A Protein Phosphatase-1γ1 Isoform Selectivity Determinant in Dendritic Spine-associated Neurabin*

Protein phosphatase-1 (PP1) catalytic subunit isoforms interact with diverse proteins, typically containing a canonical (R/K)(V/I)XF motif. Despite sharing ∼90% amino acid sequence identity, PP1β and PP1γ1 have distinct subcellular localizations that may be determined by selective interactions with PP1-binding proteins. Immunoprecipitation studies from brain and muscle extracts demonstrated that PP1γ1 selectively interacts with spinophilin and neurabin, F-actin-targeting proteins, whereas PP1β selectively interacted with GM/RGL, the striated-muscle glycogen-targeting subunit. Glutathione S-transferase (GST) fusion proteins containing residues 146–493 of neurabin (GST-Nb-(146–493)) or residues 1–240 of GM/RGL (GST-GM-(1–240)) recapitulated these isoform selectivities in binding and phosphatase activity inhibition assays. Site-directed mutagenesis indicated that this isoform selectivity was not due to sequence differences between the canonical PP1-binding motifs (neurabin, 457KIKF460; GM/RGL, 65RVSF68). A chimeric GST fusion protein containing residues 1–64 of GM/RGL fused to residues 457–493 of neurabin (GST-GM/Nb) selectively bound to and inhibited PP1γ1, whereas a GST-Nb/GM chimera containing Nb-(146–460) fused to GM-(69–240) selectively interacted with and weakly inhibited PP1β, implicating domain(s) C-terminal to the (R/K)(V/I)XF motif as determinants of PP1 isoform selectivity. Deletion of Pro464 and Ile465 in neurabin (ΔPI) to equally space a conserved cluster of amino acids from the (R/K)(V/I)XF motif as in GM/RGL severely compromised the ability of neurabin to bind and inhibit both isoforms but did not affect PP1γ1 selectivity. Further analysis of a series of C-terminal truncated GST-Nb-(146–493) proteins identified residues 473–479 of neurabin as containing a crucial PP1γ1-selectivity determinant. In combination, these data identify a novel PP1γ1-selective interaction domain in neurabin that may allow for selective regulation and/or subcellular targeting of PP1 isoforms.

Reversible protein phosphorylation regulates a multitude of cellular processes such as muscle contraction, glycogen metabolism, mitosis, synaptic plasticity, and learning and memory (reviewed in Refs. [1][2][3][4]. The importance of phosphatases in determining the balance of cellular protein phosphorylation/ dephosphorylation reactions has been increasingly recognized in recent years. Two distinct gene families, PPP and PPM, account for most serine/threonine phosphatase activity in mammalian extracts (reviewed in Refs. 5 and 6). The PPP family member, protein phosphatase-1 (PP1), 1 consists of four mammalian isoforms (␣, ␤, 2 ␥1, and ␥2), which share greater than 80% amino acid sequence identity (7)(8)(9). Despite their high homology, PP1 isoforms have distinct tissue and subcellular distributions, suggesting that they are localized via different mechanisms and may have different cellular functions (10 -12).
Subcellular targeting and substrate specificity of PP1 are determined largely by interaction of the catalytic subunits with a diverse family of PP1-binding proteins. The prototypical PP1binding protein, G M /R GL , 3 was identified in a skeletal muscle glycogen-associated PP1 holoenzyme (PP1 glycogen ) and targets PP1 catalytic subunits to glycogen and the sarcoplasmic reticulum (reviewed in Refs. 1, 2, 5, and 13). G M /R GL has been implicated in PP1 regulation of multiple enzymes involved in glycogen synthesis and degradation (14,15). Similarly, an actin-associated PP1 holoenzyme (PP1 actin ) was isolated from brain extracts and was shown to contain two homologous PP1binding proteins neurabin and spinophilin (also called neurabin II) (16 -18). Spinophilin and neurabin have been implicated in the regulation of glutamate receptors, G proteincoupled receptors (spinophilin only), and the stabilization of the actin cytoskeleton in filopodia and in dendritic spines (19 -23). PP1␤ was identified in purified PP1 glycogen (24), whereas the PP1 actin holoenzyme is selectively enriched in PP1␥1 (18). G M /R GL , spinophilin, and neurabin, like most PP1-binding proteins, contain the canonical (R/K)(V/I)XF-binding motif which is essential for stable PP1 binding (16,25,26). A short PP1-binding peptide from G M /R GL has been crystallized with recombinant PP1␥1, revealing multiple interactions between Arg 65 , Val 66 , and Phe 68 in G M /R GL with a hydrophobic groove on the surface of the PP1 catalytic subunit opposite to the catalytic site (27). Consequently, this interaction has little direct effect on PP1 activity. However, domains outside of the canonical (R/K)(V/I)XF motif of several PP1-binding proteins play ancillary roles in PP1 binding and can variably modulate PP1 activity (28 -32).
This report describes experiments directed at understanding the molecular basis for isoform-selective subcellular targeting of PP1 catalytic subunits. The isoform selectivity of native G M /R GL and native spinophilin/neurabin was further explored using co-immunoprecipitation assays. Recombinant fragments of G M /R GL and neurabin were then used to isolate domains necessary for selectivity toward native brain PP1 catalytic subunit isoforms. Our data implicate a 6-amino acid domain (residues 474 -479) C-terminal to the canonical (R/K)(V/I)XF PP1binding motif as playing a critical role in the selective interaction of neurabin with PP1␥1.
