Synaptotagmins form a hierarchy of exocytotic Ca²⁺ sensors with distinct Ca²⁺ affinities

Relative Ca²⁺ affinities of the synaptotagmin C₂A-domains

We compared, in the same experiment, the apparent Ca²⁺ affinities of the C₂A-domains of synaptotagmins 1, 2, 3, 5, 7 and 10. These synaptotagmins were chosen because synaptotagmins 3 and 7 are the most abundant synaptotagmins after 1 and 2 (Butz et al., 1999; Sugita et al., 2001), and because synaptotagmins 3, 5, 6 and 10 form a class of closely related synaptotagmins (Fukuda et al., 1999). Synaptotagmin 6 was not studied because we were unable to produce soluble properly folded C₂-domains from this isoform. We estimated the apparent Ca²⁺ affinities of the C₂A-domains by Ca²⁺-dependent phospholipid binding, and compared two independent methods and three different buffers to control for potential artifacts. Ca²⁺-dependent phospholipid binding assays were chosen because synaptotagmins bind Ca²⁺ at physiological concentrations only in the presence of phospholipids and because Ca²⁺-dependent phospholipid binding most likely constitutes part of their physiological function (reviewed in Südhof, 2002).

We first measured the apparent Ca²⁺ affinity of the C₂A-domains with a standard resin-based assay in which immobilized glutathione S-transferase (GST) fusion proteins of the synaptotagmin C₂A-domains were incubated with radiolabeled liposomes at different Ca²⁺ concentrations (Davletov and Südhof, 1993). Figure 1 demonstrates that each synaptotagmin is characterized by a distinct Ca²⁺ affinity, as measured by this assay. The vesicular synaptotagmins 1 and 2 consistently had the lowest Ca²⁺ affinities [EC₅₀ ≈ 10–20 µM Ca²⁺ with liposomes composed of 25% phosphatidylserine (PS)/75% phosphatidylcholine (PC); Figure 1A and B]. By contrast, the plasma membrane synaptotagmins 3 and 7 exhibited the highest Ca²⁺ affinities (EC₅₀ ≈ 1–2 µM Ca²⁺; Figure 1C and E). Synaptotagmins 5 and 10 (whose localizations are unknown, but which are most closely related to the plasma membrane synaptotagmins 3 and 6) also displayed relatively high Ca²⁺ affinities (EC₅₀ ≈ 3 µM Ca²⁺; Figure 1D and F). Moreover, we observed that synaptotagmins 1 and 2 (which are both localized to synaptic vesicles but differentially distributed in brain; Ullrich et al., 1994; Marqueze et al., 1995) also differ in Ca²⁺ affinity (Figure 1A and B), with synaptotagmin 2 reproducibly exhibiting an ∼2-fold lower Ca²⁺ affinity than synaptotagmin 1. The specificity of the phospholipid binding reaction was confirmed by a point mutant in the predicted Ca²⁺ binding site of the synaptotagmin 3 C₂A-domain (D333N), which abolished Ca²⁺-dependent phospholipid binding (Figure 1C).

The distinct Ca²⁺ affinities of synaptotagmins could have important implications for the Ca²⁺-dependent regulation of neurotransmitter release. This is illustrated by the R233Q point mutation in synaptotagmin 1, which causes an ∼2-fold decrease in the Ca²⁺ affinity of synaptotagmin 1 and a similar decrease in the Ca²⁺ responsiveness of synapses (Fernandez-Chacon et al., 2001), suggesting that the Ca²⁺ affinity of vesicular synaptotagmins controls synaptic responses. In view of the potential importance of differences in Ca²⁺ affinity, we sought to confirm the measured apparent Ca²⁺ affinities of synaptotagmins with an independent assay. For this purpose we incubated soluble C₂A-domain–GST fusion proteins with liposomes at different Ca²⁺ concentrations, isolated the liposomes by centrifugation, and determined the amount of bound synaptotagmin C₂A-domains by Coomassie Blue staining of SDS gels. This assay, referred to as the liposome centrifugation assay, utilizes C₂A-domains in solution and thus avoids possible immobilization artifacts of the standard resin-based assay. Figure 2 shows that the liposome centrifugation assay produced results similar to the resin-based assay, revealing an ∼2-fold higher Ca²⁺ affinity of synaptotagmin 1 than of synaptotagmin 2, and an ∼5- to 10-fold higher affinity of the plasma membrane synaptotagmins than of vesicular synaptotagmins. A major difference between the two assays was that in the resin-based assay, a bell-shaped Ca²⁺ concentration dependence was observed for synaptotagmins 3 and 7 (Figure 1), whereas the liposome centrifugation assay exhibited no decrease in phospholipid binding at higher Ca²⁺ concentrations (Figure 2). This difference may reflect the more stringent washes used in the resin-based assay, which would also explain the inability of the resin-based assay to detect Ca²⁺ binding to the C₂B-domain (see below; Fernandez et al., 2001). The concurrence of the apparent Ca²⁺ affinities determined with the two assays confirms that the Ca²⁺ affinity is not dependent on whether the C₂A-domains are immobilized or in solution.

