The inhibitor-1 C terminus facilitates hormonal regulation of cellular protein phosphatase-1: functional implications for inhibitor-1 isoforms.

Inhibitor-1 (I-1) is a selective inhibitor of protein phosphatase-1 (PP1) and regulates several PP1-dependent signaling pathways, including cardiac contractility and regulation of learning and memory. The human I-1 gene has been spliced to generate two alternative mRNAs, termed I-1alpha and I-1beta, encoding polypeptides that differ from I-1 in their C-terminal sequences. Reverse transcription-PCR established that I-1alpha and I-1beta mRNAs are expressed in a developmental and tissue-specific manner. Functional analysis of I-1 in a Saccharomyces cerevisiae strain dependent on human I-1 for viability established that a novel domain encompassing amino acids 77-110 is necessary for PP1 inhibition in yeast. Expression of human I-1 in S. cerevisiae with a partial loss-of-function eukaryotic initiation factor-2alpha (eIF2alpha) kinase (Gcn2p) mutation permitted growth during amino acid starvation, consistent with the inhibition of Glc7p/PP1, the yeast eIF2alpha phosphatase. In contrast, human I-1alpha, which lacks amino acids 83-134, and I-1 with C-terminal deletions were significantly less effective in promoting yeast growth under starvation conditions. These data suggest that C-terminal sequences of I-1 enhance regulation of the eukaryotic eIF2alpha phosphatase. In vitro studies established that C-terminal sequences, deleted in both I-1alpha and I-1beta, enhance PP1 binding and inhibition. Expression of full-length and C-terminally truncated I-1 in HEK293T cells established the importance of the I-1 C terminus in transducing cAMP signals that promote eIF2alpha phosphorylation. This study demonstrates that multiple domains in I-1 target cellular PP1 complexes and establishes I-1 as a cellular regulator of eIF2alpha phosphorylation.

Protein phosphatase-1 (PP1) 1 is a major eukaryotic serine/ threonine protein phosphatase that controls numerous physiological processes, including protein synthesis, gene expression, the cell cycle, cardiac contractility, and neuronal signaling (1). Regulation of these diverse PP1 functions in mammalian tissues is mediated by the interaction of the PP1 catalytic subunit with a number of targeting subunits and endogenous protein inhibitors (2). The prototypic mammalian PP1 regulator, Inhibitor-1 (I-1), requires phosphorylation by protein kinase A (PKA) to inhibit PP1. I-1 is widely expressed in mammalian tissues and is a critical regulator of PP1 function in many cAMPregulated physiological processes, including cardiac contractility (3) and neuronal signaling (4). The I-1 homolog DARPP-32 (dopamine-and cAMP-regulated phosphoprotein of apparent M r 32,000) is predominantly expressed in neurons. I-1 and DARPP-32 share extensive sequence homology within an Nterminal domain that encompasses the threonine residue phosphorylated by PKA. N-terminal fragments of I-1 and DARPP-32 including the phosphorylated threonine inhibit PP1 in vitro (5,6). Such studies also highlighted a conserved tetrapeptide sequence (KIQF) that is critical for PP1 inhibition by I-1 and DARPP-32. Subsequent studies that noted the presence of homologous sequences in several PP1 regulators as well as co-crystallization of the PP1 catalytic subunit with a peptide containing an (R/K)(I/V)XF sequence established that this domain represents a conserved PP1-binding motif found in many PP1 regulators (7).
Although the N-terminal PP1 inhibitory domains of DARPP-32 and I-1 are nearly identical, the C-terminal sequences are highly divergent, and the precise role of these sequences is largely unknown. Both I-1 and DARPP-32 are phosphorylated at serine residues in their C-terminal domains, and studies of DARPP-32 have provided the clearest insights into the function of these covalent modifications. Serine phosphorylation of DARPP-32 by casein kinases I and II modulates the phosphorylation and dephosphorylation of DARPP-32 at Thr 34 and thereby regulates DARPP-32 function as a PP1 inhibitor (8,9). In contrast, our studies of I-1 showed that the C terminus plays an important role in binding other cellular proteins, specifically GADD34, a scaffolding protein that also binds PP1; this generates a heterotrimeric complex with eukaryotic initiation factor-2␣ (eIF2␣) phosphatase activity (10). We speculated that the C terminus targets I-1 to specific cellular PP1 complexes, such as GADD34⅐PP1, and that, following PKA phosphorylation, I-1 inhibits eIF2␣ phosphatase activity. In this manner, the GADD34⅐PP1⅐I-1 complex represents a mechanism by which cAMP could regulate mammalian protein synthesis.
The human I-1 gene is alternatively spliced to generate two additional mRNAs, termed I-1␣ and I-1␤ (11). Full-length I-1 differs from the I-1␣ and I-1␤ polypeptides in their C-terminal sequences. I-1␣ contains a deletion of amino acids 83-134, whereas I-1␤ results from a frameshift at amino acid 61 and possesses a unique C terminus. Until this work, the functions of these novel I-1 isoforms as PP1 inhibitors, specifically the holoenzymes found in eukaryotic cells, have not been analyzed.
