Defining the Structural Determinants and a Potential Mechanism for Inhibition of Myosin Phosphatase by the Protein Kinase C-potentiated Inhibitor Protein of 17 kDa*

Contractility of smooth muscle and non-muscle microfilaments involves phosphorylation of myosin II light chain. Myosin light chain phosphatase (MLCP) is specif-ically inhibited by the protein kinase C-potentiated inhibitor protein of 17 kDa, called CPI-17, as part of Ca 2 (cid:1) sensitization of vascular smooth muscle contraction. Phosphorylation of Thr 38 in CPI-17 enhances inhibitory potency toward MLCP over 1000-fold. In this study we mapped regions of CPI-17 required for inhibition and investigated the mechanism using deletion and point mutants. Deletion of either the N-terminal 34 residues or C-terminal 27 residues gave no change in the IC 50 of either phospho- or unphospho-CPI-17. However, further deletion to give CPI-17 proteins of 1–102, 1–89, 1–76, and 1–67, resulted in much higher IC 50 values. The results indicate there is a minimal inhibitory domain between residues 35 and 120. A single Ala substitution at Tyr 41 eliminated phosphorylation-dependent inhibition, and phospho-Thr 38 in the Y41A protein was efficiently dephosphorylated by MLCP itself. The wild type CPI-17 expressed in fibroblast-induced bundling and contraction of actomyosin filaments, whereas expression of the Y41A

to humans (3,4). The isolated PP1C can dephosphorylate multiple phospho-Ser/Thr substrate proteins. Regulatory subunits bind to PP1C and tether it to intracellular loci, thereby the PP1C activity is compartmentalized for specific substrates. In addition, the interaction of regulatory subunit can induce allosteric modulation of PP1C, narrowing down specificity and activating catalysis (5,6). The PP1C binding sequence of many of these regulatory subunits is called the RVXF motif (5). The analog peptide of the glycogen targeting subunit including a RVSF sequence was co-crystalized with PP1C and seen to contact the ␤5/6 loop, Leu 243 , Phe 257 , and Phe 293 of PP1C, some distance from the active site (5).
Specific inhibitor proteins also control cellular PP1 activity. Phosphorylation-dependent inhibitors, inhibitor-1, and DARPP-32 inhibit PP1C activity in response to phosphorylation by protein kinase A (7,8). In addition to the phosphorylation site, there is a KIQF sequence present as a PP1Cbinding motif that is necessary for the potent inhibition of PP1C by both inhibitor-1 and DARPP-32 (9,10). A synthetic 31-residue DARPP-32 peptide including both the KIQF sequence and the phosphorylation site produced phosphorylationdependent inhibition of PP1C (11). Neither inhibitor-1 nor DARPP-32 inhibited activity of PP1C associated with certain regulatory subunits containing an RVXF motif. Thus, it is thought that these inhibitors and regulatory subunits compete for binding to PP1C at the site for RVXF. Inhibitor-2 and its analog inhibitor-4 interact with PP1C using sites over their entire sequences (12,13,14). The N-terminal IKGI sequence of inhibitor-2 was suggested to contact to the ␤5/6 loop of PP1C near, but not identical to the RVXF peptidebinding site (15,5). Nonetheless, inhibitor-2 has little inhibitory activity against myosin light chain phosphatase (MLCP) that contains PP1C and a myosin targeting subunit (MYPT1) (16). The concept is that PP1C binds either regulatory subunits or inhibitor proteins, but not at the same time (17). Unlike these other inhibitor proteins, CPI-17 and its analog PHI-1 (30% overall sequence identity), can inhibit the myosin phosphatase enzyme complex as potently as monomeric PP1C. This shows that binding to the RVXF site on PP1C is not required for inhibition, nor does the MYPT1 subunit bound to this site interfere with CPI-17 inhibition (18,19). Therefore CPI-17 presents a new mode of interaction with PP1C, and this study set out to define the determinants for this interaction.
