Role of Lipid Microdomains in P/Q-type Calcium Channel (Cav2.1) Clustering and Function in Presynaptic Membranes*

Lipid microdomains can selectively include or exclude proteins and may be important in a variety of functions such as protein sorting, cell signaling, and synaptic transmission. The present study demonstrates that two different voltage-gated calcium channels, which both interact with soluble N-ethyl-maleimide-sensitive fusion protein attachment protein receptor (SNARE) proteins but have distinct subcellular distributions and roles in synaptic transmission, are differently distributed in lipid microdomains; presynaptic P/Q (Cav2.1) but not Lc (Cav1.2) calcium channel subtypes are mainly accumulated in detergent-insoluble complexes. The immunoisolation of multiprotein complexes from detergent-insoluble or detergent-soluble fractions shows that the α1A subunits of Cav2.1 colocalize and interact with SNARE complexes in lipid microdomains. The altered organization of these microdomains caused by saponin and methyl-β-cyclodextrin treatment largely impairs the buoyancy and distribution of Cav2.1 channels and SNAREs in flotation gradients. On the other hand, cholesterol reloading partially reverses the drug effects. Methyl-β-cyclodextrin treatment alters the colocalization of Cav2.1 with the proteins of the exocytic machinery and also impairs calcium influx in nerve terminals. These results show that lipid microdomains in presynaptic terminals are important in organizing membrane sites specialized for synaptic vesicle exocytosis. The cholesterol-enriched microdomains contribute to optimizing the compartmentalization of exocytic machinery and the calcium influx that triggers synaptic vesicle exocytosis.

When activated, the Ca 2ϩ channels in many different cell types mediate a rapid Ca 2ϩ influx that triggers important intracellular events (Ref. 1 and references therein). In neurons and neuroendocrine cells, the Ca 2ϩ entering through voltagegated Ca 2ϩ channels acts as the second messenger of electrical signaling and initiates regulated exocytosis and synaptic transmission.
Five types of voltage-gated calcium channels have been identified in neurons on the basis of their physiological and phar-macological properties: L, N, P/Q, R, and T (recently referred to as Ca v 1, Ca v 2.2, Ca v 2.1, Ca v 2.3, and Ca v 3) (2). These distinct calcium channels show different compartmentalization on neuronal plasma membrane, e.g. Ca v 2.2, Ca v 2.1, and Ca v 2.3 are mainly located in the nerve terminals, where they form clusters in specialized regions of the presynaptic membrane (the active zones), which also contain predocked synaptic vesicles (Refs. 1, 3, and references therein). This clustered accumulation in specific membrane regions is presumably important for the roles of calcium channels in neurotransmitter release.
It has been proposed that the formation of channel clusters may involve cytoskeletal elements and/or interactions with modular adaptor proteins (4,5), whereas another proposed mechanism is based on the formation of microdomains within the lipid bilayer that recruit specific proteins and form platforms for protein sorting and/or signal relay stations for intracellular signaling (see Refs. 6 -9 for recent reviews). Recent data have demonstrated that various components of the synaptic vesicle fusion machinery, including the soluble N-ethylmaleimide-sensitive fusion protein attachment protein receptor (SNARE) 1 proteins (10 -12), syntaxin 1 (syn1), 25-kDa synaptosome-associated protein (SNAP-25), vesicle-associated membrane protein (VAMP-2, also called synaptobrevin), and synaptotagmin1, are localized in the cholesterol-enriched microdomains of neuroendocrine and epithelial cells (13)(14)(15). These findings suggest that lipid microdomains may be involved in defining sites for neurosecretory vesicle release and may help to concentrate other membrane proteins that are important for exocytosis.
On the basis of the above, we investigated the lipid microdomain distribution of the two voltage-gated calcium channel subtypes Ca v 1.2 (Lc) and Ca v 2.1 (P/Q). These channels are particularly interesting because they both interact with SNARE proteins (Refs. 1, 16 -18, and references therein) but have different cell compartmentalization and play different roles in synaptic vesicle exocytosis. A large proportion of Ca v 1.2 channels is concentrated in the soma and at the tips of dendritic spines, where they often colocalize with the ␤2-adrenergic receptor to form postsynaptic complexes regulating synaptic transmission (19,20), whereas channels 2.1 are mainly clustered in the presynaptic terminals of many neurons, where they mediate the rapid calcium influx that triggers synaptic vesicle fusion (1,21).
