Importance of a Surface Hydrophobic Pocket on Protein Phosphatase-1 Catalytic Subunit in Recognizing Cellular Regulators*

Cellular functions of protein phosphatase-1 (PP1), a major eukaryotic serine/threonine phosphatase, are defined by the association of PP1 catalytic subunits with endogenous protein inhibitors and regulatory subunits. Many PP1 regulators share a consensus RVXF motif, which docks within a hydrophobic pocket on the surface of the PP1 catalytic subunit. Although these regulatory proteins also possess additional PP1-binding sites, mutations of the RVXF sequence established a key role of this PP1-binding sequence in the function of PP1 regulators. WT PP1α, the C-terminal truncated PP1α-(1–306), a chimeric PP1α containing C-terminal sequences from PP2A, another phosphatase, PP1α-(1–306) with the RVXF-binding pocket substitutions L289R, M290K, and C291R, and PP2A were analyzed for their regulation by several mammalian proteins. These studies established that modifications of the RVXF-binding pocket had modest effects on the catalytic activity of PP1, as judged by recognition of substrates and sensitivity to toxins. However, the selected modifications impaired the sensitivity of PP1 to the inhibitor proteins, inhibitor-1 and inhibitor-2. In addition, they impaired the ability of PP1 to bind neurabin-I, the neuronal regulatory subunit, and GM, the skeletal muscle glycogen-targeting subunit. These data suggested that differences in RVXF interactions with the hydrophobic pocket dictate the affinity of PP1 for cellular regulators. Substitution of a distinct RVXF sequence in inhibitor-1 that enhanced its binding and potency as a PP1 inhibitor emphasized the importance of the RVXF sequence in defining the function of this and other PP1 regulators. Our studies suggest that the diversity of RVXF sequences provides for dynamic physiological regulation of PP1 functions in eukaryotic cells.

(I-1) 1 and inhibitor-2 (I-2). Type-I protein serine/threonine phosphatase or protein phosphatase-1 (PP1), displays unique and potent inhibition by nanomolar concentrations of I-1 and I-2. In contrast, PP2A and other type-2 protein phosphatases are essentially resistant to these inhibitor proteins. To date, more than 50 PP1-interacting proteins have been identified (3). Some, like I-1 and I-2, function as PP1 inhibitors, whereas others represent targeting subunits that direct subcellular localization and substrate recognition by multiprotein complexes containing PP1 catalytic subunits. Remarkably, PP1 and PP2A catalytic subunits share nearly 50% primary sequence identity, which may account for their overlapping substrate specificity in vitro (1). However, all PP1 regulators thus far analyzed demonstrate the unique ability displayed by I-1 and I-2 to selectively associate with PP1 and modify its catalytic function.
Detailed structure-function analyses of I-1 (4) and its neuronal homologue, DARPP-32 (5), first highlighted the tetrapeptide sequence KIQF, which acts in conjunction with PKA phosphorylation at a conserved threonine to inhibit PP1 activity. Subsequent studies noted a homologous sequence, RVSF, in the skeletal muscle glycogen targeting subunit, G M , that was also required for PP1 binding (6). Cocrystallization of PP1 with a synthetic dodecapeptide encompassing the RVSF sequence from G M established the RVXF motif as a conserved PP1binding sequence that associates with a hydrophobic pocket on the surface of the PP1 catalytic subunit (7).
There are extensive surface interactions between PP1 and regulatory subunits, as demonstrated by alanine-scanning mutagenesis of yeast PP1 (8) and the recently resolved structure of PP1 catalytic subunit complexed with the myosin-targeting subunit, MYPT1 (9). However, the ability of RVXF-containing peptides derived from several PP1 regulatory subunits to displace G M from the glycogen-bound PP1 complex suggested a critical role for the RVXF sequence in PP1 binding (7). Single amino acid substitutions (e.g. Phe to Ala) in the RVXF motif severely impaired or abolished the ability of PP1 regulatory subunits (10) and inhibitors (4) to bind and regulate PP1. This emphasized the pivotal role played by the RVXF motif in PP1 binding and regulation. Several studies have identified RVXFcontaining PP1-binding proteins using an overlay or far-Western with an isolated PP1 catalytic subunit (11,12). Although this technique successfully identified many regulators, the pro-totypic PP1 regulators I-1 and I-2 were either weakly or not detected using this technique. This suggested that not all RVXF sequences were equivalent in PP1 binding. Direct binding studies also indicated that the unphosphorylated forms of I-1 (4) and DARPP-32 (5) bind PP1 very weakly. Although PKA phosphorylation activates these proteins as nanomolar inhibitors of PP1, the covalent modification only modestly increased their affinity for the PP1 catalytic subunit. Comparison of RVXF-containing synthetic peptides modeled on I-1 and G M in their ability to disrupt neuronal PP1 complexes (13) also highlighted their differing affinities for PP1. Substituting the PP1binding motif from nuclear inhibitor of PP1, RVTF, in place of the KIQF sequence normally found in I-1 enhanced its potency as a PP1 inhibitor, and emphasized the key role played by the RVXF sequence in defining the function of PP1 regulators (12). These studies suggested that the differences in the association of this conserved sequence with the common binding site on the PP1 catalytic subunit defined its physiological regulation of cellular PP1 regulators.