Biochemical studies show that human I-1 inhibits the yeast PP1 catalytic subunit, Glc7p, in an indistinguishable manner from human PP1 (12) and that expression of I-1 impacts only select PP1-dependent signaling pathways (13). Moreover, in Saccharomyces cerevisiae strain JC1007-97, expression of active human I-1 is essential for growth and viability (13). This provides a novel cell-based assay for examining the structurefunction of I-1 as a regulator of PP1 holoenzymes. Our work validates the use of yeast as a model eukaryotic system for analyzing the function of human I-1 and shows, for the first time, that selected C-terminal residues in human I-1 are required for effective PP1 inhibition in yeast. Analysis of tissues from wild-type (WT) I-1 and I-1-null mice established that three distinct mRNAs are transcribed from a single mouse I-1 gene and confirmed that alternative splicing yields the two additional transcripts, I-1␣ and I-1␤, in many mouse tissues. Utilizing S. cerevisiae strain Y27, in which Glc7p was previously shown to dephosphorylate eIF2␣ to allow the recovery of yeast from amino acid starvation (14), we established that human I-1␣, which lacks C-terminal residues present in fulllength I-1, is a less effective inhibitor of phosphatase activity. In vitro biochemical experiments demonstrated that the Cterminal region of I-1 enhances binding to the PP1 catalytic subunit and inhibition of phosphatase activity. Expression of I-1 in HEK293T cells demonstrated that the I-1 C terminus facilitated cAMP signals that inhibit protein translation in mammalian cells and suggested distinct roles for the newly defined isoforms, I-1␣ and I-1␤, in regulating cellular PP1 activity.
Human I-1 Expression Plasmids-All plasmids used in this study are listed in Table I. Deletions of human I-1 were generated using JZ205 as template. With the QuikChange site-directed mutagenesis kit (Stratagene), stop codons were added at amino acids 142, 123, 97, 77, and 54 of human I-1 (hI-1). The region encoding C-terminal amino acids 81-171 was excised from pGEM3ZF/I-1 using SacI and HindIII and ligated into pGEX-4T using EcoRI and XhoI. The N-terminal region of I-1 cDNA (encoding amino acids 1-80) was excised from pGEM3ZF/I-1 using NcoI and SacI and ligated into pRSET-B, which was then digested with BglII and SalI, and the resulting cDNA was inserted into pGEX-4T using BamHI and SalI. Expression constructs for I-1␣ were generated from the hI-1 cDNA by introducing NheI sites at codons 83 and 134. Following digestion with NheI, the cDNA was religated to produce a cDNA encoding hI-1␣ with one exception, with an alanine substitution at residue 84. Mammalian expression plasmids were generated using PCR primers that added BglII and SalI restriction sites to the hI-1 cDNA. The PCR product was digested with BglII and SalI and ligated into pCMV-FLAG-2 (Sigma) digested with BamHI and SalI. All cDNAs were confirmed by sequencing at the Duke Comprehensive Cancer Center DNA Sequencing Facility.
Yeast Assay for Amino Acid Starvation-The ability of hI-1 to suppress the phenotype of the gcn2-507 allele was assayed as described (14,15). Briefly, strains Y27, H1402, and H1149 were transformed with plasmids expressing hI-1 or dominant-negative GLC7. Cultures were grown at 30°C on synthetic medium lacking uracil and histidine and containing 40 mM leucine and 30 mM 3-aminotriazole (3-AT) or on control medium lacking uracil. Growth was monitored as formation of yeast colonies.
Sectoring Assay in S. cerevisiae gfa1-97-Yeast sectoring assays were performed as described (13). Briefly, the yeast strain JC1007-97 containing JZ203 was transformed with competing hI-1-expressing plasmids, and yeast cells were grown on synthetic medium lacking uracil and reduced adenine (6 g/ml). The red-and-white sectoring colonies were counted after 4 days of growth at 30°C. Deletions of hI-1 were generated using gap repair. PCR primers were designed that included the GDP promoter and various 5Ј-regions of hI-1 cDNA from JZ205. The PCR-amplified cDNAs in the linearized plasmids YEp195 (2 URA3) and YCp33 (CEN URA3) were transformed into JC1007-97, and yeast cells were grown as described (13).
Immunoblotting of Yeast Lysates-Yeast cells were grown on nonselective medium at 30°C to A 600 ϭ 0.6. The cells were centrifuged at 5000 ϫ g for 5 min. The pellet was resuspended in 500 l of ice-cold lysis buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.1% (v/v) ␤-mercaptoethanol, 0.5% (w/v) Triton X-100, 1 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, 1 g/ml aprotinin, and 1 g/ml pepstatin), mixed with an equal volume of glass beads, and vortexed at full speed for five bursts of 20 s. The lysate was then cleared by centrifugation at 20,000 ϫ g for 10 min. The lysate was analyzed by immunoblotting.
PP1 Sedimentation-Glutathione-Sepharose beads (25-l bed volume) were washed twice with Tris-buffered saline (10 mM Tris-HCl (pH 7.5) and 150 mM NaCl). GST fusion proteins were incubated with washed beads in Tris-buffered saline (total volume of 300 l) for 1 h at 4°C. Protein-bound beads were sedimented by centrifugation at 500 ϫ g and rinsed twice with Tris-buffered saline. HEK293T cell lysate (300 g of total protein) in Tris-buffered saline, 1 mM EDTA, 1 mM EGTA, and 0.1% (w/v) Triton X-100 was incubated with the GST fusion proteinbound beads for 1 h at 4°C. Beads were washed five times with 1 ml each of the same buffer, resuspended in SDS-PAGE sample buffer, and analyzed by SDS-PAGE. PP1 was detected by Western immunoblotting with anti-PP1 antibody. All PP1 sedimentation assays were performed three times with two independent preparations of recombinant I-1 protein.
