Neuronal Ca2+ Sensor 1

Overexpression of frequenin and its orthologue neuronal Ca2+ sensor 1 (NCS-1) has been shown to increase evoked exocytosis in neurons and neuroendocrine cells. The site of action of NCS-1 and its biochemical targets that affect exocytosis are unknown. To allow further investigation of NCS-1 function, we have demonstrated that NCS-1 is a substrate forN-myristoyltransferase and generated recombinant myristoylated NCS-1. The bacterially expressed NCS-1 shows Ca2+-induced conformational changes. The possibility that NCS-1 directly interacts with the exocytotic machinery to enhance exocytosis was tested using digitonin-permeabilized chromaffin cells. Exogenous NCS-1 was retained in permeabilized cells but had no effect on Ca2+-dependent release of catecholamine. In addition, exogenous NCS-1 did not regulate cyclic nucleotide levels in this system. These data suggest that the effects of NCS-1 seen in intact cells are likely to be due to an action on the early steps of stimulus-secretion coupling or on Ca2+ homeostasis. Myristoylated NCS-1 bound to membranes in the absence of Ca2+ and endogenous NCS-1 was tightly membrane-associated. Using biotinylated NCS-1, a series of specific binding proteins were detected in cytosol, chromaffin granule membrane, and microsome fractions of adrenal medulla. These included proteins distinct from those detected by biotinylated calmodulin, demonstrating the presence of multiple specific Ca2+-independent and Ca2+-dependent binding proteins as putative targets for NCS-1 action. A model for NCS-1 function, from these data, indicates a constitutive membrane association independent of Ca2+. This differs from the Ca2+ myristoyl switch model for the closely related recoverin and suggests a possible action in rapid Ca2+ signal transduction in response to local Ca2+ signals.

Among the many EF-hand-containing Ca 2ϩ -binding proteins is a family known as the recoverin/neurocalcin or neuronal calcium sensor family (1). These small Ca 2ϩ -binding proteins include members expressed only in photoreceptor cells such as recoverin (2), the best characterized of this family (3). Recoverin has a well defined function in the control of rhodopsin phosphorylation due to direct inhibition of rhodopsin kinase (3,4). Other members such as VILIP (6) neurocalcin (7), hippocalcin (8), frequenin (9), and neuronal Ca 2ϩ sensor 1 (NCS-1) (10) are highly expressed in neurons and were originally all be-lieved to be neuron-specific implying a key role in Ca 2ϩ -regulated events in the nervous system. The cellular function of many of the neuronal proteins of this family remains a mystery, but, in contrast, some insight has been gained into the role of frequenin/NCS-1 in the regulation of presynaptic function.
Frequenin is a Drosophila protein that has been demonstrated in molecular genetic and other studies to regulate synaptic neurotransmission. Overexpression of frequenin in Drosophila, in the V7 mutant (9), was found to facilitate evoked neurotransmission at the neuromuscular junction, and a direct stimulation of both spontaneous and evoked neurotransmission was found following injection of Xenopus frequenin into Xenopus spinal neurons (11). NCS-1, first identified in chickens (10) and later in rodents (12,13) and Caenorhabditis elegans (13), appears to be the frequenin orthologue expressed in those species. Recently, it has been shown that NCS-1 is not expressed solely in neurons, since it was detected in glial cells in the central nervous system (12). In addition, it is present in adrenal chromaffin cells and PC12 neuroendocrine cells, where its overexpression also results in an increase in Ca 2ϩ -regulated exocytosis but from dense-core granules (14), analogous to the effect reported for frequenin in Drosophila for synaptic vesicle exocytosis (9). It is clear, therefore, that frequenin/NCS-1 has an important regulatory role in the steps leading to Ca 2ϩ -dependent exocytosis of synaptic vesicles and dense core granules in neurons and neuroendocrine cells. The step in the exocytotic sequence at which frequenin/NCS-1 acts is, however, not known.
The biochemical processes on which NCS-1 might act in vivo are also unclear. From in vitro experiments, frequenin has been shown to activate membrane-bound guanylate cyclase, but only at low Ca 2ϩ concentration, in rod outer segments (9) and NCS-1 to inhibit rhodopsin kinase (13). These in vitro effects on photoreceptor proteins may not be relevant to the function of frequenin/NCS-1 in neurons and neuroendocrine cells. In addition, NCS-1 has been shown to activate various calmodulin targets including cyclic nucleotide phosphodiesterase, calcineurin, and nitric-oxide synthase (15). NCS-1 has also been implicated as a direct or indirect (via cGMP) activator of Ca 2ϩdependent K ϩ channels (9,15,16) and also Na 2ϩ /Ca 2ϩ exchange (17). Despite these interactions, the molecular targets for frequenin/NCS-1 action in vivo are not known for certain, and thus the mechanism by which evoked exocytosis is increased remains to be elucidated.
