Interaction of neuronal calcium sensor-1 and ADP-ribosylation factor 1 allows bidirectional control of phosphatidylinositol 4-kinase beta and trans-Golgi network-plasma membrane traffic.

We have identified a novel Ca(2+)-dependent interaction between neuronal calcium sensor-1 (NCS-1) and the GTPase ARF1. Both of these proteins are localized to the Golgi complex, and both regulate phosphatidylinositol 4-kinase IIIbeta (PI(4)Kbeta). Spatial and temporal control of phosphatidylinositol 4-phosphate levels through activation of PI(4)Kbeta is important for the recruitment of trafficking complexes to the trans-Golgi network (TGN) and vesicular traffic from this organelle. The NCS-1-ARF1 interaction and its specificity have been demonstrated through in vitro binding assays, in vitro enzyme assay, and through functional cellular assays. We show that NCS-1 can exert bidirectional effects to activate PI(4)Kbeta on its own or inhibit the activation by ARF1. NCS-1 was shown to modulate the effects of expression of ARF mutants that disrupt Golgi morphology and to recruit GDP-loaded ARF to the Golgi complex in a Ca(2+)-dependent manner. We demonstrate antagonist effects of NCS-1 and ARF on constitutive and regulated exocytosis. The NCS-1-ARF1 interaction provides evidence for functional cross-talk between Ca(2+)-dependent and ARF-dependent pathways in TGN to plasma membrane traffic.

A model proposing distinct subcellular organelle identity based upon a unique compliment of lipid and protein components is an accepted paradigm. In recent years, however, evidence from numerous sources has hinted at a greater level of compartmentalization whereby specific cellular membranes are themselves mosaics of distinctive, highly specialized and tightly regulated sub-or micro-domains (1,2). The best studied example of this domain complexity is in vesicular trafficking from the TGN 1 where multiple mixed cargoes are firstly sorted and subsequently transported to various cellular destinations. There is considerable interest in the idea that specific phosphoinositides may specify functional sub-domains within organelles such as the TGN (3,4), but the mechanisms underlying this are unknown.
The ubiquitous GTPase ARF1 has been directly implicated as an essential protein in multiple vesicular trafficking steps at both the Golgi apparatus and on endosomes (5). ARF1 is a central component of both COPI and clathrin coat assemblies, which mediate vesicular budding in the Golgi and eventual vesicle transit from the TGN (5). A related TGN-localized ARF1 activity is its ability to recruit and activate soluble phosphatidylinositol-4-OH kinase type III ␤ (PI(4)K␤) (6), an enzyme whose activity is essential to the structural integrity of the Golgi and that has been shown to exert multiple effects on protein traffic from the TGN in mammalian cells (7). ARF1 activation of PI(4)K␤ directly elevates levels of phosphatidylinositol 4-phosphate (6), a lipid species that is essential for Golgi to plasma membrane trafficking in yeast (8,9). In higher eukaryotes, the precise function of phosphatidylinositol 4-phosphate synthesis at the TGN is less well understood, although recent studies have begun to identify phosphatidylinositol 4-phosphate-selective binding proteins that might control distinct TGN trafficking events (10,11).
A second characterized activator of PI(4)K␤, is neuronal calcium sensor-1 (NCS-1) (12). Originally thought to be neuronalspecific, NCS-1 was subsequently identified in various other cell types (13)(14)(15)) and a Saccharomyces cerevisiae orthologue (Frq1) was discovered that is essential due to its ability to activate Pik1, one of the two PI(4)K enzymes in yeast (16). Pik1 is required for normal secretion from the Golgi and shows synthetic lethality with arf1 (8,9). In mammalian cells NCS-1 is a TGN resident protein (14,17) that is likely to recruit PI(4)K␤ to this locale (18,19). Overexpression of NCS-1 (13) and subsequent activation of the kinase has been shown to enhance regulated secretion from PC12 cells (20)in a Ca 2ϩ -dependent fashion (18).
