, compare A  and B). As well, GST-ROK did not phosphorylate caldesmon, which is one of the main substrates for GST-mPAK3 (Fig. 4, D  and E). In vitro phosphorylation studies confirm that chicken gizzard h-caldesmon is a better substrate for GST-mPAK3 than GST-ROK (Fig. 4F). In vitro, GST-mPAK3 phosphorylated hcaldesmon to 2 mol of phosphate/mol of protein (Fig. 4F). This explains the mono-and diphosphorylated states of caldesmon found in two-dimensional gel electrophoresis of the PAK phsophorylated Triton-skinned muscle fibers (Fig. 4C). Furthermore, the C-terminal domain of human fibroblast l-caldesmon (corresponding to chicken gizzard caldesmon amino acid residues 458 -752) is a substrate for GST-mPAK3 (Fig. 4G), but no phosphorylation of the N-terminal caldesmon domain was observed (data not shown). The C terminus of caldesmon contains multiple binding sites for actin, tropomyosin, and calmodulin.
Caldesmon inhibits the actin-activated Mg-ATPase of myosin (review see Ref. 35) and has been suggested to provide a basal inhibition of vascular tone. The force of contraction of Triton-skinned smooth muscle fibers increases upon the partial extraction of caldesmon (36) or decreases because of competitive binding of a 20-kDa actin-binding fragment of caldesmon (37). As well, a synthetic peptide of an actin-binding region of caldesmon increases force of ␤-escin skinned arterial muscle fibers at low concentrations of Ca 2ϩ (11), probably by competing with endogenous caldesmon for the actin filament. Taken together, these results suggest that reduction in caldesmon interaction with actin would increase force generation, resulting in contraction. Phosphorylation of the C terminus of caldesmon by PAK could release caldesmon inhibition of the ATPase activity resulting in augmented force development.
In conclusion, although two different Rho family-dependent kinases, PAK and ROK, are able to induce Ca 2ϩ -independent contractions in smooth muscle, they do so via different molecular mechanisms. ROK increases the steady state level of MRLC phosphorylation. PAK, on the other hand, uncouples force from MRLC phosphorylation and likely acts by phosphorylating caldesmon. The data presented are consistent with a model of PAK-induced contraction in which myosin phosphorylation is at least complemented through the regulation of thin filament proteins.