Conservation of regulatory function in calcium-binding proteins: human frequenin (neuronal calcium sensor-1) associates productively with yeast phosphatidylinositol 4-kinase isoform, Pik1.

Frequenin, also known as neuronal calcium sensor-1 (NCS-1), is an N-myristoylated Ca2+-binding protein that has been conserved in both sequence and three-dimensional fold during evolution. We demonstrate using both genetic and biochemical approaches that the observed structural conservation between Saccharomyces cerevisiae frequenin (Frq1) and human NCS-1 is also reflected at the functional level. In yeast, the sole essential target of Frq1 is the phosphatidylinositol 4-kinase isoform, Pik1; both FRQ1 and PIK1 are indispensable for cell viability. Expression of human NCS-1 in yeast, but not a close relative (human KChIP2), rescues the inviability of frq1 cells. Furthermore, in vitro, Frq1 and NCS-1 (either N-myristoylated or unmyristoylated) compete for binding to a small 28-residue motif near the N terminus of Pik1. Site-directed mutagenesis indicates that the binding determinant in Pik1 is a hydrophobic alpha-helix and that frequenins bind to one side of this alpha-helix. We propose, therefore, that the function of NCS-1 in mammals may closely resemble that of Frq1 in S. cerevisiae and, hence, that frequenins in general may serve as regulators of certain isoforms of phosphatidylinositol 4-kinase.

Frequenins are a distinct subgroup in a larger class of related N-myristoylated proteins within the superfamily of small EF-hand-containing calcium-binding regulatory proteins, whose prototype is calmodulin (for a recent review, see Ref. 12). Other subgroups within this class are recoverins, visinins, neurocalcins, hippocalcins, guanylyl cyclase-activating proteins (GCAPs), 1 and potassium channel-interacting proteins (KChIPs). A distinguishing characteristic of these proteins, aside from modification by N-terminal myristoylation, is that all possess four recognizable EF-hand motifs, yet the first (and, sometimes, the fourth) is degenerate and unable to bind Ca 2ϩ .
Frequenin has been highly conserved throughout evolution. There is 100% sequence identity among vertebrate frequenins, close to 75% identity between vertebrate and either of two invertebrate frequenins (Caenorhabditis elegans or Drosophila melanogaster), and nearly 60% identity between any of these frequenins and their ortholog (Frq1) in budding yeast (Saccharomyces cerevisiae) (1). A crystal structure of the unmyristoylated Ca 2ϩ -bound form of a mammalian (human) frequenin, which is also termed neuronal calcium sensor-1 (NCS-1), has been solved to 1.9-Å resolution (2). A model for the threedimensional structure of N-myristoylated Ca 2ϩ -bound yeast Frq1, based on analysis by NMR, is nearly superimposable (3). Three calcium ions are found bound to EF2, EF3, and EF4 in the structures of both yeast Frq1 (3) and human NCS-1 (2). This stoichiometry is consistent with titration experiments in which calcium binding to both mammalian NCS-1 (4) and S. cerevisiae Frq1 (3) was shown to be high affinity, cooperative and saturated at three Ca 2ϩ per molecule of protein.
In the structures of both human NCS-1 and yeast Frq1, the four EF-hands come in two pairs positioned in a tandem linear array on one side of the molecule. This arrangement is quite different from that found in calmodulin and troponin C, where each pair of EF-hands is situated at either end of a dumbbellshaped molecule (5). The overall topology of two pairs of EFhands in a tandem linear array is also observed in the structures of other members of the N-myristoylated, small Ca 2ϩbinding proteins, namely neurocalcin (6), recoverin (7,8), and GCAP (9). Despite this superficial similarity, both yeast Frq1 (3) and human NCS-1 (2) possess a wide hydrophobic crevice at their surface situated between the two EF-hand-containing lobes, as well as solvent-exposed N and C termini. These latter features are lacking in recoverin, neurocalcin, and GCAP.
The function of frequenin has attracted much attention ever since it was implicated in neurotransmitter release at neuromuscular junctions. In fact, the name of the molecule derives from the fact that mutants of D. melanogaster that overexpress this protein display a marked facilitation of neurotransmitter release in third instar larvae, which depends dramatically on the frequency of stimulation (10). On this basis, it was suggested that frequenin might serve as a calcium sensor to modulate synaptic activity and secretion. Likewise, in a tissue culture cell model of synaptic transmission, overexpression of rat NCS-1 in live PC12 cells potentiated release of growth hormone-containing secretory vesicles in response to agonists such as ATP (11,12). Indirect immunofluorescence indicates that NCS-1 is localized at the trans-Golgi network of animal cells, consistent with a role for frequenin in vesicle trafficking and exocytosis (2). Several lines of evidence suggest, however, that frequenin-mediated modulation of exocytosis may be exerted at the level of signal transduction, rather than through direct effects on the machinery for exocytosis. For instance, Ca 2ϩ -induced exocytosis in permeabilized PC12 cells is not potentiated by overexpression of NCS-1 (13). Similarly, NCS-1 is apparently not involved in modulation of depolarizationevoked exocytosis in PC12 cells (13). By contrast, like the enhancement of agonist-evoked exocytosis in PC12 cells by NCS-1 overexpression, agonist-induced phosphoinositide turnover is also potentiated when NCS-1 is overexpressed (13,14). These data suggest that NCS-1 might modulate some aspect of phosphoinositide metabolism in animal cells, thereby promoting, for example, agonist-activated phospholipase C-dependent signaling. In agreement with this hypothesis is the observation that NCS-1 and phosphatidylinositol (PtdIns) 4-kinase-␤ colocalize in animal cells, as judged by immunocytochemical approaches (2,14).