CPI-17 is important for control of smooth muscle and is highly expressed in mammalian vascular smooth muscle (18,20,21). Myosin phosphatase has been identified as the dominant target for CPI-17 phosphorylated at Thr 38 (22,16). Phosphorylation of Thr 38 occurs in response to stimulation of intact smooth muscle by agonists such as histamine and phenylepher-ine (23) and dephosphorylation occurs in parallel with vasodilation that is triggered by nitric oxide production (24). Recombinant phospho-CPI-17 induces myosin phosphorylation and contraction of permeabilized artery strips (25). Permeabilization of artery strips by Triton X-100 treatment depletes endogenous CPI-17 and eliminates the contractile response to protein kinase C (PKC), but reconstitution with recombinant CPI-17 restores PKC-induced contraction (26). The expression level of CPI-17 among smooth muscles correlates with sensitivity to phorbol ester-induced contraction, supporting the concept that PKC acts via CPI-17 (21). Experiments show several other purified kinases phosphorylate CPI-17 at Thr 38 , including Rho-associated coiled-coil kinase, protein kinase N, and ZIP-like kinase (27)(28)(29)(30). The results suggest that CPI-17 is a hub at the center of various signaling pathways to regulate MLCP activity and control vascular tone (27). Here we prepared a series of deletion and point-mutated CPI-17 and tested their activity against purified MLCP and in living cells to define the structures required for potent inhibition.

EXPERIMENTAL PROCEDURES
Materials-MLCP was prepared from porcine aorta smooth muscle (18). Partially purified PKC was prepared from pig brain by a modification of the method of Kikkawa et al. (31), using DE32 TM (Whatman) and Phenyl-Cellulofine TM (Seikagaku Co.) column chromatographies. The active fragment of PKC␦ was purified (27) and chicken gizzard MLC20, MLCK, and calmodulin were prepared as previously described (22). MLC20 was phosphorylated with MLCK in the presence of Ca 2ϩ /calmodulin and ATP (18). [␥-32 P]ATP and sequencegrade trypsin were obtained from Amersham Pharmacia Biotech and Promega, respectively.
Preparation of Recombinant CPI-17 Proteins-Pig CPI-17 cDNA was ligated into the pET30 vector (Novagene) to produce recombinant CPI-17 fused with N-terminal His 6 -, plus the 44-residue S-tag sequence TM (Novagene) in Escherichia coli strain BL21(DE3) (20). For Nterminal deletion mutants of CPI-17 a new NcoI site was generated by polymerase chain reaction at the desired sites in wild type CPI-17 cDNA using a mismatch sense primer. To create C-terminal deletion mutants a stop codon was introduced, followed by a new EcoRI site for ligation into the expression vector. Point mutations were produced by the Quickchange TM protocol following the manufacturer's directions (Stratagene). Sequences of all cDNA inserts were confirmed by deoxy sequencing in the biomolecular core facility at the University of Virginia. Recombinant CPI-17 and mutants were purified using nickelnitrilotriacetate resin (Qiagen) as previously described (20). Phosphorylation or thiophosphorylation of CPI-17 was carried out at 37°C using pig brain PKC and 0.1 mM ATP or 1 mM ATP␥S (20). Samples were subjected to polyacrylamide gel electrophoresis with 8 M urea (20) and stoichiometric phosphorylation or thiophosphorylation was demonstrated by a change in migration of all the protein. For mammalian expression, the cDNA insert in pET30 vector was excised and ligated into the pcDNA3 TM vector (Invitrogen) that was modified to express N-terminal FLAG TM (Sigma)-tagged fusion proteins.