In this study, our aim was to identify novel NCS-1-binding proteins. ARF1 was identified as a selective partner for NCS-1, and we were able to ascribe functional consequences of this interaction for trafficking events between the TGN and the plasma membrane. Both ARF1 and NCS-1 were capable of individually stimulating in vitro PI(4)K␤ activity, but in combination an antagonistic relationship was revealed. This antagonism was seen in studies monitoring both constitutive and regulated secretory events in the PC12 cell line. These data show an inhibitory interaction between ARF1 and NCS-1 pathways that may function to prevent simultaneous activation of PI(4)K␤ by both effectors and thereby may assist in the generation of segregated trafficking domains at the TGN. In addition, the interaction of NCS-1 and ARF will allow local changes in Ca 2ϩ concentration to exert a regulatory influence on ARFdependent processes during traffic from the TGN to the plasma membrane.

Pull Down and Identification of Binding Partners for NCS-1 from
Bovine Brain Cytosol-Bovine brain cytosol was prepared from 1 kg of whole frozen brain tissue (FirstLink UK Ltd., Birmingham, UK) (21). Recombinant GST, GST-NCS-1, and GST-L-CaBP1 were prepared as described previously (22). 10 mg of recombinant GST or GST fusion proteins were immobilized onto 3 ml of glutathione-Sepharose 4B resin (Amersham Biosciences) that had been pre-washed with cytosol buffer (25 mM Tris-HCl (pH 7.8), 50 mM KCl, 5 mM EGTA, 5 mM nitrilotriacetic acid, 1 mM dithiothreitol, 1 M free Ca 2ϩ ) by incubation for 2 h at 4°C. Brain cytosol extract was dialyzed against cytosol buffer before application to GST affinity columns (5 ml of dialyzed extract/column), and binding was allowed to proceed for 16 h at 4°C. Columns were washed extensively with cytosol buffer, and specific Ca 2ϩ -dependent binding proteins eluted using cytosol buffer without added Ca 2ϩ . An additional elution step with cytosol buffer containing 1 M NaCl was used to isolate Ca 2ϩ -independent binding interactions. Eluted protein fractions were concentrated by methanol precipitation and pellets extracted into SDS dissociation buffer (125 mM HEPES (pH 6.8), 10% (w/v) sucrose, 10% (v/v) glycerol, 4% (w/v) SDS, 1% (v/v) ␤-mercaptoethanol, 2 mM EDTA) by boiling for 5 min. Proteins were analyzed by SDS-PAGE (8 -16% Novex TM Tris-glycine gradient gel, Invitrogen) followed by staining with Colloidal Blue reagent (Novex TM , Invitrogen). Bands of interest were excised from the gel, digested with Trypsin, and processed for mass spectrometry peptide fingerprinting (23) using Micromass MALDI-MS equipment (Waters Ltd., UK). Protein identifications based on mass fingerprints of digestion generated peptides were determined using Mascot search software (Matrix Science). Protein samples from these experiments were transferred by transverse electrophoresis to nitrocellulose membranes for detection via Western blotting. Filters were blotted with pan-specific anti-ARF antibody (1:1000, 1D9 clone, AbCam, Cambridge, UK), and immunoreactivity was visualized using ECL-Plus reagents (Amersham Biosciences).