Synaptotagmin C₂B-domains exhibit similar differences in Ca²⁺ affinity

Synaptotagmins have two Ca²⁺ binding domains: the C₂A- and C₂B-domains. Initially the C₂B-domains were thought to be unable to bind to phospholipids as a function of Ca²⁺ because the resin-based assay did not detect such binding (reviewed in Südhof and Rizo, 1996). Recently, however, less stringent assays revealed that the C₂B-domain of synaptotagmin 1 specifically binds to phospholipids in a Ca²⁺-dependent manner with an apparent Ca²⁺ affinity that resembles that of the C₂A-domain (Fernandez et al., 2001). The C₂B-domain probably forms a less tight Ca²⁺–phospholipid complex than the C₂A-domain because the C₂B-domain contains only two Ca²⁺ binding sites as opposed to the three Ca²⁺ binding sites of the C₂A-domain. To test whether the C₂B-domains of synaptotagmins 3 and 7 also bind phospholipids in response to Ca²⁺ and whether the various C₂B-domains exhibit similar differences in Ca²⁺ affinity as the C₂A-domains, we examined the C₂B-domains using the liposome centrifugation assay (Figure 2). Figure 3 shows that the synaptotagmin 1 and 7 C₂B-domains, but not the synaptotagmin 3 C₂B-domain, bind phospholipids as a function of Ca²⁺. The apparent Ca²⁺ affinity of the synaptotagmin 7 C₂B-domain was ∼10-fold higher than that of the synaptotagmin 1 C₂B-domain (Figure 3), demonstrating that the C₂A- and C₂B-domains of these synaptotagmins exhibit the same difference in Ca²⁺ affinity. The lack of Ca²⁺-dependent phospholipid binding by the synaptotagmin 3 C₂B-domain, as judged by this assay, is somewhat surprising considering its sequence similarity to other C₂B-domains. It may be due either to a lack of Ca²⁺ binding by this domain (as indicated by the crystal structure; Sutton et al., 1999) or to a selective inability to bind phospholipids as a function of Ca²⁺.

The relative Ca²⁺ affinities of synaptotagmin C₂A-domains are independent of Ca²⁺ buffers

The apparent Ca²⁺ affinities for synaptotagmin C₂A-domains determined above and in previous studies (Fernandez-Chacon et al., 2001) were determined with the use of Ca²⁺/EGTA buffers, which may introduce systematic errors. Furthermore, although Ca²⁺ binding assays and secretion measurements in permeabilized PC12 cells both employ Ca²⁺/EGTA buffers, these buffers have distinct compositions (Gerber et al., 2001; Sugita et al., 2001). To better relate Ca²⁺-dependent phospholipid binding to Ca²⁺-induced secretion in PC12 cells and to evaluate the accuracy of the Ca²⁺/EGTA buffers, we directly compared three different Ca²⁺ buffers: the traditional NaCl-based Ca²⁺/EGTA buffer (Gerber et al., 2001); the K-glutamate Ca²⁺/EGTA buffer containing 2 mM Mg²⁺, 2 mM ATP and 0.1% bovine serum albumin (BSA), which is used for permeabilized PC12 cell experiments (Sugita et al., 2001); and a NaCl-based buffer, which contained only Ca²⁺ but no EGTA (see Materials and methods). All three buffer systems generally gave similar results, suggesting that the Ca²⁺/EGTA buffers are reliable (data not shown). The only major difference noted was that the synaptotagmin 1 C₂A-domain displayed a lower apparent Ca²⁺ affinity in the K-glutamate Ca²⁺/EGTA buffer than in the traditional Ca²⁺/EGTA buffer or the Ca²⁺-only buffer, probably because the ATP in the K-glutamate buffer inhibits phospholipid binding to the C₂A-domain of synaptotagmin 1 but not to that of the other synaptotagmins (data not shown).