Our earlier studies (14) that substituted C-terminal sequences from PP2A in PP1␣ highlighted the diminished ability of the chimeric phosphatase, CHRM2, to bind I-1, I-2, nuclear inhibitor of PP1, and G M . Later studies (15) showed similar deficits in CHRM2 binding to two other PP1 regulators, PP1 nuclear targeting subunit and spinophilin. Together, these studies suggested that the C-terminal sequences unique to PP1␣ played a key role in PP1 binding and regulation by cellular proteins. Attempts to reverse these regulatory defects in CHRM2 by systematic substitution of PP1-specific sequences in the PP2A-derived C terminus highlighted a PP1specific sequence, 290 MC 291 , that enhanced the potency of DARPP-32 as an inhibitor of the modified CHRM2 (15). These and other data suggested that Met 290 and Cys 291 , which line the hydrophobic RVXF-binding pocket and interact with the RVXF sequences in G M (7) and MYPT1 (9), played a key role in PP1 regulation by cellular regulators. However, substitution of 290 MC 291 in CHRM2 had little effect on its decreased inhibition by I-2. This suggested that the hydrophobic pocket on the PP1 catalytic subunit could distinguish RVXF motifs present in different regulators and thereby differentiate the ability of these proteins to regulate PP1 activity.
Current studies deleted or substituted C-terminal sequences in PP1␣ as well as specifically modifying selected amino acids in the hydrophobic pocket of PP1␣ to define the role of the surface hydrophobic pocket in the recognition of RVXF-containing proteins. Our studies provided the first direct experimental evidence that the affinity of cellular regulators for the PP1 catalytic subunit is dictated by the RVXF sequence present in these proteins, and that the docking of this RVXF sequence in the surface hydrophobic pocket is conserved in all PP1 isoforms. Implications of these findings for the physiological regulation of PP1 in eukaryotic cells will be discussed.
For purification on heparin-agarose, bacterial extract was applied to the matrix in Buffer A and washed extensively with Buffer A containing 0.1 M NaCl. PP1 catalytic subunits were eluted with Buffer A containing 0.5 M NaCl. Fractions were collected and assayed for PP1 activity. Fractions were pooled, dialyzed overnight against Buffer A, concentrated using Centricon-10 (Pall Life Sciences), and dialyzed extensively against Buffer A containing 55% (v/v) glycerol prior to storage at Ϫ20°C. Purification of PP1 catalytic subunits on MCLR-Sepharose was undertaken essentially as described by Moorhead et al. (16), and stored as described above.
Preparation of Recombinant PP1 Regulators-GST-I-1 fusion proteins were expressed, and phosphorylated as described (22), with the following modifications: bacteria were lysed by two passages through French press at ϳ19,000 p.s.i. The fusion proteins were purified via glutathione-Sepharose according to manufacturer's instructions (Amersham Biosciences). GST-neurabin-I-(374 -516) and GST-G M -(1-240) were expressed and purified as described (21). I-1 (23) and I-2 (12) were expressed and purified as previously described. 32 P-Phosphoprotein Substrates-For phosphorylase a, phosphorylase kinase (0.25 mg) was added to a solution of phosphorylase b (10 mg/ml) in 100 mM ␤-glycerol phosphate and 100 mM Tris-HCl, pH 8.2, containing 1 mM MgCl 2 , 100 M ATP, 0.2 mM CaCl 2 , 0.1% (v/v) 2-mercaptoethanol, and 250 -300 Ci of [␥-32 P]ATP, and incubated at 37°C. At 30-min intervals, aliquots (10 l) of the reaction mixture were removed and added to 200 l of 20% (v/v) trichloroacetic acid and 50 l of bovine serum albumin (10 mg/ml). The sample was placed on ice for 2 min, then centrifuged at 15,000 ϫ g for 10 min. The pellet was repeatedly washed with 20% (w/v) trichloroacetic acid prior to Cerenkov counting. Once the 32 P incorporation reached a plateau (ϳ90 min), the reaction was stopped by the addition of an equal volume of 90% saturated ammonium sulfate and placed on ice for 20 min. The mixture was centrifuged at 20,000 ϫ g for 20 min, and the sedimented protein was washed twice with ice-cold 45% saturated ammonium sulfate. The pellet was then resuspended in 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 0.1% (v/v) 2-mercaptoethanol, and dialyzed extensively against the same buffer.
For myelin basic protein (MBP), 200 g of MBP was added to a solution of 1 mM ATP, 10 mM MgCl 2 , 10 -20 Ci of [␥-32 P]ATP, and 50 mM Tris-HCl, pH 7.5. Then, 10 -20 l of purified PKA was added to the reaction and incubated at 37°C. 32 P incorporation was examined in the same manner as phosphorylase a, and as the 32 P incorporation reached a plateau, the reaction was placed in dialysis buffer (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 0.1% (v/v) 2-mercaptoethanol), with frequent changes until background 32 P diminished.
Protein Phosphatase Assays-PP1 catalytic subunits diluted in 50 mM Tris-HCl, pH 7.5, containing 1 mM MnCl 2 , 0.1% (v/v) 2-mercaptoethanol, and 1 mg/ml bovine serum albumin were incubated with [ 32 P]phosphorylase a (2 mg/ml) at 37°C for 10 min (total reaction volume of 60 l). Reaction was terminated by addition of 200 l of 20% (v/v) trichloroacetic acid and 50 l of bovine serum albumin (10 mg/ml), and the mixture was centrifuged at 15,000 ϫ g for 5 min. [ 32 P]Phosphate released into the supernatant was measured by liquid scintillation counting. Assays using MBP as substrate were performed in the same manner, using 10 M MBP per reaction.