In Vitro Protein Binding Assays-1 M GST or GST-NCS-1 were immobilized onto 50 l of glutathione-Sepharose 4B in cytosol buffer containing either 0 or 1 M free Ca 2ϩ and supplemented with either 50 M GDP or GTP␥S by incubation for 30 min at 4°C. 1 M recombinant C-terminally tagged ARF1-His 6 protein (24) was added to incubations, and binding was allowed to proceed for 2 h at 4°C. Sepharose pellets were collected by centrifugation and washed in the appropriate ϮCa 2ϩ / guanine nucleotide-supplemented cytosol buffer. Final pellets were extracted into 50 l of SDS dissociation buffer, and bound protein was analyzed by Western blotting with anti-ARF 1D9 antibody (1:1000). For analysis of the specificity of NCS-1 for ARF1 over ARF3 and binding of PI(4)K␤ to NCS-1 and ARF1, [ 35 S]methionine-labeled ARF1-HA, ARF3-HA, and HA-PI(4)K␤ proteins were generated using a TNT® quickcoupled transcription/translation system (Promega, UK) according to the manufacturer's protocol and with pcDNA-ARF1-HA, pcDNA-ARF3-HA, or pcDNA-HA-PI(4)K␤ (6, 25) as templates. Plasmids for the expression of recombinant murine ARF were prepared by PCR from authentic cDNA samples, obtained as a gift from Dr. K. Nakayama, University of Tsukuba, Japan, with PCR products cloned into pcDNA3 plasmid between the HindIII and XbaI sites with the addition of a His 6 tag at the C terminus (24). Binding assays were performed as above, and bound ARF proteins or PI(4)K␤ were detected by autoradiography of SDS-PAGE gels. For analysis of the specificity of ARF1 for NCS-1, recombinant GST, GST-NCS-1, GST-hippocalcin, GST-neurocalcin-␦, or GST-L-CaBP1 (all at 1 M) were immobilized onto beads in cytosol buffer containing either no added or 1 M free Ca 2ϩ and 50 M GDP. Non-radioactive TNT® Quick in vitro translated ARF1 protein was then added to incubations at ϳ1 M final concentration, binding was allowed to proceed for 2 h at 4°C, and bound protein was detected by immunoblotting.
Immunoprecipitation-250 l of bovine brain cytosol dialyzed against cytosol buffer was incubated with 5 g of anti-ARF 1D9 antibody, rabbit anti-NCS-1 antibody, or pre-immune NCS-1 serum (13) overnight at 4°C with constant agitation. 50 l of protein-G-Sepharose (Amersham Biosciences) in cytosol buffer was added to samples, and incubations were continued for 2 h at 4°C. Protein-G pellets were collected by centrifugation and washed in cytosol buffer. Bound proteins were analyzed by immunoblotting with either anti-NCS-1 (1:1000) or anti-ARF.
Confocal Laser Scanning Microscopy-For confocal laser scanning microscopy, fixed transfected cells were analyzed using a Leica TCS-SP-MP microscope (Leica Microsystems, Heidelberg, Germany) using a 22-m pin hole and a ϫ63 water immersion objective with a 1.2 numerical aperture. For optimal imaging of ECFP, cells were excited at 430 nm and light was collected at 450 -500 nm. Fluorescein was imaged using excitation at 488 nm and light collection at 500 -560 nm. For optimal imaging of enhanced green fluorescent protein, cells were excited at 488 nm, and light was collected at 500 -550 nm. For optimal imaging of Texas Red, cells were excited at 543 nm, and light was collected at 600 -650 nm. Images were exported as TIFF files and compiled in CorelDraw.

ARF1 Is a Specific Ca 2ϩ -dependent Binding Partner for
NCS-1-To identify novel NCS-1-interacting proteins, a GST-NCS-1 affinity purification column was prepared. NCS-1-specific binding proteins were subsequently identified by mass spectroscopic peptide fingerprinting of digested gel samples and searching of known peptide mass databases. One Ca 2ϩ -dependent NCS-1-binding protein was initially identified as either ARF1 or ARF3 (Fig. 1A) based on matching of nine peptides with 74% coverage of the protein. To confirm this data, samples from binding experiments were used in Western blot-ting with a pan-specific anti-ARF monoclonal antibody (29). Consistent with our initial identification, an ARF-reactive protein of the correct molecular mass was bound only to GST-NCS-1 and only in the presence of 1 M free Ca 2ϩ (Fig. 1B). No ARF-reactive species was purified by a GST control affinity column (Fig. 1B, GST) or by a second column containing the NCS-1-related Ca 2ϩ -binding protein L-CaBP1 (22) (Fig. 1B). To determine the identity of the ARF protein that was bound to NCS-1, a series of in vitro binding assays were used to analyze the selectivity of NCS-1 for ARF1 compared with ARF3. In binding assays with equimolar amounts of the two recombinant proteins myristoylated ARF1-His 6 was bound to GST-NCS-1, and the interaction displayed a dependence upon free Ca 2ϩ . Binding occurred in the presence of both GDP or GTP␥S and was thus insensitive to the nucleotide state of ARF1, although there was more binding in the presence of GDP (Fig. 1C). The loading control lane (Input ARF1) shows the total amount of ARF1 added to the incubation and run at the same dilution. The interaction was specific for NCS-1, because no binding of ARF1 was observable to GST (Fig. 1C). With the knowledge that NCS-1 could directly associate with ARF1 in vitro, we examined the possibility of an interaction with ARF3. To ensure that correctly post-translationally modified forms of both proteins were available, radiolabeled, C-terminal HA-tagged ARF1 and ARF3 proteins were synthesized using a coupled transcription/translation system, which is able to generate myristoylated protein products (30). Similar levels of radiolabeled ARF1 and ARF3 were produced (input, Fig. 1D) and

FIG. 1. NCS-1 and ARF1 directly interact with high specificity.
