Molecular Interactions of Yeast Frequenin (Frq1) with the Phosphatidylinositol 4-Kinase Isoform, Pik1* 210

Frq1, a 190-residue N-myristoylated calcium-binding protein, associates tightly with the N terminus of Pik1, a 1066-residue phosphatidylinositol 4-kinase. Deletion analysis of an Frq1-binding fragment, Pik1-(10–192), showed that residues within 80–192 are necessary and sufficient for Frq1 association in vitro. A synthetic peptide (residues 151–199) competed for binding of [35S]Pik1-(10–192) to bead-immobilized Frq1, whereas shorter peptides (164–199 and 174–199) did not. Correspondingly, a deletion mutant, Pik1(Δ152–191), did not co-immunoprecipitate efficiently with Frq1 and did not support growth at elevated temperature. Site-directed mutagenesis of Pik1-(10–192) suggested that recognition determinants lie over an extended region. Titration calorimetry demonstrated that binding of an 83-residue fragment, Pik1-(110–192), or the 151–199 peptide to Frq1 shows high affinity (K d ∼100 nm) and is largely entropic, consistent with hydrophobic interaction. Stoichiometry of Pik1-(110–192) binding to Frq1 was 1:1, as judged by titration calorimetry, by changes in NMR spectrum and intrinsic tryptophan fluorescence, and by light scattering. In cell extracts, Pik1 and Frq1 exist mainly in a heterodimeric complex, as shown by size exclusion chromatography. Cys-15 in Frq1 is notS-palmitoylated, as assessed by mass spectrometry; a Frq1(C15A) mutant and even a non-myristoylated Frq1(G2A,C15A) double mutant rescued the inviability of frq1Δ cells. This study defines the segment of Pik1 required for high affinity binding of Frq1.

Recognition that phosphoinositides and inositol phosphates are key regulators of many processes in eukaryotic cells has brought increased attention to the enzymes that regulate the synthesis and turnover of these molecules (reviewed in Refs. [1][2][3]. Of particular interest are the enzymes responsible for producing the various polyphosphoinositides situated on the cytosolic face of cellular membranes, which initiate several different signaling pathways by serving as highly specific recognition determinants for the selective recruitment of proteins to membranes (reviewed in Refs. 4 -7) and as the precursors for several intracellular second messengers (reviewed in Refs. 8 -10). The first committed step in the synthesis of the polyphosphoinositide, phosphatidylinositol 4,5-bisphosphate, is considered to be ATP-dependent phosphorylation of the hydrophilic myo-inositol head group of phosphatidylinositol (PtdIns) 1 at the D-4 position by PtdIns 4-kinase (ATP:1-phosphatidyl-1Dmyo-inositol 4-phosphotransferase, EC 2.7.1.67) (reviewed in Refs. [11][12][13] . The resulting product, PtdIns(4)P, can be phosphorylated on the D-5 position by PtdIns(4)P 5-kinase to generate PtdIns(4,5)P 2 , PtdIns(4,5)P 2 can be phosphorylated on the D-3 position by yet other lipid kinases, and the phosphoinositides so generated can be converted to other species by specific phosphatases and phospholipases (reviewed in Refs. 14 -17).
The first PtdIns 4-kinase to be purified to homogeneity from any organism (18), and to have the corresponding gene cloned (19,20), was Pik1 from the yeast Saccharomyces cerevisiae. Thereafter, a second isoform, Stt4, which is the product of a discrete gene, was described (21). Absence of either Pik1 or Stt4 is lethal, and overproduction of each protein cannot compensate for absence of the other, indicating that these enzymes participate in distinct cellular processes and generate discrete pools of PtdIns(4)P that are essential for yeast cell viability. Indeed, subsequent work has shown that, together, Pik1 and Stt4 account for all of the PtdIns(4)P generated in the yeast cell (22) and that Pik1 is required for vesicular trafficking in the late secretory pathway (23,24) and perhaps for cytokinesis (20), whereas Stt4 plays roles in cell wall integrity, maintenance of vacuole morphology, and aminophospholipid transport from the endoplasmic reticulum to the Golgi (25)(26)(27). The presence of Pik1-and Stt4-like isoforms is also conserved in metazoans (11,12,28).
We have shown previously that Frq1, a small calcium-bind-ing protein, co-purifies with Pik1 and is required for optimal activity of the enzyme (29). Frq1 is the yeast ortholog of a protein called frequenin, first described in Drosophila (30), but referred to as neuronal-calcium-sensor-1 (NCS-1) in mammalian cells. Members of a large subfamily of small, EF-handcontaining, calcium-binding proteins that includes frequenin (31)(32)(33)(34) are characterized by a consensus signal for N-terminal myristoylation and four Ca 2ϩ -binding sites (of which the first and, in some cases, the fourth or another contain substitutions that make them non-functional). We have shown previously that Frq1 binds three Ca 2ϩ (35). Available evidence indicates that frequenin/NCS-1 may also modulate PtdIns 4-kinase activity in animal cells (36,37). Frq1, which is itself essential for the viability of yeast cells (29), associates with membranes in a manner that depends on both the N-myristoyl group and conformational changes induced upon Ca 2ϩ binding (35). Thus, in addition to its stimulation of enzymic activity, Frq1 may contribute to the optimal function of Pik1 by assisting with its membrane recruitment, because Pik1 itself lacks any obvious membrane-targeting motifs. Indeed prior work indicated that N-myristoylation of Frq1 is important, but not essential, for its function (29). In some Ca 2ϩ -binding regulatory proteins, in addition to the N-terminal myristoyl group, palmitoylation of a cysteine residue near the N terminus is also required for efficient membrane association (38,39). Frq1 has only two Cys residues, one is near its N terminus and the other buried in the interior (35).
In this study, as a prelude to structural analysis to determine at atomic resolution how Frq1 recognizes Pik1, we have applied several independent methods to determine the affinity and stoichiometry of the Frq1-Pik1 interaction, used different approaches to delineate the sequences in Pik1 responsible for high affinity binding of Frq1 and utilized both biochemical and genetic techniques to explore the role, if any, of S-palmitoylation in the function of Frq1.

EXPERIMENTAL PROCEDURES
Strains and Growth Conditions-Yeast strains used in this study are listed below in Table I. Standard rich (YP) and synthetic complete (SC) media (40) were supplemented with carbon sources (either 2% Glc, or 2% Raf/0.2% Suc, or 2% Gal/0.2% Suc, as indicated) and with appropriate nutrients for the selection and maintenance of plasmids. Yeast was cultivated at 30°C, unless otherwise noted. Conventional methods for DNA-mediated transformation and other genetic manipulations of yeast cells were used (41).
Plasmid Construction-Plasmids were constructed using standard methods for the manipulation of recombinant DNA (42). Escherichia coli strain DH5␣ (43) was used for routine manipulation and propagation of plasmids. Unless otherwise indicated, all PCR reactions were performed using Pfu DNA polymerase (Stratagene, La Jolla, CA). All recombinant plasmids were verified by dideoxy chain termination sequencing.
Production of Radiolabeled Proteins by Coupled in Vitro Transcription and Translation-Vectors for in vitro production of mRNA were constructed, as follows. A fragment containing the FRQ1 coding sequence, generated by PCR and containing an NcoI site overlapping the translation initiation codon and a BamHI site introduced 3Ј to the stop codon, was inserted into pBAT4 (46), which had been cleaved with NcoI and BamHI, yielding pBAT4 -FRQ1. A PCR fragment was amplified from pET23d-PIK1 (10 -192) using appropriate primers to introduce a SmaI site 5Ј to the coding sequence and a HindIII site 3Ј to codon 192 of the PIK1 open reading frame, cleaved with SmaI and HindIII, and ligated into pBAT4 that had been cleaved with the same enzymes, generating pBAT4 -PIK1 (10 -192). 35 S-Labeled proteins were produced by coupled in vitro transcription and translation in the presence of [ 35 S]Met and [ 35 S]Cys (PerkinElmer Life Sciences, Boston, MA) using the TNT TM coupled reticulocyte lysate system (Promega, Madison, WI), according to the manufacturer's instructions. Translation mixtures were clarified by centrifugation (10 min, 4°C) at maximum speed in a microcentrifuge. If not used immediately, the resulting supernatant fractions were mixed with an equal volume of glycerol and stored at Ϫ20°C.