To assay PP1 inhibition by I-1, recombinant I-1 was first incubated with 0.3 mg/ml PKA, 50 mM Tris-HCl, pH 7.5, 0.2 mM ATP␥S, and 2 mM MgCl 2 at 30°C for 72 h. The thiophosphorylated I-1 was dialyzed into 50 mM Tris-HCl, pH 7.5, 0.005% (v/v) Brij-35, and 0.1% 2-mercaptoethanol. Thiophosphorylated I-1 was briefly incubated with PP1 before the addition of phosphorylase a. In assays for PP1 inhibition by I-2, the I-2 protein was preincubated with PP1 for 20 min at 37°C, and targeting subunits were preincubated with PP1 for 5 min at 37°C prior the initiation of the phosphatase assay.
PP1 Sedimentation-GST fusions of PP1 inhibitors and/or targeting subunits were incubated with glutathione-Sepharose (25-l bed volume) equilibrated in Tris-buffered saline (TBS) (50 mM Tris-HCl, pH 7.5, 150 mM NaCl) for 1 h at 4°C. The beads were washed twice with TBS, and incubated for 1 h at 4°C with bovine serum albumin (1 mg/ml). They were then washed twice with TBS, and recombinant PP1 was added for 1 h at 4°C. The beads were washed four times with NETN-250 (250 mM NaCl, 1 mM EDTA, 10 mM Tris-HCl, pH 7.5, and 0.5% (v/v) Nonidet P-40). Bound proteins were eluted with SDS sample buffer, and subjected to SDS-PAGE on 12% (w/v) polyacrylamide gels. Proteins were electrophoretically transferred to polyvinylidene difluoride membranes, which were stained with Ponceau S to visualize the proteins. Membranes were blocked in 4% milk (w/v) in TBS plus 0.05% Tween, and PP1 detected by immunoblotting with an anti-PP1 antibody. The protein bands were quantified by scanning using Quantity One software (Bio-Rad).
PP1 Far-Western Blots-PP1 overlays were performed as described (11), with the following modifications. Protein staining using Ponceau S verified equivalent protein loading. The bound digoxigenin-labeled PP1 was detected by immunoblotting with an anti-digoxigenin antibody (horseradish peroxidase-linked) and ECL reaction (PerkinElmer Life Sciences). To ensure linearity of the overlay assays in each case, the ECL reaction was exposed to film for 2, 15, and 30 s.
For these assays, GST-I-1 was phosphorylated using PKA (0.2 g/ml), 100 M ATP, 1 mM MgCl 2 , at 37°C for 90 min. Phospho-GST-I-1 was dialyzed into 50 mM Tris-HCl, pH 7.5, 0.005% (v/v) Brij-35, and 0.1% (v/v) 2-mercaptoethanol. GST was similarly phosphorylated using PKA for control pull-down assays. PP1 binding to GST alone was defined as nonspecific and subtracted from that bound to GST fusion proteins containing known PP1 regulators. Protein concentration was determined using the Bio-Rad protein assay with BSA (Pierce) as the standard.
Analysis of PP1 Structure-The program Rasmol 2.6 (24) was used to generate the PP1 structure, as well as the programs Mage and Prekin (25), and interaction dots were generated with Molprobity (26) and Probe (27). The latter four programs were obtained from kinemage.biochem.duke.edu.

Purification of Recombinant PP1 Catalytic Subunits-
Emerging studies (21,28,29) showed that PP1 regulators displayed selectivity for distinct PP1 isoforms, which differed largely in their C-terminal sequences. Prior studies (30,31) suggested that extensive deletions of C-terminal sequences impaired or destabilized PP1␣ activity. Although recombinant PP1␣-(1-297) demonstrated phosphorylase phosphatase activity equivalent to that of full-length PP1␣, PP1␣-(1-276) was not expressed in bacteria. To focus on the role of the RVXF-binding site conserved in all PP1 isoforms, we expressed PP1␣-(1-306) (PP1306), which eliminated the majority of C-terminal residues unique to this PP1 isoform. We also analyzed PP1␣ and CHRM2, which contains PP1␣-(1-273) fused to the C-terminal 43 amino acids from bovine PP2A␣ catalytic subunit, previously shown to generate an active phosphorylase phosphatase that was resistant to many PP1 regulators (14). In addition, we expressed PP1306 with the single amino acid substitutions L289R, M290K, and C291R to modify the RVXF-binding pocket, which introduced residues more commonly found in type-2 phosphatases.