Synaptotagmins in PC12 cells

As a first approach to testing the functional consequences of the different Ca²⁺ affinities of synaptotagmins, we employed neuroendocrine PC12 cells as a model system. These cells were chosen because they have been productively used in studying Ca²⁺-triggered exocytosis (Ahnert-Hilger et al., 1987; Walent et al., 1992; McFerran et al., 1998; Avery et al., 2000) and are known to express at least synaptotagmins 1, 3 and 7 (Shoji-Kasai et al., 1992; Mizuta et al., 1994; Sugita et al., 2001). Previous studies have shown that synaptotagmin 1 is present on secretory vesicles in neuroendocrine cells, while synaptotagmin 7 is localized to plasma membranes (Perin et al., 1991; Sugita et al., 2001). The localization of synaptotagmin 3, however, is unclear. Some studies in pancreatic β-cells (the only endocrine cell where it has been studied) reported synaptotagmin 3 on secretory vesicles (Mizuta et al., 1997; Brown et al., 2000; Gao et al., 2000), although a more recent study detected it on the plasma membrane (Gut et al., 2001), similar to its localization in brain (Butz et al., 1999). To address this discrepancy with an independent approach, we examined in PC12 cells the localization of transfected synaptotagmin 3–EYFP fusion protein in comparison with other synaptotagmins. This approach was employed because the low abundance of synaptotagmin 3 made detection of the endogenous protein difficult, and because the direct fluorescence of EYFP avoids possible artifacts due to indirect immunofluorescence procedures. Figure 4 shows that synaptotagmin 3 is quantitatively deposited into the plasma membrane, which is labeled with the fluorescent dye FM4-64. The localization of synaptotagmin 3 is identical to that of synaptotagmin 7–EYFP, whereas synaptotagmin 1–EYFP is exclusively present in intracellular vesicles (Figure 4). These findings suggest that synaptotagmins 3 and 7 are general plasma membrane proteins in neurons and endocrine cells.

Inhibition of Ca²⁺-dependent norepinephrine secretion from permeabilized PC12 cells by synaptotagmin C2-domains

We loaded PC12 cells with radioactive norepinephrine, permeabilized the cells by freeze–thawing and triggered exocytosis by addition of Ca²⁺ at increasing concentrations. In the absence of protein additions or in the presence of only GST, we observed a bell-shaped Ca²⁺ response curve for exocytosis from the permeabilized PC12 cells (control in Figure 5A). This curve was similar to the Ca²⁺-concentration dependence of phospholipid binding by synaptotagmins 3 and 7 as measured by the resin assay (Figure 1; Sugita et al., 2001). Addition of the synaptotagmin 1 GST–C₂A-domain fusion protein to the PC12 cells caused a small but significant inhibition of exocytosis at high Ca²⁺ concentrations (Figure 5A). A similar but lower amount of inhibition was also observed for the C₂A-domain of synaptotagmin 2 (Figure 5B), consistent with its lower apparent Ca²⁺ affinity (Figures 1 and 2). The C₂A-domain of synaptotagmin 3, however, almost abolished Ca²⁺-triggered exocytosis at all Ca²⁺ concentrations (Figure 5C). This inhibition is similar to the effect we previously observed with the C₂A-domain of synaptotagmin 7 (Sugita et al., 2001). Inhibition by the C₂A-domain of synaptotagmin 3 required Ca²⁺ binding to the C₂A-domain because the Ca²⁺ binding site mutant of the C₂A-domain (D333N), which is unable to mediate Ca²⁺-dependent phospholipid binding (Figure 1C), had no effect on exocytosis (Figure 5D). Finally, the C₂A-domains of synaptotagmins 5 and 10 had an intermediate effect and significantly inhibited Ca²⁺-triggered exocytosis only at higher Ca²⁺ concentrations (Figure 5E and F).