Phosphorylase Phosphatase Assay-Phosphorylase phosphatase assays were performed in 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 0.1% (v/v) ␤-mercaptoethanol with PP1 catalytic subunit (0.02 units) purified from rabbit skeletal muscle and hI-1 proteins. I-1 was phosphorylated using purified bovine cardiac PKA catalytic subunit (16). Two independent preparations of each I-1 fusion peptide were phosphorylated. The stoichiometry of phosphorylation was established by "back-phosphorylation" of all phosphorylated I-1 polypeptides using fresh PKA and [ 32 P]ATP and analysis of 32 P incorporation to ensure 1 mol/mol phosphorylation of all GST-I-1 proteins.
The assays were initiated by [ 32 P]phosphorylase (final concentration of 2 mg/ml), and the mixture was incubated for 30 min at 37°C. Assays were terminated by addition of 200 l of 20% (w/v) trichloroacetic acid and 50 l of bovine serum albumin (10 mg/ml) and subjected to centrifugation at 15,000 ϫ g for 10 min. [ 32 P]Phosphate release was analyzed by liquid scintillation counting. All assays were performed at least three times in triplicate.
Expression of I-1 in Cultured Mammalian Cells-HEK293T cells (grown to near confluency in 6-well plates) were transfected with 2.0 g of FLAG-hI-1 DNA, 2.0 g of FLAG-hI-1(T35A) DNA, 2.0 g of FLAG-hI-1-(1-77) DNA, or 2.0 g of vector control (pCMV-FLAG-2) and 4 l of LipofectAMINE (Invitrogen) according to the manufacturer's instructions. Cells were then treated with 0.1% Me 2 SO or 5 M forskolin. After 6 h, cells were washed with phosphate-buffered saline and lysed in 500 l of radioimmune precipitation assay buffer (10 mM sodium phosphate (pH 7.5), 150 mM NaCl, 1% (w/v) Nonidet P-40, 0.5% (w/v) deoxycholate, and 0.1% (w/v) SDS) for 15 min on ice. Cells were scraped, and the lysates were cleared by centrifugation at 20,000 ϫ g for 3 min. The cleared lysates were then run on SDS-polyacrylamide gels and analyzed by immunoblotting. Immunoblots were quantified using Scion Image Beta 4.0.2. I-1 phosphorylation was normalized to total I-1 and then reported as -fold increase relative to control Me 2 SO-treated cells (n ϭ 4). Phosphorylation of eIF2␣ was normalized to total eIF2␣ and reported as -fold change relative to cells expressing the vector control (n ϭ 4).

RESULTS
Numerous PP1 complexes have been identified in eukaryotic cells and are composed of PP1 catalytic subunits associated with regulatory subunits (17). The regulatory subunits target PP1 to subcellular organelles and/or direct the dephosphorylation of selected phosphoproteins. The PP1 catalytic subunit is also inhibited by several PP1-selective inhibitor proteins, such as I-1 (18). Biochemical studies show that nanomolar concentrations of PKA-phosphorylated I-1 inhibit the PP1 catalytic subunit isolated from rabbit skeletal muscle, but are less effective in inhibiting the glycogen-bound complex composed of PP1 bound to G M (19). Analysis of I-1-null mice suggests that I-1 preferentially regulates a subset of PP1 complexes that control vascular and cardiac muscle contractility (3) and neuronal signaling (20), whereas other PP1 functions, such as glycogen metabolism (21), are essentially unaffected in the mutant mouse. The molecular basis for I-1 selectivity for specific PP1 complexes remains unknown in part because of the lack of suitable cellular assays to investigate the structure-function of I-1 as an inhibitor of PP1 function in eukaryotic cells.
Structure-Function of I-1 in Yeast-Our previous studies showed that hI-1 potently inhibits Glc7p, the budding yeast PP1 catalytic subunit, in vitro (12). Moreover, the overexpression of hI-1 in yeast showed that some Glc7p-regulated pathways, such as mitosis and gene transcription, are inhibited, whereas others, such as glycogen metabolism and meiosis, are essentially unaffected (13). This raised the possibility that I-1 expression in yeast could be used to define the molecular basis by which I-1 targets cellular PP1 complexes. I-1 dependence in yeast was also used in a genetic screen to identify novel Glc7p (PP1)-regulated pathways and yielded strain JC1007-97, in which the presence of active hI-1 is required for transcription of the GFA1 (glutamine-fructose-6-phosphate transaminase-1) gene and yeast growth (13). This provided a novel assay for evaluating the mechanism of action of I-1 as a regulator of eukaryotic PP1 complexes since only an expression vector carrying a hI-1 construct with full PP1 inhibitory activity could complement the loss of existing WT hI-1. To analyze the structure-function of I-1, yeast containing JZ203 (ADE3 hI-1) was transformed with various I-1 constructs (pDW series). The loss of the plasmid-borne ADE3 marker, which maintains the gfa1-97 yeast as red colonies, occurs only when pDW plasmids express functional PP1 inhibitors, and this results in the conversion of red-to-white colonies (Fig. 1). Counting the white and/or red-and-white sectoring colonies provided a measure of relative PP1 inhibitory activity of the replacement I-1 polypeptide. In this regard, the plasmid expressing hI-1-(1-171) (JZ205) resulted in significant red-to-white conversion, with Ͼ70% white colonies. I-1(T35A), expressing full-length hI-1 lacking Thr 35 , whose phosphorylation by PKA is required for PP1 inhibition, failed to complement the loss of ADE3 hI-1, and all colonies remained red. The C-terminal truncations I-1-(1-142) and I-1-(1-123) were also less effective inhibitors of the PP1 holoenzyme, which regulates GFA1 transcription, and showed significantly reduced red-and-white sectoring. In contrast, I-1-(1-97), I-1-(1-77), and I-1-(1-54) failed to complement hI-1, and largely red colonies were observed, consistent with the retention of the ADE3 hI-1 plasmid. Internal deletions that eliminated amino acids 77-97 and 97-110 also failed to complement hI-1, indicating a lack of PP1 inhibition. In contrast, hI-1 with the deletion of amino acids 110 -124 was active as a PP1 inhibitor, as determined by the sectoring assay. None of the aforementioned I-1 polypeptides were functional when Thr 35 was substituted with alanine (T35A), consistent with the absolute requirement of this phosphorylated threonine for PP1 inhibition. Interestingly, substitution of Thr 35 with aspartic acid (T35D) to mimic threonine phosphorylation also failed to support the function of I-1 polypeptides in yeast (Fig. 1, table). These data suggest that, in contrast to in vitro studies in which PKA-phosphorylated I-1-(1-54) and I-1-(1-54)(T35D) inhibited the PP1 catalytic subunit (5), C-terminal amino acids 77-110 are required for effective inhibition of the PP1 complex(es) that regulate GFA1 expression in the budding yeast.