Previous studies utilized heparin-agarose to separate PP1 from type-2 protein serine/threonine phosphatases (32) and purify recombinant PP1 catalytic subunit expressed in E. coli (18). Complete purification of PP1, however, required multiple chromatographic steps and reduced the yields of this protein (18). Comparison of a recombinant PP1 catalytic subunit, such as CHRM2 (Fig. 1B), purified from bacterial extracts using either heparin-agarose or affinity chromatography with MCLR immobilized on Sepharose (16), demonstrated that MCLR-Sepharose yielded an essentially single polypeptide (Ͼ95% purity) as judged by SDS-PAGE (Fig. 1B, lane 3). By comparison, chromatography on heparin-agarose (Fig. 1B, lane 2), although eliminating Ͼ90% of bacterial proteins, yielded a preparation of CHRM2 that still contained other proteins. MCLR (IC 50 Ͻ 1 nM) bound PP1 tightly, requiring strong chaotropic agents (e.g. 3 M NaSCN) to elute the phosphatase. However, following dialysis, PP1 preparations purified by either MCLR-Sepharose or heparin-agarose displayed similar phosphorylase phosphatase activity; their specific activities reflected the relative content of the 37-kDa polypeptide representing PP1 catalytic subunit identified by anti-PP1 immunoblots (data not shown). This was consistent with previous findings of Campos et al. (33), which demonstrated that chromatography on MCLR-Sepharose provided preparations of purified recombinant PP1 catalytic subunits with excellent specific activity.
Using MCLR-Sepharose, we purified recombinant PP1␣ and several PP1 mutants to near homogeneity. We first assessed the relative enzymatic activity of the purified phosphatases using two different substrates, phosphorylase a and MBP (Table I). Compared with PP1306, PP1␣ showed slightly reduced activity against both substrates. In contrast, CHRM2 showed either similar or slightly increased activity. These findings may be consistent with earlier observations (34) that the extreme C-terminal sequences specific to PP1␣ partially inhibited or suppressed its enzymatic activity. Thus, the deletion of amino acids 306 -330 in PP1306 or substitution of PP2A C-terminal sequences in CHRM2 may increase the intrinsic activity of these PP1 catalytic subunits. Further introduction of the point mutations L289R, M290K, or C291R resulted in a slight reduction of activity when compared with PP1306, but the activity of these mutant catalytic subunits was either equal to or higher than that of PP1␣ against both phosphoprotein substrates.
Toxin Sensitivity of WT and Mutant PP1␣ Catalytic Subunits-Most xenobiotic inhibitors bound at the catalytic site of the PP1 catalytic subunit (35)(36)(37). Thus, sensitivity of mutant PP1␣ subunits to one or more toxins provided evidence of structural alterations at or near the catalytic site. Compared with PP1␣ (MCLR IC 50 ϭ 1.1 Ϯ 0.2 nM), deletion of PP1␣-specific C-terminal sequences in PP1306 (MCLR IC 50 ϭ 0.7 Ϯ 0.1 nM) had no discernable effect on its sensitivity to either MCLR or okadaic acid. By comparison, CHRM2, which showed a similar IC 50 for MCLR (38), was much more sensitive to okadaic acid than PP1␣ or PP1306 (Table I and Fig. 2A). MCLR is an equipotent inhibitor of PP1 and PP2A. In contrast, okadaic acid demonstrates an IC 50 for PP2A, which is ϳ100-fold lower than that for PP1 (38). These data confirmed that the ␤12-␤13 loop played a key role in defining the sensitivity of PP1 to okadaic acid (14). Thus, PP1306, which retains the ␤12-␤13 loop of PP1␣, was inhibited by okadaic acid in a manner indistinguishable from PP1␣ ( Fig. 2A). By comparison, CHRM2, which incorporates the ␤12-␤13 loop of PP2A, displayed nearly 10-fold greater sensitivity to okadaic acid compared with either PP1␣ or PP1306. Further introduction of L289R, M290K, or C291R in PP1306 had only a modest impact on the sensitivity of PP1 catalytic subunits to okadaic acid. Essentially identical results were obtained with PP1 catalytic subunits purified on heparin-agarose (data not shown), demonstrating that the potentially harsh conditions used to elute PP1 from MCLR-Sepharose did not modify the catalytic sites of these phosphatases.
Inhibition of WT and Mutant PP1␣ Catalytic Subunits by Inhibitor Proteins-PP1 is characterized by its unique sensitivity to the mammalian proteins I-1 and I-2. Although nanomolar concentrations of I-2 or PKA-phosphorylated I-1 inhibit PP1 activity (38), high micromolar concentrations of these pro-  teins have no effect on PP2A activity (4). In this regard, sensitivity of PP1306 to I-1 and I-2 was essentially identical to that of PP1␣, indicating that the deletion of the C-terminal 24 amino acids had no effect on PP1 recognition by these protein regulators. However, as noted in previous studies (14), CHRM2 with the PP2A C terminus showed more than a 10-fold reduction in its IC 50 for I-1 and a nearly 500-fold reduced IC 50 for I-2 compared with either PP1␣ or PP1306 (Table I). This defect could result from the altered ␤12-␤13 loop, previously shown to dictate sensitivity from PP1 to I-1 and I-2, or the absence of critical amino acids constituting the surface hydrophobic pocket that binds RVXF motifs present in I-1 (14, 23) and I-2 (39).
Although some modifications of the RVXF-binding pocket, L289R and M290K, showed modest decrements in IC 50 values for I-1 and I-2, C291R showed a dramatic loss in its sensitivity for I-1. The nearly 10-fold decrease in IC 50 for I-1 was almost equal to that seen with CHRM2 (Table I and Fig. 2B). This suggested that the substitution of a single amino acid, C291R, in PP1306 attenuated PP1 regulation by I-1 to the same extent as removal of 53 amino acids from the PP1␣ C terminus and their replacement by 42 amino acids of the PP2A C terminus, which generated CHRM2. This highlighted the critical importance of the surface hydrophobic pocket for PP1 regulation by I-1.