The PC12 cell experiments suggest that the Ca²⁺ sensor responsible for triggering PC12 cell exocytosis is activated by low micromolar Ca²⁺ concentrations, which precisely match the apparent Ca²⁺ affinities of synaptotagmins 3 and 7, and is inhibited by the C₂A-domains from these synaptotagmins. The Ca²⁺ binding measurements described above showed that the C₂B-domains of synaptotagmins 1 and 7 exhibited similar differences in Ca²⁺ affinities, raising the question of whether this also applies to the PC12 cell inhibitions. To test this, we measured the effects of the synaptotagmin 1, 3 and 7 C₂B-domains on PC12 cell exocytosis. Figure 6 shows that the synaptotagmin 1 and 3 C₂B-domains had little effect on Ca²⁺-triggered exocytosis, whereas the synaptotagmin 7 C₂B-domain effectively inhibited exocytosis at all Ca²⁺ concentrations. These results again agree well with the Ca²⁺-dependent phospholipid binding measurements (Figure 3), indicating that there is an overall correlation for synaptotagmins in their apparent Ca²⁺ affinities as measured by Ca²⁺-dependent phospholipid binding and their effects in PC12 cells.

We had previously found that the synaptotagmin 7 C₂A-domain inhibits Ca²⁺-triggered exocytosis at all Ca²⁺ concentrations except very low ones (Sugita et al., 2001), and a similar trend was observed for the synaptotagmin 3 C₂A-domain (Figure 5C). Interestingly, the synaptotagmin 7 C₂B-domain exhibited the opposite behavior, in that inhibition of exocytosis was most effective at low Ca²⁺ concentrations (Figure 6C), suggesting that exocytosis at very low Ca²⁺ concentrations is primarily mediated by this C₂-domain, and is not due to calmodulin. This suggestion is also supported by the finding that Sr²⁺, a divalent cation that is unable to activate calmodulin (Chao et al., 1984) but does bind to synaptotagmins, strongly activates exocytosis in PC12 cells (see below) and that calmodulin antagonists had no effect on exocytosis under our assay conditions (data not shown).

Sr²⁺ as a Ca²⁺ mimetic for synaptotagmins and PC12 cell exocytosis

As an independent approach to test the potential Ca²⁺ sensor function of synaptotagmins in exocytosis, we examined the ability of Sr²⁺ to substitute for Ca²⁺. Differently to previous studies (Li et al., 1995b), the current experiments were performed using Sr²⁺/EGTA buffers to control the free Sr²⁺ concentration more precisely, to eliminate potential interference by contaminating Ca²⁺ and to allow a direct comparison of Ca²⁺ binding and of Ca²⁺ triggered exocytosis in PC12 cells. Figure 7 shows that at all Sr²⁺ concentrations examined, the C₂A-domains of vesicular synaptotagmins 1 and 2 did not bind to phospholipids, whereas synaptotagmins 3 and 7 exhibited similar apparent Sr²⁺ affinities in both buffers (EC₅₀ ≈ 30–50 µM). The resin-based and the liposome centrifugation assays for measuring the apparent Sr²⁺ affinity gave identical results (Figure 7). The specificity of the reactions was confirmed with the synaptotagmin 3 mutant D333N, which, similarly to Ca²⁺-dependent phospholipid binding (Figure 1C), was unable to support Sr²⁺-dependent phospholipid binding (data not shown).

We next measured the response of permeabilized PC12 cells to increasing Sr²⁺ concentrations and examined the effects of recombinant synaptotagmin C₂A-domains on this response. As before, GST alone was employed in all experiments as a negative control analyzed in parallel with the test proteins. Under control conditions, Sr²⁺ was as effective as Ca²⁺ in triggering exocytosis (Figure 8). A bell-shaped Sr²⁺ concentration curve was measured similarly to the Ca²⁺ concentration curve, although much higher Sr²⁺ concentrations (100–300 µM) were required for maximal effect. Similarly to Ca²⁺-triggered exocytosis, the synaptotagmin 1 C₂A-domain had only a limited effect on Sr²⁺-triggered exocytosis, with a moderate inhibition at high Sr²⁺ concentrations (Figure 8A). In contrast, the wild-type synaptotagmin 3 C₂A-domain completely abolished Sr²⁺-triggered exocytosis (Figure 8B), but the mutant synaptotagmin 3 C₂A-domain (which is unable to form Sr²⁺-dependent phospholipid complexes; see Figure 7) had no effect (Figure 8C). The synaptotagmin 7 C₂A-domain was even more effective in inhibiting exocytosis than the synaptotagmin 3 C₂A-domain, and depressed release below non-stimulated control levels (Figure 8D).