The results of the study show that Ca v 2.1 but not Ca v 1.2 channels are widely distributed in lipid microdomains, in which they colocalize and interact with SNARE protein complexes as shown by immunoisolation. Altering the organization of synaptosomal membrane lipid microdomains by means of the cholesterol-complexing drug, methyl-␤-cyclodextrin (M␤CD), alters the distribution of Ca v 2.1 and inhibits their interactions with SNAREs. Furthermore, cholesterol depletion reduces calcium influx in the synaptosomes in response to depolarization. Taken together, these data support the conclusion that lipid microdomains in presynaptic membranes are important for defining organized sites in which specific voltage-gated calcium channel subtypes and SNAREs are colocalized.

Rat Brain Subcellular Fractionation
Rat brain fractionation was carried out by means of differential centrifugation, essentially as described previously (23,24). Briefly, the dissected cerebral cortices were homogenized in 4 mM HEPES-NaOH, pH 7.3, containing 0.32 M sucrose and protease inhibitors (2 g/ml pepstatin, 2 g/ml aprotinin); the total homogenate was centrifuged for 10 min at 800 ϫ g, and the postnuclear supernatant was collected and centrifuged as described previously (23) to yield a pellet corresponding to the synaptosomal fraction (P2) and a supernatant (S2). The S2 containing the remaining membrane-bound vesicles and organelles of the total homogenate was centrifuged at 165,000 ϫ g for 2 h to yield a high-speed supernatant corresponding to the cytosol and a pellet (P3) enriched in cell body organelles (24).
M␤CD Treatment-The P2 samples (660 g of proteins) were resuspended in 250 l of buffer A supplemented with protease inhibitors, incubated with or without 10 or 30 mM M␤CD (final concentration) for 40 min at 37°C and then recentrifuged. The supernatants were completely removed, and the synaptosomal pellets were resuspended in the appropriate buffer for cholesterol reloading or Triton X-100 solubilization and gradient analysis.
Cholesterol Reloading-The cholesterol and M␤CD inclusion complexes were prepared as described by Klein et al. (25). Briefly, 15 mg of cholesterol dissolved in methanol:chloroform (2:1) were added to 10 ml of a 5% M␤CD solution and incubated for 90 min at 80°C. After M␤CD treatment, the P2 samples (500 g) were resuspended in 320 l of buffer A containing 10 mM cholesterol/M␤CD complexes and incubated for 30 min at 37°C. After reloading, the P2 samples were centrifuged, and the pellets were resuspended in buffer A for further analysis. Cholesterol concentrations were determined in aliquots of the untreated (control), M␤CD-treated, and M␤CD/cholesterol complextreated P2 samples. The aliquots were centrifuged, the synaptosomal pellet was resuspended in a lysis buffer (0.5% Triton X-100, 0.5% deoxycholic acid-sodium salt, 150 mM NaCl, and 20 mM Tris-HCl, pH 7.2), and the amount of cholesterol was determined using the Infinity cholesterol reagent according to the Sigma protocol. The protein con-centrations in the M␤CD and cholesterol/M␤CD-treated samples were determined using the Bio-Rad Protein Assay.

Detergent Solubilization and Sucrose Flotation Gradients
The proteins from the untreated P2 and P3 samples, or the P2 samples treated with saponin, M␤CD, or cholesterol/M␤CD, were adjusted in 250 l of buffer A (supplemented with protease inhibitors) containing a final concentration of 1% (w/v) Triton X-100 (protein: detergent ratios 1:1.25 or 1:3.6). The samples were then incubated for 30 min on ice. The detergent-insoluble complexes were separated by centrifugation on discontinuous gradients as described previously (15,26). Briefly, 250 l of buffer A containing 2.4 M sucrose was added to 250 l of each sample, which was then placed in a centrifuge tube and overlaid with 1 ml of 0.9 M, 0.5 ml of 0.8 M, 1 ml of 0.7 M, and 1 ml of 0.1 M of sucrose solutions (all prepared in buffer A). The discontinuous gradients were centrifuged at 335,000 ϫ g for 16 h using a rotor SW 55 Ti (Beckman Instruments), after which 500-l fractions were taken. The pellets were resuspended in 500 l of buffer A. Forty to 60 l of each fraction were analyzed by means of SDS-PAGE and Western blotting. The sucrose concentration in each fraction was determined by refractometry; the mean values of seven gradients from three different experiments are shown.

