A Novel Transmembrane Ser/Thr Kinase Complexes with Protein Phosphatase-1 and Inhibitor-2*

Protein kinases and protein phosphatases exert coordinated control over many essential cellular processes. Here, we describe the cloning and characterization of a novel human transmembrane protein KPI-2 (Kinase/Phosphatase/Inhibitor-2) that was identified by yeast two-hybrid using protein phosphatase inhibitor-2 (Inh2) as bait. KPI-2 mRNA was predominantly expressed in skeletal muscle. KPI-2 is a 1503-residue protein with two predicted transmembrane helices at the N terminus, a kinase domain, followed by a C-terminal domain. The transmembrane helices were sufficient for targeting proteins to the membrane. KPI-2 kinase domain has about 60% identity with its closest relative, a tyrosine kinase. However, it only exhibited serine/threonine kinase activity in autophosphorylation reactions or with added substrates. KPI-2 kinase domain phosphorylated protein phosphatase-1 (PP1C) at Thr320, which attenuated PP1C activity. KPI-2 C-terminal domain directly associated with PP1C, and this required a VTFmotif. Inh2 associated with KPI-2 C-terminal domain with and without PP1C. Thus, KPI-2 is a kinase with sites to associate with PP1C and Inh2 to form a regulatory complex that is localized to membranes.

into pVP16 vector, which contains the Gal4 activation domain (created by Dr. S. Hollenberg, Fred Hutchinson Cancer Center, Seattle, WA). The screening was done by using the large-scale, sequential transformation method (30). 2 ϫ 10 6 clones were screened. Positive clones were tested first for expression of the HIS3 gene (Hisϩ) by growth of the clones on the plates lacking histidine (SD/-Trp/-Leu/-His). The positives of those were tested for expression of the reporter gene, lacZ, using an assay for ␤-galactosidase activity. Clones were rescued by electroporation into Escherichia coli HB101 and grown on M9 plates lacking leucine, which allowed for analysis of positives by transformation tests and DNA sequencing.
For protein-protein interaction, we used an alternative yeast twohybrid system. Inh2 or PP1C␣ gene was fused to the Gal4 DNA-binding domain in the pGBT10 vector, and KPI-2 C-terminal wild type (residues 1099 -1503), AA-mutant, or IB-4 (KPI-2 fragment from two-hybrid screen) was fused to the Gal4-activation domain in a pVP16 vector. Both bait and prey plasmids were cotransformed into HF7c cells. Protein-protein interaction was determined by checking the growth of clones on the plate lacking histidine (SD/-Trp/-Leu/-His).
Cloning and Construction of Different Expression Vectors-mRNA was isolated from HeLa cells with a QuickPrep TM Micro mRNA purification kit (Amersham Biosciences). N-and C-terminal portions of KPI-2 cDNA were synthesized separately using a One-Step RT-PCR kit (Qiagen). Primers were designed according to the sequence of KIAA1079 cDNA, and the following primers were used (Fig. 1A): for N-terminal, forward primer containing BamHI site 5Ј-TAT AAT GGA TCC ACC ATG CCG GGG CCG CCG GCG TT-3Ј, reverse primer containing HindIII site 5Ј-TGT GTT CTT TGC TGG ACA ATG AAG CTT TTA GTA AGT-3Ј; for C-terminal, forward primer containing HindIII site 5Ј-ACT TAC TAA AAG CTT CAT TGT CCA GCA AAG AAC ACA-3Ј, reverse primer containing XhoI site 5Ј-CAC TCG AGG TCC TTT TCT CCG TCT TCG CTG CTT CC-3Ј. The amplified C terminus was subcloned into a pCMV myc-tagged vector with an HindIII-XhoI site. The full-length KPI-2 was created by inserting the N-terminal portion into the plasmid containing the C-terminal portion at the BamHI-HindIII site. The construct of KPI-2 full-length was named pCMV-KPI-2-full-length. The HindIII site was repaired to original sequence by using a Stratagene QuikChange mutagenesis kit according to the manufacturer's protocol. pCMV-KPI-2-N-terminal was constructed by inserting an N-terminal fragment (1-703 residues, amplified by PCR) into pCMV myc-tagged vector with a BamHI-XhoI site.
