Regulation of presynaptic strength by controlling Ca²⁺ channel mobility: effects of cholesterol depletion on release at the cone ribbon synapse

Abstract

Synaptic communication requires proper coupling between voltage-gated Ca²⁺ (Ca_V) channels and synaptic vesicles. In photoreceptors, L-type Ca_V channels are clustered close to synaptic ribbon release sites. Although clustered, Ca_V channels move continuously within a confined domain slightly larger than the base of the ribbon. We hypothesized that expanding Ca_V channel confinement domains should increase the number of channel openings needed to trigger vesicle release. Using single-particle tracking techniques, we measured the expansion of Ca_V channel confinement domains caused by depletion of membrane cholesterol with cholesterol oxidase or methyl-β-cyclodextrin. With paired whole cell recordings from cones and horizontal cells, we then determined the number of Ca_V channel openings contributing to cone Ca_V currents (I_Ca) and the number of vesicle fusion events contributing to horizontal cell excitatory postsynaptic currents (EPSCs) following cholesterol depletion. Expansion of Ca_V channel confinement domains reduced the peak efficiency of release, decreasing the number of vesicle fusion events accompanying opening of each Ca_V channel. Cholesterol depletion also inhibited exocytotic capacitance increases evoked by brief depolarizing steps. Changes in efficiency were not due to changes in I_Ca amplitude or glutamate receptor properties. Replenishing cholesterol restored Ca_V channel domain size and release efficiency to control levels. These results indicate that cholesterol is important for organizing the cone active zone. Furthermore, the finding that cholesterol depletion impairs coupling between channel opening and vesicle release by allowing Ca_V channels to move further from release sites shows that changes in presynaptic Ca_V channel mobility can be a mechanism for adjusting synaptic strength.

synaptic release begins with the opening of Ca²⁺ channels followed by the diffusion of Ca²⁺ to the release machinery and fusion of synaptic vesicles. In many neurons, the spread of cytoplasmic Ca²⁺ is limited to nanodomains within a few hundred nanometers from an open Ca²⁺ channel. As a consequence, Ca²⁺ channels are often located very close to presynaptic release sites (Adler et al. 1991; Bucurenciu et al. 2008; Fedchyshyn and Wang 2005; Jarsky et al. 2010). For example, the L-type voltage-gated Ca²⁺ (Ca_V) channels that regulate release from cone photoreceptors in the vertebrate retina are positioned within 50–100 nm of vesicle release sites at the base of the synaptic ribbon (Lasansky 1973; Mercer et al. 2011a; Raviola and Gilula 1975). Although clustered, Ca_V channels at the base of the ribbon are not fixed in position. Single-particle tracking techniques used to visualize the movements of individual Ca_V channels in cone synaptic terminals have revealed that these channels move continuously within a confined area slightly larger than the base of the ribbon (Mercer et al. 2011b).

Changes in the lateral membrane mobility of postsynaptic AMPA receptors can tune synaptic strength by altering the ability of receptors to enter or exit the synapse (Heine et al. 2008; Petrini et al. 2009; Rust et al. 2010). For example, diffusion of undesensitized AMPA receptors into the synapse facilitates recovery from synaptic depression (Heine et al. 2008). In the present study, we asked whether synaptic strength might also be regulated by changes in the lateral membrane mobility of presynaptic Ca_V channels. To test this idea, we manipulated Ca_V channel confinement area by depleting cholesterol from the cone terminal using methyl-β-cyclodextrin (MβCD) or cholesterol oxidase (COase). We confirmed that depleting cholesterol from the presynaptic membrane expanded this confinement area, suggesting that cholesterol-rich lipid rafts help to confine Ca_V channels at the cone ribbon synapse (Cooper and McLaughlin 1984; Mercer et al. 2011b). Furthermore, by allowing channels to move further away from release sites, depletion of membrane cholesterol weakened the coupling between Ca_V channel opening and vesicle fusion at cone synapses. These results indicate that changes in presynaptic Ca_V channel mobility can provide a mechanism for regulating synaptic strength.

MATERIALS AND METHODS

Animal care and use.

Experimental procedures were approved by the University of Nebraska Medical Center Institutional Animal Care and Use Committee. Male and female adult aquatic tiger salamanders (Ambystoma tigrinum, 18–25 cm in length; Kons Scientific, Germantown, WI and Charles Sullivan, Nashville, TN) were maintained on a 12:12-h day-night cycle and were killed 1–2 h after the beginning of subjective night. Salamanders were decapitated with heavy shears and immediately pithed.