Compelling evidence for a role of frequenin in phosphoinositide metabolism was first found by a combination of genetic and biochemical studies in the yeast S. cerevisiae (1). It was shown that the yeast ortholog of frequenin, Frq1, functions as a Ca 2ϩsensing subunit of the yeast PtdIns 4-kinase ␤ isoform, Pik1. Pik1 is a key enzyme in the control of vesicular trafficking in the late secretory pathway (15)(16)(17). Absence of Pik1 in S. cerevisiae cells (a pik1⌬ null mutation) is lethal (18), and pik1-ts mutants exhibit severe protein trafficking defects and accumulate morphologically aberrant Golgi membranes at the restrictive temperature (15)(16)(17). Likewise, frq1 mutants are inviable, but overexpression of PIK1 suppresses the lethality of frq1 cells, indicating that Pik1 is the sole essential target for Frq1 in yeast cells (1). Yeast Frq1 co-purifies stoichiometrically with Pik1 and is required for optimal activity of the enzyme and may also assist in membrane recruitment of Pik1 (1,3,19). Frq1 and Pik1 form a 1:1 complex, and the binding site for Frq1 has been mapped to residues 125-169 within the N terminus of Pik1 (20). The binding of Frq1 to Pik1 appears to be mediated primarily by hydrophobic interactions in a Ca 2ϩ -independent manner, as judged by a variety of different biochemical studies (1,20).
Given that the primary and tertiary structures of frequenins have been highly conserved from yeast to man, we sought to investigate in this study how closely the function of human NCS-1 resembles that of yeast Frq1. In a first approach, we examined whether human NCS-1 is able to functionally substitute for the absence of Frq1 in yeast and, if so, whether Ca 2ϩ binding is required for the ability of human NCS-1 to rescue the inviability of Frq1-deficient yeast cells. Second, we used biochemical approaches, including the two-hybrid method in vivo and pull-down assays and monitoring of intrinsic tryptophan fluorescence in vitro, to assess whether human NCS-1 can interact with yeast Pik1. If so, we sought to determine whether NCS-1 associates at the same site on Pik1 as Frq1, then to delineate the sequence determinants responsible for this binding using alanine-and lysine-scanning mutagenesis, and finally to measure the strength and stoichiometry of this binding.
NCS-1 and PIK1 Cloning, Expression, and Purification-Expression and purification of NCS-1 have been described previously (2). DNA encoding PIK1 was amplified by PCR using yeast genomic DNA as the template with primer sequences, derived from the data base entry for PIK1 (GenBank TM /EBI accession no. NC_001146), designed to permit subcloning into pET22b (Novagen) that was linearized with NdeI and SacI, generating the bacterial expression vector pET-PIK1. pET-PIK1 was introduced into E. coli strain BL21(DE3)RIL (Stratagene) by DNAmediated transformation. The resulting cells were grown in Luria-Bertani medium containing ampicillin (100 g/ml) and chloramphenicol (50 g/ml) at 37°C until the culture reached an A 600 nm of 0.7. Next, Pik1-His 6 expression was induced by addition of isopropyl-1-thio-␤-Dgalactopyranoside to a final concentration of 0.5 mM. Induced cells were harvested by centrifugation, resuspended in 20 ml of lysis buffer (20 mM Tris/HCl, pH 7.9, 300 mM NaCl) per 500 ml of bacterial culture and subsequently treated with lysozyme (final concentration 100 g/ml) for 10 min on ice. Dithiothreitol (DTT) and N-laurylsarcosine were added to final concentrations of 5 mM and 1.5% (v/v), respectively. After sonication (1 min, level 4, 50% duty cycle), the lysate was cleared of cellular debris by centrifugation (10,000 ϫ g, 10 min, 4°C). The resulting supernatant solution was adjusted to a final concentration of 3% Triton X-100, filtered (0.45-m pore size) and loaded onto a 1-ml Ni 2ϩ -saturated nitrilotriacetic acid (NTA)-agarose column (Qiagen). After washing with buffer A (20 mM Tris/HCl, pH 7.9, 300 mM NaCl, 20 mM imidazole), bound protein was eluted with buffer B (20 mM Tris/HCl, pH 7.9, 300 mM NaCl, 250 mM imidazole). Pik1-His 6 -containing fractions were dialyzed against 20 mM Tris/HCl, pH 7.9 and concentrated with the help of centrifugal filter devices (Millipore) until a final concentration of ϳ5 mg/ml had been reached.
In Vitro Pull-down Assays-NCS-1 (0.7 g), expressed in and purified from bacteria (2), was incubated at 4°C either alone or with 7 g of the different Pik1-His 6 derivatives in a final volume of 0.5 ml of YLB (50 mM Tris/HCl, pH 7.2, 100 mM NaCl, 1 M CaCl 2 , 1 mM DTT) on a roller drum for 1 h, and then 20 l of a 50% slurry of Ni 2ϩ -saturated NTAagarose beads (Qiagen) that had been preblocked in YLB containing 1% (w/v) bovine serum albumin were added. After an additional 1 h of incubation at 4°C, the beads were washed three times (1 ml each) in wash buffer (20 mM Tris/HCl, pH 7.9, 300 mM NaCl, 10 mM ␤-mercaptoethanol, 1 M CaCl 2 , 50 mM imidazole). Subsequently, bound proteins were eluted from the beads in 30 l of sodium dodecyl sulfate-polyacrylamide gel (SDS-PAGE) sample buffer (28), resolved by electrophoresis in 15% SDS-PAGE gels, and analyzed by immunoblotting using appropriate antisera or staining with Coomassie Brilliant Blue.
Yeast Strains, Growth Conditions, and Media-Routine growth and manipulation of S. cerevisiae strains were performed as described previously (29). Conventional methods for DNA-mediated transformation of yeast cells were used (30). Strain YKBH1 has been described elsewhere (1). All yeast strains used in this study are listed in Table I.
Construction of Yeast Strains-Heterozygous diploid strain YKBH1 (frq1::HIS3/FRQ1) was transformed with 2-m DNA-based (high copy) plasmids carrying either wild-type or mutant versions of human NCS-1 or KChIP2 under the control of the strong inducible GAL1 promoter. Transformants were subjected to conditions appropriate to induce sporulation (1% potassium acetate, 0.022% (w/v) raffinose) and incubated for 72 h at 26°C. Subsequently, ascospores were dissected with the help of a micromanipulator on YPGal plates and grown at 26°C for 96 h.
