Differential Use of Myristoyl Groups on Neuronal Calcium Sensor Proteins as a Determinant of Spatio-temporal Aspects of Ca 2 (cid:1) Signal Transduction*

The localizations of three members of the neuronal calcium sensor (NCS) family were studied in HeLa cells. Using hippocalcin-EYFP and NCS-1-ECFP, it was found that their localization differed dramatically in resting cells. NCS-1 had a distinct predominantly perinuclear localization (similar to trans-Golgi markers), whereas hippocalcin was present diffusely throughout the cell. Upon the elevation of intracellular Ca 2 (cid:1) , hippocalcin rapidly translocated to the same perinuclear compartment as NCS-1. Another member of the family, neurocalcin (cid:2) , also translocated to this region after a rise in Ca 2 (cid:1) concentration. Permeabilization of transfected cells using digitonin caused loss of hippocalcin and neurocalcin (cid:2) in the absence of calcium, but in the presence of 10 (cid:3) M Ca 2 (cid:1) , both proteins translocated to and were retained in the perinuclear region. NCS-1 localization was un-changed in permeabilized cells regardless of calcium concentration. The localization of NCS-1 was unaffected by mutations in all functional EF hands, indicating that its localization was independent of Ca 2 (cid:1) . A minimal myristoylation motif (hippocalcin-(1–14)) fused to EGFP resulted in similar perinuclear of I/ Not all hippocalcin and EYFP sequences the first 14 codons of the hippocalcin sequence, the 10 codons required myristoylation. this construct sequence nonmyristoylatable hippocalcin fusion constructs

Calcium is a widely used intracellular signal that regulates many different cellular processes in a very specific manner. An aspect that has received considerable attention in recent years is the varied use of Ca 2ϩ signal mechanisms that determine spatio-temporal aspects of the changes in Ca 2ϩ concentration (1). The existence of highly localized changes in Ca 2ϩ concen-tration in addition to global Ca 2ϩ changes has been increasingly highlighted, and these local Ca 2ϩ signals are likely to contribute to the specificity of Ca 2ϩ actions (2). The specificity of Ca 2ϩ signaling is also the result, in part, of the existence of many different Ca 2ϩ -binding proteins, which act as Ca 2ϩ sensors in the transduction of Ca 2ϩ signals. Localization of Ca 2ϩ sensors could be a significant factor, and the effect of localization of Ca 2ϩ sensors on signal transduction has been examined for two such proteins. Calmodulin responds to global Ca 2ϩ changes by binding Ca 2ϩ and then subsequently to target proteins; it can also translocate into the nucleus (3)(4)(5). Protein kinase C isoforms also show specific patterns of translocation to the plasma membrane (6). Analysis of the factors that determine the localization of other Ca 2ϩ sensors is likely to be crucial for further understanding of Ca 2ϩ signal transduction, but has so far been little studied.
The neuronal calcium sensor (NCS) 1 proteins are a family of high affinity Ca 2ϩ -binding proteins that can sense Ca 2ϩ elevations above resting Ca 2ϩ concentration in the range of 0.1-1.0 M (7). The NCS family includes proteins expressed only in retinal photoreceptors (recoverin (Ref. 8) and the GCAPs (Ref. 9)) or in neurons, in some cases most highly in particular cell types (hippocalcin (Ref. 10), neurocalcins/VILIPs (Ref. 11)) or with more widespread expression (NCS-1/frequenin (Refs. 12 and 13)). All the NCS family members share four characteristic EF hands and a myristoylated N terminus. In the NCS proteins, three (or only two in some family members) of the four EF hands are capable of binding calcium and the EF hand nearest the N terminus (EF1) is altered so that it cannot bind Ca 2ϩ (7). The myristoyl tail is a fatty acid post-translationally added to the NCS protein. Its function has been extensively studied in recoverin (14). In recoverin, the myristoyl group is buried in a nonpolar pocket formed by EF1 and the surrounding region in the Ca 2ϩ -free state (15). When recoverin binds calcium, there is a large conformational change that ejects the myristoyl group from its pocket (16). This exposed hydrophobic tail is then free to interact with the nonpolar cell membranes or hydrophobic protein domains. This exposure of the myristoyl tail on calcium binding is referred to as a "calcium/myristoyl switch" and considered to be primarily a mechanism by which the protein can translocate from the cytosol to intracellular membranes (14,16). The structural features required for the Ca 2ϩ /myristoyl switch are apparently conserved in all NCS proteins. The presence of this Ca 2ϩ -mediated membrane-association process was, therefore, suggested for the other members of the neuronal calcium sensor protein family (16). There are biochemical data from disrupted cells for some of the family members in support of this mechanism (17,18). A clear understanding of the localization and translocation of these proteins in response to calcium will allow the generation of models of how these proteins modulate and integrate calcium signals. There has, however, been little in vivo work to demonstrate the Ca 2ϩ /myristoyl switch during elevation of Ca 2ϩ concentration within live cells, and so kinetic aspects of the proposed membrane translocation are unknown.