Phosphatase Assay-MLCP activity was assayed at 25°C by measuring the initial rate of dephosphorylation of 2 M phospho-MLC20 or phospho-CPI-17 as a substrate. Assays were performed with 0.5 nM MLCP and various concentrations of recombinant CPI-17 in the presence of 50 mM MOPS-NaOH, pH 7.2, containing 50 mM NaCl, 0.1 mM EGTA, 1 mM MnCl 2 , 1 mM dithiothreitol, 1 mM benzamidine, 0.1 mg/ml bovine serum albumin, and 0.02% Brij-35. Dephosphorylation was initiated by addition of phospho-MLC20 or phospho-CPI-17 to the mixture. At 5, 10, 20, and 30 min 10-l samples were transferred into tubes containing about 10 mg of solid urea and 2 l of 2-mercaptoethanol to terminate the reaction. The phospho-and dephospho-MLC20 or CPI-17 were separated on urea-PAGE gel (32,25) and stained with Coomassie Blue G-250 solution. The ratio of the band intensity of phospho/dephospho-MLC20 or CPI-17 was determined by use of a densitometer (ATTO). The time course of dephosphorylation was linear under these conditions and the rate of dephosphorylation was measured with each concentration of recombinant CPI-17 and plotted as relative MLCP activity against the activity without CPI-17, set as 100%.
Limited Proteolysis-Limited proteolysis was performed using 32 Pphosphorylated H 6 S-CPI-17. Wild type and Y41A proteins (0.2 mg/ml) were phosphorylated for 10 min at 30°C with the active fragment of PKC␦ and 0.1 mM [␥-32 P]ATP (2.4 Ci/nmol). The PKC was inactivated by addition of 10 M GF109203X (Calbiochem). Samples (19 l) were mixed with 1 l of trypsin solution and incubated for 30 min at 23°C. Trypsin was inactivated by addition of 2 mM Pefabloc TM (Boehringer) and 10 l of the sample was subjected to SDS-PAGE. The tryptic fragments were visualized by staining with Coomassie Blue G-250 and autoradiography.
Immunofluorescence-Rat embryo fibroblasts (REF52) were cultured at 37°C in 5% CO 2 with Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum. Immunofluorescence was carried out as previously described (16). Briefly, the REF52 cells on a coverslip were transiently transfected for 24 h with 1 g of mammalian expression vectors of FLAG-tagged CPI-17 and its mutants, using NovaFEC-TOR TM transfection reagent (Venn Nova, LLC., Pompano Beach, FL). After fixation and blocking, cells were incubated with anti-FLAG antibody (M2 clone, Sigma, 1:500), and then the primary antibody and actin microfilaments were stained with a mixed solution of Texas Red-conjugated anti-mouse IgG secondary antibody (Calbiochem) and fluorescein-labeled phalloidin (Sigma). Images were acquired with a Nikon Microphot FXA/SA fluorescence microscope using a Hamamatsu C4742-95 CCD camera. The camera was controlled by Open Lab software (Improvision), and images were processed in Photoshop 5.5 (Adobe).

Definition of a Minimal Inhibitory Domain in CPI-17-
We prepared a series of deletion and point-mutated CPI-17 as His 6 -tagged recombinant proteins (Fig. 1). The inhibitory potency of phospho-, thiophospho-, and unphospho-forms of these proteins was tested using MLCP prepared from pig aorta. All unphosphorylated proteins had essentially the same potency as unphospho-wild type, with IC 50 over 1 M (Fig. 2). Phosphorylation or thiophosphorylation at Thr 38 was catalyzed by pig brain PKC. The deletion mutant (35-147) required longer reaction time to achieve stoichiometric phosphorylation (data not shown). Either thiophosphorylation or phosphorylation at Thr 38 of wild-type CPI-17 was sufficient to produce full potency (Fig. 2, Table I), as described previously (18). We tested the inhibitory potency of CPI-17 with Glu substituted for Thr 38 , T38E (Fig. 2). The negative charge of Glu did not mimic a phosphate group on Thr 38 , suggesting that a phosphate or a thiophosphate moiety on Thr 38 is necessary for potent inhibition.