A, Coomassie Blue-stained SDS-PAGE gel highlighting a protein from bovine brain cytosolic fractions that was specifically purified by a GST-NCS-1 affinity matrix in the presence of 1 M free Ca 2ϩ (duplicate lanes are shown for each affinity column/condition tested) and that was identified by peptide mass fingerprinting as ARF1 or ARF3 (arrowhead). B, Western blot of samples probed with pan-specific anti-ARF antibody. An ARF-reactive protein was bound only to a GST-NCS-1 matrix and only in the presence of free Ca 2ϩ . No binding of ARF protein to a GST control column or to a related small Ca 2ϩ -binding protein (L-CaBP1) was detected. C, Western blotting of samples from direct in vitro binding assays examining ARF1-His 6 association with GST control protein or GST-NCS-1 in the presence or absence of 1 M free Ca 2ϩ . ARF1 bound to GST-NCS-1 only in the presence of free Ca 2ϩ and did not bind to GST controls. D, autoradiograph of samples from binding assays where [ 35 S]methionine-labeled ARF1 and ARF3 proteins prepared by in vitro transcription and translation were incubated with GST control protein or GST-NCS-1 in the presence and absence of 1 M free Ca 2ϩ . A detectable signal was observed only for ARF1 binding to GST-NCS-1 in the presence of free Ca 2ϩ . No binding of ARF1 was detected to GST control protein, and no binding of ARF3 was observed to any GST proteins analyzed. E, anti-HA Western blot of in vitro translated ARF1-HA binding to a series of Ca 2ϩ -binding protein GST fusion constructs and to GST control protein in the presence or absence of 1 M free Ca 2ϩ . ARF1 binding was detectable only to GST-NCS-1 in the presence of free Ca 2ϩ , and no binding was detected to any other GST fusion proteins tested or to GST control protein.
incubated with GST control protein or GST-NCS-1. A significant proportion of the input ARF1 was bound to GST-NCS-1 in the presence of free Ca 2ϩ (Fig. 1D). This interaction was specific, because no binding of ARF1 was observed to GST controls. No binding of ARF3 to GST-NCS-1 was detected. These data suggest that there is a highly selective binding of ARF1 to NCS-1.
NCS-1 belongs to the NCS family of closely related EF handcontaining proteins (12,31). We examined whether NCS-1 was a unique ARF1-interacting NCS protein by analyzing the binding of in vitro translated ARF1 to a series of NCS proteins fused to GST (Fig. 1E). ARF1 was only efficiently bound by GST-NCS-1 in the presence of free Ca 2ϩ (Fig. 1E, lane 4), and no binding was detectable to any of the other Ca 2ϩ -binding proteins examined.