Bacterial Expression and Purification of (His) 6 -tagged Proteins-The Pik1 and Frq1 constructs containing a C-terminal His 6 tag were expressed in E. coli strain BL21(DE3) (Novagen, Madison, WI) and purified using Ni 2ϩ -saturated NTA-agarose (Qiagen, Valencia, CA) under denaturing conditions according to the manufacturer's specifications. To confirm their identity and purity, proteins recovered by binding to Ni 2ϩ -saturated NTA-agarose were resolved by SDS-PAGE and visualized by staining with Coomassie Brilliant Blue and by immunoblotting. Protein concentration was determined by the dye-binding method of Bradford (47) using a commercial kit (Bio-Rad, Inc., Hercules, CA) with bovine serum albumin as the standard.
Purification of Frq1-(His) 6 from Yeast Cells-Protease-deficient strain, BJ2168 (Table I), transformed with YEp352-GAL-FRQ1-(His) 6 , was grown in SC-Raf lacking uracil at 30°C to mid-exponential phase and induced by addition of Gal (2% final concentration). After incubation for 6 h at 30°C, cells were collected by centrifugation and resuspended in an equal volume of distilled water. The cell suspension was frozen by dripping into liquid nitrogen, and the resulting pellets were crushed with a pestle in a precooled mortar. The resulting frozen cell powder was dissolved in lysis buffer (5 mM imidazole, 145 mM NaCl, 50 mM Na-PO 4 (pH 7.5); 20 ml/liter yeast culture) containing a mixture of protease inhibitors (Complete TM , Promega, Madison, WI). The crude lysate was clarified by centrifugation at maximum rpm in a microcentrifuge at 4°C for 15 min and then at 72,000 ϫ g for 90 min in a L8 -80M ultracentrifuge (Beckman-Coulter Inc., Fullerton, CA). The resulting supernatant fraction was then applied to a Ni 2ϩ -saturated NTA-agarose column (1.5-ml bed volume) that had been pre-equilibrated with three volumes of lysis buffer. After washing with 10 bed volumes of lysis buffer and 6 volumes of wash buffer (20 mM imidazole, 145 mM NaCl, 50 mM Na-PO 4 (pH 7.5)), the bound Frq1-His 6 was eluted with 3 volumes of elution buffer (120 mM imidazole, 145 mM NaCl, 50 mM Na-PO 4 (pH 7.5)). The eluate was concentrated by ultrafiltration through an anisotropic membrane (3-kDa cut-off, Centricon YM-3, Amicon, Beverly, MA) until a final concentration of 0.5 g/ml was reached. To confirm identity and purity, the resulting fraction was resolved by SDS-PAGE and visualized by staining with Coomassie Brilliant Blue and by immunoblotting.
In Vitro Protein Binding Assays-Prior to use, Ni 2ϩ -saturated NTAagarose beads used in protein binding and peptide competition experiments were pre-blocked by incubation for 30 min in 8 volumes of buffer A (10 mM imidazole, 100 mM NaCl, 1 M CaCl 2 , 1 mM dithiothreitol, 50 mM Tris-HCl (pH 7.4)) containing 5 mg/ml ovalbumin at room temperature. All in vitro binding assays were carried out at 4°C. Radiolabeled Frq1, prepared by coupled in vitro transcription and translation, was mixed with an equal volume of a slurry of pre-blocked Ni 2ϩ -saturated NTA-agarose beads in buffer A and incubated on a roller drum for 30 min. The beads and any nonspecifically bound radioactivity were removed by brief sedimentation in a microcentrifuge, and the resulting pre-cleared supernatant fraction was collected. Aliquots (400 l) of the pre-cleared fraction were mixed either with 40 l of pre-blocked Ni 2ϩsaturated NTA-agarose beads, as a control for background binding, or with Ni 2ϩ -saturated NTA-agarose beads on which had been immobilized Pik1-(10 -192)-(His) 6 or its deletion derivatives (30 g of protein/40 l of beads) and incubated for 1 h on a rollerdrum. The beads were collected by centrifugation for 15 s in a microcentrifuge and washed three times with buffer A. Bound proteins were eluted from the beads in 50 l of buffer A containing 300 mM imidazole, and samples of the resulting eluate were resolved by SDS-PAGE on a 12% gel and visualized by autoradiography.
Synthetic peptides corresponding to Pik1-(174 -199), Pik1-(164 -199), and Pik1-(151-199) were prepared by standard solid phase peptide synthesis (using Fmoc chemistry) on an automated synthesizer (Model ABI 431A, PerkinElmer Life Sciences-Applied Biosystems, Foster City, CA), purified by high-performance liquid chromatography, and confirmed by electrospray ionization mass spectrometry. A slurry (30 l), of either pre-blocked Ni 2ϩ -saturated NTA-agarose beads or the same beads pre-coated to saturation with purified Frq1-His 6 , was mixed with 500 l of 2ϫ-concentrated buffer A, 290 l of H 2 O, and 100 l of either an aqueous solution of the indicated peptide in 10% acetonitrile or H 2 O containing 10% acetonitrile as a control. After preincubation of the samples on a roller drum for 30 min, 80 l of 35 S-labeled Pik1- (10 -192), produced by coupled in vitro transcription and translation, was added to each tube, and the mixture was incubated for a further 1.5 h. The beads were collected by sedimentation in a microcentrifuge and washed twice with buffer A. Bead-bound radioactivity was solubilized by boiling in SDS-PAGE sample buffer (48), and the resulting eluate was subjected to SDS-PAGE. The species corresponding to [ 35 S]Pik1- (10 -192) was quantitated using a PhosphorImager TM (Amersham Biosciences, Sunnyvale, CA). At each peptide concentration, the amount of radioactivity bound nonspecifically to empty beads was subtracted from the amount of radioactivity bound to the beads coated with Frq1-His 6 .
NMR Spectroscopy-15 N-Labeled and unlabeled samples of unmyristoylated Frq1 were prepared as described previously (35). Chemical synthesis of the 49-residue synthetic peptide corresponding to residues 151-199 of Pik1 was described in the preceding section. Using PCR with appropriate primers, an 83-residue segment from the N terminus of Pik1 consisting of residues , was tagged at its C terminus with an His 6 tract, inserted as an NcoI-XhoI fragment into the corresponding sites in the vector, pET23d, and expressed in E. coli strain BL21(DE3), as described above; in this construct, the Pik1derived sequence is preceded by a 2-residue leader (Met-Ala-). Pik1-(110 -192)-(His) 6 was produced primarily in inclusion bodies, which were solubilized using 6 M guanidine hydrochloride (49) and purified using Ni 2ϩ -saturated NTA-agarose chromatography, essentially as described above. Samples for NMR analysis were prepared by dissolving 15 N-labeled Frq1 (0.4 mM) with various amounts (0, 1, or 2 molar equivalents) of Pik1- (110 -192) or the synthetic peptide, Pik1-(151-199), in 0.5 ml of a 95% H 2 O/5% [ 2 H]H 2 O solution containing 10 mM imidazole (pH 6.7), 10 mM [ 2 H 10 ]dithiothreitol, and either 1 mM EDTA (Ca 2ϩ -free) or 5 mM CaCl 2 (Ca 2ϩ -bound). All NMR experiments were performed at 37°C on a Model DRX-500 or DRX-600 NMR spectrometer (Bruker Instruments, Billerica, MA) equipped with a four-channel interface and a triple-resonance probe with triple-axis pulsed field gradients as described before (35). The NMR spectra were processed and analyzed as described previously (50).