Various N-and C-terminal deleted proteins were assayed (Fig. 3). A deletion mutant  lacking the N-terminal 34 residues inhibited MLCP activity with IC 50 of 10 nM (Fig. 3). Mutants with shorter deletions at the N terminus, (10 -147) and , were essentially the same potency as wild type (Table I). The mutant protein (1-120), deleted 27 residues at the C terminus, had the same potency as full-length phospho-CPI-17 (Fig. 3) indicate that the region between 35 and 120 of CPI-17, that is conserved with the inhibitor protein PHI-1, is necessary for potent inhibition of MLCP.
Tyr 41 Is Crucial to Prevent Dephosphorylation of Phospho-Thr 38 -The sequence around Thr 38 is highly conserved between CPI-17 and PHI-1. We tested for the functions of various side chains in this region by making single Ala substitutions and assaying the mutant proteins. Consistent with results using the truncated protein , neither substitution of Lys 32 nor Arg 33 affected the IC 50 value of phospho-CPI-17 (Table I). On the other hand, single Ala substitutions of residues between Tyr 41 and Arg 44 resulted in large increases of IC 50 values (Table I and Fig. 4A). Especially the single residue mutant, the Y41A protein, completely lost phosphorylation-dependent inhibition of MLCP (Fig. 4B). Thiophosphorylation of Thr 38 in the Y41A protein increased inhibitory activity (IC 50 of 360 nM) compared with the oxo-phosphorylated version of this protein (Fig. 4B). In contrast, the IC 50 value of the thiophosphorylated (1-76) protein was not much different from that of the oxo-phosphorylated form (Table I). Thiophospho residues are more stable to hydrolysis by phosphatases compared with oxophospho residues, so we suspected that the Y41A version of CPI-17 might have been dephosphorylated, accounting in part for its weak inhibitory activity.
We tested whether the oxophospho-[Thr 38 ]CPI-17 Y41A protein was dephosphorylated by MLCP during the assay. Fig. 5 19 ]MLC20. A low concentration of okadaic acid (10 nM) had no effect on these activities, consistent with PP1 not PP2A-type activity. The results were consistent with the previous kinetic analysis that indicated mixed inhibition of MLCP by CPI-17 (22). The Ala substitution of Tyr 41 greatly enhanced dephosphorylation of phospho-Thr 38 . Surprisingly, the phospho-[Thr 38 ]CPI-17 Y41A protein was 1.4-fold better substrate for MLCP than even phospho-[Ser 19 ]MLC20 (Fig.  5). The R43A substituted protein also was moderately dephosphorylated by MLCP. The results suggest that side chains C-terminal to the phospho-Thr 38 in CPI-17 act to prevent dephosphorylation of this phospho-residue posi-   tioned at the active site, thereby producing inhibition of MLCP. The side chain of Tyr 41 is especially important in this regard because the Y41A protein of CPI-17 was so efficiently dephosphorylated as a substrate.
We tested whether Ala substitution of Tyr 41 perturbed the overall conformation of CPI-17. The Y41A protein was subjected to limited proteolysis as a probe of protein folding. Fig. 6 shows the pattern of tryptic fragments of 32 P-H 6 S-CPI-17 wild type (left) and the Y41A protein (right). Incubation with 0.05 g/ml trypsin cleaved only a small fraction of the total protein, and one fragment stained by Coomassie migrated the same as CPI-17 itself (marker lane). This was probably due to low yield cleavage of the fusion tag from the CPI-17. On the other hand incubation with 0.5 g/ml trypsin digested all of the full-length fusion protein. The predominant product was a 15-kDa fragment, smaller than CPI-17 itself (left lane). This indicated that trypsin had cleaved at least one site within the CPI-17 sequence. The fragment was 32 P-labeled (bottom panel), therefore it contained Thr 38 , the site of phosphorylation. The site(s) for trypsin cleavage were displayed identically because there was no difference between wild type and Y41A proteins detected by Coomassie staining (top) and 32 P-autoradiogram (bottom). Furthermore, in separate experiments the kinetics of modification of Lys residues in the Y41A protein with N-hydrosuccinimidyl rhodamine was identical compared with the wild type (not shown). Phosphorylation of wild type and Y41A proteins by PKC was not significantly different. Together, these results suggest that the overall folding of the Y41A protein was indistinguishable from wild-type CPI-17.