Retinal tissue preparation.

After enucleation, each eye was opened along the ora serrata, and the cornea, lens, and vitreous body were removed. The resultant eyecup was quartered and pressed gently, vitreous side down, onto a 5 × 10-mm piece of filter paper (type AAWP, 0.8-μm pores; Millipore, Billerica, MA). The filter paper and eyecup were then immersed in cold amphibian saline solution, and the retinal pigment epithelium, choroid, and sclera were carefully peeled away, leaving the retina adhered to the filter paper. The retina was sliced into 125-μm sections with the use of a razor blade tissue chopper (Stoelting, Wood Dale, IL) fitted with a no. 121-6 razor blade (Ted Pella, Redding, CA). Slices were separated from one another and rotated 90° to view the laminar stratification of the retina under a water-immersion objective (×60, 1.0 NA) on an upright microscope (E600FN; Nikon, Tokyo, Japan). Throughout these procedures, the tissue was submerged in cold, oxygenated amphibian extracellular saline solution consisting of (in mM) 111 NaCl, 2.5 KCl, 1.8 CaCl₂, 0.5 MgCl₂, 10 HEPES, and 5 glucose (pH 7.8). For experiments on dissociated photoreceptors, a whole salamander retina was isolated and incubated in 0.5 mM Ca²⁺ amphibian extracellular saline solution containing 0.2 mg/ml cysteine and 10 U/ml papain (Sigma-Aldrich, St. Louis, MO) for 25 min at 20°C. Tissue was washed twice in low-Ca²⁺ saline solution containing 1% BSA and 1 mg/ml DNase (Worthington Biochemicals, Lakewood, NJ), followed by two additional washes in low-Ca²⁺ saline. To isolate individual neurons, retinal tissue was gently triturated using a fire-polished glass Pasteur pipette and plated onto 18-mm coverslips (Deckgläser/Thermo, Braunschweig, Germany) coated with 1 mg/ml concanavalin-A (Sigma-Aldrich).

Retina tissue fixation and cholesterol imaging.

To image membrane lipids at the terminals of cone photoreceptors, we applied filipin III (Sigma-Aldrich) to label membrane cholesterol or FITC-CTX_B (Sigma-Aldrich) to label ganglioside GM1 glycoproteins associated with cholesterol-rich lipid rafts. Dissociated retinal cells were prepared as described above and fixed in 0.1 M phosphate-buffered saline (PBS) containing 4% paraformaldehyde for 30 min. Cells were then washed in 0.1 M PBS three times (10 min per wash) and bound with either 300 μg/ml filipin III for 90 min or 2 μg/ml FITC-CTX_B for 45 min. Cells were washed an additional three times with 0.1 M PBS for 10 min and placed at room temperature in darkness until dry. After a brief rinse in de-ionized H₂O, each slide was coverslipped using ProLong Gold anti-fade reagent (Invitrogen). Retinal preparations were visualized under brightfield illumination on an upright microscope (Nikon E600FN) to identify individual cones. Cones treated with filipin III were imaged using a DsQi-1 camera (Photometrics, Tucson, AZ), NIS-Elements software (Nikon), a ×60, 1.2-NA water-immersion objective (Nikon), a 380-nm excitation/525-nm emission filter (Chroma Technology, Burlington, VT), and 5-s exposure. FITC-CTX_B-stained cones were imaged through a water-immersion objective (×60, 1.0 NA; Nikon) using a laser confocal scanhead (UltraVIEW Live Cell Imaging System; Perkin Elmer, Waltham, MA) equipped with a cooled charge-coupled device camera (Orca ER; Hamamatsu, Hamamatsu City, Japan). Z-stack image series were captured at 1-μm intervals for image analysis by using an exposure of 2 s per image with a 488-nm excitation/525-nm bandpass emission filter to optimize the FITC signal. Post-image processing was performed in Adobe Photoshop (Adobe Systems, San Jose, CA) to apply pseudocolors and optimize the brightness and contrast.

Quantum dot binding and imaging.