In this study we have examined the potential Ca 2ϩ -myristoyl switch in three NCS proteins. Hippocalcin is expressed most highly in hippocampal neurons (19) and the closely related neurocalcin ␦ (VILIP-3) in cerebellar Purkinje cells (20,21). Biochemical studies with these proteins suggest Ca 2ϩ -dependent membrane interactions (17,18), but this has not been examined within intact cells. The functions of these two proteins are unknown. In contrast, NCS-1 has been shown to play crucial roles in learning and memory in Caenorhabditis elegans (22), in the release of neurotransmitters (12) and hormones (23,24), and in the regulation of voltage-gated Ca 2ϩ channels (25,26). In addition, the yeast orthologue of NCS-1 is essential for survival as a regulator of the PI 4-kinase, Pik1 (27). Unlike the other NCS proteins, NCS-1 is found outside of the nervous system (23,28) and may be a regulator of mammalian Golgiassociated PI 4-kinase ␤ in non-neuronal cells (29). Biochemical data suggest that, in contrast to other NCS proteins, NCS-1 can bind to isolated membranes in the absence of Ca 2ϩ (24). We have prepared enhanced GFP-variant constructs of these proteins to compare their Ca 2ϩ /myristoyl switch mechanisms. We demonstrate different uses of N-terminal myristoylation, which would generate distinct spatio-temporal Ca 2ϩ sensing by members of the NCS family and have assessed the kinetics of the Ca 2ϩ /myristoyl switch-dependent membrane translocation in living cells.

MATERIALS AND METHODS
Plasmids-The EYFP-tagged hippocalcin fusion construct (pHippo-EYFP) was made by the in-frame insertion of a hippocalcin sequence, amplified from rat brain cDNA, into the pEYFP-N1 vector (CLONTECH). The primers contained endonuclease sites (underlined) to facilitate this cloning. The primers were based on the rat nucleotide sequence (GenBank™ accession no. NM017122). The sense primer was designed to the 5Ј-untranslated region from Ϫ165 to Ϫ135 (5Ј-GGCCG-GCTAGCTCTTTTTGGGTCAAATGAG-3Ј; NheI) and the antisense primer with a mutated stop codon to the region ϩ735 to ϩ765 (5Ј-CCTCTCACTCGAGAACTGGGAAGCGCTGCT-3Ј; XhoI). The amplified sequence digested with the endonucleases NheI/XhoI was cloned into the vector pEYFP-N1 digested with the endonucleases NheI/SalI using standard methods. The ECFP-tagged NCS-1 fusion construct (pNCS-1-ECFP) was made by the in-frame insertion of a NCS-1 sequence, amplified from the pNCS-1 plasmid (23), into the pECFP-N1 vector (CLONTECH). The primers contained endonuclease sites (underlined). The sense primer was designed to the region from 0 to ϩ30 (5Ј-ATACGGTACCATGGGGAAATCCAACAGCAAG-3Ј; KpnI) and the antisense primer to the region from ϩ492 to ϩ522 (5Ј-ATGCGGTAC-CAATACCAGCCCGTCGTAGAGGG-3Ј; KpnI). The amplified sequence digested with KpnI was inserted into the vector pEYFP-N1 digested with KpnI using standard methods.
The sequence of all these constructs were confirmed by automated sequencing by Oswel (Southampton, United Kingdom (UK)).
Culture and Transfection of HeLa Cells-HeLa cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen, Paisley, UK) containing 5% fetal calf serum (Invitrogen) and 1% nonessential amino acids (Invitrogen), incubated at 37°C in an atmosphere of 5% CO 2 and plated onto glass coverslips in a 24-well plate at 40,000 cells/well. The cells were allowed 4 -24 h to adhere before they were transfected. The transfection reaction mixture contained 93 l of Dulbecco's modified Eagle's medium (Invitrogen), 3 l of FuGENE™ (Roche), and 4 l of plasmid DNA (250 g/l). This was incubated at room temperature for 30 min before being added dropwise to HeLa cells in a 24-well plate. The cells were maintained for 8 -96 h before being used in experiments.
Immunofluoresence Staining-Anti-NCS-1 was prepared in rabbits as described previously (23). Anti-neurocalcin ␦ was prepared as a rabbit antiserum using a similar protocol in which rabbits were immunized with purified recombinant neurocalcin ␦ expressed in Escherichia coli and affinity-purified on immobilized recombinant neurocalcin ␦. Anti-␥-adaptin was from Sigma (Poole, UK), anti-GFP from CLONTECH (Basingstoke, UK), and anti-transferrin receptor from Roche (East Sussex, UK). Cells attached to coverslips were washed twice in phosphate-buffered saline (PBS, 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM NaH 2 PO 4 ), and fixed in PBS containing 4% formaldehyde for 30 min. The cells were washed twice in PBS and incubated in blocking buffer (0.3% bovine serum albumin, 0.5% Triton X-100 in PBS) for 1 h. The blocking buffer was removed and the primary antibody added at the appropriate dilution (anti-␥-adaptin 1 in 100, anti-GFP 1 in 400, anti-transferrin receptor 1 in 400, anti NCS-1 1 in 1,000, and anti neurocalcin ␦ 1 in 100). The primary antibody was incubated for 1-2 h and removed, and the cells were washed three times in blocking buffer. The cells were then incubated for an additional 1 h with the appropriate biotinylated secondary antibody diluted to 1 in 100, (anti-rabbit for anti-NCS-1 and anti-neurocalcin ␦ primary antibodies and anti-mouse for the other primary antibodies; Amersham Biosciences, Buckinghamshire, UK). The cells were washed three times and then incubated in streptavidin-Texas Red (Amersham Biosciences) diluted 1 in 50 for 30 min. The cells were washed three times, and the coverslips were dried and mounted on antifade glycerol (glycerol/PBS 9:1, 0.25% diazabicyclo(2-2-2)octane, 0.002% p-phenyldiamine). The cells were examined on a Zeiss Universal microscope and fluorescence imaged using the following standard Zeiss filter sets: ECFP BG3ϩLP400, FT460, LP495; EYFP, BP450 -490, FT510, BP515-565; Texas Red, BP546/12, FT850, LP590.