Analysis of I-1 mRNAs in Mouse Tissues-Recent studies identified three hI-1 mRNAs in human brain and liver (11). In addition to the major transcript encoding full-length hI-1-(1-171), two additional mRNAs, termed I-1␣ and I-1␤, were noted. I-1␣, which lacks exon 5, was predicted to generate a shorter I-1 polypeptide (a total of 120 residues) lacking C-terminal residues 83-134. The I-1␤ mRNA lacks exons 4 and 5 and resulted in a frameshift, yielding a 132-amino acid polypeptide whose C terminus following amino acid 61 completely differs from those of I-1 and I-1␣ (Fig. 3A). To confirm and extend these findings, we analyzed mRNAs from tissues of WT I-1 and I-1-null mice. Oligo(dT)-primed mRNAs from mouse tissues were amplified and hybridized with probes specific for exons 4 -6 (Fig. 3B). The specificity of these probes was established as the exon 6 probe detected all three I-1 mRNAs, the exon 5 probe detected in only I-1 mRNA, and the exon 4 probe detected in both I-1 and I-1␣ mRNAs. At the exposure shown in Fig. 3B, the mRNA levels of I-1 far exceeded those of I-1␣ and I-1␤ (which were more clearly seen at longer exposures). The signal on this Southern blot depended on the amount of mRNA analyzed (Fig. 3B, lanes 28 -30). Interestingly, expression of all three mRNAs was decreased in adult testes compared with postnatal (P) days P7 and P14 and was detectable only at exposures longer than those shown in Fig. 3. Finally, the exon 6 probe showed a low signal in mouse liver, seen only by increasing the amplification cycles, as noted in rat liver (22).
Southern blotting using the exon 6 probe showed that the three I-1 mRNAs were expressed in mouse at embryonic days 9 and 15. All three mRNAs were also observed in tissues from P7, P14, and adult mice (Fig. 3B). Careful analysis showed that, in some tissues, e.g. heart, the I-1 mRNA levels were developmentally regulated, progressively decreasing from P7 to P14 to adult heart (Fig. 3C). Comparison of cortex and hippocampal mRNAs from WT I-1 and I-1-null mice established that all three mRNAs were derived from a single gene and were present in WT animals but absent in all tissues from I-1 Ϫ/Ϫ mice. Amplification of ␤-actin confirmed that equivalent amounts of total mRNA from all tissues were analyzed.
To establish the presence of I-1␣ and I-1␤ in mouse tissues, specific PCR products were amplified using primers that spanned the splice junctions (Fig. 3E). This demonstrated that I-1␣ and I-1␤ mRNAs were present in all tissues, with much lower but detectable levels in testes (at P7 and P14). However, both I-1 mRNAs appeared to be absent in adult testes and liver at all stages of mouse development. In contrast to the progres-sive decrease in I-1 and I-1␣ mRNAs in heart during development, I-1␤ mRNA levels increased from barely detected levels in P7 heart to much higher levels in P14 and adult heart. This suggests that, although derived from a single gene, alternative splicing of I-1␣ and I-1␤ mRNAs is regulated in both a developmental and tissue-specific manner.
Regulation of Yeast Protein Translation by hI-1-Although the experiments using S. cerevisiae strain JC1007-97 identified a novel functional domain in the hI-1 C terminus required for Glc7p (PP1) inhibition, the biochemical pathway required for GFA1 gene transcription and yeast growth has not been defined. Thus, the number and nature of PP1 complexes inhibited by hI-1 protein to promote GFA1 gene expression are unknown. By comparison, numerous studies have established that both the Gcn2p protein kinase and Glc7p phosphatase control the phosphorylation state of the yeast translation initiation factor eIF2␣ (Sui2p). This pathway is particularly interesting because not only does it provide useful tools to study I-1 function in yeast, but PP1 regulation of eIF2␣ phosphorylation is also highly conserved in mammals (10,23,24). Under conditions of amino acid starvation, the Gcn2p kinase is activated and phosphorylates eIF2␣ to inhibit general protein translation and subsequently activates genes, such as that encoding the GCN4 transcription factor, which promotes the expression of amino acid biosynthetic genes such as HIS4 (14). The allele gcn2-507 resulted in a partial loss-of-function Gcn2p and reduced eIF2␣ kinase activity and compromised the expression of GCN4 and HIS4 genes, therefore inhibiting the recovery from amino acid starvation (Fig. 4A). Thus, strain Y27 failed to grow on starvation medium (containing 3-AT). We anticipated that inhibition of yeast PP1 by phospho-hI-1 would enhance the function of the mutant eIF2␣ kinase and reinstate the growth of Y27 on starvation medium (14,15).