Quantum dot (QD) attachment to individual Ca_V channels at photoreceptor synapses was conducted as described previously (Mercer et al. 2011b). Briefly, retinal neurons were dissociated and incubated with a primary rabbit anti-α₂δ₄-subunit antibody (Qin et al. 2002) at a dilution of 1:1,000 in standard 1.8 mM Ca²⁺ saline solution for 3 h at 4°C. Cells were washed three times in standard saline solution and then incubated with a secondary biotinylated goat-anti-rabbit IgG (1:2,000; Sigma-Aldrich) for 1 h at 4°C. Retinal neurons were washed another three times and then conjugated to streptavidin-coated 525-nm emission QDs (20 nM; Invitrogen, Carlsbad, CA) for 15 min. After eight subsequent washes, QDs on dissociated photoreceptors were visualized on an IX71 inverted microscope (Olympus, Center Valley, PA) with a ×60, 1.45-NA oil-immersion objective (Olympus) and cooled electron-multiplying charge-coupled device camera (ImageM-1K, Hamamatsu). Epifluorescence was provided by an X-Cite 120Q light source (Lumen Dynamics Group, Ontario, Canada) controlled by a model D122 shutter (UniBlitz, Rochester, NY). QDs were selected for tracking analysis only if they were localized to photoreceptor terminals, exhibited a small size (≤4 pixels), and showed intermittent blinking; these criteria are consistent with labeling by a single QD (Alcor et al. 2009).

The location of a QD can be determined with precision exceeding the diffraction limit by fitting the fluorescence profile with a Gaussian point spread function (Alcor et al. 2009; Courty et al. 2006). From the standard deviation (SD) of displacements exhibited by immobilized QDs, we estimate the pointing accuracy (full-width half-maximum = 2.35 × SD) to be 35 nm (Mercer et al. 2011b). Videos were captured at 30 ms per frame for 400 frames using the MetaMorph imaging suite (Molecular Devices, Sunnyvale, CA). Images were adjusted to optimize QD contrast and smoothed by convolving with a 5 × 5-pixel Gaussian. QD position was then tracked using NIS-Elements (Nikon) software. From the x and y coordinates, we calculated the mean squared displacement (MSD; in μm²/s) (Saxton 1997) using the following equation:

$$\text{MSD}=\frac{1}{N-n}{\displaystyle \sum _{i=1}^{N=n}\left[{\left(X_{i+n}-X_i\right)}^2-{\left(Y_{i+n}-Y_i\right)}^2\right]}$$ (1)

We calculated the confinement area that a given Ca_V channel traverses within the presynaptic plasma membrane by fitting the data with Eq. 2:

$$\text{MSD}=\frac{L^2}{3}\left[1-\exp \left(\frac{-12Dt}{L^2}\right)\right]$$ (2)

where L² provides the surface area that a given channel traverses (in μm²).

Pharmacology and cellular cholesterol content.

To deplete membrane cholesterol, sliced or dissociated retinal preparations were incubated with 10 mM MβCD (Sigma-Aldrich) or 2 U/ml COase (Sigma-Aldrich) in standard amphibian saline solution for 1 h. To replenish membrane cholesterol, we first treated cells with 10 mM MβCD for 1 h and then applied a MβCD-cholesterol complex for 1 h, following established protocols (Abulrob et al. 2005; Pucadyil and Chattopadhyay 2005; Tekpli et al. 2008). To prepare the MβCD-cholesterol complex, cholesterol (Sigma-Aldrich) was first dissolved in 95% ethanol at 80°C to make a 90 mM stock solution. This stock solution was diluted to a final concentration of 1 mM cholesterol by being mixed with an amphibian saline solution containing 10 mM MβCD for 1 h at 20°C on a rocking platform. MβCD-cholesterol complex solutions were prepared freshly before each experiment.

We analyzed cholesterol levels in retinal membranes using the Amplex Red cholesterol assay kit (Invitrogen). Dissected eyecups were halved, and the retina was carefully isolated from the retinal pigment epithelium and choroid. Retinal tissue was then treated with MβCD or COase or treated by following the cholesterol repletion protocol described above. Samples were hydrolyzed with the cholesterol reaction buffer (0.1 M potassium phosphate, 50 mM NaCl, 5 mM cholic acid, and 0.1% Triton X-100, pH 7.4) and combined with the Amplex Red reagent (150 μM), 1 U/ml horseradish peroxidase, 1 U/ml cholesterol oxidase, and 1 U/ml cholesterol esterase for 30 min at 37°C. Fluorescence was measured using a SpectraMax M5 microplate reader (Molecular Devices), and values were normalized by retinal tissue protein concentration determined using a BCA protein assay (Pierce/Thermo Scientific, Rockford, IL). Cholesterol levels were compared with a standard curve of 0–8 μg/ml cholesterol activated by the Amplex Red assay kit (R² = 0.992). We performed the assay for each condition as a triplicate of triplicates: retina halves were harvested from three different animals, and each sample was assayed in triplicate.

Electrophysiology.