All three mutants also showed modest reductions in their sensitivity to I-2. In contrast to M290K and L289R, which showed a reduction of ϳ3and 5-fold IC 50 for I-2, C291R demonstrated a Ͼ10-fold decrease in IC 50 for I-2. On the other hand, the reduction in I-2 sensitivity in all PP1306 mutants was significantly less than that seen in CHRM2 (Table I and Fig. 2C). This suggested differences in the mode of action of I-1 and I-2 as PP1 inhibitors. Although the fusion of PP2A C terminus significantly reduced the inhibition of CHRM2 by I-2 (IC 50 1.35 M), the chimeric phosphatase still retained some I-2 binding, consistent with previous studies that suggested that multiple regions of I-2 mediated PP1 binding (39). In contrast, the activity of PP2A was unaffected by 1 M I-2 (Fig. 2C).
Association of WT and Mutant PP1␣ Catalytic Subunits with Regulatory Subunits-The diversity of RVXF sequences in PP1 regulators (3) and differences in various RVXF-containing peptides to disrupt cellular PP1 complexes (13) suggested that RVXF-containing PP1 regulators differ in their association with the surface hydrophobic pocket on PP1. To investigate PP1 binding to regulatory or targeting subunits with different RVXF sequences, we analyzed a recombinant G M , which contains the sequence RVSF (10), and a neurabin-I (Nrb) (21) polypeptide with the sequence KIKF; both sequences are required for PP1 binding. Full-length polypeptides, representing G M (1109 amino acids) and Nrb (1150 amino acids), are either poorly or not expressed in bacteria. Thus, we expressed GST-G M -(1-240) and GST-Nrb-(374 -516), which contain the key elements required for PP1 binding.
The binding of WT and mutant PP1 catalytic subunits to GST-Nrb and GST-G M was analyzed by sedimentation of these complexes bound to glutathione-Sepharose. The presence of GST fusion proteins was analyzed by protein staining with Ponceau S, and PP1 was detected by immunoblotting with an anti-PP1 antibody. PP1306 binding to GST-Nrb and GST-G M is shown in Fig. 3A. With increasing concentrations of targeting subunit, increased amounts of PP1306 were sedimented. Binding to CHRM2, L289R, M290K, and C291R was analyzed in a similar manner, and the bound PP1 catalytic subunits were quantified by densitometry. Under the conditions of this assay, binding to PP1306 was essentially saturated at 5 and 10 g of GST-G M (Fig. 3B). PP1␣ showed greater binding to GST-G M than PP1306, particularly at low concentrations (1 g) of the GST fusion protein (data not shown). Prior studies (14) suggested that CHRM2 bound more weakly than PP1␣ to GST-G M -(1-215). This difference was greatly magnified in our assays, with CHRM2 showing little or no binding to GST-G M -(1-240). All three point mutants, L289R, M290K, and C291R, were compromised in their ability to bind GST-G M . Although M290K and L289R bound GST-G M weakly, C291R, like CHRM2, failed to bind this fusion protein at all concentrations analyzed (Fig. 2B). This suggested that, as noted with I-1, the hydrophobic pocket, particularly the amino acid C291, played a critical role in PP1 binding by G M .
Surprisingly, under conditions that GST-G M failed to bind CHRM2, GST-Nrb bound CHRM2 weakly, requiring 10 g of GST-Nrb to visualize significant CHRM2 binding (Fig. 3B). This was unexpected as prior studies (15) had demonstrated no binding of CHRM2 to much higher concentrations of GSTspinophilin, a structural homologue of neurabin. This could reflect the use of a different fragment of spinophilin/neurabin-II, namely amino acids 298 -817, compared with residues 374 -516 of neurabin-I used in this study. Alternately, these differences could be attributed to the different assay conditions or some differences in the primary sequences of the PP1-binding site. Remarkably, unlike GST-G M , GST-Nrb bound both L289R and M290K almost as effectively as PP1306, indicating that these mutations did not totally disrupt the RVXF-binding pocket. By comparison, C291R showed little binding to GST-Nrb. These data suggested that Cys 291 is essential for binding to both GST-G M and GST-Nrb, but Leu 289 and Met 290 play a less critical role in binding of Nrb compared with G M .
Many PP1 targeting subunits, although promoting PP1 activity against "relevant" substrate(s), inhibited its phosphorylase phosphatase activity. This provides an alternate assay for evaluating the functional effects of PP1 binding by regulators. In contrast to G M , which only partially inhibits the phosphorylase phosphatase activity of the PP1 catalytic subunit (10), Nrb is a potent PP1 inhibitor in the same assay (21). Thus, GST-Nrb inhibited PP1␣ with an IC 50 of ϳ2 nM (Fig. 4 and Table I). By comparison, PP2A was essentially insensitive to GST-Nrb at concentrations up to 1 M (Fig. 4). PP1306 was inhibited by GST-Nrb with a modestly reduced IC 50 ϭ 7.3 Ϯ 0.6 nM. In contrast, neither CHRM2 nor C291R were inhibited by GST-Nrb at up to 1 M concentration (Fig. 4). L289R and M290K showed ϳ50-fold and 15-fold reductions in IC 50 for GST-Nrb, respectively (Table I). Although these experiments emphasized that changes in PP1 binding were not quantitatively linked to the altered enzyme activity, mutations of the RVXF-binding pocket had generally similar effects on both functions. For example, the weakened binding displayed by GST-Nrb for L289R and M290R led to 50-and 15-fold reduction in PP1 inhibitory activity. By contrast, C291R, which failed to bind GST-Nrb in pull-down assays, was unable to inhibit phosphorylase phosphatase activity of PP1 at concentrations up to 1 M.