From analysis of its biochemical properties (18,19) and structure (3), recoverin has been described as a calcium-myristoyl switch, and this was initially assumed to typify the behavior of all members of this family of proteins. Ca 2ϩ binding to two of the four EF-hand-like domains of recoverin leads to the exposure of an N-terminal myristoyl group (3) believed to allow membrane attachment (19). The movement of the myristoyl group also exposes a hydrophobic pocket that may then interact with target proteins (3). All other members of this family possess consensus myristoylation motifs (20) and, in the case of recoverin (19), neurocalcin (21,22), and hippocalcin (23), only the myristoylated and not the nonmyristoylated protein shows Ca 2ϩ -dependent binding to membranes consistent with the calcium-myristoyl switch model. In contrast, another family member, the guanyl cyclase-activating protein 2, shows the distinct property of only binding to membranes and activating its target protein at low Ca 2ϩ concentration and dissociates from membranes at micromolar Ca 2ϩ (24). Frequenin and NCS-1 possess a putative N-myristoylation motif (10 -12), but their myristoylation and Ca 2ϩ dependence of membrane binding have not been examined for either native or recombinant protein.
The aims of this study were to examine whether NCS-1 is a substrate for myristoyltransferase by bacterial co-expression with the yeast enzyme (25), to characterize myristoylated NCS-1, and to use the recombinant protein to investigate cellular functions of NCS-1. The results show that NCS-1 does indeed act as a substrate for N-myristoyltransferase, and the recombinant protein shows Ca 2ϩ -dependent conformational changes but Ca 2ϩ -independent interaction with membranes, and thus its behavior differs from recoverin. In addition, we have shown that NCS-1 does not appear to act directly on the exocytotic machinery in adrenal chromaffin cells, nor does it modulate cyclic nucleotide levels in permeabilized cells. Endogenous NCS-1 is tightly associated with cellular membranes in a Ca 2ϩ -independent manner, and we have detected multiple putative Ca 2ϩ -dependent and Ca 2ϩ -independent membrane target proteins using recombinant NCS-1.

MATERIALS AND METHODS
Plasmid Production-Total RNA was extracted from whole brain of Wistar rats using an RNeasy isolation kit (Qiagen, Surrey, UK) and cDNA synthesized with a reverse transcription system. cDNA encoding NCS-1 was amplified in polymerase chain reactions using Pfu polymerase (Stratagene, Cambridge, UK), according to the supplier's protocol, and an Omne-E dryblock thermocycler (Hybaid, Middlesex, UK). The sense and antisense primers used were 5Ј-CATCATATGGGGAAATC-CAACAGCAAGTTGA-3Ј and 5Ј-CATGGATCCCTATACCAGCCCGTC-GTAGAGG-3Ј, respectively. These were based on the nucleotide sequence of rat NCS-1 (Ref. 13; GenBank TM accession no. L27421) and incorporated NdeI and BamHI restriction sites (underlined) to facilitate subcloning of the product into the pET-5a expression vector (Promega, Southampton, UK). The nucleotide sequence was established by automated sequencing.
Myristoylated NCS-1 was purified using methods adapted from those of Zozulya et al. (26) for recoverin and neurocalcin. Upon thawing, cells were added to 1 volume of buffer A supplemented with 0.2 mM EGTA, incubated with 1 mg/ml lysozyme for 30 min, ultrasonicated, and incubated with 2 g/ml DNase for 15 min, and finally cell debris was removed by centrifugation at 100,000 ϫ g for 1 h. The cleared lysate was supplemented with 1 mM CaCl 2 and applied to a phenyl-Sepharose (Amersham Pharmacia Biotech, Uppsala, Sweden) column. The column was washed, and NCS-1 was eluted with buffer A containing 5 mM EGTA. For further purification, pooled eluted fractions containing NCS-1 were diluted with 3 volumes of water and applied to a Q-Sepharose (Amersham Pharmacia Biotech) column equilibrated with buffer B (20 mM Tris-HCl, (pH 8.0), 1 mM dithiothreitol, and 50 mM EGTA), and NCS-1 was eluted with a 0 -600 mM KCl gradient in a single peak. All chromatography was performed at room temperature (22-25°C) using an Amersham Pharmacia Biotech FPLC system. Peak fractions containing recombinant proteins were analyzed by SDS-polyacrylamide gel electrophoresis, and the pooled fractions were stored at Ϫ80°C until required. Analysis by HPLC used reverse phase chromatography with elution by a 0 -50% acetonitrile gradient.