Muscle Tissue Extraction-Muscle tissue was prepared as previously described (34). The rat or mouse muscle was excised, freeze-clamped in liquid nitrogen, and stored at Ϫ80°C until use. Frozen tissue samples were pulverized under liquid nitrogen and then resuspended in 10 volumes (w/v) of 50 mM Tris/HCl, 0.5 mM EDTA, 2 mM EGTA, 100 mM NaF, 1% Triton X-100, 0.1 mM N-p-tosyl-L-lysine chloromethyl ketone, 2 mM benzamidine, 0.5 mM PMSF, 50 mM ␤-mercaptoethanol, and 10 mg/liter leupeptin. The tissue was homogenized with 15 passes using a motorized Teflon/glass homogenizer and centrifuged at 3800 ϫ g for 30 min to isolate a soluble extract.
Co-immunoprecipitations-Freshly prepared rat or mouse musclesoluble extracts or rat forebrain whole lysates (see above) were diluted to 1 mg/ml protein in IP buffer (50 mM Tris-HCl, 0.15 M NaCl, 1 mM EDTA, 0.5% Triton X-100, 1 mM PMSF, 5 mg/liter leupeptin, 20 mg/liter soybean trypsin inhibitor, 1 mM DTT, 0.5 mM benzamidine, pH 7.5), and 1 or 2 ml (respectively) was pre-cleared with 20 l of a 1:1 slurry of protein A-Sepharose or GammaBind Plus-Sepharose (Amersham Biosciences) for 30 min at 4°C. The supernatant was mixed for 2 h at 4°C with 5-10 l of indicated rabbit or sheep antiserum or 2 g of affinitypurified mouse G M /R GL antibodies raised in rabbit (15). After addition of 20 l of protein A-Sepharose (rabbit antibodies) or GammaBind Plus-Sepharose (sheep antibodies), incubations were continued for 2 h at 4°C. Resin was collected by microcentrifugation and then washed at least three times with 5 ml of IP buffer; during the last wash, resin was transferred to a new microcentrifuge tube. Immune complexes were solubilized in SDS-PAGE sample buffer. Aliquots of immune complexes, supernatants, and initial starting material were analyzed by immunoblotting using colorimetric detection techniques (18) or enhanced chemiluminescence (Fig. 1B) (34).
Phosphatase Catalytic Subunit Preparation-A mixture of native brain protein phosphatase catalytic subunits was isolated as described previously (18,33). Recombinant PP1␤ and PP1␥1 for immunoblotting standards were generous gifts from Dr. E. Y. C. Lee (New York Medical College) (9).
Glutathione-agarose Co-sedimentation Assays-10 g of the indicated GST fusion proteins were mixed for 1 h at 4°C with 15 g of crude phosphatase catalytic subunit mixture (see above) in 14 ml (stringent condition) or 1 ml (standard condition) of binding buffer (50 mM Tris-HCl, pH 7.5, 0.2 M NaCl, 0.1% Triton X-100, 0.25 mg/ml bovine serum albumin). About 20 l of a 50:50 slurry of glutathione-agarose was added, and the incubation was continued overnight at 4°C. The resin was sedimented and washed at least three times with 5-ml aliquots of binding buffer before being transferred to a 1.5-ml microcentrifuge tube. Proteins associated with glutathione-agarose were solubilized by boiling in SDS-PAGE sample buffer, and aliquots were analyzed by immunoblotting (18,33).
Far Western Overlay Assays-Purified GST fusion proteins were resolved by SDS-polyacrylamide gel and transferred to nitrocellulose membranes, which were stained with Ponceau-S (Mallinckrodt). Membranes from preparative gels were cut into vertical 3-mm strips, with each strip containing about 10 nmol of fusion protein (0.2-0.9 g of protein). Membranes were blocked (5 min, room temperature) with 50 mM Tris-HCl, pH 7.5, 137 mM NaCl, 0.27 mM KCl, 0.1% Tween 20 (v/v), 2% milk (w/v) and then washed (5 min) with PP buffer (10 mM Tris-HCl, pH 7.5, 154 mM NaCl, 0.1% BSA (w/v), 0.1% Tween 20 (v/v)). Membranes were incubated (overnight, 4°C) with or without crude phosphatase catalytic subunit mix diluted in PP buffer (1.2 g/ml) and then were washed three times for 5 min with PP buffer at room temperature. Primary antibodies to PP1␤, PP1␥1, PP2A, or GST were diluted in PP buffer and were incubated with membranes for 1 h at room temperature. Membranes were washed three times for 5 min with PP buffer and then incubated with alkaline phosphatase-conjugated secondary antibodies (rabbit, sheep, mouse, or goat) diluted in PP buffer (30 min, room temperature), washed again, and developed with alkaline phosphatase colorimetric assay (see details in Ref. 33).