Polyacrylamide Gel Electrophoresis and Western Blotting-The cells of each well in a 24-well plate were lysed using 100 l of Laemmli buffer (Sigma, Dorset, UK)/well and boiled for 10 min. The samples were loaded onto and separated using a 15% SDS-polyacrylamide gel and then blotted onto nitrocellulose by transverse electrophoresis. The nitrocellulose membranes were washed twice in PBS and placed in blocking buffer (PBS containing 5% bovine serum albumin, 5% fetal calf serum, 5% Marvel™ milk powder, 0.5% Tween 20™) for 1 h. The PBT was removed and the primary antibody added at the appropriate dilution in PBS, 3% Marvel™ milk powder (anti-NCS-1 at 1in 1000, affinity-purified anti-neurocalcin ␦ at 1 in 400, and anti-GFP at 1 in 400 from CLONTECH, Basingstoke, UK). The primary antibody was incu-bated for 1-2 h and removed, and the membranes were washed three times. The membranes were then incubated for an additional 1 h with the appropriate horseradish peroxidase-conjugated secondary antibody diluted to 1 in 400 with PBT, (anti-rabbit for anti-NCS-1 and antineurocalcin ␦ primary antibodies and anti-mouse for anti-GFP primary antibodies; Sigma). The membranes were washed once in PBT, three times in PBS, and once with distilled water. The signal was detected using enhanced chemiluminescence (Amersham Biosciences).
Stimulation of Cells with Ionomycin and Imaging of Fixed Cells-Transfected HeLa cells were washed three times in Krebs-Ringer buffer (145 mM NaCl, 5 mM KCl, 1.3 mM MgCl 2 , 1.2 mM NaH 2 PO 4 , 10 mM glucose, 20 mM HEPES, pH 7.4). The cells were then incubated for 10 min in 1 M ionomycin in Krebs-Ringer buffer in the presence or absence of 3 mM CaCl 2 at 37°C. The Krebs-Ringer buffer was removed, and the cells were fixed in 4% formaldehyde for immunofluorescence staining and imaged using a Zeiss Universal microscope fitted with a Nikon Coolpix 995 digital camera.
Confocal Laser Scanning Microscopy on Living Cells-For confocal laser scanning microscopy, live transfected HeLa cells were examined with a Leica TCS-SP-MP microscope (Leica Microsystems, Heidelberg, Germany) using a 166.27-m pin-hole and a 63ϫ water immersion objective with a 1.2 numerical aperture. For optimal imaging of the spatial distribution of hippocalcin-EYFP, the cells were excited at 514 nm and light collected at 545-625 nm and fluorescence images were collected every 1.86 s with alternate phase-contrast images to check for changes in cell morphology. For dual imaging of hippocalcin-EYFP and NCS-1-ECFP, the cells were excited at 514 nm with light collected at 560 -600 nm for EYFP or excited at 458 nm for and light collected at 465-500 nm for ECFP detection. In experiments on the kinetics of Ca 2ϩ -dependent translocation, cells were loaded with Fura Red (Molecular Probes) by incubation in 5 M Fura Red in growth medium for 30 min. The cells were then excited at 488 nm and light collected at 625-725 nm for Fura Red emission and at 525-590 for hippocalcin-EYFP emission and images collected every 137 ms. Ionomycin was added to the bath solution a final concentration of 3 M.
Permeabilization Experiments-Transfected HeLa cells were washed three times in Krebs-Ringer buffer and incubated for 15 min in 10 M digitonin in 300 l of permeabilization buffer (139 mM potassium glutamate, 20 mM PIPES, 5 mM EGTA, pH 6.5) in the presence or absence of 10 M free calcium at 37°C. If the cells were to be fixed, the buffer was removed and 500 l of 4% formaldehyde was added per well. For the leakage experiments, the permeabilization buffer was removed and centrifuged briefly at 12,000 rpm to sediment any detached cells. The proteins were concentrated by precipitation with cold methanol (30) and resuspended in 100 l of Laemmli buffer (Sigma). The cells remaining in the wells were solubilized in Laemmli buffer. All samples were boiled for 10 min and then used for polyacrylamide gel electrophoresis and Western blotting.