Due to the apparent exclusivity in association exhibited between ARF1 and NCS-1 in vitro we next examined whether the interaction was observable in vivo. Using a specific anti-NCS-1 antibody (13) or a pan specific anti-ARF antibody, we tested if endogenous myristoylated ARF1 and NCS-1 could be co-immunoprecipitated from bovine brain cytosolic extracts (Fig. 2B). Anti-ARF antibody precipitated ARF protein along with NCS-1, and in the reciprocal experiment, anti-NCS-1 antibody precipitated NCS-1 along with ARF. An NCS-1 pre-immune serum included as a negative control failed to immunoprecipitate detectable NCS-1 or ARF protein. These data suggest that NCS-1-ARF1 can also form a complex under in vivo conditions. We compared the cellular localization of NCS-1-ECFP with HAtagged ARF1 when the proteins were co-expressed in HeLa cells (Fig. 2C). Both proteins have a dual cytosolic and membraneassociated localization. NCS-1-ECFP localized in part to a perinuclear region, which has previously been identified as the TGN (Fig. 2C, top three panels) (14, 17, 32). ARF1 protein showed co-localization with NCS-1-ECFP (Fig. 2C, top panel, overlay).
We also examined the co-localization of NCS-1 with PI(4)K␤ (Fig.  2C, bottom three panels). Although a large proportion of PI(4)K␤ showed a diffuse cytosolic staining, there was a clear concentration of the protein in a peri-nuclear region that co-localized with NCS-1-ECFP. Collectively, these in vitro and in vivo studies of ARF1 and NCS-1 binding are consistent with a direct, Ca 2ϩ -dependent interaction that is selective for ARF1 over related ARF proteins and selective for NCS-1 with respect to similar NCS proteins. Our localization studies support the binding data and indicate a potential in vivo site of interaction at the level of the TGN where NCS-1, ARF1, and PI(4)K␤ have previously been found individually to be localized.
Binding of ARF1 and NCS-1 to PI(4)K␤ and Stimulation of Enzymic Activity-We next examined whether both NCS-1 and ARF1 could directly bind to PI(4)K␤. PI(4)K␤ was prepared by in vitro transcription and translation to enable production of enzymically active protein. In direct binding assays, [ 35 S]methionine-labeled PI(4)K␤ was found to specifically interact with GST-NCS-1 in a Ca 2ϩ -dependent manner (Fig. 3A, lane 4 versus lane 3), and no binding of PI(4)K␤ was detectable with control GST protein (Fig. 3A, lanes 1 and 2). ARF1 protein was also made using in vitro transcription and translation, and its interaction with PI(4)K␤ was determined following immunoprecipitation with anti-ARF. ARF1 was found to associate directly with PI(4)K␤ in a nucleotide-dependent manner (Fig.  3B). In contrast to the binding of PI(4)K␤ alone to NCS-1-GST or ARF1, we were unable to observe simultaneous binding of PI(4)K␤ to ARF1 and NCS-1 under conditions where the NCS-1-ARF1 interaction was detected (Fig. 3C), a result suggestive of the NCS-1-ARF1 complex having impaired PI(4)K␤ binding. To ascertain the functional significance of the NCS-1-ARF1 interaction we examined the effects of both proteins on in vitro PI(4)K␤ enzymic activity (Fig. 3D). When PI(4)K␤ was incubated alone, a basal level of enzymic activity was measured,

NCS-1-ARF1 in Control of PI(4)K and Membrane Traffic
and this was increased by ϳ70% above control by the addition of NCS-1 alone in the presence but not absence of 1 M Ca 2ϩ and by ϳ125% by the addition of ARF1 alone (Fig. 3D), indicating that both of these proteins are effective positive regulators of in vitro PI(4)K␤ activity consistent with previous studies (6,33). Incubation of PI(4)K␤ with both NCS-1 and ARF1 elicited a marked reduction in enzyme activity compared with that observed when ARF1 was present alone (Fig. 3D), suggesting that a complex of ARF1 and NCS-1 was unable to effectively stimulate PI(4)K␤ activity.