The hI-1 expression plasmid (CEN I-1), like the control (CEN URA3), had no effect on the growth of strain H1402, which expresses wild-type Gcn2p, in the presence of 3-AT (Fig. 4A). In contrast, hI-1, but not the vector control, partially restored the growth of Y27, which contains a partial loss-of-function muta-
tion of GCN2, on starvation medium (at 30°C for 2 days). Neither plasmid permitted the growth of H1149, in which the GCN2 gene is disrupted, on 3-AT plates. This suggests that, by inhibiting the yeast eIF2␣ phosphatase (PP1), hI-1 facilitates the actions of the partially active mutant Gcn2p kinase and allows the recovery of yeast strain Y27 from amino acid starvation.
Expression of hI-1 using either low copy (CEN) or high copy (2) expression or expression of dominant-negative GLC7⌬ (encoding amino acids 1-207 of Glc7p) (14) rescued Y27 growth on 3-AT plates at 30°C for 4 days (Fig. 4B). Substitution of the phenylalanine in the core PP1-binding sequence (KIQF) with alanine to yield KIQA or alanine substitution of Thr 35 (1-54), failed to support growth on 3-AT plates (Fig. 4C). Yeast lysates from Y27 cells starved for 6 h were immunoblotted for phosphorylated eIF2␣ (Fig. 4D). These data show that I-1 expression in yeast resulted in increased phosphorylation of eIF2␣, consistent with the inhibition of the yeast eIF2␣ phosphatase (Fig. 4D). In contrast, expression of hI-1(T35A) or the N-terminal fragment hI-1-(1-77) had no effect on I-1 phosphorylation, which essentially resembled that seen in vector control cells. These data suggest that, in addition to the KIQF PP1-binding sequence and phosphorylated threonine, C-terminal sequences between amino acids 97 and 123 are essential for inhibition of the yeast eIF2␣ phosphatase.
Lysates from transformed yeast cells were immunoblotted with anti-I-1 antibody to confirm equivalent expression of I-1 polypeptides (Fig. 5A). Expression of I-1 on a high copy plasmid (2) resulted in an ϳ10-fold higher concentration of I-1 relative to a low copy plasmid (CEN) (Fig. 5B). Expression of a high copy I-1 N-terminal fragment (2 1-77) also resulted in a 10-fold increase in expression relative to a low copy fragment (CEN 1-77) (Fig. 5B); however, even at this higher concentration, PP1 was insufficiently inhibited to allow Y27 growth (data not shown), indicating that truncated I-1 proteins have at least a 10-fold reduction in PP1 inhibitory activity relative to fulllength I-1. Additionally, phosphorylation of hI-1 in yeast was monitored by blotting with anti-I-1(T35-P) antibody (Fig. 5C), confirming that N-terminal fragments of hI-1 were phosphorylated similarly to full-length hI-1, indicating that the loss of function caused by truncation of I-1 does not stem from lower expression or reduced phosphorylation, but from the reduced PP1 inhibition activity of N-terminal fragments.
Regulation of Yeast PP1 by hI-1␣-Splicing of the hI-1 gene partially deleted the C-terminal domain identified in Fig. 1 to yield I-1␣. To evaluate the function of I-1␣ as a PP1 inhibitor, hI-1␣ was expressed in yeast strain Y27 grown on 3-AT plates (Fig. 6A). Although WT hI-1 effectively promoted growth on starvation medium plates, the inactive mutant I-1(T35A) failed to facilitate yeast growth on starvation medium. Compared with hI-1, hI-1␣ permitted much slower growth on starvation medium. The I-1␣-expressing colonies required additional 2 days (total of 6 days at 30°C) of growth compared with hI-1-(1-171) to be visible (Fig. 6A). This indicates that hI-1␣ is a weaker PP1 inhibitor than WT hI-1 in yeast. Equal expression of I-1 peptides was confirmed by immunoblotting for I-1 (Fig. 6A).
For in vitro biochemical studies, recombinant GST-I-1␣ was expressed in bacteria, purified on glutathione-Sepharose, and stoichiometrically phosphorylated with PKA. In pull-down assays with HEK293T cell lysates, GST-I-1␣ sedimented PP1 in a dose-dependent manner. Compared with GST-I-1- (1-171), similar concentrations of GST-I-1␣ bound significantly less PP1 (Fig. 6B). This suggests that C-terminal amino acids 84 -134 also facilitate I-1 association with mammalian PP1 complexes. PKA phosphorylation of GST-I-1␣, as noted for hI-1, was necessary for effective sedimentation of purified PP1 catalytic subunits (Fig. 6C), and unphosphorylated I-1␣ failed to bind a significant amount of PP1 catalytic subunit.
Finally, like GST-I-1, phosphorylated GST-I-1␣ inhibited the phosphorylase phosphatase activity of the purified PP1 catalytic subunit (Fig. 6D). In this assay, I-1␣ showed a 3-fold reduced activity as a PP1 inhibitor. Unphosphorylated I-1␣ did not inhibit PP1 even at concentrations exceeding 1 M (data not shown). These data suggest that hI-1␣ is a weaker PP1 inhibitor than hI-1 both in vitro and in yeast.
suggests that the C terminus of I-1 is required for low affinity PP1 binding by unphosphorylated hI-1.