Preparation of Yeast Cell Extracts and Immunoblot Analysis-Yeast cells were grown at 30°C in standard rich (YP) medium, containing 2% galactose as carbon source (29). Preparation and preclearing of cellular lysates were carried out essentially as described (20). Protein concentration in each clarified extract was determined using a commercial kit (Bio-Rad) based on the dye-binding method of Bradford (31). Equal amounts of protein were resolved in 12.5% SDS-PAGE gels, transferred onto nitrocellulose membrane, and wild-type and mutant versions of NCS-1 were detected by immunoblotting using affinity-purified polyclonal NCS-1 antiserum (dilution 1:500), followed by horseradish peroxidase-conjugated goat anti-rabbit IgG conjugate (Zymed) and a commercial chemiluminescence kit (Western Lightning, PerkinElmer Life Sciences).
Intrinsic Tryptophan Fluorescence-Fluorescence spectra were recorded at room temperature using a PerkinElmer 2550B spectrofluorimeter at an excitation wavelength of 280 nm. Intrinsic fluorescence of NCS-1 (75 nM) was monitored over an emission wavelength range of 310 to 400 nm. Fluorometric titrations were done in 50 mM HEPES, pH 7.5, 0.1 M KCl, 1 mM CaCl 2 , and 1 mM DTT, using a 5 ϫ 10-mm black cuvette. Fluorescence spectra were recorded before and after the addition of 2.5-40 l of Pik1(120 -192) (final concentration 20 -640 nM). In parallel, tryptophan emission spectra of comparable fluorescence intensity were collected at 340 -360 nm to correct for dilution and unspecific effects of Pik1(120 -192) absorbance. The Pik1 peptide itself does not contain any tryptophan residues and, therefore, should not contribute any fluorescence to affect these measurements. Binding data were calculated from the observed changes in fluorescence intensity accompanying the addition of Pik1(120 -192), using methods described previously (32). Data analysis and curve fitting were done with the help of KaleidaGraph software (Synergy Software). For K D measurements, titrations were performed at protein concentrations below the K D value. Values were determined assuming that all proteins present were in their native and functional state and not corrected for the possible presence of inactive species.

Inviability of S. cerevisiae frq1⌬ Cells Is Rescued by Human
NCS-1-Frequenins seem to be conserved at the level of both primary and tertiary structure from yeast to humans. Frq1, the apparent S. cerevisiae ortholog of NCS-1, binds to and is essential for the function of Pik1, a PtdIns 4-kinase, and this role is the only indispensable cellular function of Frq1 (1). Whether NCS-1 directly interacts with a PtdIns 4-kinase in vertebrate cells, and whether this association is physiologically relevant, has not been demonstrated unequivocally, although considerable circumstantial evidence has been obtained in support of this conclusion (14,33,34). As one approach to address this question, we examined whether NCS-1 is able to functionally substitute for Frq1 in S. cerevisiae cells. To this end, a heterozygous diploid strain, YKBH1 (frq1::HIS3/FRQ1) (Table I), was transformed with a LEU2-marked 2-m-based (high copy) vector expressing human NCS-1 cDNA under control of the inducible GAL1 promoter. The resulting transformants were sporulated, and ascospores were dissected on YP plates containing galactose to induce NCS-1 expression and then incubated at a temperature (30°C) normally optimal for yeast cell growth. The majority of dissected tetrads yielded four viable spores (Fig. 1A, left panel), and His ϩ Leu ϩ spores were obtained at the expected frequency. Thus, NCS-1 is able to functionally replace Frq1 in yeast cells lacking any endogenous Frq1.
In marked contrast, expression from the same vector of KChIP2 cDNA, which encodes a related, small, Ca 2ϩ -binding protein (35), was unable to rescue the inviability of the frq1 spores, as indicated by the recovery of only two viable spores in every tetrad examined (and all the viable spores recovered were His Ϫ , indicating that they must carry the wild-type FRQ1 allele) (Fig. 1A, right panel). As demonstrated by immunoblotting of extracts of the initial diploid strain (YKBH1) carrying the YEp351GAL-KChIP2 plasmid with polyclonal anti-KChIP2 antibodies, the protein was made under the conditions used ( Fig. 1B, lower panel). Hence, the inability of KChIP2 to complement the lethality of frq1 cells was not the result of its lack of expression in yeast cells To investigate the roles of N-myristoylation and Ca 2ϩ binding for NCS-1 function in vivo, the ability of mutant versions of NCS-1 to rescue the inviability of frq1 cells was tested in the same fashion. Both a non-myristoylated mutant, NCS-1(G2A) and a mutant in which the Ca 2ϩ binding affinity of the three functional EF-hands had been crippled, NCS-1(E81V T117A T165A) (2), rescued the lethality of the frq1 mutation, indicating that neither N-myristoylation nor Ca 2ϩ binding are essential for NCS-1 function (Fig. 1A, middle panels). These results were not unexpected because it had been demonstrated previously that N-myristoylation of Frq1 is dispensable for its function and that Frq1 associates with Pik1 tightly even in the absence of Ca 2ϩ (1). As judged by immunoblotting with polyclonal anti-NCS-1 antibodies, absence of N-myristoylation increased the steady state level of NCS-1 in both the initial diploid strain (YKBH1) (Fig. 1B, upper panel) and in the derived frq1 cells carrying the same plasmid (Fig. 1C). 2 Conversely, the Ca 2ϩ binding-defective mutant was reproducibly expressed at a level distinctly lower than normal NCS-1 in both Human NCS-1 Substitutes for S. cerevisiae Frq1 YKBH1 (Fig. 1B, upper panel) and in the derived frq1 cells carrying the same plasmid (Fig. 1C), suggesting that Ca 2ϩ binding helps stabilize the protein.