Tryptophan Fluorescence Emission Spectra-Purified recombinant NCS-1 (1 M) in 20 mM HEPES (pH 7.4), 139 mM NaCl, 2 mM ATP, 5 mM nitrilotriacetic acid, and 5 mM EGTA was excited at room temperature at a wavelength of 280 nm, and the emission spectra from 290 to 410 nm were detected using a Perkin-Elmer LS-5 luminescence spectrometer. The free Ca 2ϩ concentration was increased by incremental additions of CaCl 2 , and the subsequent spectra were similarly recorded.
Preparation of Membranes from Rat Brain-Whole brains of Wistar rats were homogenized in HEPES (pH 7.2), 140 mM sucrose, 70 mM potassium acetate, 1 mM dithiothreitol, and 230 M phenylmethylsulfonyl fluoride and centrifuged at 1600 ϫ g for 10 min at 4°C. The supernatant was removed and centrifuged at 100,000 ϫ g for 60 min at 4°C, and the subsequent pellet was resuspended in 125 mM Na 2 CO 3 (pH 11.5). After incubation at 4°C for 30 min, membranes were obtained by centrifugation at 100,000 ϫ g for 60 min at 4°C; resuspended in 20 mM HEPES (pH 7.4), 100 mM NaCl, and 2 mM EDTA; and stored at 4°C until required.
Determination of NCS-1 Association with Membranes-Membranes were washed twice by centrifugation at 13000 ϫ g for 10 min at room temperature, with 20 mM HEPES (pH 7.4), 139 mM NaCl, 2 mM ATP, 5 mM nitrilotriacetic acid, and 5 mM EGTA and finally resuspended in the same buffer with added MgCl 2 and CaCl 2 to give a calculated free Mg 2ϩ concentration of 2 mM and free Ca 2ϩ concentration of 0 or 10 M. After incubation at room temperature for 30 min with 3 H-labeled NCS-1 in a total volume of 500 l, the membranes were washed twice in the appropriate calcium buffer, and duplicate aliquots were subjected to scintillation counting. Background values for NCS-1 bound in the absence of brain membrane were subtracted.
Determination of Endogenous NCS-1 Leakage from Permeabilized Chromaffin Cells-Bovine chromaffin cell cultures were prepared and maintained as described previously (27). Cells were permeabilized in 300 l/well permeabilization buffer (20 mM PIPES (pH 6.5), 139 mM potassium glutamate, 5 mM EGTA, 2 mM ATP, and 2 mM MgCl 2 ) containing 20 M digitonin (Novabiochem, Nottingham, UK). After the indicated times, the supernatants were centrifuged at 13,000 ϫ g for 3 min to remove cell debris, and then both supernatants and cells were precipitated with an equal volume of methanol at Ϫ20°C to concentrate protein samples before solubilization in 300 l of SDS dissociation buffer and boiling. Samples were separated on a 15% SDS-polyacrylamide gel and analyzed by immunoblotting as above.
Assay of Catecholamine Secretion from Permeabilized Chromaffin Cells-For secretion assays, the cells were permeabilized as above for 10 min (step 1). Subsequent incubations were carried out in 20 mM HEPES (pH 7.4), 139 mM NaCl, 2 mM ATP, 5 mM nitrilotriacetic acid, and 5 mM EGTA in the presence of various amounts of CaCl 2 and MgCl 2 to give the desired free Ca 2ϩ concentration and a calculated free Mg 2ϩ concentration of 2 mM. The cells were then incubated for 15 min in the presence or absence of recombinant NCS-1 in the absence of Ca 2ϩ (step 2). Cells were then challenged in step 3 with buffer containing no Ca 2ϩ or up to 10 M free Ca 2ϩ , again in the presence or absence of recombinant NCS-1, and catecholamine released over a 40-min period was determined by means of a fluorometric assay (28). All experiments were performed at room temperature.