pFastBac-His 6 -KPI-2-kinase plasmid was constructed for use in the Bac-to-Bac Baculovirus Expression System (Invitrogen). The KPI-2 fragment, including a kinase domain (residues 94 -600), was amplified by PCR and subcloned into pFastBac HTb vector with a BamHI-EcoRI site to form a recombinant pFastBac donor plasmid. All the constructs are shown in Fig. 1B. All DNA sequences above were confirmed by double-stranded DNA sequencing in the Biomolecular Research Facility of the University of Virginia.
Northern Blot Analysis-Northern blotting was performed using Clontech human Multiple Tissue Northern (MTN TM ) Blots according to the manufacturer's instructions (Clontech). The membrane was probed with 32 P-labeled cDNA corresponding to the kinase domain (280 -1800 nt) of KPI-2. The membrane was stripped and reprobed with cDNA encoding the C-terminal domain (3295-4509 nt). After washing, the membrane was exposed to x-ray film with an intensifying screen for 72 h at Ϫ70°C.
Cell Culture, Transfection, Immunoprecipitation, and Western Blotting-COS7, HeLa, and HEK293T cells were cultured in Dulbecco's modified Eagle's Medium supplemented with 10% heat-inactivated newborn calf serum (Invitrogen). Cells were grown in 10-cm plates in a humidified incubator at 37°C and 5% CO 2 and subcultured every 2-3 days. Cells were transfected by using FuGENE 6 reagent according to the manufacturer's instructions. After 24 h of transfection, the cells were harvested and lysed with lysis buffer (50 mM Tris/HCl, pH 8.0, 150 mM NaCl, 50 mM NaF, 1% Nonidet P-40, 20 mM ␤-glycerophosphate, 1 mM Na 3 VO 4 , 1 mM dithiothreitol, 0.1% 2-mercaptoethanol, 1 mM Pefa-bloc-sc, 10 g/ml leupeptin, and 10 g/ml pepstatin) for 30 min on ice. The lysates were clarified by centrifugation at 10,000 ϫ g for 10 min. For Western blotting, equal amount of proteins were subjected to SDS-PAGE and immunoblotted with specific antibodies. For immunoprecipitation, the lysates were incubated with anti-FLAG M2-agarose affinity gel for 1 h at 4°C. The beads were washed three times with the lysis buffer and then subjected to SDS-PAGE and immunoblotted with anti-FLAG, anti-PP1C, or anti-Inh2 antibody.
Preparation of Cell Membranes-COS7 cells transfected with pCMV-KPI2-N-terminal for 24 h were washed once with phosphate-buffered saline, and scraped in ice-cold buffer containing 10 mM Tris/HCl, pH 8.0, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 1 mM Pefabloc-sc, 10 g/ml leupeptin and 10 g/ml pepstatin. The cells were disrupted by homogenization, and the homogenates were clarified by centrifugation at 1000 ϫ g for 10 min at 4°C. The pellet was homogenized again, and the supernatants were pooled and then centrifuged at 4000 ϫ g for 10 min. The supernatants were mixed with an equal volume of the same buffer containing 0.25 M sucrose and centrifuged at 100,000 ϫ g for 1 h. The supernatants were transferred to new tubes, and concentrated by using Centricon Centrifugal Filter Devices (Millipore). The membrane pellets were solubilized with lysis buffer containing 1% Nonidet P-40. The protein concentration was measured, and an equal amount of protein was subjected to SDS-PAGE and immunoblotted with anti-myc antibody.
Expression and Purification of Recombinant Kinase Protein-The donor plasmid pFastBac-KPI-2 kinase-(94 -600) was transformed into DH10Bac-competent cells for transposition into Bacmid. The recombinant Bacmid DNA was identified by PCR and transfected into Sf9 cells and incubated for 5 days at 27°C. The recombinant baculoviruses were harvested to obtain the P1 viruses and then amplified with P1 viruses to produce high titer P2, P3 viruses. The virus titer was measured by the cell culture center of the University of Virginia.