Remarkably, even a quadruple mutant, NCS-1(G2A E81V T117A T165A), also supported the growth of cells lacking Frq1 (Fig. 1A). This result was somewhat surprising because it had been shown before that Ca 2ϩ binding induces major structural changes in Frq1 that contribute independently of N-myristoylation to its ability to associate with membranes (3). However, the NCS-1(E81V T117A T165A) protein does have detectable residual Ca 2ϩ binding activity (2). Moreover, just as observed with otherwise wild-type NCS-1, lack of N-myristoylation significantly increased the apparent steady state level of the NCS-1(E81V T117A T165A) protein in both YKBH1 (Fig. 1B, upper panel) and in the derived frq1 cells carrying the same plasmid (Fig.  1C). These properties could account for the observed ability of the quadruple mutant to suppress the inviability of frq1 cells.
As one means to distinguish whether all of the NCS-1 variants retained full function, we examined their ability to rescue the inviability of frq1 cells under the somewhat stressful condition of elevated temperatures (35, 36, and 37°C) and when the level of expression was reduced (on glucose medium, which represses expression from the GAL1 promoter). The frq1 strains carrying each of the plasmids (His ϩ Leu ϩ cells), derived from the tetrad dissections of YKBH1 carrying the corresponding plasmid, were streaked out to single colonies on either YPGal and YPD plates and incubated at various temperatures. An isogenic yeast strain, YKBH7 (frq1::HIS3) (Table I), expressing FRQ1 from an URA3-marked high copy plasmid, was used as a positive control. When grown on YPGal plates, there were only modest differences in growth between wild-type NCS-1 and its mutant versions at all temperatures tested up to 36°C ( Fig. 1D and Table II). However, when the cells were grown on YPD, only NCS-1 (YTS47) and NCS-1(G2A) (YTS48), but not NCS-1(E81V T117A T165A) (YTS51) or NCS-1(G2A E81V T117A T165A) (YTS53), were able to support growth at 30°C (Fig. 1D and Table II). At higher temperatures, even NCS-1 and NCS-1(G2A) were unable to support vigorous growth, especially compared with the robust growth displayed by frq1 cells expressing authentic Frq1 ( Fig. 1D and Table II). Nevertheless, NCS-1 and the quadruple mutant NCS-1(G2A E81V T117A T165A) displayed very similar expression levels ( Fig. 1, B and C), yet distinctly different phenotypes ( Fig. 1D and Table II), arguing that the differences in their observed behavior cannot be attributed to differences in their level of expression. Thus, by these criteria, we conclude, first, that the loss of Ca 2ϩ binding clearly compromises the function and/or structural integrity of NCS-1 at higher temperatures, which can be largely compensated for by elevated expression. Second, although NCS-1 is clearly able to substitute for Frq1, even wild-type NCS-1 is not completely equivalent to native yeast Frq1.
NCS-1 Binds Specifically to the N Terminus of Pik1-Given that NCS-1 can replace Frq1 in vivo, NCS-1 is presumably able to associate with Pik1, and appropriately modulate its activity and localization. To test this prediction, we first used the yeast two-hybrid method to determine whether NCS-1 can interact with Pik1 in S. cerevisiae cells. It had been shown previously that the first 200 residues or so of Pik1 are necessary and sufficient for binding Frq1 (1) and, more recently, that residues 125-169 of Pik1 comprise the Frq1-binding site (20). To assess whether this region of Pik1 was able to associate with NCS-1, an N-terminal fragment of Pik1, Pik1 (10 -192), was tested for its ability to interact with NCS-1. To this end, the coding sequences for Pik1 (10 -192) and NCS-1 were introduced into appropriate vectors (pAS2-1 and pGAD424) to generate chimeras of each with the Gal4 DNA-binding domain (BD) and the Gal4 transcriptional-activation domain (AD), respectively. The resulting constructs were introduced, alone or in appropriate pairs, into a yeast reporter strain, CG-1945 (Table I), by DNAmediated transformation. Interaction was indicated by the ability of the CG-1945 cells to grow on an appropriate selective medium (SDϪTrpϪLeuϪHis). Robust growth was observed when the "bait" was Gal4BD-Pik1 (10 -192) and the "prey" was Gal4AD-NCS-1 ( Fig. 2A). Nearly as vigorous growth was observed when the chimeras were reversed (Gal4BD-NCS-1 and were prepared and samples (190 g total protein) subjected to SDS-PAGE in 12.5% gels. Next, proteins were transferred onto nitrocellulose membrane and subjected to immunoblot analysis with affinity-purified anti-NCS-1 (upper panel) or crude anti-KChIP (lower panel) antisera. C, cells derived from His ϩ Leu ϩ spores from A were grown in liquid YPGal medium to mid-exponential phase. Lysates were prepared, cleared of cell debris and vesicular content, and subjected to SDS-PAGE (250 g total protein each). After transfer onto nitrocellulose, NCS-1 was detected using affinity-purified polyclonal anti-NCS-1 antiserum. D, His ϩ Leu ϩ spores derived from the various NCS-1-transformants were plated on solid YPGal (upper panel) or YPD (lower panel) media and tested for growth at 30°C (left column) and 36°C (right column), respectively. Yeast strain YKBH7 (frq1⌬::HIS3) carrying an URA3-marked multicopy FRQ1-expression plasmid was used as a control.