GDP for GTP, and therefore it remains permanently inactive in the cytoplasm and sequesters ARF-guanine nucleotide exchange factors (36). Both of these mutant ARF1 proteins elicit alterations in Golgi morphology with ARF1-Q71L causing the appearance of distended Golgi cisternae and ARF1-T31N the total fragmentation of the Golgi apparatus (35). The effects of these mutant proteins were examined following transfection with constructs bearing a C-terminal HA tag. When ARF1-Q71L was expressed alone in HeLa cells, enlarged Golgi structures were clearly apparent (Fig. 4A). When NCS-1 was coexpressed with ARF1-Q71L, a reversal of mutant Golgi morphology to an essentially wild type was observed (Fig. 4, A  and D). GDP-locked ARF1-T31N is unable to switch to an activated state and consequently fails to bind to cellular membranes, a defect that is likely connected to Golgi fragmentation in cells overexpressing this mutant. When overexpressed in HeLa cells, ARF1-T31N exhibited a diffuse distribution throughout the cytosol (Fig. 4B). In a significant proportion of cells shown to be co-expressing both ARF1-T31N and NCS-1, ARF1-T31N instead was localized to a peri-nuclear, Golgi compartment (Fig. 4, B and D). Expression of NCS-1 not only re-established normal Golgi morphology in these cells but also appeared to recruit ARF1-T31N from the cytosol to this domain (Fig. 4B). This suggests that NCS-1 did not simply sequester the ARF mutant but can also recruit the GDP-loaded form of ARF1 to membranes. In these experiments, NCS-1 and either ARF1-Q71L or ARF1-T31N were co-localized at the Golgi when co-expressed, consistent with the in vitro capacity of NCS-1 to associate with both GTP␥S and GDP loaded ARF1 (Fig. 1C). A possible explanation for the reversal of mutant ARF1 phenotypes by NCS-1 would be if expression of NCS-1 modified mutant ARF1 protein expression. To check this possibility, transfected HeLa cells were used in Western blotting to examine ARF protein overexpression in the presence and absence of NCS-1, and it was found that co-expression of NCS-1 had no significant effect on ARF protein levels compared with control transfected cells (Fig. 4E). To test whether the ARF1-NCS-1 interaction is Ca 2ϩ -dependent within cells, as suggested by the in vitro binding data, the effect on ARF1-T31N-expressing cells of co-expressing an NCS-1 mutated in the three functional EF hands (NCS-1(2-4) (17)) was examined. This mutant NCS-1 protein is still able to associate with the TGN when expressed alone (Fig. 4). In cells co-expressing NCS-1(2-4) and ARF1-T31N, the distribution of the ARF1 mutant was similar to that in cells expressing this ARF mutant alone (Fig. 4, C and D), and NCS-1(2-4) itself remained diffusely distributed within the cells consistent with maintained dispersal of the Golgi by ARF1-T31N. These data indicate therefore that Ca 2ϩ binding by NCS-1 is essential for the NCS-1-ARF interaction that leads to reversal of the ARF1-T31N mutant phenotype and recruitment of this GDP-loaded ARF1 to the Golgi.
ARF1 Abolishes NCS-1-mediated Stimulation of Agonistevoked Regulated Secretion from PC12 Cells-NCS-1 has been implicated in the positive control of regulated secretion through its stimulation of PI(4)K␤ (18,20). ARF1, as an essential and ubiquitous regulator of Golgi trafficking, including the genesis of nascent constitutive and regulated secretory vesicles (37,38), is involved in both pathways. To assess the functional consequences of an ARF1-NCS-1 interaction we examined the effects of overexpression of both proteins on ATP-driven regulated human growth hormone (hGH) release from transfected PC12 cells. When overexpressed, NCS-1 elicited around a 40% enhancement of ATP-triggered secretion compared with control, pcDNA, and transfected cells (Fig. 5A). ARF1 overexpression, in contrast, elicited no significant alteration in hGH secretion in this system. When co-expressed, the presence of ARF1 antagonized the NCS-1 enhancement of secretion. The changes in evoked hGH release may be due to alterations in the pathway FIG. 5. ARF1 inhibits the stimulation by NCS-1 of regulated human growth hormone secretion from PC12 cells. A, PC12 cells were transfected with the indicated pcDNA-based NCS-1-ARF1 expression constructs in combination with a human Growth Hormone (hGH) reporter plasmid. Cells were stimulated with 300 M ATP, and released hGH was assayed using an enzyme-linked immunosorbent assay. Secreted hGH levels were normalized to those observed in cells expressing control empty pcDNA3.1(Ϫ) vector (100%). B, total expressed growth hormone levels were determined, and values were normalized to total hGH measured in control, pcDNA3.1(Ϫ)-expressing cells (100%). C, PC12 cells were transfected with pcDNA-NCS-1 or ARF1-HA alone or in combination, and expressed NCS-1 and ARF1 proteins were detected by Western blotting.