To measure the effects of membrane cholesterol depletion on photoreceptor vesicle release, we recorded simultaneously from cone photoreceptors and synaptically connected horizontal cells using the whole cell patch-clamp configuration. Whole cell recording electrodes were pulled using a PP-830 vertical puller (Narishige USA, East Meadow, NY) from borosilicate glass pipettes (1.2-mm outer diameter, 0.9-mm inner diameter, with an internal filament; World Precision Instruments, Sarasota, FL) with tips 2 μm in diameter and resistance values between 12 and 18 MΩ. Both recording electrodes were positioned using Huxley-Wall micromanipulators (Sutter Instruments, Novato, CA). Cones were voltage-clamped at −70 mV using an OptoPatch amplifier (Cairn Research, Faversham, UK), and second-order neurons were voltage-clamped at −60 mV with an AxoPatch 200B amplifier (Molecular Devices). Both amplifiers were controlled using an Axon Instruments Digidata 1322 interface and pClamp 9.2 software (Molecular Devices). We selected cones that required modest holding currents (<200 pA) to voltage clamp the cell at the steady holding potential of −70 mV. The cone pipette solution contained (in mM) 40 cesium glutamate, 50 cesium gluconate, 9.4 tetraethylammonium chloride (TEACl), 3.5 NaCl, 1 CaCl₂, 1 MgCl₂, 9.4 MgATP, 0.5 GTP, 5 EGTA, and 10 HEPES (pH 7.2). The presence of glutamate in the patch pipette maintains vesicle release and postsynaptic responses for up to 20 min after the initial whole cell rupture into a photoreceptor (Bartoletti et al. 2010). We measured cone synaptic output by recording glutamatergic EPSCs in horizontal cells with a pipette solution that contained (in mM) 90 cesium gluconate, 10 TEACl, 3.5 NaCl, 1 CaCl₂ 1 MgCl₂, 10 MgATP, 0.5 GTP, 5.0 EGTA, and 10 HEPES (pH 7.2). Vesicle release was elicited by applying a strong depolarizing voltage step to cone photoreceptors from −70 to −10 mV for 100 ms using a P/8 leak subtraction protocol. The liquid junction potential was estimated from the calculator in pClamp 9.2 to be −13 mV for the cone pipette solution and −16 mV for the horizontal cell pipette solution. As a vehicle control for cholesterol repletion experiments, we recorded from cones and second-order neurons after treating retina slices with 1:100 dilution of 95% ethanol-amphibian saline solution. These experiments showed no detectable change in cone Ca_V current (I_Ca; n = 7) or EPSCs in horizontal cells (n = 3) compared with control pairs.

We calculated the number of Ca_V channel openings contributing to I_Ca and the number of miniature EPSCs (mEPSCs) contributing to EPSCs by deconvolution as described in Bartoletti et al. (2011). Deconvolution of P/8 leak-subtracted I_Ca was performed using OriginPro 8 software (North Hampton, MA). To deconvolve single-channel events during I_Ca evoked by a step to −10 mV, we used a single Ca_V channel amplitude of 0.3 pA with decay time constant of 1.1 ms to represent variable channel open times (Bartoletti et al. 2011; Thoreson et al. 2000). The number of channel openings was scaled for an average of 13 ribbons per cone (Bartoletti et al. 2010; Pang et al. 2008). EPSCs were deconvolved using an average mEPSC waveform with amplitude of 5.7 pA and charge transfer of 15.5 pC (Cadetti et al. 2008). Output from a single ribbon contributes ∼46 pA to the peak amplitude of the horizontal cell EPSC evoked by a step to −10 mV applied to a cone (Bartoletti et al. 2010). The number of ribbon contacts between a cone and postsynaptic horizontal cell was therefore estimated by dividing the peak amplitude of the EPSC by 46 pA/ribbon. Release events were shifted forward by 300 μs to compensate for the latency to the beginning of the EPSC caused by an instantaneous rise in Ca²⁺ concentration (Duncan et al. 2010). Results of deconvolution were low-pass filtered using a fast-Fourier transform filter with 5–10 points of smoothing. The rate of Ca_V channel openings was divided by the rate of release events to calculate the number of vesicle fusion events per channel opening at each time point.