Expression and Localization of NCS Proteins in HeLa
Cells-To determine the localization of the neuronal calcium sensing proteins NCS-1 and hippocalcin in living cells, constructs were made that fused the C terminus of these proteins to the fluorescent proteins ECFP and EYFP to produce the plasmids pNCS-1-ECFP and pHippo-EYFP, respectively. Vec- , or an anti-neurocalcin ␦ antibody with cross-reactivity for hippocalcin (D). The samples were lysates from cells transfected with pNCS-1-ECFP, pHippo-EYFP, pECFP-N1, and pEYFP-N1. A polypeptide was detected with both the anti-GFP and anti-NCS-1 antibodies at the expected size (60 kDa) in cells transfected with pNCS-1-ECFP. A polypeptide was also detected with both the anti-GFP and anti-neurocalcin ␦ antibodies at the expected size (60 kDa) in cells transfected with pHippo-CFP. This demonstrated that both fusion proteins were being expressed correctly. For both the controls, pEYFP-N1 and pECFP-N1, the only band detected with any of the antibodies was a 27-kDa band, corresponding to EYFP or ECFP, using the anti-GFP antibody. tor-derived sequence produced a 10-or 13-amino acid linker between the NCS-1 protein and hippocalcin, respectively, and the fluorescent protein. These constructs were transfected into HeLa cells, which do not express detectable levels of the proteins endogenously, and the expressed proteins observed in living and fixed cells. HeLa cells were chosen as a convenient cell type to allow examination of the basic localization and translocation mechanism of these proteins within a cellular environment. To demonstrate their correct expression, Western blots were carried out on samples from transfected cells. Fig. 1 shows that the expressed fusion proteins were of the expected size (ϳ60 kDa) and recognized by both anti-GFP and specific antibodies. Anti-NCS-1 was used for cells transfected with pNCS-1-ECFP. Expression of Hippo-EYFP could be demonstrated using an antiserum prepared against the closely related protein neurocalcin ␦ (96% identical (Ref . 7)). The 60-kDa band representing the fusion proteins was not seen in samples from cells transfected with the control vectors pEYFP-N1 and pECFP-N1 but instead a polypeptide band corresponding to the free fluorescent proteins. Fig. 2 shows the localization of proteins encoded by pNCS-1-ECFP, pHippo-EYFP, and constructs expressing the free fluorescent proteins ECFP and EYFP. The location patterns were not dependent upon expression levels of the proteins and were indistinguishable in cells transfected for between 8 and 96 h and with varying levels of fluorescence intensity at each time point. The free fluorescent proteins were found in the cytoplasm but were particularly concentrated in the nucleus (Fig.  2, A and B). The hippocalcin-EYFP fusion protein was present in the cytoplasm and also in the nucleus but to a lesser extent than ECFP or EYFP (Fig. 2C). The NCS-1 fusion protein, in contrast, had a radically different distribution. The fluorescence of NCS-1-ECFP was at its most intense in a perinuclear region to one side of the nucleus, whereas the nucleus was devoid of fluorescence (Fig. 2D). Additionally, NCS-1 ECFP was also associated with plasma membrane, usually in a patchy distribution. The presence of the ECFP and EYFP tags did not influence the localization of the NCS proteins, as sim-ilar localization patterns were seen with wild-type protein detected by immunofluorescence (shown below). In addition, similar localization patterns were seen in living and fixed cells. Previous work has shown that endogenous NCS-1 in COS cells is also localized to a perinuclear compartment that overlaps partially with the trans-Golgi network marker ␥-adaptin (28). Fig. 3 (A-C) shows that ECFP-NCS-1 also partially co-localizes with ␥-adaptin. In contrast, ECFP-NCS-1 did not significantly co-localize with the transferrin receptor used as a marker of endosomal compartments (Fig. 3, D-F).
Effect of Ca 2ϩ Elevation on Localization of NCS Proteins-The effect of a global elevation in intracellular Ca 2ϩ concentration on the localization of the NCS fusion proteins was assessed by treating cells with 1 M ionomycin in the presence of external Ca 2ϩ . The NCS-1 fusion protein did not change its localization in response to an increase in intracellular Ca 2ϩ levels (Fig. 4, A and B). This suggests that, at both resting and raised levels of Ca 2ϩ , NCS-1 has its myristoyl tail exposed allowing membrane association. In contrast, following Ca 2ϩ elevation, the hippocalcin fusion protein translocated from its apparent diffuse cytosolic localization to a predominantly perinuclear pattern (Fig. 4, C and D). This perinuclear localization appeared to be very similar to that seen with the NCS-1 fusion protein at resting Ca 2ϩ levels. A more diffuse localization was also present, spreading out to the cell periphery consistent with an additional patchy plasma membrane localization. The Ca 2ϩdependent translocation of hippocalcin-EYFP was rapidly reversed when the external buffer was exchanged for a buffer without added Ca 2ϩ (data not shown). In cells transfected with a construct expressing an untagged version of neurocalcin ␦, another member of the NCS family that is closely related to hippocalcin (7), a similar diffuse localization was seen in unstimulated cells using immunofluorescence to detect the protein in fixed cells (Fig. 4E). Translocation of neurocalcin ␦ to the perinuclear region plasma membrane structures was also observed following an increase in intracellular Ca 2ϩ by ionomycin treatment (Fig. 4F). The apparent translocation of hippocalcin and neurocalcin ␦ from the cytosol to membranes is consistent with the proposed Ca 2ϩ /myristoyl switch mechanism and suggests that these proteins only become membrane-associated at cytosolic Ca 2ϩ concentrations above those in resting cells.