Immunoisolation of Ca v 2.1 and SNARE Complexes
The immunocomplexes from the sucrose gradient fractions were isolated by means of immunoprecipitations or immunoaffinity columns. For the immunoisolation of the proteins colocalized in lipid microdomains, particular care was taken to preserve the properties of the microdomains. All of the procedures were carried out on ice or at 4°C. Aliquots of the gradient fractions were adjusted to 0.3 M sucrose with buffer A without detergent and incubated with anti-syn1, VAMP-2, or SNAP-25 monoclonal antibodies prebound to protein G beads. As a control, equal sample amounts were immunoprecipitated with nonimmune mouse IgG prebound to protein G beads. After overnight incubation at 4°C with gentle mixing, the beads were centrifuged at 2000 ϫ g for 3 min, washed three times with buffer A containing 0.3% Tween 20, and resuspended in Laemmli's sample buffer (27). The immunocomplexes and aliquots (1:5) of the samples used for immunoisolation (inputs) were analyzed by SDS-PAGE.
For the immunoisolation of the protein complexes, immunoaffinity column purification or immunoprecipitations were carried out after solubilization of the lipid microdomains. Equal aliquots (1.5 ml) of fractions 3 and 7 were diluted (to 6 ml) in buffer A containing 0. 3 M sucrose and 1% Triton X-100 and then incubated for 30 min on ice, followed by 15 min at 28°C to solubilize the lipid microdomains. The samples were incubated overnight at 4°C with anti-syn1 antibodies covalently conjugated with protein G-Sepharose (Amersham Pharmacia Biotech). The beads were then washed three times with 0.2% octylglucopyranoside in PBS, and the protein complexes were eluted with 0.2 M glycine, pH 2.2, containing 0.2% octylglucopyranoside, neutralized with potassium phosphate, and separated on polyacrylamide gels.
The proteins in the gels were stained with SYPRO Orange (Bio-Rad) and analyzed using a VersaDoc imaging system (Bio-Rad) or transferred to nitrocellulose filters for immunoblotting.

Immunoblotting and G M1 Detection
The samples underwent SDS-PAGE on 6, 10, or 12% polyacrylamide gels and were then blotted onto nitrocellulose (NC) membranes (Schleicher & Schuell, Dassel, Germany) with 0.45-m pores. The Western blots were analyzed as described previously (24). For all antibodies except anti-␣1A and ␣1C, the blots were blocked overnight in 5% not-fat milk in Tris-buffered saline (TBS), washed in a buffer containing 5% non-fat milk and 0.3% Tween 20 in TBS (immunoblotting buffer), incubated for 2 h with the primary antibody diluted in immunoblotting buffer, and then washed and incubated with the appropriate peroxidase-conjugated secondary antibodies. After another series of washes, peroxidase was detected using chemiluminescent substrates (Pierce). The antibodies directed against the calcium channel subunits were diluted from 1:400 to 1:2000 in 1% bovine serum albumin, 0.1% Tween 20 in TBS. To detect ganglioside G M1 , the samples were separated on 15% polyacrylamide gels, transferred to NC membranes, and incubated with peroxidase-coupled choleratoxin subunit B diluted 1:100,000 in immunoblotting buffer. After extensive washing, peroxidase was detected using chemiluminescent substrates. Alternatively, 0.5-1 l of each gradient fraction was spotted directly onto NC membranes, which were then probed with peroxidase-coupled choleratoxin subunit B as described above.

I-Labeled -Conotoxin MVIIC Binding
125 I-Labeled -CTxMVIIC binding was performed essentially as described previously (28). Because sucrose can inhibit toxin binding in a dose-dependent manner (29), 100 l of each fraction were brought to 1 ml in a buffer containing a final concentration of 0.3 M sucrose, 0.1% bovine serum albumin, 100 mM NaCl, 0.1 mM EGTA, 0.1 mM EDTA, and 10 mM HEPES-NaOH, pH 7.4 (buffer B). Each sample was incubated with 0.3 nM of 125 I-labeled -CTxMVIIC for 40 min at 37°C and then filtered through GF/B filters presoaked in 1% polyethyleneimine. After washing in buffer B, the radioactivity of the filters was counted in a gamma counter. Each point was evaluated in triplicate, with nonspecific 125 I-labeled -CTxMVIIC binding being evaluated for each group by means of the parallel incubation of samples in the presence of an excess of unlabeled toxin (100 nM).