Assay of cAMP and cGMP Production in Permeabilized Chromaffin Cells-Cells were permeabilized and treated as above. For assay of cAMP, the incubation medium was removed after step 3, mixed with an equal volume of 10 mM Tris-HCl (pH 7.5) and 8 mM EDTA, boiled for 2 min, sonicated for 15 s, and centrifuged at 13,000 ϫ g for 3 min. Levels of cAMP in the subsequent supernatants were determined using the appropriate enzyme immunoassay kit (Amersham Pharmacia Biotech) according to the manufacturer's instructions. For the assay of cGMP, 1 ⁄10 volume of lysis buffer (5% dodecyltrimethylammonium bromide in 0.05 M sodium acetate, pH 5.8) was added to the wells at the end of the incubation in step 3, and samples were taken for assay of cGMP after the acetylation of samples using an enzyme immunoassay kit (Amersham Pharmacia Biotech).
Characterization of Binding of Biotinylated Proteins to Adrenal Medulla Subcellular Fractions-Cytosol, chromaffin granule membrane, and microsome fractions were prepared from bovine adrenal medulla by standard techniques (29). For biotinylation of recombinant NCS-1, ADP-ribosylation factor 1 (ARF1), or purified bovine brain calmodulin (Sigma), the proteins (200 g/ml) were mixed with 6-(biotinamidocaproylamindo) N-hydroxysuccinimide ester (Sigma) from a stock solution in Me 2 SO to give a 100-fold molar excess of biotinylating reagent. Incubation was performed at room temperature for 2 h and stopped by the addition of a final concentration of 100 mM glycine. To remove excess biotin, biotinylated proteins were dialyzed against 20 mM HEPES, 139 mM NaCl, 2 mM ATP, 5 mM EGTA, 5 mM nitrilotriacetic acid, pH 7.4. The concentration of the biotinylated proteins was determined by taking A 280 and converting to mg/ml based on the predicted molar extinction coefficient. Biotinylated probes were stored at Ϫ20°C until use. Proteins separated on SDS gels were blotted onto nitrocellulose membranes for analysis. Membranes were incubated in blocking solution (the above buffer plus 5% milk, 5% bovine serum albumin, 5% fetal calf serum, and 0.5% Tween 20) at room temperature for 2 h with three changes and then incubated overnight at 4°C with biotinylated probes at 3.6 -8 g/ml in blocking solution (with the addition of MgCl 2 and CaCl 2 to obtain the appropriate free Ca 2ϩ concentration and 2 mM free Mg 2ϩ ). Membranes were washed four times in buffer plus the appropriate free Ca 2ϩ concentration and 0.5% Tween 20 and incubated with Streptavidin-horseradish peroxidase at 1:400 in blocking solution for 30 min. The membranes were washed four times as above, and bound biotinylated fragments were detected using enhanced chemiluminescence (Amersham Pharmacia Biotech). In other experiments, binding of biotinylated proteins to bovine brain calcineurin (Sigma) was assayed using a similar protocol.

RESULTS
Expression of Myristoylated NCS-1-NCS-1 has a potential N-terminal myristoylation site (10 -12), related to the consensus motif recognized by N-myristoyltransferase (20). To determine whether NCS-1 can be a substrate for N-myristoyltransferase, NCS-1 was co-expressed in bacteria with the yeast enzyme, and [ 3 H]myristic acid was added to the growth medium to allow detection of myristic acid incorporation. Bacterial cultures expressing the enzyme alone revealed no labeled polypeptides following an overnight exposure of the fluorography. In contrast, when both plasmids for N-myristoyltransferase and for NCS-1 were present, an intensively labeled polypeptide of the expected size for NCS-1 was detected by fluorography after induction of protein expression with isopropyl-1-thio-␤-D-galactopyranoside (Fig. 1A), and labeled NCS-1 could be purified (see below), suggesting that NCS-1 is indeed an effective substrate for the NMT enzyme, which is known to show unique specificity for myristate.
For bulk expression of myristoylated NCS-1 protein, expression was induced in bacterial cultures grown with the addition of 5 g/ml myristic acid to the medium. The NCS-1 was expressed in high levels as a soluble protein, and after bacterial cell lysis, the NCS-1 was purified by Ca 2ϩ -dependent chromatography on phenyl-Sepharose. This single step purification gave a substantial purification of NCS-1 from the bacterial lysate (Fig. 1B). In some cases, further purification was carried out by ion exchange chromatography, but this was not routinely required and did not markedly improve the purity of the NCS-1. On both phenyl-Sepharose and ion exchange chromatography, NCS-1 resolved as a single but asymmetric peak with a long tail. Since it was possible that not all of the protein would be myristoylated, further analysis was carried out to determine the relative amounts of myristoylated and nonmyristoylated protein in the pool of purified protein. Analysis of purified NCS-1 by reverse-phase HPLC revealed two discrete peaks both of which contained NCS-1 when analyzed by SDS-PAGE (Fig. 1C). The first corresponded to the position of [ 3 H]myristate-labeled NCS-1 in other experiments, so the myristoylated pool of NCS-1 from Fig. 1C amounted to 58% of the total NCS-1.