Activity of Wild-type and Mutant CPI-17 in Living Cells-FLAG-tagged CPI-17 proteins were expressed in REF52 cells and Fig. 7 shows the resulting distribution of actin microfilaments. The wild type CPI-17 induced prominent bundling of actin microfilaments along the periphery, as previously reported (16). Cells expressing wild type looked to have a shorter axis, consistent with the inhibition of MLCP by ectopic CPI-17 to enhance phosphorylation of myosin and contractility of microfilaments (16) (Fig. 7A). The bundling and contraction of microfilaments also occurred in cells expressing the deleted protein (1-120) that inhibited purified MLCP with the same IC 50 value as wild type (Fig. 7B and Table I). The shorter fragment (1-89) of CPI-17, with lower potency against purified MLCP did not induce any changes in actin microfilaments (Fig.  7C), consistent with the IC 50 value that was 22-fold higher compared with wild type (Table I). These results showed the sensitivity of the living cell assay of CPI-17 activity. The Y41A protein of CPI-17 had no effects on organization of actin microfilaments (Fig. 7D). Equivalent levels of FLAG-protein expression for each of these proteins was confirmed by immunofluorescence using anti-FLAG antibody (Fig. 7, E-H). The tagged CPI-17 appeared concentrated into foci near the periphery of the cell, as we previously reported (16). Both an active and an inactive truncated form of CPI-17 seemed to concentrate in the nucleus (Fig. 7, F and G), however, this did not seem to impair the ability to inhibit MLCP in situ. We concluded that wild type and (1-120) versions of CPI-17 but neither the (1-89) nor the Y41A mutant CPI-17 inhibited MLCP in living fibroblasts. DISCUSSION In mammalian cells there are multiple inhibitor proteins specific for PP1, such as inhibitor-1 (33), inhibitor-2 (34), DARPP-32 (35), and inhibitor-3 (36), but multisubunit forms of PP1 seem resistant to the action of these inhibitor proteins. Therefore, it is generally thought that the PP1 catalytic subunit binds the inhibitor proteins after being released from regulatory subunits. Consistent with this concept, evidence indicates that inhibitor-1 and DARPP-32 use a KIQF sequence to dock at the RVXF-binding site used by various PP1 regulatory subunits (9). In contrast, CPI-17 has no sequence resembling RVXF and among the PP1 inhibitor proteins only CPI-17 and its analog PHI-1 inhibit the MLCP holoenzyme (18,25,16). The interaction of MYPT1 docked to PP1C via a KVKF sequence modulates the active center of PP1C to increase the specific activity against phosphomyosin LC20 (6). This allosterically activated PP1C in MLCP is recognized and inhibited by phospho-CPI-17.
The deleted versions of CPI-17, (1-120), and , define a minimal functional domain that inhibited MLCP about the same as wild type CPI-17, both in biochemical assays (Table I) and in living fibroblasts (Fig. 7). Deletion of C-terminal residues beyond 1-120 greatly decreased the potency of inhibition. For example, the (1-89) mutant inhibited MLCP with the same IC 50 as the (1-102) mutant that was Ͼ20-fold higher compared with wild type. This leads us to propose a specific requirement for residues 103-120 in the inhibitory domain. Consistent with this idea, the CPI-17 sequence between 90 and 102 shows only 28% identity to PHI-1, whereas 56% of the residues are identical between positions 103 and 120. The N-terminal 67-residue fragment (1-67) showed little phosphorylation dependence for MLCP inhibition, even though it contained phospho-Thr 38 and the neighboring YDRR sequence (see below). The results suggest that residues in the C-terminal region of the inhibitory domain play a role in converting CPI-17 to its high affinity form in response to Thr 38 phosphorylation. Thus the two ends of the inhibitory domain (35-120) are required for phosphorylationdependent inhibition by CPI-17.