For protein expression, Sf9 cells were infected with recombinant baculoviruses, and cells were collected 60 h later by centrifugation. The cell pellet was resuspended in lysis buffer (20 mM imidazole-HCl, pH 7.0, 20 mM potassium phosphate, 150 mM NaCl, 1% Nonidet P-40, 0.1% ␤-mercaptoethanol, 1 mM dithiothreitol, and proteinase inhibitors as above), snap-frozen in liquid nitrogen, then subsequently thawed on ice. The samples were sonicated to lyse all the cells. After centrifugation, the supernatants were incubated with nickel-nitrilotriacetic acid resin for 1 h at 4°C and transferred to a column. The column was washed with buffer containing 20 mM imidazole-HCl, pH 7.0, 20 mM potassium phosphate, 300 mM NaCl, 10% glycerol, 0.1% ␤-mercaptoethanol, 1 mM dithiothreitol, and proteinase inhibitors. The proteins were eluted with 150 mM imidazole, and fractions of 0.5 ml were collected. An aliquot of 10 l from each fraction was subjected to SDS-PAGE and analyzed by Coomassie Blue staining and anti-His immunoblotting. The fractions containing recombinant kinase were pooled and were further purified by Mono Q anion-exchange chromatography using an NaCl gradient from 100 to 600 mM. A lysate from Sf9 cells transfected with wild type baculovirus was processed in parallel.
Kinase Assay-Autophosphorylation was performed by incubating purified recombinant KPI-2 kinase at 30°C for 20 min in 20 l of reaction buffer containing 20 mM Hepes, pH 7.4, 1 mM MnCl 2 , 10 mM MgCl 2 , 5 mM ␤-glycerolphosphate, 0.1 mM Na 3 VO 4 , 2 mM dithiothreitol, 0.4 mM Pefabloc-sc, and 100 M [ 32 P]ATP (10 Ci). The reaction was terminated by the addition of 6ϫ SDS sample buffer and boiling for 5 min. The samples were resolved by SDS-PAGE, and the gel was stained with Coomassie Blue. The phosphorylation was detected by autoradiography and quantitated by excising the band corresponding to the protein and measuring the radioactivity with a scintillation counter. The substrate phosphorylation used 0.25 mg/ml myelin basic protein (MBP), Histone H1, recombinant His 6 -Inh2, or poly(Glu:Tyr) (4:1) under the same conditions. The eluant from Sf9 cells transfected with wild type baculovirus was used as negative control for all the kinase assays.
Phosphoamino Acid Analysis-Following kinase reaction and autoradiography, the band corresponding to the KPI-2 kinase and MBP were excised. The samples were digested by trypsin overnight followed by acid hydrolysis for 1 h in 6 M HCl at 100°C. Phosphoamino acid analysis was carried out using one-dimensional thin layer electrophoresis at pH 2.5 (31). The phosphoamino acids were detected by autoradiography. Phosphoamino acid standards were included in the samples, and their locations were determined by Ninhydrin staining.
PP1C Binding Assay-Both GST-KPI-2-wt and GST-KPI-2-AA-mut were expressed in E. coli DH5␣ and purified with glutathione-Sepharose beads. The GST-KPI2-wt or AA-mut fusion protein (2 g) was bound to glutathione-Sepharose beads and incubated with HeLa cell lysates at 4°C for 1 h as described previously (32). The beads were washed extensively, and PP1C binding was detected by immunoblotting with anti-PP1C antibody. The content of the GST fusion protein was determined by immunoblotting with anti-GST antibody.
Phosphorylase Phosphatase Assay-PP1C activity was assayed by the release of 32 P phosphate from 32 P-labeled phosphorylase A as described in Shenolikar and Ingebritsen (33). Activity of PP1C purified from rabbit skeletal muscle was determined in a reaction mixture (40 l) containing 20 mM MOPS, pH 7.4, 2 mM MgCl 2 , 1 mg/ml bovine serum albumin, and 15 M 32 P-labeled phosphorylase A at 30°C for 15 min. Acid-soluble 32 P was analyzed by liquid scintillation counting.