Ca 2ϩ Binding Properties of Recombinant NCS-1-To establish the functional nature of the bacterially expressed myristoylated NCS-1, its ability to bind Ca 2ϩ , leading to Ca 2ϩ -dependent conformational change, was assessed. First, the effect of Ca 2ϩ on migration of NCS-1 on SDS-PAGE was examined. As shown for other related Ca 2ϩ -binding proteins such as Drosophila frequenin (9), NCS-1 underwent a conformational change in the presence of Ca 2ϩ that resulted in faster migra- tion on SDS-PAGE ( Fig. 2A). In addition, conformational change could be demonstrated due to increasing Ca 2ϩ concentration from the monitoring of tryptophan fluorescence emission. NCS-1 possesses two tryptophans at positions 30 and 103 (10 -12), and previous work using nonmyristoylated bacterially expressed NCS-1 showed an increase in peak emission due to tryptophan fluorescence when Ca 2ϩ was added (30). With the myristoylated NCS-1, a similar increase in peak emission at about 340 nm (Fig. 2B) was detected at low Ca 2ϩ concentrations (Ͻ1 M), but a second effect, not previously observed, was also seen at higher Ca 2ϩ in which the peak emission declined (Fig. 2C) to the original levels and concomitant peak broadening occurred toward higher wavelengths (Fig. 2B). These data suggest two distinct Ca 2ϩ -dependent conformational changes in NCS-1 consistent with the existence of at least two distinct Ca 2ϩ -binding sites. The Ca 2ϩ dependence of the two conformational changes differed by almost 1000-fold.
Ca 2ϩ -independent Binding of NCS-1 to Brain Membranes-Closely related Ca 2ϩ -binding proteins such as recoverin (19), hippocalcin (23), and neurocalcin (21,22) bind to membranes when myristoylated but not in the nonmyristoylated state. This binding is strictly Ca 2ϩ -dependent, and the endogenous proteins can be stripped from membranes by Ca 2ϩ chelators. We therefore determined whether the recombinant NCS-1 protein would show Ca 2ϩ -dependent interaction with cellular membranes in the presence of physiological (millimolar) Mg 2ϩ concentration. It should be noted that we previously observed that endogenous NCS-1 can be tightly associated with adrenal medulla membranes in the presence of Ca 2ϩ chelator (14), and even after washing with chelator and subsequently with carbonate buffer at pH 11.5, endogenous membrane-associated NCS-1 was still detectable (Fig. 3A). To specifically determine binding of myristoylated NCS-1, the recombinant protein was prepared as a [ 3 H]myristate-labeled protein (Fig. 3B). Incubation of carbonate-washed rat brain membranes with recombinant [ 3 H]NCS-1 resulted in further binding of myristoylated NCS-1 in a Ca 2ϩ -independent manner to the membranes (Fig.  3C). No differences in the extent of binding were seen over the Ca 2ϩ range of 0 -100 M. These data show that NCS-1 has a pattern of membrane interaction distinct from closely related family members and that does not fit the recoverin Ca 2ϩ / myristoyl switch model.

NCS-1 Does Not Directly Stimulate Ca 2ϩ -dependent Exocytosis in Permeabilized Chromaffin
Cells-In order to examine whether NCS-1 exerts its stimulatory effects on Ca 2ϩ -regulated exocytosis directly via interaction with the exocytotic machinery, we used digitonin-permeabilized adrenal chromaffin cells. This is a well characterized model system for the analysis of dense core granule exocytosis, which allows exchange of soluble proteins within the cells and demonstration of stimulatory effects of various added soluble proteins (31,32). The possibility that bacterially expressed NCS-1 might stimulate exocytosis after the addition to permeabilized cells was examined. Initially we determined the cellular content of endogenous NCS-1 following permeabilization. Within 15 min of digitonin permeabilization, about 50% of NCS-1 was lost from the cells, but the remainder was retained in the cells following a longer permeabilization time even in the absence of Ca 2ϩ (Fig. 4A). When permeabilized cells were incubated with 100 g/ml (5 M) bacterially expressed myristoylated NCS-1, increased levels of NCS-1 were detected, after prolonged washing of the cells in either 0 or 10 M Ca 2ϩ , due to retention of the added protein with an approximate 3-fold increase compared with control cells at the end of the incubations (Fig. 4B). For analysis of exocytosis, the cells were permeabilized, preincubated for 15 min with or without NCS-1, and stimulated with 0 or 10 M free Ca 2ϩ . In some experiments, NCS-1 was also included in the stimulation step. The concentration of NCS-1 was based on maximally effective concentrations of the nonmyristoylated protein in published in vitro assays (16). Under no conditions did exogenously added NCS-1 stimulate exocytotic release of catecholamine. No effect was seen on the Ca 2ϩdependence of exocytosis based on a lack of effect at maximal (10 M Ca 2ϩ ) or a low (1.2 M) Ca 2ϩ concentration (Fig. 4C). Similar results were obtained with two separate batches of myristoylated NCS-1 and also with nonmyristoylated His 6tagged protein. These findings argue against a direct effect of NCS-1 on the exocytotic machinery and suggest that its effects in intact cells may be due to regulation of signal transduction pathways.