In vitro phosphorylase phosphatase assays using purified rabbit skeletal muscle PP1 showed that phosphorylated GST-I-1-(1-171) and GST-I-1-(1-80) inhibited PP1 activity in a dosedependent manner (Fig. 7D). Phosphorylated GST-I-1-(1-80) was nearly 10-fold weaker than phosphorylated GST-I-1-(1-171) as a PP1 inhibitor. The C-terminal fragment GST-I-1-(81-171) showed no measurable activity as a PP1 inhibitor, consistent with its inability to directly bind PP1 (data not shown). These data suggest that PP1 inhibition is mediated primarily through the N-terminal domain of I-1, which contains the PP1binding motif KIQF and PKA-phosphorylated Thr 35 , but that the presence of the C-terminal 90 amino acids greatly enhances the activity of I-1 as a PP1 inhibitor. In the absence of sufficient purified GST-I-1-(1-61) protein to undertake similar studies, the above data suggest that hI-1␤ is likely to be a weaker PP1 inhibitor than either I-1␣ or full-length hI-1- (1-171).
Regulation of eIF2␣ Dephosphorylation by I-1 in Mammalian Cells-To demonstrate the necessity of the C terminus of I-1 to regulate PP1-dependent signaling in mammalian cells, we expressed I-1 fragments in HEK293T cells, a cell line that does not express detectable I-1, and monitored the phosphorylation state of eIF2␣. Expression of I-1 had little effect on eIF2␣ phosphorylation in untreated cells (Fig. 8A; quantified in Fig.  8B). However, when cells were treated with forskolin, which elevates intracellular cAMP, phosphorylation of I-1 and I-1-(1-77), but not I-1(T35A), was greatly increased (7.9 Ϯ 1.6-and 5.6 Ϯ 1.5-fold, respectively) relative to untreated I-1-expressing cells ( Fig. 8A; quantified in Fig. 8C). Forskolin treatment of hI-1-expressing cells resulted in an ϳ70% increase in eIF2␣ phosphorylation. In contrast, forskolin had no effect on eIF2␣ phosphorylation in cells expressing either the N-terminal fragment I-1-(1-77) or the inactive mutant I-1(T35A) (Fig. 8, A and  B). These data indicate that full-length I-1 is required for cAMP-induced eIF2␣ phosphorylation in mammalian cells and suggest that deletion of C-terminal sequences greatly impairs the ability of I-1 to inhibit the mammalian eIF2␣ phosphatase. DISCUSSION PP1, a serine/threonine protein phosphatase, is ubiquitously expressed in mammalian tissues and displays broad in vitro substrate specificity. In cells, however, PP1 shows considerable specificity that is dictated by the presence of Ͼ40 regulatory or targeting subunits, which direct its subcellular localization and/or substrate recognition (2). In addition, a number of PP1specific inhibitors have been identified (18). Growing evidence suggests that, although the PP1 inhibitors potently inhibit the activity of isolated PP1 catalytic subunits, they show greater selectivity for PP1 holoenzymes containing various targeting subunits (10,25,26). PKA-phosphorylated I-1 inhibits the PP1 catalytic subunit with an IC 50 of 1 nM, but even higher concentrations of phospho-I-1 have little effect of the activity of the glycogen-bound phosphatase, a complex of PP1 and G M , the skeletal muscle glycogen-targeting subunit (19), suggesting that this complex is unlikely to be regulated by I-1 under  CEN I-1), and inactive I-1 (CEN I-1 T35A) were expressed in Y27 yeast and grown on control and selection (3-AT) media at 30°C for 6 days. The schematic shows the structures of the three I-1 polypeptides analyzed. Cells were lysed and analyzed by immunoblotting (IB) with anti-I-1 antibody to demonstrate equal expression. B, phosphorylated GST-I-1 (GST-I-1-P), phosphorylated GST-I-1␣ (GST-I-1␣-P), and GST (g protein) were coupled to glutathione-Sepharose beads and used to sediment PP1 from HEK293T cell lysates. Bound PP1 was eluted with SDS-PAGE loading buffer and detected by immunoblotting with anti-PP1 antibody. C, 5 g of phosphorylated and unphosphorylated GST-I-1␣ were used to deplete purified PP1 catalytic subunits from solution. Residual PP1 activity was analyzed using phosphorylase a as substrate. D, increasing concentrations of phosphorylated GST-I-1-(1-171) (छ) and phosphorylated GST-I-1␣ (•) were used to inhibit the PP1 catalytic subunit assayed using phosphorylase a as substrate. Error bars indicate S.E. physiological conditions. In contrast, I-1 can inhibit the GADD34⅐PP1 complex in vitro with the same IC 50 as the free catalytic subunit (10), suggesting that eIF2␣ phosphorylation is a potential physiological target of I-1 regulation. The PKCactivated inhibitors PHI-1 and PHI-2 (PP1 holoenzyme inhibitor) show equivalent activity in vitro against the PP1 catalytic subunit and the glycogen-and smooth muscle myosin-bound phosphatases (26). CPI-17, another PKC-activated PP1 inhibitor, potently inhibits the myosin-bound phosphatase, but is essentially ineffective as an inhibitor of the glycogen-bound phosphatase (27). The molecular basis by which endogenous PP1 inhibitors target some and not other PP1 complexes remains unknown.