Identification of Novel Kinase KPI-2-To identify binding
proteins for Inh2, we employed yeast two-hybrid analysis with human Inh2 (residues 1-197) as bait. One of six independent clones was a fragment of an uncharacterized protein encoded by the human open reading frame of KIAA1079 (NM-011575). The fragment identified in the Inh2 screen, called IB-4, was the region of residues 1344 -1450, including the PP1C binding motif VTF (Fig. 1A).
We cloned the cDNA from HeLa cells by using RT-PCR and named it KPI-2 (Kinase/Phosphatase/Inhibitor-2). Our KPI-2 cDNA clone was 4512 nucleotides in length encoding a 1503amino acid residue protein (GenBank TM accession number AY130988) and matched with human chromosome 7q21.3-q22.1. The KIAA1079 sequence differs from KPI-2 by missing 94 nucleotides that were thought to be an intron with a mismatched splicing junction.
The domain structure of KPI-2 protein is shown in Fig. 1A. There are two predicted transmembrane helices (available at www.enzim.hu/hmmtop/) that extend between amino acid residue ranges 11-29 and 46 -63 near the N terminus. There is one kinase domain (residues 137-407) (www.kinase.com), which contains an ATP binding motif (residues 143-168) and has about 60% sequence identity with mouse apoptosis-associated tyrosine kinase (AATYK) (34 -37). In the C-terminal region, there is a predicted PP1C binding motif KKAVTFFD that contains key Val and Phe at amino acid residue 1355 and 1357. This site was recovered in the two-hybrid clone IB-4.
Tissue-specific Expression of KPI-2-To examine the expression of KPI-2 in different tissues, we probed a Clontech Human Multiple Tissue Northern (MTN TM ) blot with a KPI-2 kinase domain (280 -1800 nt). KPI-2 mRNA was detected as a single band at size ϳ10 kb, which was expressed predominantly in skeletal muscle, with low level expression in brain and pancreas (Fig. 1C). The result was confirmed by reprobing the same membrane with a cDNA corresponding to the KPI-2 Cterminal domain (3295-4509 nt, not shown). Because KPI-2 is expressed mostly in skeletal muscle, we checked by RT-PCR the expression of KPI-2 in C2C12 cells, a mouse myoblast cell line. Consistent with the Northern blot, KPI-2 was expressed in C2C12 cells, but its expression level did not change before and after induction of muscle differentiation (not shown).
KPI-2 Is a Membrane Protein-The full-length KPI-2 protein and its N-terminal (1-703 residues) were expressed in HEK293T cells and detected by Western blotting with anti-myc antibody. Recombinant KPI-2 protein was ϳ210 kDa, and the N-terminal-(1-703) was an ϳ90-kDa protein as expected from their sequences ( Fig. 2A). Because of the predicted transmembrane helices in the N terminus, we expected the KPI-2 protein to be membrane-bound. To test this, we expressed KPI-2 N-terminal-(1-703) in COS7 cells, and prepared soluble and membrane fractions from these cells. The myc-tagged KPI-2-(1-703) protein was recovered entirely in the membrane fraction solubilized by 1% Nonidet P-40 and did not appear in the soluble fraction (Fig. 2B). Phospholemman was used as a plasma membrane protein marker, and protein phosphatase 2A catalytic subunit was used as cytosolic protein marker. The results show that KPI-2 N-terminal-(1-703) is a membrane protein. Fusion of KPI-2 residues 7-139 to green fluorescent protein (GFP) also resulted in predominant distribution of fusion protein into membrane fraction (not shown). This suggests that the predicted transmembrane helices are sufficient for targeting KPI-2 or fusion proteins to membranes.