NCS-1 Does Not Modify Cyclic Nucleotide Levels in Perme-
abilized Chromaffin Cells-Exocytosis induced by Ca 2ϩ can be regulated by cyclic nucleotides, and NCS-1 has been suggested to regulate cyclic nucleotide levels due to a stimulation of cyclic nucleotide phosphodiesterase (16) and frequenin to activate membrane-bound guanylate cyclase (9). We investigated, therefore, whether exogenous NCS-1 would affect cyclic nucleotide levels in permeabilized chromaffin cells. Neither NCS-1 nor Ca 2ϩ alone had any effect on cGMP levels when assayed in the presence of the phosphodiesterase inhibitor IBMX, ruling out any activation of guanylate cyclase (Fig. 5A). The effect of NCS-1 on cAMP levels was assayed in the absence of IBMX to allow examination of effects either on phosphodiesterase or on adenylate cyclase activity. In the absence of IBMX (Fig. 5B) or in its presence (not shown), cAMP levels were elevated due to Ca 2ϩ , but no effect of NCS-1 on cAMP levels in the absence or presence of Ca 2ϩ was seen.
Identification of Putative NCS-1-binding Proteins in Adrenal Subcellular Fractions-The persistent association of NCS-1 with cellular membranes in the presence of Ca 2ϩ chelators suggests that NCS-1 may make Ca 2ϩ -independent interactions with at least some of its target proteins. In order to examine this possibility and to investigate the nature of NCS-1-binding proteins, we established an assay based on the binding of biotinylated NCS-1 to proteins separated by SDS-PAGE. In initial experiments, multiple putative binding partners were detected using biotinylated NCS-1 overlays, and, therefore, to establish the specificity of this binding it was compared with that with a distinct biotinylated protein of similar molecular mass, the myristoylated GTP-binding protein ARF1, also expressed as a recombinant protein (33). In these experiments, the ARF1 was not activated by the addition of GTP and, therefore, was used simply as a control for nonspecific binding of a biotinylated probe. From a comparison of overlays, a series of bands, particularly of less than about 30 kDa, were detected by both probes in a microsome fraction (Fig. 6A) and are, therefore, nonspecific. NCS-1, however, detected a series of specific bands in the absence or presence of Ca 2ϩ (Fig. 6A, left and center panels) not detected by biotinylated ARF1 (Fig. 6A, right panel) particularly in granule membrane and microsome fractions. In addition, other bands were detected by biotinylated NCS-1 only in the presence of 10 M Ca 2ϩ in cytosol, granule membrane, and microsome fractions (Fig. 6A), indicating the presence of specific Ca 2ϩ -dependent as well as Ca 2ϩ -independent NCS-1-binding proteins in adrenal subcellular fractions.
Similar results were found in three separate experiments. We compared binding proteins detected in the absence or presence of 10 M Ca 2ϩ by biotinylated NCS-1 and biotinylated calmodulin. As illustrated for binding to microsomes in 10 M Ca 2ϩ , some bands were detected by both probes, but the two Ca 2ϩbinding proteins also specifically detected distinct proteins (Fig. 6B).