Unexpectedly Tyr 41 was discovered to be a key residue in the inhibitory domain of CPI-17, and when mutated to Ala it converted CPI-17 from an inhibitor to a substrate of MLCP. This mutation must subtly but significantly alter interaction with the active site of PP1C. Wild-type phospho-CPI-17 was slowly dephosphorylated by MLCP, suggesting that the phospho-Thr 38 of CPI-17 directly interacts with the active center of MLCP as a substrate and as an inhibitor. Previous kinetic analysis showed mixed inhibition of MLCP by CPI-17 (22). The thiophospho-Thr 38 -Y41A protein that should be relatively resistant to hydrolysis still was only a relatively poor inhibitor of MLCP, implying that the Tyr 41 side chain has specific functions of its own required for inhibition. We propose that the side chain of Tyr 41 contacts a site on PP1C to enhance affinity and also acts to protect the phospho-Thr 38 from hydrolysis. We propose the mechanism of CPI-17 inhibition of MLCP is arrested catalysis.
Questions remaining are: where does Tyr 41 contact the surface of PP1C? How does this interaction perturb the geometry of the catalytic center to prevent hydrolysis of phospho-Thr 38 ? A definitive answer may require solution of the three-dimensional structure of the phospho-CPI-17⅐PP1C complex. Nonetheless, our data indicate that the residues to the N-terminal side of the phospho-Thr 38 (i.e. Lys 32 and Arg 33 ) are not critical for inhibition, whereas residues to the C-terminal side of the phosphorylated residue (i.e. Tyr 41 , Asp 42 , Arg 43 , and Arg 44 ) have an important role in producing potent inhibition. The sequence YDRR in CPI-17 is critical for high affinity inhibition and for arresting dephosphorylation of phospho-Thr 38 . The corresponding sequence in PHI-1 is YDRK. For comparison, the sequence at the C-terminal side of phospho-Ser 19 in MLC20 has no charges, but a cluster of hydrophobic residues ( 19 SNV-FAMF). The docking site of phospho-MLC20 on PP1C is still obscure, but phospho-CPI-17 may contact PP1C at the same site, because the Y41A protein of CPI-17 was dephosphorylated as well as MLC20. Thus, the difference in sequences following the phosphorylation site between CPI-17 and MLC20 seems to determine whether a phospho-protein is a substrate or an inhibitor. We imagine that the YDRR sequence of CPI-17 stabilizes the phospho-CPI-17⅐PP1C complex corresponding to an ES (enzyme-substrate) complex, without allowing bond scission. This scenario contrasts with results with inhibitor-1 and DARPP-32, where residues C-terminal to the phospho-Thr are dispensable but residues to the N-terminal side, including the KIQF docking motif, are critical to produce potent inhibition (9,11). The differences in structural determinants between PP1 inhibitor proteins may explain how CPI-17, but neither inhibitor-1 nor DARPP-32 can inhibit MLCP.
In collaboration with Dr Kainosho's group, 2 we recently solved the solution NMR structure of the CPI-17 inhibitory domain defined here. An anti-parallel helix pair is formed between residues 46 -63 and 105-115. This would explain how the 103-120 region is important for function, as shown here by deletion analysis. An N-terminal loop including phospho-Thr 38 and the adjacent YDRR sequence (residues 41-44) are exposed on the surface for binding at the active site of MLCP. A direct interaction of CPI-17 with the active site, without docking at the RVXF site on PP1C, would enable it to inhibit the MLCP holoenzyme. Just as this CPI-17 structure can recognize and inhibit the exposed face of MLCP, if other PP1 inhibitor proteins recognize PP1C while associated with certain regulatory subunits (37), they could play a role in selective regulation of other individual PP1 holoenzymes.