Fluorescence Spectroscopy-The effect of Pik1 peptides on the intrinsic tryptophan fluorescence emission of Frq1, excited at 290 nm, was measured (at 300 -420 nm) using a SPEX fluorometer (Jobin Yvon Inc., Edison, NJ). Neither Pik1-(151-199) nor Pik1-(110 -192) contain any tryptophan, and they do not contribute any fluorescence under these conditions. Titrations were performed using 5 M Heats of dilution, determined by titrating Frq1 into same buffer alone, were subtracted from the raw titration data before data reduction and analysis. A single-site model was used to fit the titration data for Pik1-(151-199) binding to Frq1. A two-site model was needed to describe the multiphasic interaction of Pik1-(110 -192) with Frq1.
Preparation of Yeast Cell Extracts, Immunoprecipitation, and Immunoblot Analysis-Lysis conditions and centrifugation for the preparation of clarified yeast cells extracts for immunoprecipitation were as described previously (29). Samples (1 mg of total protein) of extract were diluted into lysis buffer (220-l final volume) and mixed with 40 l of a 1:1 (settled beads:buffer) suspension of protein G-/protein A-coupled agarose beads (Oncogene Research Products, Boston, MA) and 2 l of an irrelevant mouse antibody (anti-6-His mAb, BAbCO, Richmond, CA) and incubated at 4°C on a roller drum for 1 h. Beads were removed by brief centrifugation, and the precleared supernatant fraction was transferred to a new tube. Samples of the precleared fraction were then incubated overnight at 4°C with either 2 l of ␣-c-Myc mAb 9E10 (52), 1.5 l of ␣-Frq1, or 1.5 l of pre-immune serum from the same rabbit, which have been described before (29). To capture the soluble immune complexes, a fresh aliquot (30 l) of the suspension of protein G-/protein A-coupled agarose beads was added and the mixture was incubated on the roller drum for 45 min at 4°C. Bead-bound immune complexes were collected by brief centrifugation in a microcentrifuge, washed four times with lysis buffer (1 ml each), resuspended in SDS-PAGE sample buffer, and solubilized by boiling in a water bath for 5 min. After removal of the beads by centrifugation, equal volumes of the resulting supernatant fractions were resolved by SDS-PAGE and analyzed by immunoblotting with appropriate antibodies.
Analytical Size Exclusion Chromatography-Size exclusion chromatography was performed on a Superdex 200 column (Amersham Biosciences, Piscataway, NJ) using an fast protein liquid chromatography apparatus (650E Advanced Protein Purification System, Waters, Milford, MA) at a constant flow rate of 0.1 ml/min in 50 mM Tris-HCl (pH 7.6) containing 150 mM NaCl. Protein standards (either High Molecular Weight Gel Filtration Calibration kit from Amersham Biosciences, Piscataway, NJ, or Gel Filtration Standard from Bio-Rad, Hercules, CA) were prepared according to the manufacturer's specifications and loaded onto the column (90-ml bed volume). Fractions (0.75 ml) were collected, and the elution volumes (V e ) determined by measuring the protein concentration of each fraction by the Bradford method, as described above. The column void volume (V 0 ) was assessed using blue dextran 2000. K av values for each protein were calculated using the equation, and plotted semilogarithmically against the corresponding molecular weight. To examine the apparent molecular weight of native Pik1⅐Frq1 complexes in yeast cell extracts, strain YPH499, transformed with pRS314-GAL-mycPIK1, was grown to A 600 nm ϭ 0.6 in SCRaf-Trp at 30°C and induced by addition of Gal (2% final concentration) and incubated for 1.5 h. Induced cells were collected by centrifugation, washed once with distilled water, resuspended in ice-cold 50 mM Tris-HCl (pH 7.6) containing 150 mM NaCl and protease inhibitors (Complete TM , Promega, Madison, WI). The washed cells were broken by 10 pulses (1 min each) of vigorous vortex mixing with glass beads. The resulting crude extract was clarified by centrifugation in a TL-100 tabletop ultracentrifuge (Beckman Coulter Inc., Fullerton, CA) at 49,000 ϫ g for 30 min at 4°C. The protein concentration of the clarified extract was determined, and 250 l (ϳ2.5 mg of total protein) was loaded onto the gel filtration column. Fractions (0.75 ml) were collected, and protein was concentrated by precipitation with 10% trichloroacetic acid in the presence of 0.15% deoxycholate for 10 min at room temperature. The precipitates were dissolved in 30 l of SDS-PAGE sample buffer, and the remaining acid was neutralized by addition of 5 l of an unbuffered saturated solution of Tris. The resulting fractions were split into two, resolved by SDS-PAGE, and analyzed separately by immunoblotting with either anti-c-Myc or anti-Frq1 antibodies, respectively.
Mass Spectrometry-Mass measurements were performed by electrospray-ionization mass spectrometry using a Model 3000 ion trap mass spectrometer (Bruker Instruments, Billerica, MA). Prior to determining its mass spectrum, each peptide or protein was desalted by microbore reversed-phase high-performance liquid chromatography.

Residues 164 -192 of Pik1 Are Necessary for the Binding of
Frq1-We previously demonstrated that Frq1 associates tightly with Pik1 and that a fragment of the N-terminal domain of Pik1 comprising residues 10 -192 is sufficient to mediate this interaction (29). We also noted before (19) that this segment of Pik1 contains a sequence element (residues 35-110), distinct from the catalytic domain per se, that is weakly conserved among PtdIns 3-kinase, PtdIns 4-kinases, and even more distantly related enzymes that appear to be protein kinases, such as the Tor proteins. This motif has been referred to subsequently as the "lipid kinase unique domain" (LKU) (12,28,53). Because Frq1 binds to the region of Pik1 that contains the LKU, we suggested that the motif itself might be the binding site for this regulatory protein (29).
To test this hypothesis directly and to begin to define the minimal region in Pik1 responsible for Frq1 binding, we constructed three deletion derivatives of the Frq1-binding Pik1-(10 -192) fragment. One deletion (⌬31-79) removed the most conserved core of the LKU motif; the other two deletions were truncations that removed 29 residues (⌬164 -192) and 67 residues (⌬126 -192), respectively, from the C-terminal end (Fig.  1A). Full-length Pik1- (10 -192) and each of the three deletions were tagged with a C-terminal His 6 tract, expressed in and purified from E. coli, and immobilized on Ni 2ϩ -saturated NTAagarose beads. Solubilization and analysis by SDS-PAGE followed by staining with Coomassie Brilliant Blue demonstrated that equivalent amounts of each construct were affixed to the beads (Fig. 1B, lower panel). The beads were then incubated with [ 35 S]Frq1, prepared by coupled in vitro transcription and translation, washed, and subjected to SDS-PAGE, and the amount of radiolabeled Frq1 bound was analyzed by autoradiography. As observed previously using unlabeled Frq1 (29), Pik1-(10 -192) bound radiolabeled Frq1 avidly, and there was little or no nonspecific binding to empty beads (Fig. 1B, upper  panel). Strikingly, the 49-residue deletion lacking the heart of the LKU motif showed no detectable diminution in its ability to bind [ 35 S]Frq1. In marked contrast, even the shortest (29residue) C-terminal truncation completely ablated Frq1 binding. This result suggested that residues in the region 164 -192 of Pik1 are necessary for the high affinity binding of Frq1. Moreover, these data indicated that the LKU motif does not mediate the interaction between Pik1 and Frq1.
Residues 151-199 of Pik1 Are Sufficient for Binding to Frq1-To determine if the region of Pik1 from 164 to 192, or some sub-domain of it, was sufficient for Frq1 binding, we used an approach involving competition by synthetic peptides corresponding to different portions of Pik1 ( Fig. 2A). For these binding experiments, the arrangement was reversed. In this case, [ 35 S]Pik1- (10 -192), prepared by coupled in vitro transcription and translation, was incubated with either empty Ni 2ϩ -saturated NTA-agarose beads (as a control for nonspecific binding) or the same beads coated with purified Frq1-His 6 , preincubated in the absence or the presence of increasing concentrations (1 nM to 20 M) of the competing peptides. We found that the 36-residue peptide corresponding to residues 164 -199 was unable to block the binding of [ 35 S]Pik1- (10 -192) to immobilized Frq1 (Fig. 2B), and, not surprisingly, a shorter 26-residue peptide, residues 174 -199, was also ineffective (data not shown). Thus, although residues 164 -192 of Pik1 are necessary for Frq1 binding, they are not sufficient. Indeed, we found that a longer 49-residue peptide, corresponding to residues 151-199, was able to prevent the binding of [ 35 S]Pik1- (10 -192) to Frq1 in a dose-dependent manner (Fig. 2B). Hence, we conclude that the Frq1-binding site includes residues in Pik1 upstream of position 164.