Ca 2ϩ Influx
P2 samples were resuspended in a buffer containing 145 mM choline chloride, 5 mM KCl, 1.2 mM CaCl 2 , 1.2 mM MgCl 2 , 2.4 mM NaH 2 PO 4 , 10 mM glucose, and 10 mM HEPES-NaOH, pH 7.4 (buffer C). Aliquots (660 g/250 l) were incubated with or without M␤CD (10 or 30 mM) for 40 min at 37°C. Cholesterol and protein concentrations were determined in aliquots of untreated (control) and M␤CD-treated samples as described above. 45 Ca 2ϩ uptake was initiated by adding 320 l of prewarmed basal medium (buffer C) or depolarizing buffer (62.5 mM choline chloride, 87.5 mM KCl, 1.2 mM CaCl 2 , 1.2 mM MgCl 2 , 2.4 mM NaH 2 PO 4 , 10 mM glucose, and 10 mM HEPES-NaOH, pH 7.4) containing 3 Ci/ml of [ 45 Ca]Cl 2 . When Cd 2ϩ or EGTA was used, they were included in the depolarizing buffer at final concentrations of 100 and 10 M, respectively. 45 Ca 2ϩ uptake was terminated after 15 s by means of the addition of 1 ml of ice-cold buffer C containing 100 M CdCl 2 and immediate filtration through Whatman GF/B filters, followed by washes with ice-cold buffer C supplemented with Cd 2ϩ . The radioactivity of each filter was counted using a beta counter, and the data were expressed as nanomoles of 45 Ca 2ϩ /mg of protein.

Electron Microscopy
The synaptosomes (P2) were fixed with 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4, for 2 h at 4°C, post-fixed with 1% osmium tetroxide for 1 h at room temperature, washed extensively, dehydrated in ethanol series, and then embedded in Epon. Sections of a pale-gold, silver interference color were obtained with a Reichert Ultracut E ultramicrotome, collected on copper grids, counterstained with a saturated aqueous solution of uranyl acetate and Reynold's lead citrate, and finally examined using a Philips CM10 transmission electron microscope.

Calcium Channels and SNAREs in Detergent-resistant Membranes Isolated from
Synaptosomes-Over the last 10 years, various procedures have been developed for isolating the detergent-insoluble complexes obtained by treating cells or membranes with weak detergents at low temperature. These detergent-resistant complexes identify lipid microdomains that are thought to resemble the self-organized lipid-protein complexes present in the membrane before detergent treatment (Refs. 6 -9 and references therein).
We analyzed the lipid microdomains in rat brain fractions enriched in nerve terminal (synaptosome) or cell soma membranes following well-established procedures. The pellets enriched in synaptosomes (P2) were separated from the homogenates of rat brain cortices by means of differential centrifugation (23). The supernatants containing the remaining membranes and organelles of the cell soma (including the endoplasmic reticulum and Golgi apparatus) (24) were centrifuged at high speed to obtain P3 pellets. Low-density, Triton X-100-insoluble complexes prepared from P2 and P3 were then separated from the soluble proteins using flotation equilibrium sucrose density gradients as described previously (15,26). It has been shown previously that this method separates the Triton X-100-insoluble complexes (which peak in fractions 3 and 4) from the fully soluble and cytoskeletal proteins (which peak in fractions 7-8). In line with previous data (26), the soluble fraction marker transferrin receptor was only detected in the very high-density fractions (fraction 8 in the P2 gradients; fractions 7 and 8 in the P3 gradients; Fig. 1). On the contrary, the low-density marker G M1 was detected mainly in the low-density fractions (peaking in fractions 3-4).
We then examined the gradient distribution of the molecular components of the exocytic machinery and the ␣ subunits of Ca v 2.1 and Ca v 1.2 (Fig. 1). The proteins from the fractions collected from P2 or P3 gradients were separated on 12% gels, transferred onto NC filters, and probed with polyclonal antibodies directed against synaptotagmin1 (22) and munc-18-1 (15) or commercially available monoclonal antibodies against the SNARE proteins syn1, SNAP-25, and VAMP-2. As shown in  Fig. 1, significant amounts of SNAREs and synaptotagmin1 were found to float in the low-density fractions (mainly in fractions 3-4) collected from the Triton X-100-treated P2 or P3 samples, although larger amounts were detected in the highdensity fractions, as reported previously for neuroendocrine PC12 cells (13). However, unlike in PC12 cells, a small but significant amount of the syn1-accessory protein munc-18-1 was also found in the low-density fractions of the P2 gradients (Fig. 1).
Not all of the proteins known to play a role in exocytic machinery function floated. In line with previous results (13), immunoblotting did not detect synaptophysin (which interacts with VAMP-2 in synaptic vesicle membrane) or complexins (cytosolic proteins that selectively bind to assembled SNARE complexes) (10) in the low-density fractions of the P2 gradients (data not shown).