triggering exocytosis. It is also possible that granule biogenesis or packaging of hGH into regulated secretory vesicles may have been modified by NCS-1 expression but there was no significant difference between any of the transfection conditions tested in the total amount of hGH synthesized (Fig. 5B). Similar levels of ARF1 and NCS-1 were expressed in singly or doubly transfected cells (Fig. 5C), and so the antagonism of the NCS-1 stimulation of hGH release by ARF1 was not due a reduction in NCS-1 expression in co-transfected PC12 cells.
NCS-1 Stimulates Constitutive Traffic in PC12 Cells, an Effect Antagonized by Co-expressed ARF1-We also investigated a potential role for the interaction between NCS-1 and ARF1 on constitutive trafficking from the TGN to the plasma membrane. In these experiments we examined effects of NCS-1 and ARF1 expression on the constitutive trafficking of a green fluorescent protein-tagged form (28) of the temperature-sensitive VSV-G protein (tsO45-VSV-G), which can be trapped in a temperature-dependent manner at either the endoplasmic reticulum (40°C) (39,40) or the TGN (20°C) (41). PC12 cells were transfected with control empty vector, NCS-1, ARF1, or both NCS-1-ARF1, and co-transfected tsO45-VSV-G was allowed to accumulate at the TGN by incubation at 20°C. Following a temperature shift to 32°C (to allow transit of tsO45-VSV-G to the plasma membrane) cells were fixed at various time points, and the amount of VSV-G trafficked to the plasma membrane was quantified as a fraction of the total cellular VSV-G (Fig. 6). Expression of NCS-1 elicited a marked increase in both the initial rate of VSV-G transport and the total amount of VSV-G trafficked to the cell surface at all time points examined compared with control transfected cells. ARF1 overexpression elicited no discernable effect on total VSV-G trafficked over the time course of the experiment compared with control transfected cells, but when ARF1 was co-expressed with NCS-1, the stimulation of constitutive trafficking observed with NCS-1 alone was abolished. The effects of NCS-1 and ARF1 on VSV-G trafficking were not a consequence of variations in VSV-G protein expression, the levels of which were similar under all transfection conditions tested (data not shown). DISCUSSION ARF1 is a multifunctional regulator of vesicular traffic. This small GTPase has been identified as an essential component during the formation of both coatomer (COPI) and clathrincontaining vesicular coat complexes (42)(43)(44), which control intra-Golgi traffic in addition to TGN to plasma membrane and TGN to endosome/lysosome pathways (5,45). ARF1 may possess an even more extensive functional repertoire and can interact with multiple proteins (5). The Ca 2ϩ sensor protein NCS-1 (12) has been implicated in a number of cellular roles, which appear to rely on an association with the lipid kinase PI(4)K␤ (16,20,33,46), an enzyme also influenced by ARF1 (6). NCS-1, PI(4)K␤, and ARF1 are all present in cytosolic and membrane-associated pools in cells, with all three being found on the Golgi, including the TGN (6,14,17,19,47,48), and so could potentially interact within this compartment.
In this study our aim was to isolate novel binding partners for NCS-1 in the hope that this would provide a more complete appreciation of the cellular functions regulated by this protein.
We identified ARF1 as an NCS-1-interacting protein and confirmed this in a series of in vitro and in vivo binding assays showing a Ca 2ϩ -dependent, nucleotide-independent association between the two proteins. The interaction was, however, independent of the myristoylation state of either of the proteins. The discrimination in binding between ARF1 and ARF3 may be surprising given the high level of similarity of these proteins. Distinct binding interactions for ARF1 and -3 have not previously been observed (5). The seven amino acid differences between the two proteins are, however, highly conserved with ARF1 and -3 being identical from fish and amphibians to humans suggesting that there must be a physiological significance for these differences.