The results from three independent trials measuring the ability of the 49-residue peptide (151-199) to compete for the binding of [ 35 S]Pik1- (10 -192) to bead-immobilized Frq1-His 6 were quantitated, after correcting for the minimal background binding of [ 35 S]Pik1- (10 -192) to empty beads at each peptide concentration (Fig. 2C). The resulting data were then normalized to the amount of radioactivity bound in the absence of peptide and fitted to the following equation: where K I is the apparent dissociation constant for peptide binding, Frq⅐Pik(ϩPep) is the amount of radiolabeled Pik1- (10 -192) bound to Frq1 at a given peptide concentration and Frq⅐Pik(ϪPep) is the amount of radiolabeled Pik1- (10 -192) bound to Frq1 in the absence of peptide and a, b, and c are arbitrary constants. Assuming a stoichiometry for binding of [ 35 S]Pik1- (10 -192) to immobilized Frq1-His 6 of 1:1, K I for binding of the 49-residue peptide (151-199) to Frq1 was 1.
Structural Characterization of Frq1 Target Complexes Using NMR-NMR spectroscopy was used to examine whether any conformational changes occurred in Frq1 upon binding to either the 49-residue synthetic peptide (151-199) or to a somewhat larger 83-residue fragment of Pik1, Pik1-(110 -192), prepared by expression in and purification from bacterial cells (see "Experimental Procedures"). Two-dimensional NMR experiments ( 1 H-15 N HSQC) were performed on samples of uniformly 15 N-labeled Ca 2ϩ -free and Ca 2ϩ -bound forms of Frq1, prepared as described previously (35), in the presence and absence of saturating amounts of either the 151-199 peptide or the Pik1-(110 -192) fragment. Initial attempts to prepare NMR samples of Frq1 complexed with the 151-199 peptide were unsuccessful, because simple addition of the peptide to a solution of Frq1 at a concentration (0.4 mM) sufficient for NMR analysis resulted in irreversible denaturation and aggregation. In contrast, when a dilute solution of Ca 2ϩ -bound Frq1 (20 M) was added slowly to an equal volume of a dilute solution of the Pik1 peptide or Pik1 fragment (20 M), a soluble complex was produced that could then be concentrated more than 20-fold to yield a stably soluble sample adequate for NMR studies. This approach was only successful in preparing complexes with Ca 2ϩ -bound Frq1. Mixtures of Ca 2ϩ -free Frq1 with the Pik1 peptide or Pik1 fragment were much less soluble and, hence, were not characterized further.
Two-dimensional NMR ( 1 H-15 N HSQC) spectra for Ca 2ϩbound Frq1 in the absence and presence of Pik1-(110 -192) are shown in Fig. 3. The spectrum of Ca 2ϩ -bound Frq1 alone (Fig.  3A) was analyzed previously (35). Free Ca 2ϩ -bound Frq1 contains fewer peaks than the expected number of amide groups in the protein (180 versus 220), apparently because some of the amide resonances have extremely weak intensity and therefore escape detection. Moreover, the wide range of NMR peak intensities suggested that Ca 2ϩ -bound Frq1 exists as a somewhat heterogeneous population of species. In agreement with this conclusion, dynamic light scattering measurements performed on the sample of free Ca 2ϩ -bound Frq1 used for the NMR analysis was indicative of significant polydispersity, suggest-  In marked contrast, the NMR spectrum of Ca 2ϩ -bound Frq1 in the presence of a saturating amount of Pik1-(110 -192) (Fig.  3B) looked quite different from that of free Ca 2ϩ -bound Frq1, suggesting that Pik1 binding induces conformational changes in Frq1. The NMR spectrum of Ca 2ϩ -bound Frq1 in the presence of the 151-199 peptide looked essentially identical to that of the Frq1⅐Pik1-(110 -192) complex (data not shown). The spectrum of each complex exhibits significantly sharper peaks compared with those in the spectrum of Ca 2ϩ -bound Frq1 alone, suggesting, first, that Frq1 is monomeric in each complex. Second, in the spectrum of the complex, some new peaks are observed that are not seen in the spectrum of free Ca 2ϩbound Frq1 (compare Fig. 3, A and B). The total number of observable peaks in the spectra of the complexes of Ca 2ϩ -bound Frq1 with either the 151-199 peptide or the Pik1-(110 -192) fragment are now very close to the expected number of amide resonances (218 versus 220). Third, the intensities of all peaks in the spectra of the complexes are much more uniform than those of free Ca 2ϩ -bound Frq1, suggesting that both complexes are stable and homogeneous. In agreement with this conclusion, dynamic light scattering measurements performed on the sample of Ca 2ϩ -bound Frq1 at a saturating concentration of Pik1-(110 -192) that was used for NMR analysis indicated a monodisperse scattering profile with a molecular mass of ϳ35 kDa, consistent with a monomer of Ca 2ϩ -saturated Frq1 The sample of Ca 2ϩ -bound Frq1 complexed with the 151-199 peptide that was used for NMR analysis also exhibited a monodisperse light scattering profile indicative of a homogeneous species. However, unexpectedly, despite the smaller size of the 49-residue peptide (5.6 kDa), the molecular mass of this complex was somewhat larger (ϳ36 kDa) than that observed for the complex of Ca 2ϩ -bound Frq1 with the Pik1-(110 -192) fragment, suggesting the species present was monomeric Ca 2ϩsaturated Frq1 bound to two molecules of the 151-199 peptide. In agreement with this interpretation, the changes in the NMR spectrum of Ca 2ϩ -bound Frq1 induced by the presence of the 151-199 peptide saturated upon addition of 2 molar equivalents (data not shown), consistent with formation of a high affinity peptide⅐Frq1 complex with a stoichiometry of 2:1. Moreover, the NMR spectrum of Ca 2ϩ -bound Frq1 in the presence of 1 molar equivalent of the 151-199 peptide (data not shown) resembled the composite that would be expected from equal parts of the spectra (Fig. 3, A and B) for free and fully complexed Ca 2ϩ -bound Frq1. Thus, it seems that in the presence of 1 molar equivalent of the 151-199 peptide, half of the Frq1 molecules are peptide-bound and half are unbound. Furthermore, these observations indicate that the population of peptide-bound Frq1 molecules exchanges with the unbound species only very slowly, such that the dissociation rate must be slower than the time scale of NMR chemical shift, as expected for a high affinity complex. On this basis, we estimate that the dissociation constant for binding of the 151-199 peptide to Ca 2ϩ -bound Frq1 in solution must be in the nanomolar range.
Pik1-Frq1 Interaction Monitored by Fluorescence Spectroscopy-The 190-residue Frq1 polypeptide contains just two tryptophan residues (Trp-30 and Trp-103). The fluorescence emission of Trp is very sensitive to its surrounding chemical environment and, hence, provides a well-documented method for probing structural changes in a protein (54). Hence, we used the change in the intrinsic Trp fluorescence of Frq1 as an independent means both to monitor the effects of and to quantitate the binding of Frq1 to the Pik1-derived fragment and the synthetic peptide (see increased the emission intensity by 30% and caused the emission maximum to shift very slightly toward the blue. These results indicate that, upon binding of the Pik1 sequences, conformational changes in Frq1 occur that cause one or both of its Trp residues to become more constrained and/or to enter a less solvent accessible and more non-polar environment. The emission intensity of Frq1 increased linearly with the addition of the Pik1-(110 -192) fragment in the range from 0 to 1 molar equivalents, and no further increase was observed when more than 1 equivalent was added. Thus, in agreement with the conclusions drawn from the NMR and light scattering analysis, the stoichiometry of the complex of Pik1-(110 -192) with Ca 2ϩbound Frq1 is 1:1. In contrast, the emission intensity of Frq1 increased linearly with the addition of the 151-199 peptide in the range from 0 to 2 molar equivalents, and no further increase of intensity was observed when more than 2 molar equivalents of peptide was added. Again, this result agrees with the conclusions drawn from the NMR and light scattering analysis, which indicated that the stoichiometry of the complex of the 151-199 peptide with Ca 2ϩ -bound Frq1 is 2:1.