In fixed cells the localization of hippocalcin-EYFP after ionomycin treatment was similar to the perinuclear localization of NCS-1-ECFP. The extent of co-localization of the two proteins in living cells was examined directly using laser scanning confocal microscopy of cells transfected to express both hippocalcin-EYFP and NCS-1-ECFP. In resting cells, hippocalcin-EYFP showed a diffuse localization (Fig. 5A). In contrast, NCS-1-ECFP was associated with the perinuclear compartment and also had a patchy distribution on the plasma membrane (Fig.  5B). Following treatment with ionomycin for 2 min, hippocalcin-EFYP was observed to have translocated (Fig. 5D). A pro-portion of hippocalcin-EYFP remained in the nucleus but in the rest of the cell showed almost complete co-localization with NCS-1-CFP (Fig. 5, D-F).
The data from intact cells on the effects of raising Ca 2ϩ with ionomycin are consistent with membrane association of NCS-1-ECFP even at low Ca 2ϩ concentration and Ca 2ϩ -dependent translocation of cytosolic hippocalcin-EYFP and neurocalcin ␦ to membranes. To confirm this interpretation, the localization of the proteins was examined after digitonin permeabilization to allow soluble proteins to leak out of the cells. Transfected cells were permeabilized with 10 M digitonin in a buffer containing either 0 or 10 M free Ca 2ϩ . At 10 M free Ca 2ϩ concentration, the hippocalcin fusion protein translocated to the perinuclear compartment as seen with ionomycin treatment (Fig. 6). Cells transfected with the pNCS-1-ECFP also retained their perinuclear fluorescence after permeabilization in 10 M Ca 2ϩ . These results showed that the perinuclear localization was because of the fusion proteins associating with intracellular structures. When transfected cells were permeabilized in a buffer containing 0 Ca 2ϩ , essentially all fluorescence was rapidly lost from pHippo-EYFP-transfected cells except a small amount that remained trapped in the nucleus. The hippocalcin fusion protein was, therefore, essentially completely cytosolic at 0 Ca 2ϩ levels and free to diffuse from the permeabilized cells. Permeabilization of pNCS-1-CFP-transfected cells in 0 Ca 2ϩ buffer did not affect the fluorescence pattern. The NCS-1 fusion protein remained attached to the perinuclear membrane construct regardless of the presence or absence of Ca 2ϩ (Fig. 6). The data from examination of fluorescence were confirmed by biochemical analysis using Western blotting of cells and medium containing leaked proteins (Fig. 7). The majority of the NCS-1-ECFP was retained in the permeabilized cells irrespective of Ca 2ϩ concentration (Fig. 7). In contrast, the majority of hippocalcin-YFP leaked into the medium at 0 Ca 2ϩ but the protein was fully retained in cells at 10 M free Ca 2ϩ (Fig. 7). Retention was half-maximal at ϳ0.2-0.3 M free Ca 2ϩ (data not shown). Similar Ca 2ϩ -dependent translocation to the perinuclear region (Fig. 6, E and F) and Ca 2ϩ -dependent retention in permeabilized cells (Fig. 7) was seen in cells transfected with untagged neurocalcin ␦ with expression detected using anti-neurocalcin ␦ antiserum. The extent of retention of the NCS proteins in the presence of Ca 2ϩ (Fig. 7) suggested that the majority of the expressed proteins were myristoylated.
Requirement for Ca 2ϩ Binding and Myristoylation for Perinuclear Localization of NCS Proteins-The data so far showed that localization of NCS-1-ECFP to the perinuclear compartment was independent of Ca 2ϩ concentration in intact and permeabilized cells. To confirm that Ca 2ϩ binding was not required for the localization of NCS-1, use was made of constructs encoding untagged wild-type NCS-1 and a version with all three functional EF hands (EF2-4) mutated to eliminate Ca 2ϩ binding. These mutations (E84Q, E120Q, and E168Q) replaced conserved acidic residues in the EF hands required for Ca 2ϩ coordination (18). Cells were fixed after permeabilization, and NCS-1 localization was determined by immunofluorescence staining. The untagged wild-type protein was found to be localized to the perinuclear compartment in the presence or absence of Ca 2ϩ (Fig. 8, A and B) as previously seen for NCS-1-ECFP. The same localization pattern was seen with the NCS-1(EF2-4) mutant, indicating a complete lack of dependence on Ca 2ϩ binding for the localization of NCS-1 (Fig. 8, C and D).