A well characterized calmodulin-binding protein is the 61-kDa subunit of calcineurin (34). NCS-1 has been shown to activate calcineurin in vitro (16), so we compared the Ca 2ϩ dependence of binding of NCS-1 and calmodulin to purified calcineurin in the overlay assay. No binding was detected with the ARF1 control. With calmodulin, binding to calcineurin was completely Ca 2ϩ -dependent (Fig. 7). In contrast, while binding of NCS-1 to calcineurin was increased at 10 M Ca 2ϩ it was also detectable at 0 Ca 2ϩ . Calcineurin may account for one of the polypeptides identified in adrenal fractions at about 60 kDa that did not show marked Ca 2ϩ -dependence of NCS-1 binding (Fig. 6). DISCUSSION The functional importance of frequenin and its orthologue NCS-1 in the regulation of calcium-dependent exocytosis of synaptic vesicles and dense-core granules in vivo has been established (9,11,14), but it is not clear whether overexpression of frequenin/NCS-1 modifies some aspect of membrane excitability, Ca 2ϩ homeostasis, or some other aspect of signal transduction or if it has a direct effect on the exocytotic machinery. For further examination of NCS-1 action and identification of its specific biochemical targets, we have expressed and characterized recombinant myristoylated NCS-1. Following co-expression with yeast N-myristoyltransferase, about 50% of the NCS-1 became myristoylated. This is a lower proportion than for other family members (19,(21)(22)(23) but still considerably higher than for other expressed myristoylated proteins such as the GTP-binding protein ARF (33), for example.
The bacterially expressed NCS-1 protein appeared to be active as a Ca 2ϩ -binding protein based on changed migration on SDS-PAGE in the presence of Ca 2ϩ and Ca 2ϩ -induced changes in tryptophan fluorescence. Previous work on NCS-1 has demonstrated a change in tryptophan fluorescence, due to a Ca 2ϩdependent conformational change, that consisted of an increase in peak emission (30). In contrast, with myristoylated NCS-1, we saw a more complex pattern of changes consistent with two conformational changes of differing Ca 2ϩ sensitivity. Similar differences in the effect of Ca 2ϩ on tryptophan fluorescence for myristoylated versus nonmyristoylated species have been reported for recoverin (18). The initial increase in peak emission occurred over the range of Ca 2ϩ concentration from 0.1 to 1.0 M, and this is consistent with previous work on direct assay of Ca 2ϩ binding to NCS-1. These data have shown binding over this Ca 2ϩ range due to two sites with high cooperativity. NCS-1 contains three potential functional EF-hand domains with an additional N-terminal domain lacking key residues required for Ca 2ϩ coordination (9). The changes in tryptophan fluorescence  7. Binding of biotinylated NCS-1 and calmodulin to calcineurin. Using biotinylated protein overlays, the ability of NCS-1, calmodulin, or ARF1 to bind to purified bovine brain calcineurin (0.1 g/track) in the absence or presence of 10 M Ca 2ϩ was compared. No binding was detected with ARF1 or with calmodulin in the absence of Ca 2ϩ . Binding to the 61-kDa subunit is shown.
we observed at higher Ca 2ϩ concentrations may reflect Ca 2ϩ binding to a third low affinity site not detected by direct Ca 2ϩ binding. Further interpretation of the Ca 2ϩ -induced changes in tryptophan fluorescence will require information on structural change during Ca 2ϩ -binding. Nevertheless, these data confirm that the recombinant myristoylated NCS-1 is a functional Ca 2ϩ -binding protein.
An effect of overexpression of a protein on neurotransmission in vivo as seen for frequenin (9) could be due to effects on many different aspects of presynaptic function. We have previously demonstrated that overexpression of NCS-1 in PC12 cells resulted in an increase in the extent of dense core granule exocytosis evoked by ATP acting on a purinergic receptor (14), indicating that the effect of frequenin/NCS-1 was not limited to synaptic vesicle exocytosis. These data also show that NCS-1 is not normally present in limiting amounts in intact PC12 cells, since increased levels of expression can exert a stimulatory effect. In contrast, when the overexpressing PC12 cells were permeabilized prior to stimulation by calcium, no effect of NCS-1 overexpression on exocytosis was detected. These findings may be consistent with NCS-1 acting on some early step in stimulus-secretion coupling in intact cells that would be bypassed in permeabilized cells. Alternatively, the possibility could not be ruled out that permeabilization resulted in loss of soluble NCS-1, explaining the difference in results between intact and permeabilized cells. We set out, therefore, to test these possibilities by directly manipulating cellular NCS-1 levels after permeabilization.