We then investigated the flotation gradient distribution of Ca v 2.1 and Ca v 1.2 channels. The blots were probed with commercially available polyclonal antibodies directed against the ␣1A or ␣1C subunits. Before being used on the gradient fractions, the antibodies were immunoblot tested on P2 membrane proteins, in which they recognized bands of the expected molecular size; in 6% polyacrylamide gels, the anti-␣1A antibody recognized a major band of ϳ190 kDa and a very minor polypeptide of ϳ210 kDa, which appears after longer exposure (data not shown) and is consistent with the results of previous work on rat brain and chromaffin cell membranes (30,31). The antibody against the ␣1C subunit detected three bands of ϳ190, 200, and 220/230 kDa (data not shown), as reported previously (19). When the blots were immunostained with irrelevant rabbit IgG as a control, no bands of the same molecular sizes as those revealed by the specific anti-␣ 49 subunit antibodies were immunodetected.
Analysis of the flotation gradient fractions using the anti-␣1A or ␣1C antibodies revealed that the Ca v 2.1 ␣1A subunit (immunodetected as the polypeptide of ϳ190 kDa) was almost exclusively present in the lower-density fractions (fractions 3-4) of the P2 gradients (Fig. 1), whereas the ␣1C subunit was mainly detected in the high-density fractions, although a minor portion was also immunodetected in the lighter fractions (Fig.  1). Finally, the amounts of ␣1A and ␣1C found in the Triton X-100-treated P3 samples were much less than those detected in the P2 membranes, and only very faint immunoreactive bands were detected in the high-density fractions.
To analyze further the presence of Ca v 2.1 in the gradient fractions and confirm the immunoblotting results, we used a detection system based on 125 I-labeled -CTxMVIIC binding. The toxin-binding experiments were performed in gradient fractions collected from the Triton X-100-treated P2 samples, and the data of multiple experiments (n ϭ 3) were averaged. As shown in Fig. 2, the majority of the toxin labeling was again found in fractions 3-4, whereas almost no toxin binding was revealed in fraction 7. These results are similar to those obtained by means of immunoblotting and further demonstrate that the antibodies recognized ␣1A subunits. The proportionally higher amount of 125 I-labeled -CTxMVIIC binding activity in fraction 4 (in comparison with the immunoblotting distribution shown in Fig. 1) may be attributable to slight variations in the different experiments and the limited linear range of the enhanced chemiluminescence technique.
Taken together, these data strongly suggest that Ca v 2.1 and a portion of the proteins of the exocytic machinery are localized in lipid microdomains isolated from synapses.
Effects of Cholesterol Perturbation on Ca v 2.1 and SNARE Protein Distribution in Detergent-insoluble Fractions-Because cholesterol is critical for the formation and stability of lipid microdomains, we analyzed the effects of cholesterol perturbation on the buoyancy of SNARE proteins and ␣1A subunits using saponin and M␤CD. Saponin interacts with cholesterol and efficiently disrupts lipid microdomains, whereas M␤CD is capable of including cholesterol in its hydrophobic pocket, thus perturbing the organization of lipid microdomains (25,32). Furthermore, M␤CD/cholesterol complexes can be used to load membranes with exogenous cholesterol (25).
When P2 samples were incubated with M␤CD, the effect on cholesterol levels was dose dependent; 10 mM M␤CD partially modified the amount of cholesterol (19.4 Ϯ 8% versus control, n ϭ 4), but 30 mM reduced the level of endogenous cholesterol to 60% (60.7 Ϯ 4.8%, n ϭ 6) without inducing any significant change in protein concentration (data not shown). We therefore used 30 mM M␤CD in the subsequent experiments. As shown in Fig. 3, in comparison with the mock-treated (control) samples, the G M1 marker in the P2 samples treated with M␤CD shifted to the higher-density gradient fractions. Peak G M1 reactivity was found in fractions 2-4 of the control sample but in fractions 4 -6 of the M␤CD-treated samples, thus indicating that the organization of the detergent-resistant membranes was altered by the cholesterol-chelator (Fig. 3). In line with this, ␣1A also shifted to the higher-density fractions: in the control P2 samples, the pore-forming subunit of the channel was immunodetected in fractions 3 and 4 but was mainly found in fraction 5 of the parallel samples treated with M␤CD and analyzed by means of gradient centrifugation (Fig. 3, A and B). Similarly, M␤CD treatment shifted the SNARE proteins from fractions 3-4 to the bottom of the gradients, with a substantial aliquot being detected in fraction 5 (Fig. 3B).