To determine whether myristoylation of the NCS proteins was required for their localization and translocation, constructs were made with the second glycine residue, essential for myristoylation (31), mutated to an alanine in both the NCS-1 and hippocalcin fusion proteins. Fig. 9 shows that intact cells transfected either with pHippo(G2A)-EYFP or pNCS-1(G2A)-ECFP showed a diffuse fluorescence pattern compatible with cytosolic localization of the proteins (Fig. 9, A and C). No indication of translocation was seen in response to elevated intracellular Ca 2ϩ levels. In addition, when the cells were permeabilized, even in the presence of 10 M free Ca 2ϩ , the unmyristoylated fusion proteins rapidly leaked out of the cells with only low levels of fluorescence remaining in the nuclei (Fig. 9, B and D). These data indicate that myristoylation was essential for the perinuclear localization of NCS-1 and for Ca 2ϩdependent translocation of hippocalcin.
To determine the degree to which the exposure of the myristoyl tail was sufficient for perinuclear membrane localization, a construct was made that fused the myristoylation sequence taken from hippocalcin (hippocalcin- (1-14)) to EGFP. This was expressed in HeLa cells, and its localization is shown in Fig. 9E. The distribution of the hippocalcin-(1-14)-EGFP protein appeared to be much the same as the perinuclear localization of the NCS proteins. There was strong perinuclear fluorescence together with some possible plasma membrane localization. This indicates that the myristoyl tail can be the sole determinant for intracellular targeting of these proteins. Using the permeabilization assay in 0 Ca 2ϩ solution, it was found that the hippocalcin-(1-14)-EGFP remained in the cells, concentrated in the perinuclear region (Fig. 9F). This demonstrated that the myristoyl group accounted for all the intracellular targeting seen with the NCS-1 and hippocalcin fusion proteins but not the Ca 2ϩ dependence of hippocalcin localization.
Kinetics of the Ca 2ϩ /Myristoyl Switch in Living Cells-To examine the kinetics of the Ca 2ϩ -dependent translocation of hippocalcin-EYFP, live cells were imaged using confocal laser scanning microscopy and fluorescence was monitored following addition of ionomycin (Fig. 10). Cells initially showed a diffuse fluorescence indicative of cytosolic and nuclear localization. Within a few seconds of ionomycin addition, translocation of hippocalcin-EYFP occurred with the appearance of a bright punctate spots with a perinuclear localization. In addition, hippocalcin-EYFP also appeared to a lesser extent associated with the plasma membrane in distinct patches (arrows in the last frame of Fig. 10), confirming the earlier interpretation from epifluorescence microscopy. The time course of translocation was similar to that previously seen for the Ca 2ϩ -dependent C 2 domain of PKC␥ (6) or the pH domain of PLC␦ (32). Translocation of hippocalcin-EYFP was also observed

FIG. 5. Hippocalcin-EYFP and NCS-1-ECFP show similar localizations in co-transfected HeLa cells stimulated with ionomycin.
HeLa cell were co-transfected to express both hippocalcin-EYFP and NCS-1-ECFP and living cells examined using confocal laser scanning microscopy. Before ionomycin treatment hippocalcin-EYFP (A) was diffusely distributed, whereas NCS-1-ECFP (B) was concentrated in the perinuclear compartment and on plasma membrane patches with little overlap with hippocalcin-EYFP (C). After treatment with ionomycin for 2 min, hippocalcin-ECFP (D) and NCS-1-ECFP (E) were both localized to the same regions except that hippocalcin-EYFP was also seen in the nucleus. Extensive co-localization was evident in the overlapped image (F). The scale bar represents 20 m. under more physiological conditions following stimulation with histamine and was reversible several minutes after the initial translocation. 2 The relationship of kinetics of the Ca 2ϩ /myristoyl switch for membrane translocation to changes in cytosolic Ca 2ϩ concentration was examined in cells expressing hippocalcin-EYFP and loaded with the fluorescence Ca 2ϩ indicator Fura Red. Dual imaging using confocal microscopy indicated that changes in hippocalcin-EYFP localization could be measured in parallel with an expected decrease in Fura Red emission because of Ca 2ϩ elevation following ionomycin addition. Measurement at high time resolution (1 frame/137 ms) allowed a comparison of the time course of hippocalcin-EYFP translocation and changes in cytosolic Ca 2ϩ concentration (Fig. 11). Cytosolic Ca 2ϩ concentration was monitored over the whole cell. Translocation was assessed by monitoring the fluorescence of 9 spots/cell where translocation occurred and averaging the time courses for each cell. Cytosolic Ca 2ϩ concentration began to increase in a homogeneous manner over the whole cell immediately upon ionomycin addition. In contrast, hippocalcin-EYFP translocation began only after a lag period of ϳ2 s (2.17 Ϯ 0.13 s from 4 cells). The time course of translocation, which then followed, could be fitted with a single hyperbolic function, indicating a first order reaction, with a half-time of ϳ12 s.