Energetics of Pik1-Frq1 Interaction-As yet another independent means to examine the association of Frq1 with its apparent binding site in Pik1, we used isothermal titration calorimetry, which also permitted assessment of the energetics of the binding interaction between Ca 2ϩ -bound Frq1 and either the Pik1-(110 -192) fragment (Fig. 4, A and B) or the 151-199 peptide (Fig. 4, C and D). These calorimetric titrations were conducted at 25°C in 10 mM HEPES (pH 7.4). In both cases, the heat change was endothermic, indicating that the binding reaction is largely entropically driven, most consistent with desolvation of one or both partners and a hydrophobic interaction between them. After correction for the heats of dilution (see "Experimental Procedures") and normalization, the concentration dependence of the absorbed heats were plotted for the interaction of Ca 2ϩ -bound Frq1 with the Pik1-(110 -192) fragment (Fig. 5B) and with the 151-199 peptide (Fig. 5D). Once the ratio of Ca 2ϩ -bound Frq1 to the Pik1-(110 -192) fragment exceeded 1:1, the heat no longer changed, again consistent with a one-to-one complex. However, the binding of Frq1 to Pik1-(110 -192) was multiphasic and was fit to a two-site model, yielding dissociation constants (K D1 ϭ 62 nM and K D2 ϭ 200 nM) and positive (non-favorable) enthalpies of binding (⌬H 1 ϭ ϩ5.4 kcal/mol and ⌬H 2 ϭ ϩ30 kcal/mol) for the two sites, respectively. The multiphasic binding of Ca 2ϩ -bound Frq1 to the Pik1-(110 -192) fragment might be explained, in part, by the fact that we observed, using dynamic light scattering, that the starting solution of the Pik1-(110 -192) fragment alone displayed some polydispersity. This analysis indicated that more than 25% of the free Pik1-(110 -192) fragment existed in a pre-aggregated form, which might account for the highly endothermic and poorer affinity phase of binding we observed (K D2 ϭ 200 nM and ⌬H 2 ϭ ϩ30 kcal/mol). By contrast, the bulk of the population of the Pik1-(110 -192) molecules was monomeric and presumably accounts for the major, higher affinity phase of binding (K D1 ϭ 62 nM and ⌬H 1 ϭ ϩ5.4 kcal/mol). The binding of Ca 2ϩ -bound Frq1 to the 151-199 peptide appeared homogeneous and was fitted to a one-site model, yielding a K D ϭ 140 nM, and a ⌬H for binding ϭ ϩ9.5 kcal/mol. In these titrations, the heat no longer changed when the ratio of Ca 2ϩ -bound Frq1 to the 151-199 peptide reached 0.5, again consistent with a complex containing two peptides bound per Frq1 molecule.
Size Determination of Native Pik1⅐Frq1 Complexes by Size Exclusion Chromatography-Despite the finding that the Pik1-(110 -192) fragment formed a 1:1 complex with Ca 2ϩ -bound Frq1, the fact that the smaller 151-199 peptide could form 2:1 complexes with Ca 2ϩ -bound Frq1, raised the possibility, albeit remote, that in vivo Frq1 might serve to bridge two Pik1 molecules and thereby promote formation of enzyme dimers. To determine if Pik1 and Frq1 form stable complexes in cell extracts and to estimate their apparent molecular mass, analytical size exclusion chromatography was performed on extracts of yeast cells (strain YPH499) overproducing an epitope-tagged derivative of Pik1, which is otherwise quite an inabundant protein (18). For this purpose, the yeast cells carried a low copy number (CEN) plasmid expressing from the Gal-inducible GAL1 promoter full-length Pik1 containing an in-frame Nterminal c-Myc epitope. This construct is able to fully complement the inviability of a pik1⌬ mutant, even when cells are propagated under conditions (Glc as the carbon source) that repress the GAL1 promoter. 2 Expression of myc-Pik1 was induced by shifting the cells to Gal-containing medium for a brief period and lysates were prepared under non-denaturing conditions in a buffer lacking detergent. The proteins in the resulting lysates were resolved by size exclusion chromatography on a Superdex 200 column that had been pre-calibrated several times with a void volume (V o ) marker and protein standards of known molecular mass (see "Experimental Procedures"), which did not differ significantly in their elution behavior from run to run. The protein content of the fractions obtained from chromatography of the cell extracts were examined by subjecting samples to SDS-PAGE followed by immunoblotting with appropriate antibodies (mouse anti-c-Myc mAb 9E10 to detect myc-Pik1 and rabbit polyclonal antibodies to detect Frq1). Frq1 eluted in three peaks with K av values of 0.06 (Peak I, Fraction 44), 0.28 (Peak II, Fraction 60), and 0.61 (Peak III, Fraction 84) (see Fig. S2 in the Supplemental Material). When plotted on the calibration curve that was prepared from the K av values calculated from the experimentally determined elution volumes (V e ) of the standards, these three peaks correspond to molecular masses of 724 (Peak I), 162 (Peak II), and 18 kDa (Peak III), respectively. Pik1 eluted in two peaks, both of which were virtually superimposable with the two largest Frq1-containing peaks (I and II). Given the calculated molecular masses of Frq1 (22.1 kDa), Pik1 (119.9 kDa), and myc-Pik1 (122.5 kDa), our results are most consistent with the view that Peak III represents free monomeric Frq1, that Peak II represents a 1:1 complex of myc-Pik1 and Frq1 (calculated molecular mass of 144.6 kDa), and that Peak I represents the myc-Pik1⅐Frq1 complex associated with membrane fragments, because no detergent was included in the buffers and native N-myristoylated Frq1 tends to associate with membranes (35). The distribution and apparent stoichiometry of the myc-Pik1⅐Frq1 complexes observed in these experiments undoubtedly reflect the situation in cells in vivo for two reasons. First, the same elution profile was observed when extracts were prepared from untransformed YPH499 cells and the endogenous Pik1 was de-tected using polyclonal anti-Pik1 antisera (29,45), although the signal was weaker (data not shown). Second, the abundance of endogenous Frq1 exceeds that of endogenous Pik1 in all yeast strains examined to date 2 and, clearly, even when myc-Pik1 was overexpressed.
Site-directed Mutagenesis of the Frq1 Binding Region in Pik1-Comparison of Pik1 to the sequences of type III PtdIns 4-kinases from other organisms indicated that the region of Pik1 that is necessary and sufficient for Frq1 binding (residues 151-192) shows weak, but detectable, conservation. To determine whether any of the most conserved residues are critical for Frq1 binding, we generated derivatives of the Frq1-binding fragment, Pik1-(10 -192)-(His) 6 , carrying a variety of site-directed mutations. These mutants included two double mutants, Pik1 (10 -192; P181A,V183A) and Pik1 (10 -192; R188A,R189A), a triple mutant, Pik1(10 -192; L175A,P181A,V183A), and a quadruple mutant Pik1 (10 -192; E154A,N155A,V156A,-P158A). The resulting constructs were expressed in and purified from E. coli, immobilized on Ni 2ϩ -saturated NTA-agarose bead, and their ability to bind [ 35 S]Frq1 in vitro was examined, as described above (see Fig. 1). None of the four mutants proteins showed any significant reduction in their capacity to bind radiolabeled Frq1 under these conditions (data not shown). In two of the mutants, charged or polar side chains (Glu-154 and Asn-155, and Arg-188 and Arg-189, respectively) were replaced with the hydrophobic residue, Ala. Likewise, all of the other substitutions replaced more bulky hydrophobic side chains with the less bulky, but nevertheless non-polar, Ala residue. Thus, all of the mutations, although perturbing individual resides, did not dramatically change the overall hydrophobic character of this region. The fact that Frq1 binding was not affected by these alterations provides a further indication that the association of Frq1 with Pik1 is largely a hydrophobic interaction, fully consistent with the data from calorimetry (Fig. 4) and fluorescence spectroscopy (see Fig. S1 in the Supplemental Material). Moreover, preliminary NMR analysis of the NOE patterns for the complex of 15 N-labeled Pik1-(110 -192) with Ca 2ϩ -bound Frq1 indicates that a 13-residue segment in this region (Ala-157 to Ala-169) that contains no charged residues (and was not altered significantly in any of the site-directed mutants we generated) makes intimate contact with Frq1 (see "Discussion").