To investigate whether the effects on SNARE and ␣1A buoyancy were attributable to cholesterol removal, M␤CD-treated P2 samples were incubated with M␤CD/cholesterol complexes. After reloading, the cholesterol level was very similar to that observed in the control samples (98%). Analysis of G M1 in the gradient fractions treated with M␤CD/cholesterol complexes revealed a distribution pattern that was similar to that observed in the controls (compare Figs. 3 and 1), thus suggesting that the procedure at least partially restored lipid microdomain organization. Aliquots of SNARE proteins were also recovered in gradient fractions 3-4 (Fig. 3), once again with a distribution similar to that observed in the control P2 samples (see Fig.  1). On the contrary, cholesterol reloading had only a slight effect on the distribution of ␣1A, and only a very small portion of the Ca v 2.1 subunits was recovered in the lighter fractions of the gradient (mainly in fraction 4); the majority of the protein was still detected in fraction 5 (Fig. 3). These results may indicate that cholesterol depletion modifies the physical prop- erties of ␣1A subunits to such an extent that they cannot be fully restored by reloading, or that the association of ␣1A with lipid microdomains requires other lipids removed by M␤CD that cannot be restored by means of cholesterol reloading.
Given the partial solubilization achieved with M␤CD (possibly because of the incomplete removal of cholesterol from the P2 membranes), we incubated the samples with 1% saponin before Triton X-100 treatment. Two different protein:saponin ratios were used (see "Experimental Procedures"), and in both cases, G M1 was found in the very high-density fractions (Fig. 4), thus demonstrating that saponin affected the organization of the lipid microdomains more than the M␤CD treatment (Fig.  3). In line with the distribution of G M1 , the SNARE proteins and ␣1A were found mainly at the bottom of the gradients (Fig.  4). These results indicate that efficient disruption of lipid microdomains completely solubilizes Ca v 2.1 subunits and SNAREs.

Ca v 2.1 and Exocytic Complex Proteins Colocalize in Lipid
Microdomains: Perturbed Cholesterol Levels Affect Ca v 2.1 and SNARE Colocalization-We next investigated whether Ca v 2.1, SNAREs, synaptotagmin1, and munc-18-1 colocalize in lipid microdomains and whether cholesterol extraction by M␤CD affects the possible colocalization and interaction of ␣1A with the proteins of the exocytic machinery. To this end, we immunoisolated protein complexes from the fractions obtained from P2 samples that were untreated or incubated with 30 mM M␤CD and separated on flotation gradients performed in parallel (see Fig. 5). In the immunoisolation procedure, we took care to preserve the physical properties of the lipid microdomains (see "Experimental Procedures"). ␣1A and/or SNARE complexes were immunoisolated by means of immunoprecipitation with monoclonal antibodies directed against syn1 or VAMP-2 from: 1) a pool of fractions 3 and 4 (3-4) containing detergent-resistant complexes; 2) a pool of fractions 5 and 6 (5-6) containing the majority of ␣1A subunits after M␤CD treatment; and 3) fraction 7 containing the fully soluble proteins (Fig. 5A, input). In the absence of M␤CD, substantial amounts of ␣1A were coimmunoisolated from the lipid microdomain-containing fractions (3)(4) together with SNAP-25, VAMP-2, and synaptotagmin1 (Fig. 5, B and C). After M␤CD treatment, although present with SNAREs in fractions 5-6 (see input), the ␣1A subunit was not coimmunoisolated with the use of anti-syn1 or VAMP-2 antibodies (Fig. 5, B and C). Similar results were obtained by means of immunoprecipitation with anti-SNAP-25 antibodies (data not shown).
When immunoprecipitations were carried out using nonimmune mouse IgG, neither the proteins of the exocytic complex nor the Ca v 2.1 ␣1A subunits were immunodetected by Western blotting using specific antibodies (data not shown).
Taken together, these data suggest that Ca v 2.1 and SNAREs are contained within the same detergent-insoluble complex, but this colocalization is not maintained after partial extraction of cholesterol with M␤CD.