Expression and purification of recombinant proteins

All recombinant proteins were expressed as bacterial GST fusion proteins in pGEX-KG (Guan and Dixon, 1991). Most vectors have been described previously (Sugita et al., 2001) except for the synaptotagmin 3 plasmids, which were constructed by PCR using standard methods (Sambrook et al., 1989). GST fusion proteins were purified essentially as described (Guan and Dixon, 1991; Sugita and Südhof, 2000; Sugita et al., 2001) and C₂B-domain proteins were additionally treated to remove the bacterial contaminants that stick to these domains (Ubach et al., 2001). For this purpose, extracted proteins were incubated with 1500 U/l benzonase (Novagen Inc.) for 1 h at room temperature before binding to glutathione–agarose. The agarose was packed into a polypropylene column with a paper disc column (Quick-Sep, Isolab) and additionally washed five times with 5 ml of phosphate-buffered saline (PBS), 1 M NaCl, 1% Triton X-100, 1 mM EDTA and 0.1 g/l phenylmethylsulfonyl fluoride (PMSF), followed by six washes with 5 ml of PBS, 1 M NaCl, 1 mM EDTA and 0.1 g/l PMSF.

Phospholipid binding assays

These were performed essentially as described using GST fusion proteins either immobilized on glutathione–agarose in a resin assay (Davletov and Südhof, 1993; Li et al., 1995a,b) or in solution with a centrifugation assay (Fernandez et al., 2001) with three different buffers. Buffer A: NaCl-based Ca²⁺/EGTA buffer (50 mM HEPES–NaOH pH 6.8, 0.1 M NaCl, 4 mM Na₂EGTA, and total Ca²⁺ or Sr²⁺ to produce the indicated concentrations of free Ca²⁺ or Sr²⁺); buffer B: K-glutamate-based Ca²⁺/EGTA buffer (120 mM K-glutamate, 20 mM K-acetate, 2 mM EGTA, 20 mM HEPES–NaOH pH 7.2, and total Ca²⁺ or Sr²⁺ to produce the indicated concentrations of free Ca²⁺ or Sr²⁺); buffer C: NaCl-based Ca²⁺-only buffer [50 mM HEPES–NaOH pH 6.8; 0.1 M NaCl passed through AG MP-50 resin (Bio-Rad) to eliminate Ca²⁺ contamination after resin ion conversion to Na⁺ from H⁺, with Ca²⁺ or Sr²⁺ added directly to the buffer from 1 M stock solution (Fluka Chemical Corp., NY) to achieve the indicated total Ca²⁺ or Sr²⁺ concentration]. For composition of Ca²⁺/EGTA and Sr²⁺/EGTA buffers, the respective amounts of total Ca²⁺ or Sr²⁺ needed to achieve a defined free Ca²⁺ or Sr²⁺ concentration were calculated with EqCal software (Biosoft, Ferguson, MI) using buffer constants given by the program or published by Martell and Smith (1974). All buffers were prepared in high-resistance MilliQ water using a 1 M Ca²⁺ standard solution (Fluka Chemical Corp.).

Resin assay. PS and PC (1.75 mg total; Avanti Polar Lipids, AL) were dissolved in chloroform, mixed in a 1:3 weight ratio including <0.01% ³H-PC (Amersham Pharmacia Biotech, NJ) and dried under a stream of nitrogen. Dried lipids were resuspended in Ca²⁺-free buffer A or B (10 ml) by vigorous vortexing for 1 min, sonicated for 5 min in a waterbath sonicator (model G112PIG; Laboratory Supply Co. Inc., NJ; output: 80 kc, 80 W) and centrifuged (15 min at ∼5000 g) to remove aggregates. Binding assays contained ∼25 µg of recombinant protein, with 1 g protein/l wet glutathione beads. Beads were equilibrated in 0.1 ml of the respective binding buffers (buffer A or B with 8.75 µg of phospholipids and 0.025 µCi of ³H-PC). Binding reactions were incubated for 10 min at 30°C with vigorous shaking, briefly centrifuged, washed three times with 0.8 ml of the respective binding buffers and phospholipid binding quantified by scintillation counting of the beads (Beckman LS6000SC; Beckman Instruments Inc., CA).