Role of the RVXF Sequence in I-1 Function-The above studies suggested that RVXF sequences in various PP1 regulators bound differently within the surface hydrophobic pocket, and thus, were influenced variably by substitutions of different amino acids in the RVXF-binding pocket. To further test this hypothesis, we analyzed WT I-1 KIQF and a mutant I-1 RVTF . The mutant I-1 RVTF was created by substitution of the RVTF sequence derived from nuclear inhibitor of PP1, a PP1-binding protein that can be visualized readily by overlays (12). We utilized an overlay assay in which PP1-binding proteins were subjected to SDS-PAGE, and following electrophoretic transfer to polyvinylidene difluoride membranes, partially renatured. Incubation of membranes with soluble digoxigenin-conjugated PP1 catalytic subunits, followed by immunoblotting with antidigoxigenin antibody, allowed for direct comparison of PP1 binding to several different target proteins. However, due to the denaturation-renaturation involved in PP1 overlays, this assay favored the detection of PP1 regulators containing RVXF motifs (21).
GST-Nrb and GST-G M both showed dose-dependent binding by digoxigenin-coupled PP1306 (Fig. 5). By comparison to GST-Nrb and GST-G M , the PKA-phosphorylated GST-I-1 (GST-I-1-P) bound PP1306 very weakly, requiring 10-fold higher protein for detectable PP1 binding. At similar exposures as Nrb and G M , PP1306 binding to GST-I-1-P was undetectable. Substitution of the RVTF sequence in GST-I-1 increased PP1306 binding to GST-I-1 RVTF -P, albeit the mutant I-1 still bound PP1306 100-fold weaker than GST-G M . The unphosphorylated GST-I-1 and GST-I-1 RVTF , like GST alone, failed to bind any of the PP1 catalytic subunits (data not shown).
The binding of PP1␣ and mutant PP1 catalytic subunits to several regulators is summarized in Table II. These data showed that, in overlays, PP1306 bound more effectively than PP1␣ to all regulators analyzed. As noted in pull-down assays, compared with PP1306, all three mutants, L289R, M290K, and C291R, showed reduced binding to GST-G M . C291R also bound GST-Nrb more weakly than either L289R or M290K. The already weak PP1 binding to GST-I-1-P was essentially abolished by all three mutations in the RVXF-binding pocket. However, substitution of the RVTF sequence in I-1 allowed low but detectable binding by M290K. In general, results obtained with overlays (Table II) paralleled those seen with pull-down assays (Fig. 3). Together, these data suggested that not only the context, namely the parent PP1 regulator, but also the actual RVXF sequence played key roles in defining the affinity of PP1 for cellular regulators. DISCUSSION The primary structure of PP1 demonstrates Ͼ80% sequence identity from plants to animals, making PP1 one of the most highly conserved proteins in evolution (40). Consistent with its evolutionary conservation, PP1 regulates many critical functions in eukaryotic cells, including transcription, translation, metabolism, cell growth, and differentiation. Regardless of its species origin, PP1 can be readily distinguished from other protein serine/threonine phosphatases by its unique ability to be inhibited by the mammalian inhibitor proteins I-1 and I-2. Both of these PP1 regulators utilize multiple domains to bind and regulate PP1 activity. For example, in addition to the N-terminal KIQF sequence, PKA phosphorylation at threonine-35 is critical for PP1 inhibition by I-1 (4). More recent studies suggested that C-terminal sequences in I-1 also played a role in PP1 binding and regulation (41). Up to five different regions of the I-2 protein are thought to participate in PP1 regulation (39,42). These and other studies have fostered the hypothesis that cellular regulators have evolved a combinatorial mechanism, utilizing multiple interaction domains, some of which are common to a subset of regulators, to modulate cellular PP1 functions.
Our prior studies utilized a number of different ways to identify the structural determinants on the PP1 catalytic subunit that defined its regulation by cellular proteins. Such studies included the analysis of random mutations (23), surface "charged-to-alanine" substitutions (42), a "core" PP1 catalytic  5. PP1306 binding to PP1 regulators. PP1 overlays using digoxigenin-conjugated PP1 catalytic subunits were undertaken as described under "Experimental Procedures." The panels show a representative overlay demonstrating PP1306 binding to increasing concentrations of GST-Nrb, GST-G M , PKA-phosphorylated GST-I-1, and PKAphosphorylated GST-I-1 RVTF . PKA-phosphorylated GST was used as control. The upper panel shows protein staining with Ponceau S, and the lower panels shows immunoreactivity to anti-digoxigenin on film exposed for 30 s. Similar experiments were undertaken using other PP1 catalytic subunits and the results are summarized in Table II. subunit (amino acids 41-269) that lacked the N-and C-terminal sequences unique to specific PP1 isoforms and a chimeric PP1␣ (CHRM2), in which the C-terminal PP1 sequences were substituted with those from PP2A (14). Together, these studies highlighted the ␤12-␤13 loop (amino acids 269 -282) overhanging the catalytic site as a key determinant defining the sensitivity of PP1 to selected toxins and endogenous PP1 inhibitors, specifically I-1 and I-2. Comparative studies of PP1␣ and CHRM2 also suggested that the C terminus played a role in the affinity of PP1 for targeting subunits such as G M (14), PP1 nuclear targeting subunit and spinophilin (15), all of which either bound poorly or failed to bind CHRM2. These results were confirmed in the current studies, which emphasized the significant decrements in CHRM2 binding to not only G M but also neurabin and I-1.