Deletion of the Frq1-binding Site Compromises Pik1 Function in Vivo-In contrast to the substitution mutations, truncations of Pik1-(10 -192) did greatly impair Frq1 binding (see Fig. 1). Therefore, to determine whether the region of Pik1 found to be necessary and sufficient for Frq1 binding in vitro is also required for association of Frq1 with Pik1 in vivo and plays a role in the physiological function of this enzyme, we generated a mutant allele, pik1 (⌬152-191), in which a small 2 T. Strahl, unpublished observations.  Table I) overexpressing either full-length myc-Pik1 (lanes 1-3), myc-Pik1(⌬152-191) (lanes 4 -6), or untagged Pik1(⌬152-191) (lanes 7 and 8) were grown and lysed as described under "Experimental Procedures." Samples of the extracts were subjected to immunoprecipitation with either rabbit polyclonal anti-Frq1 antiserum (lanes 3, 6, and 8) or, as a control, preimmune serum from the same rabbit (lanes 1 and 4), or with mouse ␣-c-Myc mAb 9E10 (lanes 2, 5, and 7). The resulting immune complexes were washed, solubilized, resolved by SDS-PAGE, and visualized by immunoblotting with the anti-c-Myc mAb.
Before testing its phenotype, we wanted to confirm that the pik1(⌬152-191) mutation does not adversely affect the catalytic activity of the enzyme per se but does compromise the ability of Frq1 to bind to Pik1. For this purpose, we prepared extracts from yeast cells expressing from the GAL promoter on CEN plasmids versions of either wild-type Pik1 or Pik1-(152-191) tagged at the N terminus with the c-Myc epitope, or, as a negative control, untagged Pik1- (152-191). These extracts were then subjected to immunoprecipitation with either anti-Frq1 antibodies or, as a control, with pre-immune serum from the same rabbit. The resulting immune complexes were resolved by SDS-PAGE, transferred to filters, and the amount of Pik1 that co-immunoprecipitated determined by immunoblotting with anti-Myc mAb. Samples of the same extracts were also directly immunoprecipitated with the anti-Myc mAb to determine the level of production of the tagged Pik1 proteins, and to examine their activity. First, the c-Myc epitope tag allowed for highly sensitive, specific, and background-free detection, because no signal was observed after immunoprecipitation of the extract expressing untagged Pik1(⌬152-191) with either the anti-Myc mAb or the anti-Frq1 antibodies (Fig. 5,  lanes 7 and 8). Second, as expected, wild-type myc-Pik1 was co-immunoprecipitated by the anti-Frq1 antibodies, but not by the preimmune serum (Fig. 5, lanes 1 versus 3). Moreover, the amount of myc-Pik1 co-immunoprecipitated with Frq1 was similar to the amount of myc-Pik1 that could be captured by direct immunoprecipitation with the anti-Myc mAb (Fig. 5,  lanes 2 versus 3). In marked contrast, even though copious amounts of myc-Pik1(⌬152-191) could be immunoprecipitated directly with the anti-Myc mAb (Fig. 5, lane 5), only a trace of myc-Pik1(⌬152-191) was co-immunoprecipitated by the anti-Frq1 antibodies (Fig. 5, lane 6), although this level was above that seen in the pre-immune serum control (Fig. 5, lane 4). Nonetheless, Pik1(⌬152-192) clearly interacted with Frq1 much less efficiently than wild-type Pik1. In contrast, when immune complexes obtained with anti-Myc mAb 9E10 were assayed for PtdIns 4-kinase catalytic activity, by methods that will be described in greater detail elsewhere, the specific activity of myc-Pik1(⌬152-191) was quite comparable to that for myc-Pik1. 2 To examine the consequences of defective Frq1 binding on the function of Pik1 in vivo, a heterozygous pik1⌬::LEU2/PIK1 diploid strain (YES10; Table I) was transformed with a CEN vector, pRS314-myc-pik1(⌬152-191), expressing Pik1(⌬152-191) under the control of the GAL1 promoter, or with the same vector expressing wild-type myc-PIK1, or the same empty vector (as a control). The transformants then were induced to sporulate, and the resulting tetrads were dissected on medium containing Gal as the carbon source and grown at 30°C. As expected, the diploids transformed with the empty vector only yielded tetrads in which two spores (PIK1) were viable and two spores (pik1⌬::LEU2) were inviable (see Fig. S3 in the Supplemental Material), indicating that the absence of Pik1 function is lethal, as shown before (19). In contrast, the diploids transformed with the vector expressing normal Pik1 yielded many tetrads in which three or all four spores were viable, indicating that the plasmid-borne copy of PIK1 was able to rescue the inviability of the pik1⌬::LEU2 spores. Under the same conditions, the diploids transformed with the vector expressing pik1(⌬152-191) also yielded tetrads in which three or four spores were viable. Thus, when overexpressed from the strong inducible GAL1 promoter, Pik1(⌬152-191), was able to complement the pik1⌬::LEU2 null mutation. Nevertheless, one way to determine if a gene product is not operating optimally is to test its ability to function under conditions of stress. Indeed, we found that pik1⌬::LEU2 spores expressing Pik1(⌬152-191) as the sole source of the enzyme were compromised in their growth at 36°C and unable to grow at all at 37°C, whereas pik1⌬::LEU2 spores expressing normal Pik1 from the same vector grew robustly at both temperatures (data not shown). Thus, even when highly overexpressed, Pik1(⌬152-191) is not fully functional under the stressful condition of elevated temperature, presumably because its efficient association with Frq1 is required for its optimal function.
We have demonstrated before that, when PIK1 is highly overexpressed, it is able to support the growth of frq1⌬ cells, which are otherwise inviable when Pik1 is present at its normal level (29). In other words, Frq1 becomes completely dispensable when Pik1 is highly abundant. Therefore, we tested whether Pik1(⌬152-191) was still able to rescue the inviability of pik1⌬::LEU2 cells when it was expressed at a lower level rather than at the high level that results from expression from the GAL1 promoter. For this purpose, we constructed a CEN plasmid, pRS314-pik1(⌬152-191), that expressed Pik1(⌬152-191) from the native PIK1 promoter and introduced it into the pik1⌬::LEU2/PIK1 heterozygous diploid. Even at this lower level of expression, diploids expressing Pik1(⌬152-191) yielded tetrads with three and four viable spores at essentially the same frequency as diploids expressing normal Pik1 (Fig. 6A). However, as observed before, pik1⌬::LEU2 cells expressing Pik1(⌬152-191) were unable to grow above 35°C (Fig. 6B). Thus, under the moderate stress of elevated temperature, efficient association of Frq1 with Pik1 becomes essential for the function of the enzyme, regardless of its level of expression.