Ca v 2.1 Interacts with SNARE Complexes in Lipid Microdomains-Because it has been widely demonstrated that there are physical interactions between the proteins of the SNARE complex and the pore-forming subunits of calcium channel subtypes (see Refs. 16 -18 for recent reviews), we investigated whether SNAREs and ␣1A interact in the lipid microdomains of the synaptosomal fraction. To do this, protein complexes were isolated from fractions 3 or 7 after complete solubilization (see "Experimental Procedures") using anti-syn1 affinity columns or immunoprecipitations and analyzed by means of SDS-PAGE and Western blotting ( Fig. 6 and data not shown). The gels were stained with SYPRO Orange before protein transfer to NC filters to analyze the total pattern of the proteins coisolated with anti-syn1 antibodies. This procedure revealed the presence in both fractions of two major polypeptides of about 36 and 25 kDa, corresponding to syn1 and SNAP-25 (as also verified by means of immunoblotting, data not shown), and a large number of other molecules (Fig. 6A). Immunoblotting was used to investigate the presence of other components of the exocytic complex and ␣1A subunits in the immunoisolated proteins. As shown in Fig. 6B, VAMP-2, synaptotagmin1, and ␣1A were immunoisolated with syn1 from fraction 3, but interestingly, no significant amount of munc-18-1 was immunoprecipitated. On the contrary, the complexes isolated from fraction 7 contained substantial amounts of munc-18-1, as well as SNAREs and synaptotagmin1. Because interactions between syn1 and munc-18-1 or the cognate SNAREs are mutually exclusive (33), these results suggest that, in lipid microdomains, syn1 is largely engaged in complexes with the cognate SNARE, synap-totagmin1, and Ca v 2.1 to form the molecular machinery for neurotransmitter release (Refs. 10, 16, 17, and references therein). The association of munc-18-1 with the membrane regardless of syn1 binding is in line with the data reported by Garcia et al. (34), who demonstrated that a pool of membraneassociated munc-18-1 does not interact with syn1.
Role of Lipid Microdomains in Nerve Terminal 45 Ca 2ϩ Influx-We next examined whether modified cholesterol levels in synaptic membranes have effects on nerve terminal calcium influx. P2 samples were incubated with 10 or 30 mM M␤CD or left untreated. Electron microscopy showed that the drug treatment did not lead to any major alterations in nerve terminal morphology (Fig. 7, A and B). When the efficiency of the calcium influx was monitored, we found that depolarizationevoked calcium uptake was affected in the M␤CD-treated synaptosomes in a dose-dependent manner and was clearly visible after treatment with 30 mM. The Cd 2ϩ channel blocker reduced the evoked uptake of calcium in both the control and M␤CDtreated synaptosomes, and more sustained inhibition was obtained when EGTA was included in the depolarization buffer (data not shown), thus indicating that the calcium influxes occurred via voltage-gated calcium channels.

DISCUSSION
The aim of this study was to investigate the mechanisms involved in the localization and clustering of Ca v 2.1 subtypes in presynaptic membranes. Our experiments provide the first evidence that Ca v 2.1 calcium channels are largely accumulated in lipid microdomains isolated from synaptosomes, in which they colocalize with proteins of the exocytic complex. These conclusions are supported by a number of experimental findings: i) the distribution of the large majority of ␣1A subunits in the low-density, detergent-resistant membrane fraction; ii) the shift in the buoyant density of ␣1A after alterations in membrane cholesterol levels with saponin or M␤CD (a criterion chosen to confirm the association of a protein with lipid microdomains enriched in cholesterol); and iii) the immunoisolation of detergent-insoluble complexes containing SNAREs, syn-aptotagmin1, and ␣1A.
The accumulation in detergent-resistant membranes appeared to be more specific for Ca v 2.1; the behavior of the 1C subunit of Ca v 1.2 was different. The widespread distribution of ␣1A in lipid microdomains may play a role in the trafficking and subcellular compartmentalization (presynaptic versus post-synaptic membranes) of the channel. A recent study has shown that the interaction of ␣1A subunits with SNARE proteins is an important requirement for the efficient nerve terminal localization of Ca v 2.1 channels (35). Therefore, our results showing the codistribution of ␣1A and SNARE proteins suggest that lipid microdomains function as platforms for the recruitment and presynaptic localization of Ca v 2.1 channels and proteins of the exocytic machinery. This hypothesis is further supported by the observation that disrupting lipid microdomain organization inhibits the colocalization of Ca v 2.1 channels and SNAREs. However, specific localization in lipid microdomains is not sufficient to explain the subcellular dis- tribution of Ca v 2.1 channels because accumulating evidence indicates that lipid microdomains are also present in dendrites and post-synaptic membranes (36), and so it can be expected that other mechanisms are involved in the targeting of Ca v 2.1 channels. In particular, it has been shown that the long carboxyl-terminal splice variants of Ca v 2.2 and Ca v 2.1 specifically interact with the Mint and CASK components of the so-called tripartite complex present in the presynaptic region (5), and with ␣-neurexins, a family of pre-synaptic surface proteins thought to play an important role in the organization of presynaptic compartments (37,38). Specific sequence domains and interactions with scaffolding proteins may therefore also be involved in the presynaptic targeting of Ca v 2.1 channels.