RESULTS
Native Spinophilin, Neurabin, and G M /R GL Display PP1 Isoform Selectivity-Previous studies have shown that the brain PP1 actin holoenzyme, which contains spinophilin and neurabin, is selectively enriched in PP1␥1 (18). In contrast, the PP1␤ isoform has been shown to associate with the skeletal muscle PP1 glycogen holoenzyme containing G M /R GL (24), although the PP1 isoform selectivity of this complex has not been addressed. To further examine the isoform selectivity of PP1-binding proteins in brain and skeletal muscle, PP1␤ and PP1␥1 were specifically immunoprecipitated from whole tissue extracts, and the presence of spinophilin, neurabin, and G M /R GL was determined by immunoblotting. Neurabin (not shown) and spinophilin selectively co-immunoprecipitated with PP1␥1 (but not PP1␤) from rat forebrain lysates (Fig. 1A, left panel), extending previous observations using actin-enriched forebrain extracts (18). Furthermore, PP1␥1 immune complexes from mouse (not shown) and rat (Fig. 1A, right panel) skeletal muscle also contained spinophilin, despite the fact that spinophilin is expressed at much lower levels in muscle as compared with brain (compare brain and muscle applied lanes in Fig. 1A). Neurabin is not expressed in muscle (17). Spinophilin was not detected in PP1␤ immunoprecipitates from muscle extracts (Fig. 1A), even though there is 7-fold less PP1␥1 expressed in muscle relative to brain, whereas PP1␤ is expressed at similar levels in muscle and brain (Fig. 1, C and D). In contrast, immunoprecipitates of rat (Fig. 1A) or mouse (not shown) muscle PP1␤, but not PP1␥1, contained G M /R GL . Likewise, immu-noprecipitates of G M /R GL from mouse skeletal muscle contained PP1␤ but not PP1␥1 (Fig. 1B). As a control, PP2A immunoprecipitates from either rat brain or muscle tissue did not contain appreciable amounts of PP1␤, PP1␥1, spinophilin, neurabin, or G M /R GL (Fig. 1A). Moreover, PP2A was not precipitated with antibodies to any of the other proteins investigated here (data not shown). In combination, the present immunoprecipitation data significantly extend previous reports that native spinophilin and neurabin selectively interact with PP1␥1, whereas native G M /R GL selectively interacts with PP1␤.
Recombinant Fragments of Neurabin and G M /R GL Selectively Interact with Specific, Native PP1 Isoforms-GST fusion proteins containing the PP1-binding domains of G M /R GL (residues 1-240) and neurabin (residues 146 -493) were expressed in bacteria and purified (Fig. 2B). These proteins are referred to as GST-G M -(1-240) and GST-Nb-(146 -493) in the remainder of this report. Selective interactions between native mammalian PP1 isoforms and these fusion proteins were investigated using three complementary methods: 1) far Western overlay assays, 2) glutathione-agarose co-sedimentation assays, and 3) PP1 isoform inhibition assays. We chose to use native, rat brain PP1 isoforms for these assays because of the difficulties in expressing fully functional recombinant isoforms, especially in regard to the strength of their interactions with PP1-binding proteins (33,36). Because PP1␤ and PP1␥1 immunodetection protocols were optimized for comparable sensitivity, immunoblotting demonstrated that preparations of protein phosphatase catalytic subunit preparation contained approximately equimolar amounts of the two PP1 isoforms, as well as containing other phosphatases, such as PP2A (Fig. 2C) (33). This preparation was used for the far Western overlay and glutathione-agarose co-sedimentation assays, which assess stable, di-FIG. 1. G M /R GL and spinophilin make PP1 isoform specific interactions in muscle and brain tissue. A, rat forebrain or gastrocnemius muscle extracts were incubated with sheep antibodies to PP1␤ or PP1␥1 or mouse antibodies to PP2A. Aliquots of isolated immune complexes (P, 50% of total), immune supernatants (S, about 0.5% of total), and applied extracts (App., 20 g, 1% of total) were immunoblotted for the indicated proteins (see "Materials and Methods"). Sp, spinophilin; NS, nonspecific. B, mouse skeletal muscle extracts were incubated with rabbit antibodies recognizing mouse G M /R GL and immune complexes analyzed as in A. C, rat gastrocnemius muscle or rat whole brain (ϳ10 g) whole lysates were resolved on an SDS-PAGE gel and transferred to nitrocellulose membrane. After staining with Ponceau-S to permit normalization for protein loading (lower panel), membranes were probed for PP1␤ and PP1␥1. D, the relative abundance of the PP1 isoforms in brain and muscle (C) was compared by digitally scanning the immunoblots using Image J software (National Institutes of Health). The amount of PP1 isoform was normalized to the total amount of protein loaded, and the bar graph reports the -fold difference between brain and muscle for both PP1 isoforms (mean Ϯ S.E. of three experiments). rect interaction of PP1 catalytic subunits with the recombinant GST fusion proteins that may or may not affect PP1 activity. They differ in that GST fusion proteins undergo a denaturation/renaturation cycle for the overlay assays: consequently, this assay may favor detection of interactions that require shorter segments in the primary sequence of the fusion protein.
For PP1 isoform activity assays, PP1␤ and PP1␥1 were immunoisolated from the phosphatase catalytic subunit preparation (see "Materials and Methods"). Activity assays are performed in the continuous presence of the PP1-binding proteins and so may detect transient interactions that are not observed in the other assays. However, this method requires that the interaction affects catalytic activity.
Differences in the PP1-binding Motif Are Not Responsible for Isoform-selective Binding-To determine whether the specific sequence of the (R/K)(V/I)XF PP1-binding motif in neurabin contributed to PP1␥1 selectivity, the 457 KIKF 460 sequence in neurabin was mutated to RVSF, the specific PP1-binding sequence found in G M /R GL (Fig. 2B). The mutated GST-neurabin fusion protein, GST-Nb(KIKF3 RVSF), was analyzed using binding and inhibition assays. Despite the fact that GST-Nb(KIKF3 RVSF) contains the essential PP1-binding motif of G M /R GL , it maintained PP1␥1-selective interactions in co-sedimentation and overlay assays (Fig. 3). The binding to GST-Nb(KIKF3 RVSF) was generally weaker as compared with GST-Nb-(146 -493), possibly due to an attenuated affinity for PP1. GST-Nb(KIKF3 RVSF) also inhibited PP1␥1 Ͼ 20-fold more potently than it inhibited PP1␤, but both inhibition curves were significantly shifted to the right as compared with GST-Nb-(146 -493) (Fig. 4C). Thus, although replacement of the native neurabin 457 KIKF 460 sequence with RVSF residues from G M /R GL reduces the ability of neurabin to bind PP1 catalytic subunits, this domain is not a primary determinant for PP1 isoform selectivity by neurabin.