To confirm the results of the two-hybrid method and to examine the binding of NCS-1 to Pik1 in greater detail and under conditions that could be more readily manipulated, we next carried out in vitro binding assays using NCS-1 and Pik1(10 -192)-His 6 , both expressed in and purified from E. coli. Pik1(10 -192)-His 6 was immobilized on Ni 2ϩ -saturated NTAagarose beads, and the beads were incubated with purified NCS-1 for 1 h. After extensive washing, bound proteins were eluted from the beads by adding SDS-PAGE sample buffer, resolved by SDS-PAGE, blotted onto a nitrocellulose membrane, and the amount of NCS-1 in each eluate was assessed by immunostaining with rabbit ␣-NCS-1 antiserum and horseradish peroxidase-conjugated goat anti-rabbit immunoglobulin. As observed in previous studies with yeast Frq1 (1,20), there was little or no nonspecific binding of human NCS-1 to empty beads (Fig. 2B, lane 4). In contrast, when the beads were first decorated with Pik1(10 -192)-His 6 , NCS-1 bound with high effi-ciency (Fig. 2B, lane 1). Moreover, this binding could be effectively competed by addition of a 10-fold molar excess of yeast Frq1 (Fig. 2B, lane 2), but not by an equivalent amount of KChIP2 (Fig. 2B, lane 3). These data confirm that NCS-1 is able to associate specifically and directly with the N terminus of Pik1 in the absence of any other yeast protein and demonstrate that the site bound must be congruent, or overlap substantially, with that normally occupied by Frq1. Moreover, as also observed for Frq1-Pik1 interaction, association of NCS-1 with Pik1 was not dependent on Ca 2ϩ because equivalent binding of NCS-1 to Pik1(10 -192)-His 6 -coated beads was observed in either the presence of 1 mM Ca 2ϩ or when a 2-fold molar excess of EGTA was present (Fig. 2C).
Mapping sufficient for its efficient interaction with Frq1 (20). To delineate the region within Pik1(10 -192) (designated "Fragment a" or Fa) responsible for its association with NCS-1, we first generated several overlapping deletion derivatives (Fig. 3A). One fragment (Fb) represented the most proximal portion of the Pik1 N terminus (residues 10 -107) and contained a sequence element referred to as the "lipid kinase unique domain," which is conserved among PtdIns 3-and PtdIns 4-kinases (18, 36 -39). The second fragment (Fc) corresponded to the central region (residues 53-144) of Pik1 (10 -192), and the third fragment (Fd) contained the C-terminal segment (residues 100 -192) of Pik1 (10 -192) (Fig. 3A). Interaction of these fragments with NCS-1 was examined, first, using the yeast two-hybrid reporter assay. For this purpose, the coding sequence for each fragment was introduced into vector pAS2-1, creating the corresponding chimeras with the Gal4BD. The resulting baits were tested for their ability to interact with Gal4AD-NCS-1 (in pGAD424-NCS-1). Only full-length Pik1(10 -192) (Fa) and its C-terminal end (Fd), corresponding to residues 100 -192, were able to associate with NCS-1, as judged by growth of the reporter strain CG-1945 (Fig. 3B).
To narrow down what portion within Fd was sufficient to mediate interaction with NCS-1, two additional Gal4BD derivatives of Pik1 were made: Fe (residues 65-172), which lacks the C-terminal end (residues 173-192); and Ff (residues 145-192), which lacks the N-terminal end (resides 100 -144) (Fig. 3A). Both constructs supported readily detectable growth of the reporter strain on selective medium when co-expressed with Gal4AD-NCS-1 (Fig. 3B); however, Ff gave distinctly better growth than Fe. Thus, although residues 145-172 of Pik1 are both necessary and sufficient for its association with NCS-1 in vivo, residues distal to 172 may contribute somewhat to the stability or conformation of this fragment in vivo.
As an independent means to verify these conclusions, four of the fragments (Fa, Fb, Fe, and Ff) were tagged with C-terminal His 6 tracts and the corresponding derivatives were expressed in and purified from E. coli. Ni 2ϩ -saturated NTA beads coated with these purified fragments were then used for in vitro pulldown assays. After incubation with purified NCS-1 (or with purified Myc epitope-tagged Frq1, as a positive control) and thorough washing, the bead-bound proteins were resolved by SDS-PAGE and analyzed by immunoblotting with appropriate antibodies to detect NCS-1 and Frq1 and by staining with Coomassie Brilliant Blue to determine the quantity of each His 6 -tagged fragment that had been immobilized on the beads. As observed before, little or no NCS-1 bound to empty beads (Fig. 3C, lane 5). Likewise, Fb, which did not interact with NCS-1 in the two-hybrid assay, was also unable to bind NCS-1 in vitro (Fig. 3C, lane 2). Similarly, Fc was also unable to bind NCS-1 in the in vitro pull-down assay (data not shown). As expected, NCS-1 bound avidly to Fa. Most importantly, and in agreement with the results obtained by the two-hybrid method, both Fe and Ff were able to bind NCS-1 with an affinity similar to NCS-1 binding by Fa (Fig. 3C, lanes 3 and 4). Essentially identical results were obtained for the binding of Frq1 to the same set of fragments (Fig. 3D).