Cys-15 Is Neither S-Palmitoylated in Vivo nor Required for Frq1 Function-One apparent role for the association of Frq1 with Pik1 is to promote its association with membranes, and we have shown previously that the N-terminal myristoyl group of Frq1 is important, but not essential, for this function (29,35). First, we found that, like wild-type Frq1, a mutant, Frq1(G2A), that cannot be and is not myristoylated is capable of rescuing the inviability of frq1⌬ cells; however, unlike overexpression of wild-type Frq1, overexpression of Frq1(G2A) is unable to rescue the temperature-sensitive lethality of pik1-11 ts cells (29). Second, we found that, like normal Frq1, Frq1(G2A) still associates with membranes, but does so much less efficiently than normal Frq1 in either the absence or presence of Ca 2ϩ (35). One explanation for both of these observations is that Frq1 might carry a second lipophilic modification that partially compensates for the absence of the N-myristoyl group. Moreover, there are precedents for S-palmitoylation of Cys residues situated near the N terminus of other small Ca 2ϩ - MATa leu2 trp1 ura3-52 prb1-1122 pep4-3 prc1-407 gal2 (59) binding regulatory proteins (38,39). Indeed, Frq1 has only two Cys residues, and although one (Cys-38) appears to be buried, based on our NMR-derived structural model (35), the other (Cys-15) is located near the N terminus. We took two independent approaches to address whether S-palmitoylation has any role in the function of Frq1. First, we generated a Frq1(C15A) single mutant and found that its properties were indistinguishable from those of wildtype Frq1. 3 Next, we generated a Frq1(G2A,C15A) double mutant and examined its ability to rescue the temperature-sensitive lethality of frq1-1 ts cells (strain YKBH4, Table I) when expressed from the native FRQ1 promoter in a multicopy vector. The empty vector was unable to permit cell growth at the non-permissive temperature, as expected, whereas both normal FRQ1 and frq1(G2A,C15A) expressed from the same vector suppressed the temperature-sensitive phenotype of the frq1-1 ts cells (Fig. 7A). Likewise, when expressed from the same vector, expression of both normal Frq1 and Frq1(G2A,C15A) was able to restore viability to otherwise inviable spores carrying a frq1⌬ null mutation, whereas the empty vector did not (see Fig.  S4A in the Supplemental Material). Even when expressed at a lower level from the native FRQ1 promoter on a CEN vector, Frq1(G2A,C15A) was able to rescue the inviability of frq1⌬ spores at essentially the same frequency as wild-type Frq1 expressed from the same plasmid, whereas the empty vector did not (Fig. 7B). Thus, as judged by these criteria, Frq1 that is unmyristoylated and non-palmitoylatable (at Cys-15) showed no impairment in its physiological function in vivo.
Second, to determine whether Frq1 is S-palmitoylated at either Cys-15 or Cys-38 in vivo, we tagged the 3Ј-end of the FRQ1 gene with a sequence encoding an in-frame His 6 tag, expressed the protein product in yeast cells, purified the protein to apparent homogeneity by chromatography on Ni 2ϩ -saturated NTA-agarose, and performed electrospray ionization mass spectrometry on the purified protein. The calculated molecular mass for Frq1-His 6 with an N-terminal myristoyl group (but lacking any palmitoyl moiety) is 23,839.90 Da. After deconvolution of the spectrum of mass species observed (see Fig.  S4B in the Supplemental Material), we found a single homogenous component with a molecular mass of 23,838.18 Da. We conclude, therefore, that N-myristoylation at Gly-2 is the only lipophilic modification present in native Frq1 and, correspondingly, that neither Cys-15 nor Cys-38 is S-palmitoylated. Thus, S-palmitoylation is not required for the function of Frq1, as also confirmed by our mutagenesis and phenotypic tests. DISCUSSION We showed previously using primarily genetic means that the sole essential target of the small Ca 2ϩ -binding protein, Frq1, in the yeast S. cerevisiae is the PtdIns 4-kinase, Pik1 (29). At the biochemical level, Frq1 is present in a stoichiometric amount in preparations of Pik1 (18) that were purified more than 25,000-fold by ammonium sulfate fractionation followed by chromatography on five different columns in buffers lacking Ca 2ϩ and containing chelator (EDTA). Thus, Frq1 is constitutively bound to Pik1, even in the absence of Ca 2ϩ , and should be considered a non-catalytic subunit of the enzyme. Indeed, we demonstrated, using an assay in which the substrate (PtdIns) was displayed in detergent micelles, rather than in authentic biological membranes, that the presence of Frq1 is required for optimal activity of the enzyme (29). As judged by cell fractionation experiments and other methods, Frq1 associates with the membrane-containing particulate fraction in a manner that depends on both its N-terminal myristoyl group and Ca 2ϩinduced changes in the protein (35). Thus, in vivo, Frq1 associated with Pik1 presumably also assists in targeting Pik1 to membranes in a Ca 2ϩ -dependent manner. In our prior work, based on both genetic and biochemical approaches, we showed that the region of Pik1 responsible for Frq1 binding seemed to 3 I. G. Huttner, unpublished results.

FIG. 6. Efficient association of Frq1
with Pik1 is required for growth at elevated temperature. A, a complementation test was used to assess the function of overexpressed Pik1(⌬152-191). Heterozygous pik1⌬::LEU2/PIK1 diploid strain YES10 (Table I) was transformed with either the empty TRP1-marked CEN vector (left panel), or the same vector expressing wild-type Pik1 from the native PIK1 promoter (middle panel), or the same vector expressing Pik1(⌬152-191) from the native PIK1 promoter (right panel) and subjected to conditions that induce sporulation, and samples of the resulting tetrads (six to ten shown) were dissected. Viability of the four spores (A-D) was assessed at 30°C on medium containing Gal as the carbon source. B, a representative Leuϩ Trpϩ spore expressing myc-Pik1(⌬152-191) (left) and a representative Leuϩ Trpϩ spore expressing normal Pik1 (right), obtained as in A, were tested for growth on SCGal-Leu-Trp at the indicated temperatures.
reside within the first ϳ200 residues of the protein. In this present study, we sought to further define the Frq1-binding site and to characterize the nature of this interaction in greater detail.
We generated deletions in an His 6 -tagged 183-residue fragment of Pik1, Pik1-(10 -192)-(His) 6 , that, when immobilized on Ni 2ϩ -saturated NTA-agarose beads, is able to capture [ 35 S]Frq1 from solution with high affinity. Using this in vitro binding method, we found that the region from 164 to 192 was essential for the observed interaction, whereas the segment from residue 31 to 79 was totally dispensable. However, using competition assays with synthetic peptides, we found that the 164 -192 region of Pik1 was not sufficient for Frq1 binding. A peptide corresponding to residues 164 -199 was not able to prevent the binding of [ 35 S]Pik1- (10 -192) to bead-immobilized Frq1-His 6 , whereas a longer peptide corresponding to residues 151-199 did compete. Thus, residues N-terminal to the 164 -192 region are also important for high affinity binding of Frq1 to Pik1. It also appears that the segment corresponding to residues 80 -109 is dispensable for the interaction of Frq1 with Pik1, because a smaller 83-residue fragment of Pik1, Pik1-(110 -192), bound to Frq1 with high affinity and a 1:1 stoichiometry, as judged by NMR, intrinsic Trp fluorescence, titration calorimetry, and light scattering.
Revealingly, the 151-199 peptide also bound to Frq1 avidly, however, as judged by the same criteria (NMR, intrinsic Trp fluorescence, titration calorimetry, and light scattering), the stoichiometry of peptide:Frq1 binding was 2:1, unlike that of the Pik1-(110 -192) fragment. Moreover, as judged by calorimetric measurements in solution, the association of Frq1 with the 83-residue Pik1-(110 -192) fragment was tighter (apparent K d ϭ 62 nM) than its association with the 49-residue 151-199 peptide (apparent K d ϭ 140 nM). When measured by competition for the binding of [ 35 S]Pik1- (10 -192) to bead-bound Frq1-His 6 , the K i for the 151-199 peptide was 1 M. However, in this assay format and given that a tracer level of the radioactive probe was used, competition was only observed when the amount of peptide added approached that of the large excess of  Table I) carrying either an empty URA3marked multicopy vector (YEp352) (upper sector), or the same vector expressing wild-type FRQ1 from the native FRQ1 promoter (lower right sector), or the same vector expressing a non-myristoylated and unpalmitoylatable mutant, frq1(G2A C15A) (lower left sector), were incubated on SCGlc-Ura plates at the indicated temperatures. B, heterozygous frq1⌬::HIS3/ FRQ1 diploid strain YKBH1 (Table I) was transformed with either an empty URA3marked low copy number (CEN) vector (pRS316), or the same vector expressing normal FRQ1 from the native FRQ1 promoter, or the same vector expressing frq1(G2A C15A) from the native FRQ1 promoter and subjected to conditions that induce sporulation, and samples of the resulting tetrads (five to seven shown) were dissected. Viability of the four spores (A-D) was assessed at 30°C on medium containing Glc as the carbon source.
immobilized Frq1 present. Nevertheless, the 151-199 peptide may be missing residues upstream of 151, which are present in Pik1- (110 -192), that also contribute to maximal binding affinity.