A Role for Sequences Flanking the PP1-binding Motif in Isoform-selective Binding-Because the specific sequence of the canonical PP1-binding motif of neurabin does not dictate PP1 isoform selectivity, domains flanking the PP1-binding motif were hypothesized to play a role. To address this issue, chimeric neurabin/G M fusion proteins were created using GST-G M -(1-240) and GST-Nb-(146 -493) as parent constructs. Both fusion proteins were constructed to preserve the PP1-binding motif of neurabin ( 457 KIKF 460 ) but alternating the N-and C-terminal flanking regions giving rise to GST-G M /Nb and GST-Nb/G M (Fig. 2B). Although GST-Nb/G M exhibited weaker binding to PP1 isoforms than either GST-G M -(1-240) or GST-Nb-(146 -493), it selectively bound PP1␤ over PP1␥1 in both interaction assays (Figs. 3). In addition, there was a weak but statistically significant, selective inhibition of PP1␤ by GST-Nb/G M (Fig. 4D). These data suggest that neurabin residues 146 -460 are not sufficient to make significant PP1␥1 interactions but that G M /R GL residues 69 -240 may contain PP1␤specific binding determinants. Moreover, the combination of Methods"). Purified GST or indicated GST fusion proteins (ϳ5 nmol each) were analyzed by SDS-PAGE, transferred to nitrocellulose, and immunoblotted using antibodies to GST. Full-length fusion proteins are indicated with asterisks. C, purified, bacterially expressed PP1␤ and PP1␥1 (5-50 ng as indicated) were analyzed by SDS-PAGE, transferred to nitrocellulose, and immunoblotted for the respective PP1 isoform (left). A native rat brain protein phosphatase catalytic subunit (PPC) preparation was fractionated by preparative SDS-PAGE and transferred to nitrocellulose membranes. Membranes were cut into 3-mm vertical strips and immunoblotted for PP1␤, PP1␥1, or PP2A (right).
neurabin residues 146 -460 and G M /R GL residues 69 -240 is not sufficient for potent inhibition and binding of PP1.
Role of Residues C-terminal to the PP1-binding Motif in PP1 Isoform Binding-The data presented so far suggest that do- FIG. 3. Analysis of GST fusion proteins in glutathione agarose co-sedimentation and far Western overlay assays. A, the indicated GST fusion proteins were incubated with a crude protein phosphatase catalytic subunit mixture under stringent conditions (14 ml of total volume), and the resulting complexes were sedimented with glutathione-agarose. After washing, bound proteins were solubilized in SDS-PAGE sample buffer and analyzed (ϳ30% of total per lane) by immunoblotting with antibodies to PP1␤, PP1␥1, PP2A C , or GST. A sample of the protein phosphatase catalytic subunit mixture (PPC Std.), representing 5% of the input, was analyzed in parallel. B, GST or the indicated GST fusion proteins were analyzed via SDS-PAGE and transferred to nitrocellulose. Membrane strips (ϳ10 nmol of protein/strip) were incubated with or without native protein phosphatase catalytic subunits (PPC) and then with PP1␤ (␤) or PP1␥1 (␥1) antibodies (see "Materials and Methods"). An additional strip of membrane for each fusion protein was incubated with GST antibodies to determine where full-length fusion protein migrated, as indicated by the arrows. M -(1-240), GST-Nb-(146 -493), and GST fusion protein chimeras. Activities of native PP1␤ (gray lines) and native PP1␥1 (black lines) were assayed in the presence of various concentrations of GST fusion proteins depicted in Fig. 2B. Each data point represents the mean Ϯ S.E. of 2-10 observations. Curves were fit to the inhibition data using the standard "One-site competition" equation included with Prism (GraphPad). Statistical comparison (analysis of variance) of PP1␤ and PP1␥1 inhibition curves for GST-Nb/G M confirmed that they were significantly different (p Ͻ 0.001). GST alone (2 M) did not significantly inhibit PP1␤ or PP1␥1 (data not shown). mains C-terminal to the PP1-binding motif play a critical role in determining PP1 isoform selectivity of GST-Nb-(146 -493) and GST-G M -(1-240). To provide additional insight into the neurabin domains that may contribute to PP1 binding, amino acid sequences surrounding the PP1-binding motifs of neurabin and G M /R GL were compared. We focused on neurabin residues 436 -479, because this domain was previously shown to be sufficient for PP1␥1 selectivity (18). Notably, neurabin contains three proline residues N-terminal to the PP1-binding motif and an additional proline residue C-terminal to the PP1binding motif, whereas G M /R GL does not contain any proline residues in this region (Fig. 5A). Consistent with the idea that these proline residues may play a role in PP1␥1 selectivity, the corresponding sequence of spinophilin, another PP1␥1-selective protein (Fig. 1) (18), also contains multiple proline residues highly conserved with neurabin (not shown). It seemed possible that structural constraints imposed by these proline residues may play a role in the PP1␥1 selectivity of neurabin. Therefore, single prolines in GST-Nb-(146 -493) were mutated to alanine (P436A, P447A, P453A, and P464A), and the mutated proteins were assayed for PP1 isoform selectivity. The single P453A and P464A mutations each reduced the potency for inhibition of both PP1␥1 and PP1␤ by 3-to 5-fold but did not affect PP1 isoform selectivity (Fig. 5B, top and middle panels). The P436A and P447A mutations, either individually or in combination, had no significant impact on the inhibition of these PP1 isoforms (data not shown). Similarly, co-sedimentation assays failed to detect noticeable changes in PP1 isoform binding following mutation of any of these proline residues (data not shown). In combination, these data suggest that proline residues nearest the canonical PP1-binding motif are not essential for isoform selectivity.