Mapping the NCS-1-binding Residues in Pik1 by in Vitro Mutagenesis-Binding of Frq1 to Pik1 seems to depend largely on hydrophobic interactions (20), and both NCS-1 (2) and Frq1 (3) have a wide hydrophobic crevice at their surface. These considerations suggested that hydrophobic side chains within the 145-172 segment of Pik1 might contribute primary contacts that mediate the docking of NCS-1. As one means to characterize the chemical nature of the interaction of NCS-1 with Pik1(10 -192), we performed Ala and/or Lys scanning mutagenesis at 39 positions within the NCS-1-binding region of Pik1 (10 -192) (Table III and Fig. 4A). To examine the ability of these mutants to bind purified NCS-1, each was tagged at its C terminus with His 6 and expressed and purified from E. coli and used for the in vitro pull-down assay, as described above. None of the Ala substitution mutations tested was impaired in its ability to interact with NCS-1 ( Fig. 4B and Table III). In contrast, replacement of some of the same residues with Lys either reduced somewhat (S164K and, perhaps, I170K) or almost totally eliminated (L160K and A171K) NCS-1 binding (Fig. 4B and Table III). These effects were most dramatically illustrated for double mutants. Pik1 (10 -192; L160A S164A) bound NCS-1 with an affinity indistinguishable from normal Pik1 (10 -192), whereas Pik1(10 -192; L160K S164K) showed no detectable binding of NCS-1 whatsoever (Fig. 4C). These observations confirm in striking fashion that the hydrophobic character of residues in this region of Pik1 are crucial for NCS-1 binding. In addition, given the spacing of the residues that appear to be most critical for NCS-1 binding, as judged by the Lys scanning mutagenesis, the results would be most easily interpreted if these residues projected from the same face of an ␣-helix, suggesting that part of the NCS-1 binding site in Pik1 (at least that situated between residues 160 and 171) may adopt an ␣-helical conformation and, therefore, that NCS-1 binds to one side of this ␣-helix. Recent NMR studies on complexes of Frq1 with Pik1-derived peptides indicate that a similar 13-residue stretch of Pik1 (residues 157-169) comprises part of the Frq1binding site (20) and adopts an ␣-helical conformation. 3 Monitoring of Pik1-NCS-1 Interaction by Fluorescence Spectroscopy-To detect possible structural changes induced in NCS-1 upon its binding to Pik1, we monitored the intrinsic tryptophan fluorescence of NCS-1. The NCS-1 polypeptide contains just two Trp residues: Trp 30 , adjacent to EF1, and Trp 103 , neighboring EF3. Both of these Trp residues are invariant features of frequenins from yeasts (e.g. S. cerevisiae, Schizosaccharomyces pombe, Candida albicans), invertebrates (e.g. C. elegans, D. melanogaster), non-mammalian vertebrates (e.g. 3 J. B. Ames, personal communication.

TABLE III
In vitro binding of NCS-1 to alanine and lysine scanning mutants of Pik1 (10 -192)-His 6 ϩ, binding to NCS-1 is comparable to wild-type Pik1(10 -192)-His 6 ; Ϫ, no detectable binding of NCS-1; ϩ/Ϫ, significantly impaired binding to NCS-1; ND, not determined; NA, not applicable. Xenopus laevis), and mammals (e.g. mice, rats, humans). For these studies, we used Pik1(120 -192) because it contains the NCS-1 binding site, but lacks any Trp residue, and, as a consequence, should not contribute to the fluorescence emitted at the wavelengths examined. To monitor structural changes in NCS-1 induced by its interaction with Pik (120 -192), corresponding changes in the intensity of the fluorescence emission were monitored. Upon excitation at 280 nm, the fluorescence emission spectrum of free Ca 2ϩ -bound NCS-1 exhibited a maximum at 340 nm. Addition of a 2-fold molar excess of Pik1(10 -192)-His 6 increased the emission at 340 nm by 20% and caused the emission maximum to shift by ϳ5 nm to the blue (Fig. 5A). The change in the intensity of the fluorescence at 340 nm as a function of increasing concentration of Pik1(120 -192) was used to measure the apparent affinity of NCS-1 for the peptide. As described in detail under "Experimental Procedures," a saturation binding curve can be calculated from the observed changes in fluorescence intensity (Fig. 5B). From these binding data (three independent trials), one can calculate both the Hill coefficient, which was 1.01, and a dissociation constant (K D ϭ 0.15 Ϯ 0.01 M). These results further corroborate that NCS-1 binds to Pik1(120 -192) with high affinity and non-cooperatively, most likely indicative of a 1:1 stoichiometry for the complex.

DISCUSSION
As described here, we have shown that human NCS-1 is able to substitute for the function of endogenous yeast Frq1. We showed previously that frog frequenin is also able to replace Frq1 in yeast cells (1). Thus, proteins that appear to be homologs on the basis of sequence criteria are indeed true functional orthologs. In marked contrast, we showed before that a closely related, small, N-myristoylated Ca 2ϩ -binding protein, bovine recoverin, was unable to substitute for Frq1 in yeast (1). Similarly, in this study, we have shown that yet another highly related, small, Ca 2ϩ -binding protein closely related to NCS-1, human KChIP2, is also unable to replace Frq1 in yeast. KChIP has been implicated as a modulator of certain types of potassium channels (35). Recently, however, it has been reported that NCS-1/frequenin itself can affect potassium channel activity, at least as judged by a Xenopus oocyte microinjection assay (40). Nevertheless, the fact that overexpression of neither human KChIP2 nor bovine recoverin can compensate for the loss of Frq1 in S. cerevisiae, whereas frog frequenin and human NCS-1 can, argues strongly that in an appropriate in vivo context, the physiological functions of even rather closely related members of the family of small, N-myristoylated, Ca 2ϩbinding proteins can be readily distinguished. In this regard, our results indicate that the in vivo role of both Frq1 and NCS-1 required for the maintenance of viability, at least in yeast cells, is not modulation of potassium channel function. Our findings suggest further that, in their endogenous context, different members of the family of small, N-myristoylated, Ca 2ϩ -binding proteins serve discrete and independent biological functions. Indeed, as we demonstrated here by two independent means (the two-hybrid method in vivo and a competition binding assay in vitro), NCS-1 can bind directly to the N  (10 -192). A, linear depiction of the primary sequence of Pik1 (residues 147-171) in single-letter code. Residues found to be sensitive to mutation (see also Table III) are marked by asterisks. B, immunoblot analysis of the interaction between NCS-1 and wild-type (wt) or mutant versions of Pik1 (10 -192). NCS-1 was incubated with various Pik1 (10 -192) derivatives, immobilized on Ni 2ϩ -agarose beads or, as a control, with empty beads (Control). After washing, the beads were stripped of bound proteins by adding SDS buffer and samples were resolved by SDS-PAGE. Pik1 (10 -192) and its derivatives were visualized by staining with Coomassie Brilliant Blue (lower panel) to assess possible differences in their expression levels, and bound NCS-1 was detected by immunoblot analysis using ␣-NCS-1 antiserum (upper panel). C, in vitro binding of NCS-1 to wild-type Pik1(10 -192) (wt), Pik1(10 -192; L160A S164A) or Pik1 (10 -192; L160K S164K). Experimental procedures were identical to B. Immunoblot analysis of NCS-1 is shown in the upper panel and Coomassie Blue staining of Pik1 (10 -192) in the bottom panel.
terminus of Pik1, like Frq1, whereas KChIP2 cannot, providing a straightforward explanation for how NCS-1 is able to genetically complement the frq1 mutation and why KChIP2 does not.