The sequence homology between the PP1 inhibitors I-1 (a total of 171 amino acids) and DARPP-32 (a total of 204 amino acids) resides in their N-terminal 50 amino acids, which contain a conserved KIQF PP1-binding motif and the PKA-phosphorylated threonine (5,28). The precise role of their diverse C termini is still poorly understood. In vitro biochemical evidence suggests that covalent modification of the unique DARPP-32 C-terminal domain allows for modulation of the phosphorylation and dephosphorylation of the critical threonine and thus fine tunes the activity of DARPP-32 as a PP1 inhibitor (8,9,29). In contrast, our previous work suggests that the I-1 C terminus, which is dispensable for binding to the isolated PP1 catalytic subunit in vitro, associates with a number of other cellular proteins, including GADD34 (10). Recent studies suggest that the specificity of some PP1 inhibitors may be dictated by their ability to bind both PP1 catalytic and targeting or regulatory subunits. Formation of such heterotrimeric complexes may dictate the specificity of CPI-17 for the myosin phosphatase (27), inhibitor-2 for the PP1⅐neurabin-I (25) and PP1⅐Nek2 complexes (30), and I-1 for the GADD34⅐PP1 complex (10). These studies suggest that distinct domains of these PP1 inhibitors bind the PP1 catalytic subunit and the targeting or regulatory subunit.
Recent studies suggest that alternative splicing of the hI-1 . The cells were then lysed and immunoblotted for total protein expression (anti-FLAG antibody), phospho-I-1, total eIF2␣, and phospho-eIF2␣. B, quantification of eIF2␣ phosphorylation was normalized to total eIF2␣ content and is shown as fold change compared with cells expressing the empty vector. *, p Ͻ 0.05 for several independent experiments (n ϭ 4) analyzed using Student's t test; NS, not significant. C, I-1 phosphorylation was also quantitated and normalized to total I-1 content and is shown as -fold change compared with untreated cells expressing full-length I-1. **, p Ͻ 0.01 for several independent experiments (n ϭ 4) analyzed using Student's t test.
gene generates three distinct mRNAs that encode PP1 inhibitors with different C-terminal sequences (11). Alternative splicing that excises exon 5 generates I-1␣, which differs from full-length I-1-(1-171) by the deletion of amino acids 83-134. Splicing of exons 4 and 5 yields I-1␤, which retains the Nterminal 61 amino acids of I-1, but also has an additional 72 amino acids that share no homology with any known protein.
The conservation of this splicing event was not examined. We observed the same splicing pattern in mice (Fig. 3) and cloned I-1␣ from a Xenopus embryo cDNA library. 2 As all three I-1 isoforms retain the sequences previously shown to represent the minimal PP1 inhibitory domain (11), but both splice variants lacked the novel C-terminal domain identified in our yeast experiments. The novel variants were predicted to inhibit PP1 activity, but the precise physiological role of I-1 isoforms remains unknown.
We analyzed mRNAs from tissues of WT I-1 ϩ/ϩ and I-1 Ϫ/Ϫ mice and established the presence of three distinct mRNAs encoding I-1, I-1␣, and I-1␤ in most WT mouse tissues except liver, where very little I-1 mRNA was detected, consistent with previous studies that required extensive RT-PCR to visualize the very low levels of I-1 mRNA in rat liver (22). Previous immunoblotting studies also showed that, in contrast to other tissues, rat and mouse livers express undetectable amounts of I-1 protein (31). No I-1 mRNAs were amplified from tissues of I-1-null mice, confirming that all three mRNAs are from a single I-1 gene. Previous work showed that both I-1 mRNA (32) and protein (33) are postnatally expressed in rat brain, with low expression seen at P0, increasing thereafter to peak at P7 and finally declining to low steady-state levels in adult brain. In contrast, our studies showed a significant amount of all three I-1 mRNAs in mouse embryos at embryonic days 9 and 15. This was consistent with immunohistochemistry of whole mouse embryos that showed the presence of I-1 protein in many embryonic tissues, including high level expression in the mouse mesothelium (34). Analysis of mouse heart showed that I-1 and I-1␣ mRNA levels progressively decreased from P7 to adult, whereas I-1␤ mRNA levels increased over the same period, suggesting that I-1 mRNA expression and splicing may be regulated in both a developmental and tissue-specific manner.