FIG. 4. Selective inhibition of native PP1 isoforms by GST-G
The PP1-binding motif of neurabin and Pro 464 are located N-terminal to a cluster of amino acids that is conserved in G M /R GL , but it is located two residues closer to the PP1-binding motif (black/gray boxes in Fig. 5A). Thus, deletion of Pro 464 and Ile 465 from neurabin (⌬PI) realigns the neurabin cluster as in G M /R GL as well as removing the potential structural constraint imposed by the proline residue. The ⌬PI mutation caused a substantial rightward shift of the inhibition curves for both PP1 isoforms (ϳ100-fold for PP1␥1) as compared with GST-Nb-(146 -493) (Fig. 5B, bottom). However, it was clear that the ⌬PI mutant protein retained PP1␥1-selective inhibitory properties. Similarly, binding of PP1␥1 to the ⌬PI mutant protein in cosedimentation assays was reduced to less than 5% of the level observed with wild type, but binding of PP1␤ was not consistently detected (data not shown). Thus, the ⌬PI mutation had a substantially more deleterious affect on the binding of PP1 isoforms than point mutation of Pro 464 to Ala. These data may indicate that Ile 465 plays a critical role in the interactions of neurabin with PP1. Alternatively, specific spacing between the PP1-binding motif and residues C-terminal to Pro 464 in neurabin may be essential for effective binding of PP1. However, the ⌬PI mutant retained selectivity for PP1␥1.
The point and deletion mutations directed at proline residues in GST-Nb-(146 -493) further supported the notion that domains C-terminal to the canonical PP1-binding motif in neurabin play a significant role in the interactions with PP1. However, they did not provide insight into the mechanism for PP1 isoform selectivity. As an additional approach to reveal potential mechanisms for PP1 isoform selectivity, a truncation mutagenesis strategy was employed. GST-Nb-(146 -493) was subjected to sequential C-terminal truncations producing GST-Nb-(146 -479), GST-Nb-(146 -473), GST-Nb-(146 -460) (terminating immediately after the canonical PP1-binding motif), and GST-Nb-(146 -453) (lacking the PP1-binding motif). In phosphatase inhibition assays (Fig. 5C, top), GST-Nb-(146 -479) was less efficacious at inhibiting either PP1␥1 or PP1␤ activity (ϳ70% maximum inhibition) as compared with GST-Nb-(146 -493) (ϳ95% maximum inhibition), although the potency of inhibition was decreased only ϳ3-fold for both isoforms (Table I). Thus, GST-Nb-(146 -479) retained the ϳ20-fold selectivity for inhibition of PP1␥1. Further truncation yielding GST-Nb-(146 -473) dramatically decreased the potency for inhibition of PP1␥1 but had only a modest effect on the potency for inhibition of PP1␤ or the maximum extent of inhibition, such that GST-Nb-(146 -473) displayed no PP1 isoform selectivity (Fig.  5C, middle panel, and Table I). The potency for inhibition of both PP1 isoforms was further reduced in parallel by additional C-terminal truncation to yield GST-Nb-(146 -460), which lacks all residues C-terminal to the PP1-binding motif (Fig. 5C). The GST-Nb-(146 -453) protein lacks the canonical PP1-binding motif and inhibits only about 20% of the activity at a 1.5 M concentration (Table I). In combination, the inhibition assays suggest that multiple domains C-terminal to the PP1-binding motif affect PP1 activity, but that residues 474 -479 of neurabin play an important role in PP1␥1-selective inhibition.
To investigate whether neurabin residues 474 -479 also play a role in stable and selective binding of PP1␥1, the C-terminal truncation mutants were analyzed by co-sedimentation and overlay assays (Fig. 6). Co-sedimentation assays were per-formed under standard conditions using higher protein concentrations than in previous experiments (Fig. 3A), yielding more readily detectable binding of PP1␤ to GST-Nb-(146 -493) (Fig.  6A). Nevertheless, binding of PP1␤ was 2-to 3-fold less than that binding of PP1␥1. In addition, GST-Nb-(146 -493) displayed similar selectivity for binding of PP1␥1 over PP1␤ in overlays. Truncation to generate GST-Nb-(146 -479) had no significant impact on the binding of either PP1␤ or PP1␥1 in either the co-sedimentation or overlay assays. However, further truncation to GST-Nb-(146 -473) resulted in an almost complete loss PP1␥1 binding, but less that a 2-fold diminution of PP1␤ binding, such that this protein is now PP1␤-selective in these assays. No significant binding of either isoform to GST-Nb-(146 -460) or GST-Nb-(146 -453) was detected. Moreover, GST-G M -(1-240) displayed the expected strong PP1␤ selectivity, and PP2A C did not bind to any of these GST fusion proteins. Qualitatively similar results were obtained in co-sedimentation assays performed under more stringent conditions (data not shown). In combination, these data indicate that residues 474 -479 of neurabin play a critical role in the stable binding of PP1␥1, without having substantial impact on the weaker binding of PP1␤. DISCUSSION Prior work revealed that the postsynaptic proteins spinophilin and neurabin are selectively associated with PP1␥1 over PP1␤ in F-actin-enriched subcellular fractions from brain (18). Here we extend this observation by showing similar selectivity in the context of both whole brain and whole muscle lysates (Fig. 1A). This is significant, because spinophilin and PP1␥1 are expressed at much lower levels in muscle such that PP1␥1 selectivity was evident despite a 7-fold elevated ratio of PP1␤ to PP1␥1 (Fig. 1, C and D). In addition, we demonstrate that G M /R GL , a striated-muscle-specific PP1-binding protein, preferentially binds PP1␤ and not PP1␥1 (Fig. 1, A and B). Although PP1␤ has been identified in PP1 glycogen holoenzymes containing G M /R GL (24), the PP1 isoform selectivity of G M /R GL has not previously been reported.