Increases in mammalian NCS-1/frequenin in organisms with a nervous system enhance neurotransmitter release at neuromuscular junctions (10,41) and at synapses of the central nervous system (42). There are data showing that frequenin action promotes both agonist-induced potentiation of N-type Ca 2ϩ channels (43) and activity-dependent facilitation of currents through P/Q-type Ca 2ϩ channels (42), suggesting, first, that frequenin contributes to synaptic facilitation at a variety of different synapses and, second, that enhanced Ca 2ϩ entry might explain the increase in synaptic vesicle release when NCS-1 is overexpressed. The apparent modulation of oocytesexpressed A-type K ϩ channel activity by NCS-1 (40) might also represent an indirect effect of changes in cellular Ca 2ϩ . On the other hand, it has been reported that overexpression in bovine chromaffin cells of a purported dominant-negative mutant, NCS-1(E120Q), increased (rather than decreased) non-L-type Ca 2ϩ channel currents (44). In addition to the evidence for some regulatory role of NCS-1 that affects the function of ion channels, NCS-1 has also been implicated in Ca 2ϩ -signaling pathways that result in receptor desensitization by G-proteincoupled receptor kinases (45)(46)(47). Correspondingly, in PC12 cells, NCS-1 overexpression increases phosphoinositide turnover in response to agonists, such as UTP and bradykinin, that are known to activate receptors that stimulate phospholipase C, presumably resulting in inositol trisphosphate-dependent Ca 2ϩ release, thereby leading to increased Ca 2ϩ -dependent events and enhancement of exocytosis (13). Taken together, these data suggest that the effects of NCS-1/frequenin on ion channel activity, Ca 2ϩ signaling, exocytosis/neurotransmitter release, and the efficiency of synaptic function, are all indirect. All of these effects could be explained if NCS-1 controlled the supply of phosphoinositides (48), a premise based largely upon what is known about the molecular function of Frq1, the NCS-1 ortholog in budding yeast (S. cerevisiae).
It has been shown that Frq1 is bound tightly to, and serves as a non-catalytic subunit of, the essential PtdIns 4-kinase isoform, Pik1 (1). At restrictive temperature, pik1 ts mutants display defects in the delivery of secretory proteins from the Golgi apparatus to the plasma membrane (15)(16)(17), suggesting a role for Pik1-derived phosphoinositides in a late stage of the secretory pathway. In agreement with this finding is the demonstration that pik1 mutants accumulate abnormal Golgi structures, termed "Berkeley bodies," at the restrictive temperature. Frq1, which is itself essential for yeast cell viability, co-purifies stoichiometrically with Pik1 from yeast cell extracts and is required for optimal activity of the enzyme (1). Moreover, Frq1 is able to interact with biological membranes via both its N-myristoyl group and via residues exposed upon the conformation changes induced by Ca 2ϩ binding (3). Thus, Frq1 might assist with recruitment of Pik1 to membranes, because Pik1 itself lacks any obvious membrane targeting motifs. Pik1 is the yeast ortholog of mammalian PtdIns 4-kinase ␤ (PtdIns4K␤), an enzyme also found to associate primarily with the Golgi apparatus (36). Hence, it is conceivable that NCS-1 regulates PtdIns4K␤ in mammalian cells in a manner analogous to the modulation of Pik1 activity by Frq1 in yeast. Indeed, over time, there have been reports that NCS-1 and PtdIns4K␤ can be co-immunoprecipitated from lysates of several tissue culture cell lines, including bovine chromaffin cells, Madin-Darby kidney cells, and COS-7 cells (14, 49, 50), suggesting that NCS-1 and PtdIns4K␤ might interact directly. Consistent with this notion, overlapping immunocytochemical localization also has been reported for NCS-1, the small GTPase, ARF1, and PtdIns4K␤ in COS-7 cells (2) and for NCS-1 and PtdIns4K␤ in newborn cultured dorsal root ganglia neurons (51).
All frequenins, including NCS-1, are myristoylated at their N termini (12), allowing them to transiently interact with biological membranes. Irreversible covalent attachment of myristate requires an N-terminal Gly residue (52). When the nonmyristoylatable NCS-1(G2A) mutant was present as the sole source for frequenin in frq1 cells, we could detect no phenotypic differences from frq1 cells expressing wild-type NCS-1. Thus, N-myristoylation of NCS-1 is not essential for its function, at least in yeast cells, in good agreement with the previous finding that a Frq1(G2A) mutant also restores viability to frq1 cells (1).
Frq1 and NCS-1 are both able to coordinate three Ca 2ϩ ions per molecule with high affinity (2,3). In prior work, based on both biochemical and genetic approaches, optimal membrane association of Frq1 was found to depend on both Ca 2ϩ -binding and the N-terminal myristoyl group (3), but no mutations in the EF-hand domains of Frq1 were generated to investigate the role of Ca 2ϩ binding in its cellular function. In this regard, we found that NCS-1(E81V T117A T165A), in which the Ca 2ϩ binding affinity of all three functional EF-hand domains has been greatly impaired (2), was still able to support growth of frq1 cells. This finding was somewhat unexpected because it was shown by a variety of biophysical techniques that Ca 2ϩ binding induces major structural changes in Frq1 (3) and, presumably, also in NCS-1. However, NCS-1(E81V T117A T165A) does display residual Ca 2ϩ binding activity (2), which might explain how it was still able to function when it was expressed at a high level from the GAL1 promoter. However, at low levels of expression, NCS-1(E81V T117A T165A) was not capable of restoring viability in frq1 cells under conditions where normal NCS-1 did. By this criterion at least, the lack of efficient Ca 2ϩ binding does compromise the function, folding, and/or stability of NCS-1 and/or in its ability to assist with the targeting of Pik1 to membranes.