The intriguing 2:1 stoichiometry observed for binding of the 151-199 peptide suggests that the Pik1-binding site in Frq1 may be, in some sense, bipartite. Indeed, preliminary HSQC spectra for [ 15 N]Pik1-(110 -192) bound to Ca 2ϩ -saturated Frq1 indicate that contact is made with two short segments in the Pik1 fragment of 12 residues (Phe-125 to Gln-136) and 13 residues (Ala-157 to Ala-169), respectively, both of which assume an ␣-helical conformation in the complex. 4 The intervening 20 residues (Thr-137 to Val-156) appear to be unstructured 4 and presumably form a loop that does not interact with Frq1. The 13-residue segment is comparable in length to the short helices and other sequence elements (e.g. IQ motif) recognized by the well-characterized Ca 2ϩ -binding regulatory protein, calmodulin (55). Furthermore, this 13-residue segment (-APALVLSSMIMSA-) is comprised exclusively of non-polar and uncharged residues, in agreement with the evidence we obtained from titration calorimetry and intrinsic Trp fluorescence that the association between Pik1-(110 -192) is primarily a hydrophobic interaction. An interaction between another myristoylated Ca 2ϩ -binding regulatory protein and a largely hydrophobic element in its target has recently been described. The "CIB" (calcium-and integrin-binding) protein binds tightly to a 15-residue ␣-helical sequence (-LVLAMWKVGFFKRNR-) in the cytoplasmic tail of the integrin ␣IIb chain (56). This sequence bears some resemblance to the 13-residue Frq1-binding segment of Pik1.
We derived a model for the three-dimensional structure of Ca 2ϩ -bound yeast Frq1 based on NMR analysis in solution (35). A very similar structure for the Ca 2ϩ -bound form of its human ortholog, NCS-1, was determined by x-ray analysis and refined to 1.9-Å resolution (36). Both structures revealed that frequenin has two tightly folded domains that pack against each other to form an accessible crevice lined primarily with hydrophobic side chains. This hydrophobic cleft would seem the mostly likely site for binding of the 13-residue hydrophobic segment in the Frq1-binding region of Pik1. By contrast, the 12-residue segment in the Pik1-(110 -192) fragment (-FQVARRVLNNLQ-) that also appears to associate with Frq1 contains both polar and charged residues, the most striking feature of which is a tandem pair of Arg residues (Arg-129 and Arg-130). We note that, although the 151-199 peptide lacks this basic hydrophilic sequence found in Pik1-(110 -192), there is a 12-residue segment at the C-terminal end of the 151-199 peptide that has some similarity (-ESQGRRQKAFVF-), including a tandem pair of Arg residues (Arg-188 and Arg-189). Thus, it is possible that the 151-199 peptide binds to Frq1 with a 2:1 stoichiometry because one copy of the peptide uses the 13-residue hydrophobic motif (Ala-157 to Ala-169) near its N-terminal to correctly occupy the corresponding hydrophobic pocket in Frq1, and another copy of the peptide uses the basic segment near its C-terminal end to artifactually occupy the site in Frq1 that would normally recognize the 12-residue hydrophilic motif (Phe-125 to Gln-136). Indeed, our site-directed mutagenesis experiments indicated that the Arg-188/Arg-189 pair is not required for the binding of Frq1 to Pik1- (10 -192), consistent with the view that normally it is the upstream Arg-129/Arg-130 pair that fulfills this function.
The fact that a single Frq1 was able to bind two 151-199 peptides raised the possibility that one role of Frq1 in vivo might be to promote dimerization of Pik1. However, the fact that the Pik1-(110 -192) fragment bound to Pik1 with a 1:1 stoichiometry made such a dimerization scenario unlikely. Nonetheless, there is precedent for protein dimerization being induced by the binding of small, EF-hand type, Ca 2ϩ -binding proteins, most notably the oligomerization of the SK class of potassium channels promoted by calmodulin binding (57). Hence, we examined the nature of the native Frq1⅐Pik1 complexes found in yeast cell extracts by size-exclusion chromatography. Consistent with the binding of Pik1-(110 -192) to Frq1 in vitro, we found that the bulk of the Pik1⅐Frq1 complexes in yeast cell extracts (Peak II) had an apparent molecular mass most consistent with a 1:1 complex. However, we did observe a fraction (Peak I) that contained both Pik1 and Frq1 that had a much larger apparent size. Because the extracts were prepared in the absence of detergent, and Frq1 has a propensity to interact with membranes, this fraction most likely represents the 1:1 Pik1⅐Frq1 complex associated with small vesicles or membrane fragments. On the other hand, we cannot rule out the possibility that this fraction indeed represents a higher order Pik1⅐Frq1 oligomer or a novel complex of Pik1⅐Frq1 that contains other tightly associated polypeptides.
We demonstrated previously that Frq1 lacking its N-terminal myristoyl group retains a residual capacity to associate with membranes in a Ca 2ϩ -dependent manner (35). This behavior could be due to greater exposure of previously buried hydrophobic side chains upon Ca 2ϩ -induced conformational change. Alternatively, however, it was possible that some other lipophilic substituent is also present in Frq1 that becomes more solvent exposed when Ca 2ϩ binds to the protein. In other small, EF-hand type, Ca 2ϩ -binding proteins, such as a flagellar regulator in Trypanosoma cruzi (38) and the novel mammalian potassium channel regulator, KChIP2 (39), Cys residues located near the N terminus are S-palmitoylated and are important for the ability of those proteins to associate with membranes. However, using mass spectrometry we found that Frq1 purified from yeast is fully N-myristoylated but not palmitoylated on either of its two Cys residues (Cys-15 and Cys-38). Moreover, site-directed mutagenesis of the most N-terminal Cys (Cys-15) did not compromise the function of Frq1 in vivo, nor did it exacerbate the effects of a mutation (G2A) that prevents N-myristoylation of Frq1. Therefore, S-palmitoylation is not involved in the physiological function of Frq1.
Taken together, our data support the conclusion that a single molecule of Frq1 docks onto a single molecule of Pik1 and does so by binding to a site that includes as its core a 13-residue hydrophobic sequence (Ala-157 to Ala-169). As expected if this region is critical for high affinity binding of Frq1 to Pik1, we found that a deletion mutation, pik1 (⌬152-191), which removed the 13-residue hydrophobic motif, produces a protein that, in contrast to normal Pik1, co-immunoprecipitates very inefficiently with Frq1, yet retains catalytic activity. The lack of efficient Frq1 binding to Pik1 appears to compromise the function of the enzyme because, again in contrast to wild-type Pik1, Pik1(⌬152-191) was unable to support the growth of pik1⌬ yeast cells at elevated temperature. We do not know if the binding modality we have defined for the Pik1-Frq1 interaction is conserved in the interaction of metazoan frequenins with PtdIns 4-kinase-␤ and/or any other targets. However, in collaborative studies, which will be described in greater detail elsewhere, we have delineated the site in yeast Pik1 bound by human NCS-1 using in vitro pull-down assays of the sort presented here and using the two-hybrid method in vivo. As observed for the interaction of Frq1 with Pik1, the most essential sequences to support the interaction of human NCS-1 with Pik1 fall between residues 145 and 172, which includes the 13-residue hydrophobic sequence (Ala-157 to Ala-169). 5 Furthermore, as also observed for Frq1 association with Pik1, residues in the region 100 -144, which includes the 11-residue hydrophilic element (Phe-125 to Gln-136), contribute to the strength of the interaction between human NCS-1 and Pik1. 5