Cholesterol depletion not only alters the distribution of ␣1A subunits but also affects calcium influx to nerve terminals. This effect may be explained by various mechanisms. The first implies the direct role of cholesterol in maintaining the correct molecular structure and function of Ca v 2.1 and possibly of other calcium channel subtypes known to be localized in presynaptic membrane (1,3). It has been reported that lipid composition affects the folding and activity (39 -41) of voltagegated K ϩ channels. The K v 1.5 and K v 1.2 channel subtypes are specifically targeted to distinct lipid microdomains (40,41), and if the cholesterol in the expressing cells is depleted with M␤CD, the steady-state inactivation kinetics of the channels shifts in the hyperpolarization direction. The function of Ca v 2.1 may therefore also be sensitive to specific chemical interactions with neighboring membrane components and the biophysical properties of the membrane environments. The disruption of lipid microdomains caused by cholesterol depletion may modify the membrane properties affecting channel activity and/or folding.
A second cause of impaired calcium influx to nerve terminals may be related to the altered distribution and inhibition of the interaction of Ca v 2.1 with SNAREs and synaptotagmin1 after cholesterol extraction with M␤CD, which negatively influence channel activities (Refs. 16 -18, 35 and references therein).
Finally, the possibility that cholesterol depletion alters the interaction of the ␣1A subunit with other important functional regulatory molecules such as neurexins cannot be excluded. Recent data have demonstrated that ␣-neurexins are essential for Ca 2ϩ -triggered neurotransmitter release, which is impaired in neurexin-knockout mice, because Ca 2ϩ currents, Ca v 2.2 and Ca v 2.1 function, and synaptic neurotransmitter release are markedly reduced despite the apparently normal number of cell-surface calcium channels. This suggests that ␣-neurexins promote functional coupling between calcium channels and synaptic vesicle exocytosis and help the organization of the presynaptic domain (38). Interestingly, experimental evidence suggests that members of the neurexin superfamily are incorporated into lipid microdomains as a necessary step for cell surface sorting (42). In the light of these data, it can be speculated that disorganization of the lipid microdomains of presynaptic membranes impairs the interaction of neurexin with Ca v 2.1 and contributes to the reduction in Ca 2ϩ influx in the nerve. Further experiments are required to define the mechanisms involved in the inhibition of Ca 2ϩ influx after alterations in the cholesterol content of presynaptic membranes.
In brief, our data demonstrate that Ca v 2.1 channels are FIG. 7. 45 Ca 2؉ influx. Equal amounts of synaptosomes were incubated for 40 min at 37°C in choline-rich medium in the absence (C, Ⅺ) or presence (C, u) of 10 mM M␤CD or of 30 mM M␤CD (C, f). At the end of incubation, the synaptosomes from the untreated samples (A) and the samples treated with 30 mM M␤CD (B) were examined by means of electron microscopy (bars, 200 nm). In C, equal aliquots from each sample were used to analyze the basal (-dep) or the depolarizationinduced (ϩdep) uptake of 45 Ca 2ϩ in the presence or absence of the channel-blocker cadmium (ϩCd). The results represent the mean value Ϯ S.E. of three triplicate experiments ‫,ء(‬ p Ͻ 0.001 versus respective control, one-way ANOVA).

FIG. 6. Ca v 2.1 interacts with SNARE complexes in lipid microdomains. In
A and B, equal volumes of fractions 3 or 7 were incubated with anti-syn1 antibody covalently conjugated to protein G-Sepharose and then eluted from the beads and analyzed by SDS-PAGE. The proteins in the gels were stained with SYPRO Orange (A, the arrows indicate the position of SNAP-25 (SN-25) and syn1 or analyzed by Western blotting (B) using antibodies directed against ␣1A, synaptotagmin1 (Syt), munc-18-1 (M-18), and VAMP-2 (VAMP). The data are representative of three separate experiments. selectively accumulated in the lipid microdomains of nerve terminals, where they associate with the proteins of the exocytic complex. This compartmentalization may play an important role in organizing presynaptic membrane domains and may contribute to the coupling of calcium influxes and neurotransmitter release. Various lines of evidence indicate that neuronal lipid microdomains play a role in protein sorting and/or processing, neurotrophic factor signaling, axon guidance, and the localization and modulation of post-synaptic neurotransmitter receptors (Ref. 36 and references therein), and recent studies investigating the function of lipid microdomains in postsynaptic membrane organization have demonstrated the relevant role of these specialized microdomains in controlling the shape and number of spines and in stabilizing surface ␣-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptors (43). However, although there is a considerable amount of information concerning postsynapses, little is known about the role of cholesterol-enriched microdomains in the organization of presynapses. Our results provide evidence that lipid microdomains also function in the molecular organization of specialized presynaptic membranes, where they play a role in the clustering of calcium channel subtypes with molecules of the exocytic machinery and their proper function.