Expression, Purification, and Autophosphorylation of Recombinant KPI-2 Kinase-We expressed the KPI-2 kinase domain (residues 94 -600) as a His 6 -tagged fusion protein in Sf9 cells. Following metal-ion affinity chromatography, one major protein of ϳ75 kDa was detected in the Coomassie Blue-stained gel (Fig. 3A, left panel), and this protein also reacted with anti-His antibody (Fig. 3A, right panel). The 75-kDa KPI-2 kinase from Sf9 cells was purified by Mono Q anion-exchange chromatography (not shown), and an autophosphorylation as-say was performed. The purified recombinant kinase was incubated with [ 32 P]ATP, and autophosphorylation was observed (Fig. 3B, left panel). We also performed a kinase assay using different substrates. Myelin basic protein (MBP) was phosphorylated efficiently, but Histone H1 and His 6 -Inh2 were relatively poor substrates under the same conditions (Fig. 3B, right  panel). The tyrosine kinase substrate poly(Glu:Tyr) (4:1) was not phosphorylated at all, even with prolonged incubation (data Total cell lysates were prepared 24 h later and subjected to 9% SDS-PAGE gel followed by immunoblotting with anti-myc antibody. The molecular standards (kDa) are indicated in the left side of the frame. B, COS7 cells were transfected with pCMV-myc-KPI2-N-terminal-(1-703), and empty vector was used as control. The soluble (S) and detergentsolubilized membrane (M) fractions were prepared by homogenization followed ultracentrifugation, 100,000 ϫ g for 60 min. Equal amounts of proteins were applied to SDS-PAGE and analyzed with anti-myc immunoblotting. The membrane was stripped and reblotted with anti-Phospholemman (membrane protein marker), and anti-protein phosphatase 2A catalytic subunit (PP2Ac) (cytosolic protein marker) antibodies.

FIG. 3. Expression of KPI-2 kinase domain and kinase assay. A,
Sf9 cells were infected with recombinant baculovirus of KPI-2 kinase-(94 -600) for 60 h. The recombinant protein was purified by metalaffinity chromatography and the lysate, unbound, and eluted fractions were analyzed by Coomassie Blue staining (left panel), and the eluant was assayed by anti-His immunoblotting (right panel). An aliquot of 10 l of each fraction was subjected to SDS-PAGE. Sf9 cells transfected with wild type baculovirus was used as control. Size standards (kDa) are the same as Fig. 2. B, purified KPI-2 kinase-(94 -600) was assayed alone (left panel) or with different substrates (0.25 mg/ml each): myelin basic protein (MBP), Histone H1 or His 6 -Inh2 (right panel) with [ 32 P]ATP at 30°C for 20 min. The reactions were terminated by the addition of 6ϫ SDS sample buffer. The phosphorylation of proteins was detected by autoradiography after SDS-PAGE. C, phosphorylated KPI-2 kinase and MBP were excised from the gel after autoradiography, digested with trypsin, and hydrolyzed in 6 N HCl for 1 h. Phosphorylated amino acids were determined by using one-dimensional thin layer electrophoresis as described under "Experimental Procedures." The left panel is autophosphorylated KPI-2 kinase, and the right panel is phosphorylated MBP. not shown). A lysate from Sf9 cells transfected with wild type baculovirus was processed in parallel, and the corresponding Mono Q eluant was used as negative control for all the kinase assays. These results showed that purified KPI-2 protein is an active kinase.
We next determined the residues phosphorylated by KPI-2 kinase. After autoradiography, the bands corresponding to the autophosphorylated KPI-2 kinase and the phosphorylated substrate MBP were excised, and phosphoamino acid analysis was carried out using one-dimensional thin layer electrophoresis. The results showed that both KPI-2 autophosphorylation and MBP phosphorylation were mostly located at serine, with a trace at threonine. No tyrosine phosphorylation was observed (Fig. 3C). Immunoblotting with anti-phospho-Thr-Pro and antiphospho-Ser antibody also showed reactivity with autophosphorylated KPI-2. This evidence shows that KPI-2 is a serine/ threonine protein kinase.