In vitro structure-function studies have thus far failed to define a role of I-1 C-terminal sequences in the inhibition of the PP1 catalytic subunit (5). Thus, current studies have focused on a cell-based assay to define the functional differences between three hI-1 isoforms as PP1 inhibitors. hI-1 is phosphorylated (at Thr 35 ) when expressed in yeast (13). Moreover, hI-1 protein effectively binds and inhibits the single yeast PP1 catalytic subunit (12). Overexpression of hI-1 inhibits some PP1-regulated events while having little effect on other events (13), suggesting that hI-1 protein targets selected yeast PP1 complexes. Finally, consistent with PP1 inhibition, the functional effects of I-1 expression are suppressed by overexpression of yeast protein kinases (13). This opened the way for a novel genetic screen to identify new I-1-and PP1-regulated pathways. One such hI-1-dependent strain, JC1007-97, was used to evaluate the structural requirements for PP1 inhibition in yeast. A sectoring assay established that the phosphorylated threonine is essential for growth of gfa1-97 suggesting a requirement for Glc7p inhibition and growth in the absence of glucosamine (13). The inactive T35A mutant failed to support gfa1-97 growth in the sectoring assay. Surprisingly, I-1 polypeptides containing a T35D substitution to mimic phosphorylation also failed to complement the loss of WT hI-1 in gfa1-97 yeast. Previous in vitro studies showed that the T35D substitution fails to activate full-length I-1 as a PP1 inhibitor (5), and deletion of the C terminus in hI-1-(1-54)(T35D) is required to inhibit PP1, albeit with a much reduced IC 50 (ϳ150 nM). These studies suggest a conformational cross-talk between the N-and C-terminal sequences of I-1 that is modulated by Thr 35 phosphorylation to enhance PP1 inhibition. It should be noted, however, that the constitutively active I-1-(1-54)(T35D) peptide has been utilized to inhibit PP1 activity in mammalian cells (35) and tissues (3). Our studies suggest that high levels of expression of this weak PP1 inhibitor may overcome its reduced affinity for the PP1 catalytic subunit, and lacking Cterminal sequences that target specific PP1 complexes, the I-1-(1-54)(T35D) peptide may be more pleiotropic than WT I-1 and inhibit a wider range of PP1 complexes. Analysis of I-1 structure-function in gfa1-97 yeast also highlighted the critical role for the I-1 C terminus, specifically amino acids 77-110, in the inhibition of the PP1 complex that activates GFA1 gene transcription in yeast. These findings are unlikely to result from gross changes in protein folding, as previous studies used circular dichroism (5) and NMR (36) to show that I-1 is largely disordered. Yet other studies established that the PP1 inhibitory activity of I-1 is resistant to boiling and acid treatment (16). Finally, synthetic peptides derived from DARPP-32 also retain potent PP1 inhibitory activity (28). Additional biochemical studies suggest that region 77-110 stabilizes I-1 binding to the isolated PP1 catalytic subunit and selected human PP1 complexes and thereby enhances PP1 inhibition.
The biochemical pathway regulating GFA1 gene expression has not been defined. Thus, the number and types of PP1 complexes inhibited by hI-1 in the GFA1 pathway remain unknown. On the other hand, the ability of Glc7p to regulate protein translation in yeast has been clearly defined, and in the Y27 strain, which contains an impaired eIF2␣ kinase (Gcn2p), Glc7p inhibition by a truncated dominant-negative GLC7 catalytic subunit permits yeast growth on amino acid starvation medium. hI-1 also allowed growth on starvation medium. Our results show that I-1 C-terminal sequences play a critical role in PP1 inhibition in yeast, but suggest that hI-1␣, which lacks amino acids 83-134, is a weaker PP1 inhibitor. Pull-down assays with recombinant GST-I-1␣ also confirmed its weaker binding to human PP1 complexes from HEK293T lysates and reduced inhibition of the isolated PP1 catalytic subunit. Cterminal truncation of I-1 also suggest that I-1␤, which shares only the N-terminal 61 amino acids with I-1␣ and I-1, is an even weaker PP1 inhibitor. These data suggest that the I-1 C-terminal sequences, while not showing direct PP1 binding, may contribute to the enhanced I-1 binding to the target phosphatase complexes.
Expression of I-1 in human HEK293T cells showed that I-1 enhanced phosphorylation of eIF2␣ in mammalian cells. Forskolin treatment of I-1-expressing cells activated I-1 via Thr 35 phosphorylation and led to increased eIF2␣ phosphorylation. Cells expressing the inactive I-1(T35A) mutant or the vector control failed to respond to forskolin, and no significant change in eIF2␣ phosphorylation was observed. Interestingly, the ability of cAMP to promote eIF2␣ phosphorylation and thereby inhibit translation has been recognized for several decades (37). Insulin induces the simultaneous dephosphorylation of I-1 (38) and eIF2␣ (39), whereas hormones, such as glucagon (40) and vasopressin (41), simultaneously increase cellular cAMP and eIF2␣ phosphorylation. As PKA does not directly phosphorylate eIF2␣ or activate eIF2␣ kinases, the mechanism by which cAMP increases eIF2␣ phosphorylation remains unknown. Our studies suggest a mechanism by which hormones that increase intracellular cAMP may activate I-1 and suppress the cellular eIF2␣ phosphatase. As seen in yeast, deletion of the I-1 C terminus impaired its ability to regulate protein translation. This may be consistent with our previous observation that I-1, via its C terminus, binds a key component of the mammalian eIF2␣ phosphatase complex, GADD34 (10). Thus, the current studies provide new evidence that the C terminus of I-1 not only enhances the affinity of I-1 for the PP1 catalytic subunit, but may also target specific PP1 complexes to regulate mammalian cell physiology. Finally, differences in PP1 binding displayed by the three I-1 isoforms, combined with predicted differences in their ability to bind specific PP1 regulatory subunits, suggest that the I-1 isoforms serve distinct roles in regulating PP1 function in mammalian cells.
In conclusion, our studies utilized novel cell-based assays in yeast and mammalian cells to highlight, for the first time, the critical role played by C-terminal sequences in regulating the potency and specificity of I-1 as a PP1 inhibitor in eukaryotic cells. Our studies provide compelling evidence for PP1 as a component of mammalian and yeast eIF2␣ phosphatases that are targeted by I-1 and suggest a role for I-1 in transducing the cAMP signals that inhibit protein translation in mammalian cells. Further work is clearly needed to identify additional cellular PP1 complexes targeted by I-1 and to establish the molecular basis for the immunity of other PP1 complexes to this protein inhibitor. Finally, the tissue-specific and developmental expression and the alternative splicing of the I-1 gene to generate three distinct I-1 isoforms differing in their Cterminal sequences may expand the functional diversity of this prototypic PP1 regulator.