Digitonin-permeabilized adrenal chromaffin cells have been shown to be a useful model to examine the effects of exogenous soluble proteins on Ca 2ϩ -dependent exocytosis, allowing demonstration of the stimulatory effects of annexin II (35), 14-3-3 proteins (32), ␣-SNAP (36), calmodulin (37), and cAMP-dependent protein kinase (38). Following digitonin permeabilization, about 50% of endogenous NCS-1 leaked from the cells. Preincubation with exogenous NCS-1 resulted in a 3-fold increase in cell-associated NCS-1, demonstrating its ability to enter the cells and remain cell-associated through all of the steps of the exocytosis assay. This association was Ca 2ϩ -independent, consistent with the ability of NCS-1 to bind to membranes in a Ca 2ϩ -independent manner. Using protocols similar to those used successfully with other proteins, we did not see any stimulatory effect of exogenous NCS-1 on exocytosis at low or high Ca 2ϩ concentrations. These results are unlikely to be due to the levels of NCS-1 remaining following permeabilization already being sufficient, since intact cell studies with overexpression clearly show that NCS-1 levels are not saturating in neurons or PC12 cells (9,11,14). These data in combination with previous work on cells permeabilized after transfection (14) argue that NCS-1 does not directly regulate the machinery for Ca 2ϩ -triggered exocytosis but may act on signal transduction pathways that control exocytosis or an early step in stimulus-secretion coupling.
Frequenin has been shown to stimulate membrane-bound photoreceptor guanylate cyclase at low but not high Ca 2ϩ concentrations in rod outer segment membranes (9). It is not known if it can stimulate the forms of guanylate cyclase expressed in neuronal or other cell types. We did not detect any changes in cGMP levels following introduction of NCS-1 into permeabilized chromaffin cells. Similarly, while NCS-1 (16) and the related protein VILIP (38) have been suggested to be potential regulators of cAMP, we did not see any effects on cAMP levels. We had examined these cyclic nucleotides, since both have been shown to modulate dense core granule exocytosis in chromaffin cells (39). It seems unlikely, therefore, that cyclic nucleotide generation is part of the pathway by which NCS-1 overexpression enhances dense core granule exocytosis.
Related members of this family of Ca 2ϩ -binding proteins have been reported to only bind to membranes in the presence of Ca 2ϩ (19,21,22,23,24), consistent with the Ca 2ϩ /myristoyl switch model for recoverin (3). In contrast, one member of the neuronal calcium sensor family, guanyl cyclase-activating protein 2, has been shown to have a reversed Ca 2ϩ dependence in that it binds membranes at low Ca 2ϩ and dissociates as Ca 2ϩ is elevated (24). NCS-1 behaves in a manner distinct from these proteins in that the myristoylated protein shows Ca 2ϩ -independent membrane binding. It is significant that endogenous NCS-1 remains membrane-associated in adrenal medullary fractions (14) and with rat brain membranes, even following extensive washing in the presence of Ca 2ϩ chelators and even with carbonate buffer to extract extrinsic membrane proteins demonstrating a tight Ca 2ϩ -independent membrane interaction. Recent work has shown that the binding of the related protein S-modulin to membranes, which is Ca 2ϩ -dependent, requires the presence of a series of charged residues at the C terminus (40). These charged residues are absent in NCS-1, suggesting a distinct mechanism for membrane interaction. These findings demonstrate, therefore, that NCS-1 makes Ca 2ϩ -independent interactions with target membrane and that Ca 2ϩ does not regulate its membrane association. Presumably, conformational change following Ca 2ϩ binding in membranebound NCS-1 will lead to activation of target membrane proteins. This situation is distinct from the Ca 2ϩ /myristoyl switch model based on the properties of recoverin.
Little information was available on the nature and number of possible target proteins for NCS-1 or other members of this family of Ca 2ϩ -binding proteins or on whether their binding proteins are distinct from or overlap with those for calmodulin. The use of biotinylated NCS-1 has now allowed the detection of several binding proteins in subcellular fractions of adrenal medulla, some of which showed Ca 2ϩ -independent and others Ca 2ϩ -dependent interactions. These were shown to be specific based on lack of detection by biotinylated ARF1. In addition, a distinct pattern of Ca 2ϩ -dependent binding partners was detected using biotinylated calmodulin. The multitude of binding partners detected is consistent with the various biochemical effects of NCS-1/frequenin that have been reported. It has been suggested that NCS-1 overlaps with calmodulin in its target protein interactions (16), but our data on biotinylated NCS-1 binding suggest that NCS-1 is unlikely to act simply as a calmodulin replacement but also has specific target proteins distinct from those for calmodulin. In the case of one common target protein, calcineurin, binding of calmodulin was strictly Ca 2ϩ -dependent, but binding of NCS-1 was not entirely Ca 2ϩdependent. These data suggest, therefore, a distinct mode of action for NCS-1 in which it is constitutively associated with membranes and certain target proteins at resting Ca 2ϩ concentration, allowing it to act as a rapid transducer of Ca 2ϩ signals in response to localized changes in Ca 2ϩ concentration.