Phosphorylation of PP1C by KPI-2 Reduces PP1C Activity-KPI-2 is a serine/threonine protein kinase with a PP1 binding motif in the C-terminal domain that would bring the kinase and phosphatase together. We checked whether KPI-2 could phosphorylate PP1C and alter its activity. Purified PP1C was incubated with or without purified KPI-2 kinase-(94 -600). Reactions contained 100 M ATP␥S, which was used instead of ATP to retard dephosphorylation of PP1C during the reaction. Phosphatase assays were performed using [ 32 P]phosphorylase a as substrate. PP1C activity was reduced 70% by reaction with KPI-2 kinase compared with incubation without kinase (Fig.  4A). PP1C is known to be inactivated by phosphorylation at Thr 320 (38 -40), so we tested if the phosphorylation by KPI-2 was located at the Thr 320 site in PP1C. PP1C was inhibited by incubation with 1 M microcystin-LR and incubated with ATP plus/minus KPI-2 kinase. Western blotting using anti-phospho-PP1C␣(Thr 320 ) antibody showed that PP1C was phosphorylated at Thr 320 by KPI-2 kinase (Fig. 4B). Thiophosphorylated PP1C did not react with the phospho-specific antibody. Thus, KPI-2 phosphorylated Thr 320 in PP1C, which reduces phosphatase activity.
KPI-2 Binding to PP1C Requires a C-terminal VXF Motif-There is a consensus PP1C binding motif VTF in the C-terminal domain of KPI-2. To check PP1C binding, we prepared GST-KPI-2-(1099 -1503) and used HeLa cell lysates as a source of PP1C in a pull-down assay. Wild type (wt) GST-KPI-2-(1099 -1503) bound PP1C, but GST alone as a control did not pull down PP1C from these same lysates. Substitution of Ala for Val 1355 and Phe 1357 in GST-KPI-2-(1099 -1503) to give an AA-mutant protein eliminated PP1C binding (Fig. 5, upper  panel). Equivalent amounts of GST and GST fusion proteins were present in the assays, shown by immunoblotting with anti-GST antibody (Fig. 5, lower panel). Similar results were obtained with GST-KPI-2-wt plus purified PP1C (not shown), showing that PP1C bound directly and did not require any other proteins for association. These results showed that a GST-KPI-2 fusion protein binds PP1C and that this binding requires the VTF motif.
Binding of VXF Motif Subunits Reduces Phosphorylase Phosphatase Activity of PP1C-GST-KPI-2-wt-(1099 -1503) and the AA-mut fusion protein or GST alone were preincubated with purified PP1C, then phosphorylase phosphatase activity was measured. As shown in Fig. 6A, GST-KPI-2-wt potently reduced PP1C activity in a dose-dependent manner over the nanomolar concentration range. The AA-mutant protein that did not bind PP1C in a pull-down assay (Fig. 5) produced no reduction in phosphorylase phosphatase activity of PP1C, like GST alone, up to 100 nM (Fig. 6A).
We compared the effects of binding different regulatory subunits to PP1C. Various amounts of GST-KPI-2- (1099 -1503), GST-G M -(1-240), or GST-M130-(1-498) fusion proteins were incubated with purified PP1C and assayed for phosphorylase phosphatase activity of PP1C. As shown in Fig. 6B, GST-KPI-2-(1099 -1503) and GST-M130-(1-498) gave identical reduction of PP1C activity, but GST-G M -(1-240), which binds PP1C in a pull-down assay over this range of concentrations (41), produced less of an effect. The results showed that different regulatory subunits could use a VXF motif to bind PP1C with about the same apparent affinity but reduce PP1C activity with phosphorylase as substrate to different levels.
Inhibitor 2 Binding to KPI-2-Does KPI-2 bind to PP1C in living cells? Does Inh2 bind to KPI-2 directly or only bind to PP1C, producing an indirect association with KPI-2? We overexpressed FLAG-KPI-2-wt- (1099 -1503), the KPI-2-AA-mut or  empty vector, and HA 3 -tagged Inh2 in COS7 cells and prepared cell extracts and immunoprecipitated them with anti-FLAG M2 agarose. PP1C bound to FLAG-KPI-2-wt but did not bind to FLAG-KPI-2-AA-mut (Fig. 7, top panel, IP). This result was completely consistent with the GST fusion protein pull-down assay (see Fig. 5). However, Inh2 bound to both FLAG-KPI-2-wt and FLAG-KPI-2-AA-mut but not to the anti-FLAG M2agarose used as a negative control (Fig. 7, bottom panel, IP). This indicated that PP1C was not required for Inh2 binding to the KPI-2 C-terminal domain. Immunoblot analysis of the cell extracts and the immunoprecipitates revealed that FLAG-KPI-2 proteins were expressed to the same level and precipitated identically (Fig. 7). Inh2 was also expressed at the same level in each sample. These results showed: 1) FLAG-KPI-2wt-(1099 -1503) protein binds to PP1C in living cells; 2) Val 1355 and Phe 1357 in KPI-2 are necessary for this association; and 3) Inh2 binds to KPI-2-(1099 -1503) with or without PP1C.