Highly specific PP1␤ and PP1␥1 antibodies were then used in three complementary assays (glutathione-agarose co-sedimentation assays, far Western overlay assays, and isolated PP1 isoform inhibition assays) to assess interactions of native PP1 isoforms with recombinant fragments of neurabin and G M /R GL expressed as GST fusion proteins. Because these complementary techniques report on different aspects of interactions of PP1 with PP1-binding proteins (see "Results"), they may yield somewhat varying results. However, that the majority of our data using GST-G M -(1-240) and GST-Nb-(146 -493) are consistent in all three assays indicates that these domains contain elements critical for selective binding and inhibition of PP1 isoforms. Moreover, this isoform selectivity recapitulates   Pooled inhibition data (see Fig. 5C) were fitted with the "One-site" competition model included with Prism software (GraphPad) to estimate the maximal inhibitory effect (bottom value) and potency (EC 50 ) of the interaction for each GST fusion protein with native PP1␤ and PP1␥1 (the "top value" was fixed at 100% for curve fitting). The "best fit" values are reported with the 95% confidence interval. Data for GST-Nb-(146 -153) and GST-Nb-(146 -460) could not be fitted with any accuracy, because technical limitations precluded analyses at higher protein concentrations. The average inhibition detected at 1.5 M of each fusion protein is also reported as the mean Ϯ S.E. of five to six observations. GST the isoform selectivity of native neurabin and native G M /R GL isoform in tissue extracts ( Fig. 1) (18). The use of native PP1 catalytic subunit isoforms rather than recombinant proteins is critical, because GST-Nb-(146 -493) inhibits recombinant PP1␥1 (50% inhibition at 200 nM (18,33)) about 100-fold less potently than it inhibits native PP1␥1 (Figs. 4 and 5 and Table  I). Thus, recombinant PP1 isoforms do not necessarily recapitulate all of the properties of the native proteins, as documented previously (36). Consequently, experiments characterizing interactions between PP1-binding proteins and recombinant PP1 isoforms should be interpreted with caution. Although the canonical PP1-binding motif is essential for both G M /R GL and neurabin to interact with PP1 (16,25,37), mutation of the PP1-binding motif of neurabin ( 457 KIKF 460 ) to a G M /R GL -like PP1-binding motif (GST-Nb(KIKF3 RVSF)) had no significant affect on PP1 isoform selectivity (Figs. 3 and 4). Therefore, amino acids critical for selectivity reside outside the essential PP1-binding motif. This observation is perhaps not surprising, because the (R/K)(V/I)XF peptide from G M /R GL interacts with residues forming the hydrophobic groove on recombinant PP1␥1 that are absolutely conserved in PP1␤ (27). However, although selectivity was unaffected, the interactions with both PP1 isoforms were modestly weakened by the 457 KIKF 460 to RVSF mutation. This is consistent with previous reports that mutation of non-essential residues within the canonicalbinding motif affect PP1 binding. Substitution of the PP1binding motif in Nuclear Inhibitor of PP1 (NIPP1) with thebinding motif from inhibitor-1 results in a mutation (KIQF to RVTF) that is similar to the neurabin KIKF to RVSF mutation reported in the present work, but the affinity of NIPP1 for PP1 increased (29). Similarly, the impact of the identity of the amino acid in the 'X ' position in a canonical (R/K)(V/I)XF motif also depends on the broader sequence context. Mutation of the 'X ' position in inhibitor-1 has little impact on PP1-binding unless a proline is inserted (29), but even conservative mutations at the 'X ' position in G M /R GL (Ser to Val or Ala) or metabotropic glutamate receptor 7b (Thr to Ala) attenuate PP1 binding (37,38). The different consequences of these seemingly similar mutations may result from the broader sequence context of the canonical PP1-binding motifs in these proteins. Taken together, these data indicate that the ability of the canonical PP1-binding motif to anchor PP1 is significantly influenced by surrounding sequences.