The C-terminal residues of the PP1 catalytic subunit also contribute to the formation of a surface hydrophobic pocket that binds the RVXF motif conserved in many PP1 regulators (7). Prior studies noted that substitution of the sequence 290 MC 291 , which lines the hydrophobic pocket, impaired PP1 regulation by DARPP-32 and to a much lesser extent, I-2 (15). Some PP1 regulators also displayed a remarkable selectivity for distinct PP1 isoforms. Thus, G M selectively bound PP1␤ in skeletal muscle (29) and neuronal neurabin complexes primar-ily contained PP1␥ 1 (21,29). While the mechanism underlying PP1 isoform selectivity of regulators is not fully understood, it most likely reflects the divergent C-terminal sequences found in the different PP1 isoforms. To specifically focus on the functional analysis of the RVXF-binding pocket that is conserved in all PP1 isoforms, we undertook a limited C-terminal deletion of human PP1␣ to yield PP1306, which lacked the PP1␣-specific sequences. PP1306 was efficiently expressed in E. coli, and rapidly purified using one-step affinity chromatography with MCLR-Sepharose. The purified PP1306 showed increased enzyme activity against both phosphorylase a and MBP when compared with PP1␣. This was consistent with earlier studies, which showed that limited proteolysis of the 37-kDa PP1 catalytic subunit yielded a 35-kDa polypeptide with increased phosphorylase phosphatase activity (43). These studies suggested that the extreme C terminus of PP1␣ modulated its enzymatic activity.
Previous studies (44) that undertook modifications of the surface hydrophobic pocket in the single yeast PP1 catalytic subunit showed that alanine substitutions of two or more amino acids in this pocket induced lethality in yeast. Many of the "lethal" PP1 catalytic subunits, assayed as immunoprecipitates, displayed significant MBP phosphatase activity. Interestingly, yeast lethality correlated with near complete loss of FIG. 6. Association of the RVXF sequence with the hydrophobic pocket on the PP1 catalytic subunit. The structure of PP1␥1 bound to a synthetic peptide (representing amino acids 63-75) encompassing the RVSF sequence from the skeletal muscle glycogen-targeting subunit G M is shown (7). A, PP1 is shown in space-fill model using Rasmol. The G M peptide is shown as a stick model in red. The amino acids that comprise the RVXF-binding pocket are shown in blue, with residues modified in this study (L289, M290, and C291) shown in cyan. B shows the interactions of Leu 289 (red), Met290 (green), and Cys 291 (purple) with the G M peptide using the Mage program. The G M peptide is shown in red, and the amino acids lining the RVXF-binding pocket are shown in gray with cyan side chains. The strength of interactions of the RVSF sequence is displayed as dots generated using the Probe program and illustrate associations with Leu 289 (red), Met 290 (green), and Cys 291 (purple).

TABLE II
Comparison of WT and mutant PP1 catalytic subunits binding to mammalian PP1 regulators PP1 binding was analyzed using overlays as described under "Experimental Procedures," and binding was assessed by chemiluminescence. In three independent experiments, PP1 binding as seen by chemiluminescence was analyzed by exposure of the overlays to film for 2, 15, and 30 s.

GST-Nrb
GST-G M GST-I-1-P GST-I-1 RVTF -P GST-P PP1␣ a ϩϩ denotes a dark band seen in 15 s; ϩϩϩ denotes a dark band seen within 2 s; ϩ indicates a dark band in 30 s; ϩ/Ϫ denotes a faint band seen in 30 s; and Ϫ denotes no bands seen after prolonged exposure. phosphorylase phosphatase activity of the mutant PP1 catalytic subunits. As regulatory subunits modify the substrate specificity of the bound PP1 (3), these data suggested the inability of mutant PP1 catalytic subunits to bind one or more regulators required for PP1-catalyzed events essential for yeast viability. To investigate the role of the hydrophobic pocket in recognition of PP1 regulators, we introduced single substitutions in the 289 LMC 291 sequence that lined the RVXF-binding pocket in PP1306, substituting amino acids more commonly found in type-2 phosphatases (40). Specifically, Leu 289 was replaced with Arg found in a plant PP2A (45), Met 290 was substituted with Lys present in mammalian PP2A and C291 with Arg found in PP2B and several PP2A-like phosphatases. The high phosphorylase a and MBP phosphatase activity of all mutant PP1306 enzymes showed that these substitutions did not alter their catalytic function. This was further verified by demonstrating that the toxin sensitivity of the PP1306 mutants was similar to that of PP1␣.