Inhibitor 2 Binding to KPI-2 Requires PP1C in Yeast-We checked Inh2 and PP1C binding to KPI-2 in yeast by performing protein-protein interaction assay based on yeast 2-hybrid. Both bait-vector (PP1C␣ or Inh2) and prey-vector (KPI-2-wt or AA-mut) were cotransformed into HF7c yeast strain, and protein-protein interaction was determined by growth of clones on the medium lacking histidine. KPI-2-wt-(1099 -1503) and IB-4 interacted with both PP1C␣ and Inh2. In contrast, the KPI-2-AA-mut did not bind either PP1C␣ or Inh2 (Fig. 8, ϪHis). PP1 was the positive control for binding to Inh2, and the PP1 glycogen targeting subunit G M -(1-240) was the positive control for binding to PP1C␣. All the strains grew well in double dropout medium as a control (Fig. 8, ϩHis). The results confirm that KPI-2 binding to PP1C requires the C-terminal VXF motif. However, Inh2 did not associate with KPI-2-AA-mut, so in yeast Inh2 binding to KPI-2 required PP1C. DISCUSSION We identified KPI-2 as a unique human protein with transmembrane helices, a Ser/Thr kinase domain and a C terminus region that binds PP1 and Inh2. By sequence analysis the nearest relative of KPI-2 is a tyrosine kinase (AATYK), which is induced during apoptosis or terminal differentiation of 32Dcl3 cells and is otherwise expressed predominantly in brain (34 -37). KPI-2 has about 60% sequence identity with AATYK within the kinase domain and shows no sequence similarity outside the kinase domain. Based on sequence alignments, Hanks and Hunter (42) proposed that protein kinase subdomains VI-VIII predict serine/threonine versus tyrosine kinase specificity. The catalytic loop in these catalytic domains is highly conserved among each of these groups of enzymes. In serine/threonine kinases, a Lys residue is The interaction between KPI-2 and Inh2 was checked using alternative yeast two-hybrid system. pGBT10-Inh2 or PP1C␣ was used as bait vector. pVP16-KPI-2-wt-(1099 -1503), the AA-mut, and IB-4 were prey vectors. Both vectors were transformed into yeast HF7c strain, and transformations were checked by growth on double dropout plate (SD/ -Trp/-Leu) (right panel). Protein-protein interaction was determined by growth on a triple dropout plate (SD/-Trp/-Leu/-His) (left panel). PP1 was the positive control for binding to Inh2. G M -(1-240) was the positive control for PP1C␣ binding.
found at the nϩ2 position in the subdomain VIb relative to the essential Asp residue, and most are DLK. In contrast, the protein-tyrosine kinases have an Arg residue instead at either the nϩ2 position or nϩ4 position, like DLR or DLAXR (42,43). There is a common DLAXR motif in KPI-2 (residues 265-269, DLALR) and its closest related kinases (AATYK and trkB) raising the expectation that KPI-2 would be a tyrosine kinase. However, our results showed that KPI-2 is a Ser/Thr kinase and does not exhibit tyrosine kinase activity, based on multiple lines of evidence. First, phosphoamino acid analysis of autophosphorylated KPI-2 showed only phospho-Ser and phospho-Thr, no phospho-Tyr. Second, immunoblotting with anti-phospho-Thr-Pro and anti-phospho-Ser antibody both showed reactivity with KPI-2. Third, KPI-2 did not phosphorylate the nonspecific tyrosine kinase substrate poly-(Glu:Tyr) (4:1). Fourth, only Ser and Thr but not Tyr were phosphorylated in MBP by KPI-2. Fifth, KPI-2 phosphorylated Thr 320 in PP1C. Therefore, KPI-2 is an unusual Ser/Thr kinase with activity different than expected from a popular structure-function analysis of kinases.