Some members of the family of small, N-myristoylated, Ca 2ϩ -binding proteins undergo a dramatic change in conformation upon Ca 2ϩ binding that extrudes the N-myristoyl group from a pocket buried in the protein and places it in a solventexposed condition, a structural rearrangement dubbed the "Ca 2ϩ -myristoyl switch" (8,(53)(54)(55)(56). Several lines of evidence indicate that frequenins do not possess this Ca 2ϩ -myristoyl switch mechanism. First, as judged by a wide range of biophysical experiments, the myristoyl group in Frq1 and NCS-1 appears to be constantly solvent-exposed (2, 3). Second, Frq1 and NCS-1 have a high affinity for Ca 2ϩ , with a K D not far above the value (ϳ100 nM) for the basal level of cellular Ca 2ϩ concentration, suggesting that a large fraction of the population of frequenin molecules might be in the Ca 2ϩ -bound state, even in resting cells. Finally, as shown here for NCS-1 and earlier for Frq1 (1), the fact that N-myristoylation is dispensable for cellular function argues against regulation of Frq1 membrane association via a Ca 2ϩ -induced myristoyl switch.
Prior work demonstrated that Frq1 interacts with the N terminus of Pik1 (1) and does so at a specific region that includes residues 125-169 (20), which lies outside of the conserved lipid kinase unique domain motif, a sequence element that still has no defined molecular function. The Frq1-binding site in Pik1 appears to consist of two subsites defined by Phe 125 -Gln 136 and Ala 157 -Ser 169 (20). The distal residues bind Frq1 independently of the proximal residues, whereas binding to residues 125-136 appears to require that contact be made with residues 157-169. The minimal NCS-1-binding element in Pik1 defined by the in vitro pull-down experiments and the in vivo yeast two-hybrid reporter assays is located between residues 145 and 172, which is almost congruent with the distal element responsible for Frq1 binding. Presence of residues 125-136 in Pik1 was not required for high affinity interaction with NCS-1. As observed for Frq1, Ca 2ϩ binding did not appear to be required for stable association of NCS-1 with Pik1. As judged by a titration binding assay in which changes in intrinsic Trp fluorescence were used as a means to measure the amount of Pik1(120 -192) required to saturate NCS-1, NCS-1 bound with an apparent K D of 150 nM, in good agreement with the measured affinity of Frq1-Pik1 complexes (Frq1-Pik1(110 -192), apparent K D ϭ 62 nM; Frq1-Pik1(151-192) apparent K D ϭ 140 nM) (20). The observed Hill coefficient of 1 for the NCS-1-Pik1(120 -192) interaction is also in agreement with the formation of the 1:1 Frq1-Pik1 complexes observed in yeast cell extracts (20).
As one approach to characterize the physicochemical nature of NCS-1 binding to Pik1, Ala and Lys scanning mutagenesis in the region of residues 147-171 was performed in the context of the Pik1(10 -192) N-terminal fragment. None of the Ala substitution mutants tested (18/39) was able to significantly alter NCS-1 binding, as judged by the in vitro pull-down assay, whereas at four positions where a change to Ala had no effect, substitution of Lys either modestly (S164K and I170K) or drastically (L160K and A171K) reduced NCS-1 binding. The four mutations that affected NCS-1 binding are located in a very hydrophobic segment of the Pik1 N terminus, 156 VAPALV-LSSMIMSAIAFP 173 , comprising exclusively non-polar and uncharged residues. Thus, our results indicate that binding of Pik1 to NCS-1 is mediated primarily by hydrophobic contacts, as observed for the interaction of Frq1 with Pik1 (20).
The crystal structure of Ca 2ϩ -saturated NCS-1 (2) and the NMR-derived model of the structure of Ca 2ϩ -bound Frq1 (3) revealed that both proteins have a wide solvent-exposed crevice, lined predominantly with hydrophobic residues, a feature that is not present in the structures of other members of the family of small, N-myristoylated, Ca 2ϩ -binding proteins, such as neurocalcin or recoverin (6,8,57). This cleft is the most likely structural element to provide the hydrophobic environment necessary for docking of the hydrophobic sequence element in Pik1 that constitutes its Frq1-(and NCS-1-) binding site. The spacing of the residues (Leu 160 , Ser 164 , Ile 170 , Ala 171 ) where a change to Lys had some effect on NCS-1 binding is reminiscent of how they would be situated if they were located on one side of an ␣-helix. Indeed, preliminary HSQC spectra for [ 15 N]Pik1(110 -192) bound to Ca 2ϩ -saturated Frq1 show that both segments which make contact with Frq1 (Phe 125 -Gln 136 and Ala 157 -Ser 169 , respectively) also assume an ␣-helical conformation in the complex (20). 3 Hence, interaction of Pik1 with either Frq1 or NCS-1 is based on a very similar, if not identical, mechanism and involves the same binding domains. Presence of the hydrophobic crevice, which, so far, seems unique to frequenins among the members of the family of small, N-myristoylated, Ca 2ϩ -binding proteins for which structures have been determined at atomic resolution, could help explain how frequenin can interact with a physiological target(s) that does not overlap with the targets of other members of the family in the same cell type. If so, the fact that KChIP2 was unable to interact with Pik1 predicts that it lacks such a hydrophobic cleft.
Given that prior work has established that the only essential role of Frq1 in yeast cells is association with and regulation of Pik1, our findings demonstrate that NCS-1 is also capable of fulfilling these functions, at least in yeast. It remains to be seen, however, whether NCS-1 directly interacts with and reg-ulates the function and/or localization of a vertebrate PtdIns 4-kinase. If so, then frequenins, in general, might serve as regulators of PtdIns 4-kinases in all cell types.