Because the specific sequence of the canonical PP1-binding FIG. 6. Binding of native PP1 isoforms to C-terminal truncation mutants of GST-Nb- (146 -493). A, the indicated GST fusion proteins (10 g) were incubated with a crude protein phosphatase catalytic subunit mixture (15 g of total protein) under standard conditions (1 ml of volume), and resulting complexes were sedimented with glutathione agarose. Top panel: after washing, bound proteins were solubilized in SDS-PAGE sample buffer and analyzed (about 30% of total per lane) by immunoblotting with the indicated antibody. A sample of the protein phosphatase catalytic subunit preparation (equivalent to 30% of the input) was analyzed in parallel (Input). The lower panel shows the Ponceau-stained nitrocellulose membrane prior to immunoblotting to reveal the amounts of GST fusion protein in each lane. The arrowhead labeled "BSA" marks contaminating bovine serum albumin from the incubation buffer. Bottom panel: immunoblots were digitally scanned and quantitated using Image J software (National Institutes of Health (NIH)); the amount of sedimented phosphatase was expressed as a percentage of the input. The bars represent the mean Ϯ S.E. of three (PP1␤ and PP1␥1) or two (PP2A) observations. B, the indicated GST fusion proteins (ϳ10 nmol of each) were analyzed by SDS-PAGE and transferred to nitrocellulose. A sample of the protein phosphatase catalytic subunit preparation was analyzed in parallel (PPC Std.). Top panel: membranes were incubated with or without (not shown) the protein phosphatase catalytic subunit preparation and immunoblotted for PP1␤, PP1␥1, or PP2A (not shown). Bottom panel: immunoblots were digitally scanned and quantitated using Image J software (NIH); the amount of bound PP1 isoform was expressed as a percentage of the protein phosphatase catalytic subunit standard. Each bar represents the mean of two observations in independent experiments. motif did not appear to dictate PP1 isoform selectivity in neurabin or G M /R GL , we hypothesized that flanking domains may play an important role in this selectivity. Two chimeras of GST-G M -(1-240) and GST-Nb-(146 -493) were made, maintaining the neurabin PP1-binding motif and swapping domains N-and C-terminal to this motif (Fig. 2B). GST-Nb/G M selectively interacted with PP1␤ over PP1␥1, albeit more weakly that GST-G M -(1-240) (Fig. 3, A and B). In contrast, GST-G M /Nb exhibited weak but PP1␥1-selective binding in overlay assays (Fig. 3B), and it inhibited PP1␥1 with reduced potency compared with GST-Nb-(146 -493) but did not bind or inhibit PP1␤ (Fig. 4E). Taken together, these data indicate that complementary domains both N-and C-terminal to the canonicalbinding motif in GST-G M -(1-240) and GST-Nb-(146 -493) are required for stable binding and potent/effective inhibition. However, the selectivity between PP1␤ and PP1␥1 isoforms appears to be primarily dictated by the more C-terminal domains. The interpretation that domains C-terminal to the canonical PP1-binding motif play a role in isoform selectivity is consistent with a recent report that selective interaction of PP1␣ with the type 1 inositol (1,4,5)-trisphosphate receptor requires a canonical PP1-binding motif plus residues C-terminal to the motif (39).
Because site-directed mutagenesis targeted at proline residues surrounding the canonical PP1-binding motif of neurabin failed to identify PP1 isoform selectivity determinants (Fig.  5B), we adopted a more general truncation approach to the domains C-terminal to the-binding motif. The initial truncation to generate GST-Nb-(146 -479) displayed comparable PP1 binding and isoform selectivity (ϳ20-fold in inhibition assays) to the parent GST-Nb-(146 -493) (Figs. 5C and 6), as expected because GST-Nb-(436 -479) also is PP1␥1-selective (18,40). The EC 50 for inhibition of both PP1␥1 and PP1␤ was increased about 3-fold ( Fig. 5C and Table I), and interestingly, increasing GST-Nb-(146 -479) from 20 nM to 1.5 M failed to increase the extent of inhibition of PP1␥1 beyond about 70%, whereas maximal concentrations of GST-Nb-(146 -493) inhibited about 95% of PP1␥1 activity (Table I). Although it was not possible to establish a true plateau for inhibition of PP1␤ by GST-Nb-(146 -479) or GST-Nb-(146 -493), curves fitted to these data indicate that the maximal extents of inhibition by these two proteins are similar to those for PP1␥1 (Table I). These data could be interpreted in two ways. First, residues within the 480 -493 domain may make specific isoform-non-selective interactions with PP1 that disrupt the active site. However, the fact that GST-Nb-(146 -479) and GST-Nb-(146 -493) bind similar amounts of the two isoforms (Fig. 6, A and B) suggests that interactions of this domain have little impact on stable binding. Thus, the alternative possibility is that residues 480 -493 form a steric block of the PP1 active site without making direct interactions. It is tempting to speculate that removal of neurabin residues 480 -493 from the PP1 active site might activate PP1 while the catalytic subunit remains associated with neurabin, although no such mechanism has yet been identified.
The removal of an additional six C-terminal amino acids to generate GST-Nb-(146 -473) had a very significant impact: stable binding of PP1␥1 was completely abrogated and the EC 50 for PP1␥1 inhibition increased about 25-fold (Figs. 5B, 6A, and 6B and Table I). In contrast, stable binding of PP1␤ was only slightly decreased and was not significantly different from either GST-Nb-(146 -479) or GST-Nb-(146 -493) in cosedimentation assays (Fig. 6A). In addition, the EC 50 for PP1␤ inhibition by GST-Nb-(146 -473) was not significantly changed as compared with GST-Nb-(146 -479), and curve fitting suggests that inhibition of both isoforms plateaus at ϳ40 -50%, not significantly different from the plateau inhibition values with GST-Nb-(146 -479) ( Fig. 5C and Table I). Thus, GST-Nb-(146 -473) displays no isoform selectivity in inhibition assays. However, GST-Nb-(146 -473) does form a selective stable interaction with PP1␤, which may result from the presence of a previously characterized weak PP1␤-selective domain N-terminal to the (R/K)(V/I)XF motif (40). In combination, these data strongly suggest that residues 474 -479 of neurabin contain a PP1␥1-selectivity determinant.