Centrifugation assay. Phospholipids (PS/PC = 25/75, w/w) in chloroform were dried as a thin layer under a stream of nitrogen. Buffers A, B or C containing 0.5 M sucrose were added to the dried phospholipid layer, vortexed for 20 min and sonicated for 5 min in a bath sonicator (model G112SP1G; Laboratory Supply Co. Inc.). After liposome formation, 4 vols of each buffer without sucrose were added and centrifuged to separate heavy liposome from free phospholipids (100 000 g for 30 min). Heavy liposomes were washed once and re-pelleted (13 000 r.p.m. for 10 min). Liposomes were resuspended in each buffer with various concentrations of free Ca²⁺ and used within 1 h. Purified soluble recombinant GST–synaptotagmin C₂-domains (6 µg) and liposomes (100 µg of phospholipids; total volume = 1 ml) were incubated for 10 min on an Eppendorf thermal mixer at 30°C and 1400 r.p.m., liposomes were re-isolated by centrifugation (20 800 g for 10 min) and washed three times with 0.5 ml of the corresponding buffers. Chloroform:methanol (1:2, v/v) solution was added into the pelleted liposomes to denature protein and dissolve phospholipids. After centrifugation (20 800 g for 15 min), the protein precipitate was resuspended in 30 µl of 2× SDS sample buffer and analyzed by SDS–PAGE and Coomassie Blue staining.

Permeabilized PC12 cell secretion assay

Secretion assays were conducted using freeze–thaw permeabilization of PC12 cells (Klenchin et al., 1998; Sugita et al., 2001). PC12 cells were maintained in 75 cm² flasks (uncoated) in RPMI1640 with 5% fetal bovine serum (heat inactivated), 10% horse serum (heat inactivated), penicillin (50 U/ml) and streptomycin (50 U/ml) at 37°C in 5% CO₂. For each assay, PC12 cells were plated in 10 cm² dishes (Costar) at 40–50% confluency. Two to three days after plating (70% confluency), PC12 cells fresh medium (total 10 ml) containing 4 µl of [³H]norepinephrine (NEN; 1 µCi/µl in stock) and 0.5 mM ascorbic acid was added. After 24 h, cells were washed with physiological saline solution (145 mM NaCl, 5.6 mM KCl, 2.2 mM CaCl₂, 0.5 mM MgCl₂, 5.6 mM glucose, 15 mM HEPES–NaOH pH 7.4), detached from the wells by pipetting a stream of Ca²⁺-free ice-cold buffer B, washed twice with 6 ml of ice-cold buffer B and resuspended in 6 ml of buffer B in a 15 ml cone tube. For permeabilization, cells were frozen overnight at –80°C, thawed by warming the tube at room temperature, 10 mM EGTA was added and the thawed cells were left on ice for 1–5 h to allow efficient extraction of soluble proteins. Resulting cell ghosts were washed three times with 3 ml of ice-cold Ca²⁺-free buffer B containing 1 g/l BSA, with centrifugations at 1000 g for 5 min. About 18–36 secretion reactions were conducted with each preparation with standard reaction mixes containing in 1.5 ml microcentrifuge tubes (total volume 0.1 ml): washed cell ghosts, 2 mM ATP, 2 mM MgCl₂, 10 µl rat brain cytosol (10 mg/ml) in buffer B with various concentrations of CaCl₂ and recombinant proteins. Reactions were incubated for 30 min at 30°C, terminated by chilling to 0°C and samples were centrifuged at 4°C for 3 min at 20 800 g. Supernatants and the pellets solubilized in 1% Triton X-100 were analyzed by liquid scintillation counting.

PC12 cell transfections

Vectors encoding EYFP fusion proteins were constructed in a CMV-based expression plasmid using standard procedures (Sugita et al., 2001). For transfection, PC12 cells were plated on polylysine-coated coverslips in RPMI medium supplemented with 10% fetal bovine serum and 5% horse serum. After 1 day in culture, cells were transfected with appropriate vectors using Tfx-50 (Promega). Forty-eight to 72 h after transfection, culture medium was replaced with 0.2 ml of minimal essential medium containing 10 µM FM4-64 (Molecular Probes Inc.) and cells were incubated at 37°C for 5–10 min. FM4-64-labeled coverslips were mounted on slides and imaged using a Bio-Rad MRC-1024 confocal microscope with a 100× (NA1.3) objective.

Miscellaneous procedures

SDS–PAGE and immunoblotting were performed using standard procedures (Laemmli, 1970; Johnston et al., 1989). Immunoblots were developed by enhanced chemiluminescence (Amersham).