Further evidence for the importance of the NCS-1-ARF1 interaction was provided from experiments where we observed reversals in mutant ARF1 phenotypes induced as a consequence of NCS-1 co-expression and which also indicated that NCS-1 can recruit soluble GDP-loaded ARF1 to the TGN in a Ca 2ϩ -dependent manner. Both NCS-1 and ARF1 are known activators of PI(4)K␤, and hence our binding studies suggested the possibility that this interaction might be involved in the regulation of kinase activity. We observed activation of PI(4)K␤ by both NCS-1 and ARF1 in isolation but unmasked an antagonistic mode of action when both proteins were present in combination with the enzyme. We detected Ca 2ϩ -dependent binding of NCS-1 to PI(4)K␤ in contrast to an earlier study suggesting Ca 2ϩ -independent binding (33). The basis of this difference is not immediately obvious, however, the existence of a Ca 2ϩ requirement would potentially allow for fine-tuning of enzyme activity through coupling to integration of localized Ca 2ϩ signals by NCS-1. The NCS-1 used for our experiments was not myristoylated, and the previous study suggested that myristoylation was required. It should be noted that the homologous interaction between yeast proteins is not dependent on myristoylation of Frq1 (49). Although a genetic interaction between ARF and Pik1 has been inferred in yeast (8) and mammalian ARF1 has been demonstrated to recruit a PI(4)K␤ activity to isolated Golgi membranes (6) and to increase activity of PI(4)K␤ (50), no evidence for a direct physical interaction between the two proteins has thus far been provided. In this study we were able to show for the first time a direct, if somewhat inefficient, binding of ARF1 to PI(4)K␤. Differences between the assay conditions used here with those of a previous study (50) may be the reason for this.
In functional analyses, monitoring regulated secretory and constitutive trafficking events, we observed a significant stimulation of both processes by NCS-1. Co-expression of ARF1 abolished NCS-1 enhancement of constitutive and regulated traffic. The enhancement by NCS-1 of regulated exocytosis is mediated through PI(4)K␤ (18,20), and so this effect of ARF1 is consistent with observations on the in vitro activation of PI(4)K␤ and with binding data indicating that ARF1-NCS-1 form a mutually exclusive complex in preference to binding PI(4)K␤. It is known from previous work that modification of PI(4)K␤ activity in cells has multiple complex effects on constitutive traffic from the TGN to the cell surface (7), suggesting that the fine control of PI(4)K␤ activity is important for normal sorting and traffic events at the TGN.
Our data suggest that ARF1 and NCS-1 in the TGN can in combination control specific sorting/trafficking events from this organelle. PI(4)K␤ is an enzyme common to both the NCS-1 and ARF1 controlled pathways, however, we envisage instances where spatial and temporal activation of the enzyme is required by either ARF1 or NCS-1 but not both. One possibility is that a direct interaction between ARF1-NCS-1 provides a means for bi-directional control of the kinase such that activation can proceed in the presence of ARF1 or NCS-1/Ca 2ϩ alone but spurious stimulation is prevented if both effectors should happen to overlap thereby maintaining domain identity and the fidelity of the associated trafficking pathway. Alternatively, NCS-1 may exert effects on other ARF1-effector interactions. The Golgi can act as a Ca 2ϩ store (51), and local release of Ca 2ϩ within the secretory pathway has been shown to be required for many traffic steps (52), including traffic between Golgi cisternae (53), but its importance at the TGN has not been studied. Elevation of Ca 2ϩ near the TGN could occur due to release of Golgi luminal Ca 2ϩ . NCS-1 is membrane-associated at the TGN in a Ca 2ϩ -independent manner and is sensitive to Ca 2ϩ concentrations above basal levels but with a submicromolar affinity (17,22) leading to the interesting possibility that, under certain physiological conditions, TGN trafficking events might be under the direct influence of local intracellular Ca 2ϩ fluxes acting through NCS-1 and its effectors or alternatively via NCS-1 interactions with ARF1.