C291R resulted in deficits in PP1 inhibition by I-1 and Nrb (Table I and Fig. 4) comparable to those seen in CHRM2, in which 53 PP1␣-specific C-terminal residues were replaced by 42 amino acids from the PP2A C terminus. In contrast, L289R and M290K had intermediate effects on PP1306 regulation by the same proteins. Direct binding analyses using either PP1 pull-down assays or overlays confirmed the critical importance of Cys 291 in binding Nrb, I-1, and G M . PP1 pull-down assays also showed that, whereas both L289R and M290K had modest effects on Nrb binding, M290K, and particularly L289R, were significantly impaired in their ability to bind G M . Interestingly, in yeast (44), the substitution C290A, corresponding to human Cys 291 , bound GAC1, the yeast homologue of G M effectively, and accumulated WT levels of glycogen. In contrast, L288A bound GAC1 poorly, and displayed a low glycogen phenotype. Whether these differences reflect the distinct RVXF motifs present in mammalian G M (RVSF) and yeast GAC1 (KNVRF), or the differing amino acid substitutions analyzed, remains unknown, but both L288A and C290A bound other yeast PP1 regulators, Reg1 and Sds22, like WT Glc7 (44). These data pointed to differences in the association of various PP1 regulators with the RVXF-binding pocket.
Our data highlighted three C-terminal amino acids, L288, M290, and C291, which bind the highly extended RVXF sequences (Fig. 6). Cys 291 and Leu 289 are particularly important, because Cys 291 makes direct contacts with the phenylalanine that is conserved in nearly all RVXF-containing PP1-binding proteins, and Leu 289 interacts with the valine or isoleucine also found in most PP1 regulators. Interestingly, x-ray crystallography of p53bp2, a known PP1-binding protein, suggested that the region encompassing the PP1-binding sequence, RVKF, is highly extended or linear (46), and is ideally suited for binding the surface hydrophobic pocket in a manner similar to that demonstrated by the synthetic G M peptide (7). Cocrystallization of PP1 with the targeting subunit MYPT1 confirmed the extended nature of the KVKF sequence bound in the surface hydrophobic pocket on the PP1␤ catalytic subunit (9). Comparison of the flanking sequences surrounding the RVXF motifs in MYPT1 and I-1 suggested that these sequences followed distinct paths across the rear surface of the PP1 catalytic subunit to approach the catalytic site on the opposing surface and thus elicit their distinct functional effects on PP1 activity (9). Recent studies also suggested that the N- (21) and C-terminal (29) sequences flanking the RVXF motif account for the selectivity of some regulators for distinct PP1 isoforms.
To establish the critical role played by the RVXF sequence in defining the affinity of PP1 for cellular regulators, we analyzed WT GST-I-1 and the mutant GST-I-1 RVTF (12). The mutant I-1 RVTF showed a Ͼ10-fold increased binding to PP1306 (Fig. 5) and was also a 3-to 5-fold more potent inhibitor than WT I-1 (12). The identity of the RVXF motif in I-2 remains controversial, but a similar substitution of RVTF in place of the proposed PP1-binding sequence, KLHY, abolished the ability of I-2 to inhibit PP1 holoenzymes. This demonstrated that the substitution of the same RVXF motif has opposing functional effects on I-1 and I-2 and further hinted that, compared with the known RVXF-containing regulators, G M , Nrb, and I-1, I-2 utilized a distinct mode of PP1 binding. On other hand, our previous studies (42) showed that an IKGI sequence in I-2 bound at a site adjacent to the hydrophobic pocket and, thus, I-2 binding was indirectly influenced by the substitutions L289R, M290K, and C291R. None of the pocket mutants elicited the dramatic loss in I-2 sensitivity seen in CHRM2, which was at least 200-fold less sensitive. This also provided evidence that I-2 differed from G M , Nrb, and I-1 in its association with PP1.
The current studies demonstrated the critical role played by the surface hydrophobic pocket in PP1 association with the three cellular regulators G M , Nrb, and I-1. They also highlighted the differing affinities of RVXF-containing PP1 regulators for the PP1␣ catalytic subunit, which may result from differences in the binding of distinct RVXF sequences with the surface hydrophobic pocket on the PP1 catalytic subunit. This diversity in PP1 interactions with different RVXF-containing regulators may be an important evolutionary strategy that allows further regulation of PP1 by physiological signals. Thus, the weak PP1 binding displayed by I-1, although enhanced by its phosphorylation by PKA, may be insufficient to displace more tightly bound PP1 regulatory subunits, such as G M and Nrb, and thus in the absence of other modifications, I-1 is unable to inhibit PP1 complexes containing these regulators. In this regard, it is worth noting that PKA also phosphorylates serines within or adjacent to the RVXF motifs in G M (47) and Nrb (48) to attenuate their association with the PP1 catalytic subunit. This in turn may facilitate the regulation of these cellular PP1 complexes by I-1 in response to hormones that elevate cAMP. Alternately, additional interactions between I-1 and regulatory subunits, such as the growth arrest and DNA damage-inducible protein GADD34 (49), may be required to circumvent the competition of these two RVXF-containing proteins for a common site on the surface of the PP1 catalytic subunit and still permit I-1 to transduce the hormonal signals that regulate the PP1/GADD34 complex and eukaryotic protein translation (41,49). In conclusion, further studies are clearly needed to investigate the contribution of all residues that make up the RVXF-binding pocket, and thus gain a full understanding of the role of the hydrophobic pocket in the binding of cellular regulators to distinct PP1 isoforms. However, our data point to a novel experimental strategy directed at modifying or eliminating PP1 association with selected proteins and thereby elucidate the physiological role of specific PP1 complexes.