The structure of KPI-2 predicts two transmembrane helices in the N-terminal region, and our results showed recombinant KPI-2 is a membrane protein. The putative transmembrane helices were sufficient to target either KPI-2-(1-703) or a GFP fusion protein to the membrane fraction. We presume the kinase and PP1C-binding domain of KPI-2 are cytoplasmic. If both of the predicted transmembrane helices span the membrane, then the N terminus would also be cytoplasmic and only a short loop would be exposed on the cell surface. It is possible that only one of the two predicted helices spans the membrane, in which case the N terminus would be extracellular. In either case, if KPI-2 participates in signaling, we imagine that ligand binding could be done by another protein on the cell surface that acts as a KPI-2 partner, forming a heterodimeric receptor. Alternatively, KPI-2 might regulate specific ion channels or transporters in the plasma membrane. We note that the immediate juxtamembrane region of KPI-2 has multiple Cys residues that could form disulfide bonds with other proteins. There is as yet no evidence for such a coreceptor or binding partner. KPI-2 mRNA showed preferential expression in skeletal muscle, but there was no change in mRNA levels between C2C12 myoblasts and myotubes. The KPI-2 function in skeletal muscle remains unknown.
KPI-2 is a serine/threonine kinase that phosphorylated PP1, a serine/threonine phosphatase, at Thr 320 and attenuated its activity. The Thr 320 site in PP1C is also phosphorylated by cdk2:cyclin B (38,39,44) as well as by Nek2 (45). PP1 activity is attenuated in each case. Both KPI-2 and Nek2 form complexes with PP1C where the kinase and phosphatase can react with one another. Phosphorylation of PP1C is underappreciated as a regulatory mechanism.
The pleiotropic function of PP1 resides in the ability of the catalytic subunit to associate with many different regulatory subunits through a VXF motif present in the majority of these subunits (13,46). In KPI-2, there is a VTF motif near the C terminus. Our results showed that KPI-2 binds PP1C both in vitro and in vivo and that the interaction requires this VTF motif. Thus, KPI-2 is a PP1 regulatory subunit, and, like other VXF subunits, it suppresses PP1C activity with phosphorylase as a substrate. Therefore KPI-2 exhibits multiple modes of regulation on PP1C: 1) membrane localization, 2) allosteric control via VXF interactions, and 3) covalent control by phosphorylation of Thr 320 .
Lastly, KPI-2 was found by yeast 2-hybrid with Inh2. Yeast two-hybrid assays showed that KPI-2 C-terminal domain interacted with PP1C and with Inh2, but mutation of the VTF motif eliminated both interactions. This suggested that Inh2 bound to KPI-2 indirectly, by interacting with PP1C that was bound to the VTF of KPI-2. We have uncovered recently other examples where PP1C binds to Inh2 and a VXF-containing subunit at the same time, through separate sites (47,48). However, in COS7 cells Inh2 bound to the wild type KPI-2 C-terminal domain as well as the AA mutant that did not associate with PP1C. We concluded that neither PP1C nor the VXF motif were required for Inh2 interaction with KPI-2. Curiously, KPI-2 appears to bind Inh2 in COS7 cells but not in yeast. One possibility to account for this difference is Inh2 phosphorylation. In mammalian cells Inh2 is fully phosphorylated at Ser 86 , Ser 120 , and Ser 121 (49), but it is not known whether these CK-II sites are phosphorylated in yeast. Another possibility is that an unknown scaffold protein in mammalian cells, not in yeast, binds Inh2 to KPI-2. Regardless, our evidence shows that KPI-2 is a novel example of a PP1 regulator that brings together a PP1 kinase and PP1 inhibitor for regulation of this phosphatase at the membrane surface. The physiological function of this multienzyme complex remains an unsolved mystery for further investigations.