DISCUSSION
It has become increasingly apparent that the control of various cellular activities involves proteins that either are tightly co-localized in signaling complexes or can translocate in a dynamic way into such complexes (33,34). Studies on protein dynamics have been complemented by increasing information on the differential use of local versus global Ca 2ϩ signals in the regulation of cell function (2). In contrast, relatively little is known about the dynamics and localization of Ca 2ϩ sensor proteins apart from calmodulin and protein kinase C (3-6). Here we have examined the localization of three members of the NCS family of Ca 2ϩ sensors and demonstrate that, under the same conditions in the same cellular context, their N-terminal myristoyl groups are used in quite different ways to determine their localization. We also provide data on the kinetics of the Ca 2ϩ -myristoyl switch mechanism in living cells. The classic example of the NCS proteins has been the well characterized retinal protein recoverin, which is believed to show Ca 2ϩ -dependent translocation to membranes via a Ca 2ϩ /myristoyl switch mechanism based on biochemical and structural analyses (14 -16), although this has not been confirmed in living cells. The differences in behavior we observed for other NCS proteins are surprising, as the features within the sequence of recoverin that determine the Ca 2ϩ /myristoyl switch mechanism including the inactivated first EF hand and hydrophobic residues that cradle the myristoyl group are highly conserved in all of the NCS proteins (15). It has recently been shown, however, that residues within EF1 of GCAP-2 are crucial for binding to guanylyl cyclase (35). The inactivated EF hand may therefore be more generally important in NCS proteins for target protein interactions.
In HeLa cells, NCS-1 was constitutively associated with, and hippocalcin and neurocalcin ␦ could become associated with, membranes of a perinuclear compartment overlapping the trans-Golgi network (shown using ␥ adaptin as the marker). All three proteins also appeared to associate with the plasma membrane. This localization was not dependent on the presence of the fluorescent tags and is likely to be the result of the preference of the myristoyl group for cholesterol-and sphingolipid-rich domains (36). Although hippocalcin and neurocalcin ␦ are expressed only in neurons, NCS-1 is expressed in many non-neuronal cell types (23,28,37). Association with a perinuclear compartment overlapping the trans-Golgi was reported for endogenous NCS-1 in COS-7 cells (28). Importantly, studies on NCS-1 in fixed brain slices indicated by the use of electron microscopy that NCS-1 was present on cisternae of the trans-Golgi complex, in addition to its presence in both pre-and post-synaptic structures (38). The Ca 2ϩ dependence of the localization of NCS-1 to the trans-Golgi was not determined in these studies, and so it remained possible that resting cytosolic Ca 2ϩ concentration was sufficient to activate a Ca 2ϩ /myristoyl switch in NCS-1 as it has a high affinity for Ca 2ϩ (24,39). We have demonstrated here that NCS-1 association with the perinuclear compartment is independent of free Ca 2ϩ and that NCS-1 does not need functional Ca 2ϩ -binding motifs for this localization. The presence of NCS-1 on the trans-Golgi network may be related to its regulation of PI 4-kinase ␤ (28, 29), which HeLa cells transfected with pHippo-EYFP were imaged using confocal microscopy, and the panels show a compilation of timelapse images at the times indicated from 0 to 68.4 s. In this series, frames were taken every 1.9 s to give optimal spatial resolution. Addition of ionomycin at time 0 was followed by translocation of hippocalcin-EYFP to perinuclear vesicular structures and to patches on the plasma membrane (arrows in the last frame). The scale bar represents 20 m.
has the same localization. The ability of other NCS proteins such as hippocalcin and neurocalcin ␦ to interact with PI 4-kinase ␤ has not been examined, and their localization in neurons has only been investigated after fixation without elevation of intracellular Ca 2ϩ . The presence of NCS-1 on the plasma membrane that we have observed would be consistent with the known involvement of NCS-1 in Ca 2ϩ channel regulation (25,26).