Neither GST-Nb-(146 -460) protein, which was truncated immediately following the canonical PP1-binding motif, nor GST-Nb-(146 -453), which lacked the canonical-binding motif and all residues C-terminal to that, was able to bind detectable amounts of PP1␤ or PP1␥1 in overlay or co-sedimentation assays (Fig. 6). Because the GST-Nb-(146 -460) construct does not bind PP1 in these assays, it supports our conclusion from the chimeric data suggesting that domains downstream of the canonical-binding motif play a role in stable association of PP1. In addition, the potency of PP1 inhibition by either construct was substantially reduced, and there was no evidence for PP1 isoform selectivity (Fig. 5C). GST-Nb-(146 -460) inhibited about 25% of PP1␤ or PP1␥1 activity at a concentration of 1.5 M (Table I), whereas 25% inhibition was detected with 20 -50 nM GST-Nb-(146 -473) (Fig. 5C). Therefore, residues 461-473 contain determinants that allow for more potent inhibition of both isoforms as well as the stable binding of PP1␤.
Combining the present data with previous reports, there appear to be at least five domains that play a role in the interactions of neurabin with PP1 isoforms. The canonical PP1binding motif, which binds to the hydrophobic groove in the catalytic subunit (27), is essential for interaction with PP1 ( Fig.  6) (16,18,40). Second, Terry-Lorenzo et al. (40) identified determinants N-terminal to the neurabin canonical PP1-binding motif that were important for the weak PP1␤ interaction (residues 421-443). The present data identify three additional domains C-terminal to the canonical PP1-binding motif that also regulate PP1 binding and activity. Residues 461-473 and 480 -493 contain determinants that are not selective for either isoform tested here, whereas residues 474 -479 interact strongly with PP1␥1 but play no detectable role in binding to PP1␤. Thus, these data support a model for multiple interactions between PP1 and neurabin (Fig. 7) that is consistent with more general models of PP1 interactions with its targeting/ regulatory proteins (28 -30). Some of the interactions with neurabin are required for stable PP1 binding (residues 461-473 and residues 474 -479), whereas others permit regulation FIG. 7. Working model of identified PP1-interacting domains in neurabin. Residues 421-493 of neurabin contain multiple domains that have been implicated in binding to PP1 isoforms. The canonical (R/K)(V/I)XF motif (residues 457-460; indicated with an oval) anchors PP1 to PP1-binding proteins (18,40). Residues 421-443 were previously implicated as a weak PP1␤-selective domain (36). The present data suggest three independent interactions C-terminal to the (R/K)(V/ I)XF motif. Residues 460 -473 make interactions that increase the potency of inhibition in an isoform-independent manner but stabilize PP1␤-selective binding. Residues 480 -493 appear to increase the maximal extent of PP1 inhibition in an isoform-independent manner, without significantly affecting the potency of the interactions. Residues 474 -479 (black box) appear to make a novel interaction that confers PP1␥1 selectivity on neurabin constructs, recapitulating the isoform selectivity observed in tissue extracts. However, the full potency of neurabin interactions with PP1 isoforms requires the presence of neurabin sequences on both sides of the (R/K)(V/I)XF motif. of PP1 activity (residues 480 -493).
This model is reminiscent of the interaction of inhibitor-2 with muscle PP1. In total, five domains in inhibitor-2 have been shown to contribute to high affinity binding and regulation of PP1 (30,41). Residues 1-114 of inhibitor-2 display only a slightly decreased affinity for PP1 as compared with fulllength inhibitor-2, but could not fully suppress PP1 activity even at 1000-fold higher concentrations (30). This also suggests that there are separate domains of inhibitor-2 that are responsible for binding versus inhibition. In addition, both DARPP-32 and inhibitor-1 contain an (R/K)(V/I)XF motif that is essential for binding, but these proteins only bind and inhibit PP1 when phosphorylated by protein kinase A 23 residues C-terminal to the PP1-binding motif (31,32). In contrast, in vitro phosphorylation of G M /R GL at serine 48 (17 residues N-terminal to the PP1-binding motif) may increase the activity of the associated PP1 toward glycogen synthase (14). Overall, there is a growing consensus that (R/K)(V/I)XF motifs anchor PP1 to its binding proteins and facilitate a diverse array of secondary interactions that play a role in modulating the overall strength of the interactions, regulating the activity of the associated catalytic subunit, and conferring PP1 isoform selectivity (28).
In summary, the present data identify for the first time a specific domain in a PP1-binding protein that plays a critical role in distinguishing between different PP1 isoforms. Although there are domains both N-and C-terminal to the canonical PP1-binding motif that regulate isoform nonspecific interactions, neurabin residues 474 -479 located 13 amino acids C-terminal to the canonical-binding motif dictate PP1␥1selective interactions with neurabin constructs. Even though the four PP1 catalytic subunit isoforms share greater than 80% amino acid sequence identity, the few amino acid sequence and/or structural differences between PP1␤ and PP1␥1, presumably in the more variable N-and/or C-terminal domains, must allow for the isoform selectivity of neurabin and G M /R GL . Although the precise physiological relevance of isoform-specific interactions is unclear, it is likely that they differentially target PP1 isoforms to distinct subcellular structures/locations. Further studies are needed to identify potentially specific roles for each PP1 isoform in cellular physiology.