The membrane association of NCS-1 and hippocalcin in HeLa cells was entirely dependent upon myristoylation. In addition, a minimal myristoylation sequence (hippocalcin- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)) was sufficient for localization to the perinuclear compartment. The latter result is compatible with previous work on the targeting of constructs bearing the myristoylation sequence of Src family kinases (40). The difference between the NCS proteins that we investigated, however, was that NCS-1 was membrane-associated even at nominally 0 Ca 2ϩ , whereas hippocalcin and neurocalcin ␦ required Ca 2ϩ elevation above resting levels to translocate from the cytosol to a membrane-associated pool. The membrane association of NCS-1 additionally did not require the presence of any functional EF hand motifs. These data suggest that, in NCS-1, the myristoyl group is exposed even in the Ca 2ϩ free form but that hippocalcin and neurocalcin ␦ have a classic recoverinlike Ca 2ϩ /myristoyl switch. Their myristoyl group would, therefore, be sequestered in the Ca 2ϩ -free form and then exposed in the Ca 2ϩ -bound form to allow the proteins to become membrane-associated. The findings on NCS-1 are consistent with biochemical data showing that recombinant FIG. 11. Kinetics of changes in cytosolic Ca 2؉ concentration and hippocalcin-EYFP translocation. HeLa cells transfected to express hippocalcin-EYFP were loaded with Fura Red and imaged by confocal microscopy. To provide optimal time resolution, frames were taken every 137 ms. The decrease in Fura Red fluorescence (top) began as soon as ionomycin was added. The lower panel shows averaged data from 9 regions of translocation in the same cell. The standard errors of the means are not shown for clarity but did not exceed Ϯ4% of the mean values. Translocation began after a 2-s lag and then followed first order kinetics (as shown by the fitted curve) with a half-time of 11.6 s. myristoylated NCS-1 became bound to membranes in the absence of Ca 2ϩ (24). In addition, analysis of the yeast orthologue FRQ1 by NMR suggested that the myristoyl group might always be solvent-exposed (41). This idea could be directly confirmed by x-ray crystallography, but so far the only structure solved is for nonmyristoylated NCS-1 in its Ca 2ϩ -bound form (28). The study of yeast FRQ also showed, from a biochemical analysis of cell fractions, that some Ca 2ϩdependent membrane association occurred even in the absence of myristoylation (41). In contrast, in intact or permeabilized HeLa cells, NCS-1 localization to the perinuclear membrane compartment was completely dependent upon its myristoylation. This may reflect differences in behavior of the proteins when examining them in a cellular context or in cell extracts and demonstrates the importance of carrying out these studies on proteins within cells.
The different usage of the myristoyl group in the NCS proteins is likely to be important for differential aspects of Ca 2ϩ signal transduction. EF hand motifs in Ca 2ϩ sensors bind Ca 2ϩ very rapidly with binding limited only by the rate of Ca 2ϩ diffusion (42). The off-rate is also rapid in Ca 2ϩ sensors but can be slowed in Ca 2ϩ -sensor and Ca 2ϩ -buffer proteins by changes in the residue at position 9 of the EF hand (42). This allows fine-tuning of the kinetics of activation and time scale of activity of EF hand proteins. The NCS proteins show the properties of fast Ca 2ϩ sensors with rapid on-and off-rates of Ca 2ϩ binding and inactivation occurring on a millisecond time scale (43,44). It would be predicted that NCS-1, with an affinity of 0.3 M, would be able to react to Ca 2ϩ concentration changes in a time course of no more than milliseconds and retain its bound Ca 2ϩ with a time constant of ϳ20 ms when Ca 2ϩ concentration falls (see similar calculations for GCAP1 (Ref. 44)). In contrast, hippocalcin and neurocalcin ␦ would respond to more global Ca 2ϩ signals and only be able to exert their regulatory effects at target membranes after translocation from the cytosol. This would also require the Ca 2ϩ elevation to last over seconds and could preclude activation of hippocalcin and neurocalcin ␦ by brief localized Ca 2ϩ pulses. NCS-1 appears, therefore, to be a protein that would be able to sense and transduce transient and local Ca 2ϩ signals occurring close to the membranes to which it is attached on a millisecond time scale, whereas other NCS proteins such as hippocalcin and neurocalcin ␦ would have a different spatio-temporal pattern of activation requiring longer lasting Ca 2ϩ elevations.
The dual imaging of both hippocalcin-EYFP localization and cytosolic Ca 2ϩ concentration in living cells provides information on the kinetics of the Ca 2ϩ /myristoyl switch and membrane translocation in vivo. Hippocalcin and neurocalcin ␦ have high affinities for Ca 2ϩ when assayed in vitro (17,18), and it is conceivable that membrane association could occur already at resting cytosolic Ca 2ϩ concentration. We have shown, however, that membrane association of hippocalcin-EYFP does require Ca 2ϩ elevation above resting levels, and that this occurs with a half-time of ϳ12s and follows a short (2 s) lag period after Ca 2ϩ elevation. The translocation kinetics were consistent with a first-order reaction. These results are consistent with a rapid conformational change in the protein on Ca 2ϩ binding followed by a diffusion-limited translocation to membranes. The time course over which hippocalcin-EYFP accumulates indicates that it would function in Ca 2ϩ signal transduction over seconds and thus be a considerably slower Ca 2ϩ sensor in vivo than NCS-1.
NCS-1 is the evolutionarily most ancient of the NCS proteins being represented in yeast (27). An additional form most similar to neurocalcin ␦ is present in nematodes, and other NCS proteins appeared later in evolution (7). It seems that an event must have occurred during evolution whereby a membrane-anchored protein NCS-1 gave rise to new forms in which the myristoyl group is sequestered in the Ca 2ϩ -free form of the protein and thus used in a radically different way. It is likely that further alterations in the use of the myristoyl group of NCS protein has arisen; it is known, for example, from biochemical assays that the photoreceptor protein GCAP-2 shows a reversed Ca 2ϩ -dependent membrane association and dissociates from membranes as Ca 2ϩ is elevated (45). This family of Ca 2ϩ sensors, therefore, provides an intriguing example of how a single post-translational modification, N-terminal myristoylation, has been adapted during evolution to allow the appearance of a diverse family of proteins able to mediate distinct spatial and temporal patterns of Ca 2ϩ signal transduction.