Exocytotic capacitance changes were measured in cones using established protocols (Bartoletti et al. 2010; Johnson et al. 2002; Rabl et al. 2005). For these experiments, the shaft near the tip of each recording electrode was lightly coated with dental wax (Integra-Miltex, Plainsboro, NJ) to reduce stray capacitance artifacts. Capacitance measurements were made using the “track-in” mode of the OptoPatch amplifier by varying membrane potential sinusoidally (500–600 Hz, 30 mV peak to peak) about the holding potential. Series resistance and membrane capacitance were compensated electronically. Track-in feedback gain was increased to its highest stable value. The amplitude of the cone capacitance response was measured 20 ms after the end of the test step to avoid gating charges and to allow time for the phase angle feedback circuitry to settle. We excluded recordings that showed pronounced poststimulus changes in series resistance.

To examine the responses of postsynaptic glutamate receptors independent from presynaptic photoreceptors, we bath-applied 0.5 mM 4-methoxy-7-nitroindolinyl-l-glutamate (MNI-glutamate; Tocris Biosciences, Ellisville, MO) and uncaged glutamate by flash photolysis using UV light derived from a xenon arc flash lamp (JML-C2 Flash Lamp System; Rapp OptoElectronic, Hamburg, Germany). Light flashes (17.5-μm diameter) centered on the horizontal cell were delivered through a quartz fiber optic mounted to the epifluorescence port of the microscope.

mEPSCs were analyzed using MiniAnalysis 6.0.7 (Synaptosoft, Decatur, GA). mEPSCs were detected by the software using an amplitude threshold of 1 pA, and each event was then evaluated individually. If necessary, the length of preceding baseline period was adjusted to improve the amplitude measurement. Double peaks were analyzed using an algorithm within Synaptosoft that extrapolates the exponential decay of the first peak.

Modeling.

Vesicle release was modeled using Mathematica (Wolfram Research, Champaign, IL) as described by Bartoletti et al. (2011). When control conditions were simulated, the immediately releasable pool of 20 vesicles was arranged in a square grid with 50-nm center-to-center spacing on the surface of a 250-nm-long ribbon in two rows of 5 vesicles on either side. Fifty-six Ca_V channels were distributed randomly on the plasma membrane. The plasma membrane, intended to approximate the observed morphology of the cone ribbon synapse (Lasansky 1973; Raviola and Gilula 1975) was constructed as an evaginating arc with 100-nm radius centered at the ribbon apex, giving way to further invaginating arcs on both sides of the ribbon with a 200-nm radius. The range over which channels were distributed was increased from 100 to 150 or 220 nm to simulate effects of COase or MβCD, respectively, on Ca_V channel domain size. A simulated step to −10 mV opened Ca_V channels to a maximal open probability of 0.35, opening in time according to the equation P_o = 0.35 × t/(t + 5 ms). Randomly chosen Ca_V channels opened for 1.1 ms. Time-dependent Ca²⁺ gradients were modeled in increments of 100 ms, with release binned into 1.1-ms intervals to reflect Ca_V channel openings (Ait-Haddou et al. 2010). The domains of all open channels were summed linearly. The Ca²⁺ dependence of release was R = 3,634[Ca²⁺]²/(2 μM + [Ca²⁺])² (Duncan et al. 2010). The K_d for BAPTA and EGTA was 170 nM with K_on for BAPTA of 6 × 10⁸ M/s and K_on for EGTA of 9 × 10⁶ M/s. The diffusion coefficient for Ca²⁺ was 220 μm²/s, and the diffusion coefficient for Ca²⁺ buffers was 20 μm²/s. We simulated the presence of a diffusion barrier by slowing the rate of Ca²⁺ diffusion and the concentration of diffusible buffer by 1:200. We also examined in the impact of reducing the strength of the diffusion barrier beyond the arciform density to 1:100. Each simulation was repeated for 1,000 trials.

Statistical analysis.

Data were compiled in Prism 4 (GraphPad Software, La Jolla, CA) and are presented as means ± SE. Statistical significance between experimental conditions was determined using Student's t-test. P values <0.05 were considered significant.

RESULTS

Synaptic cholesterol localization and quantification of membrane cholesterol content.

Ultrastructural studies by Cooper and McLaughlin (1984) showed evidence of lipid rafts at the margins of the ribbon-style active zone of photoreceptor synapses in the chick retina. To test whether cholesterol levels were also elevated in the presynaptic membranes of salamander cones, we used filipin III (300 μg/ml for 1 h) to stain isolated cones for cholesterol. As expected for ubiquitous distribution of cholesterol throughout the membrane, filipin III stained the entire cell (Fig. 1A). However, staining was more intense in the synaptic region at the base of the soma (arrows), consistent with enrichment of cholesterol in the terminal. The presence of cholesterol-rich lipid rafts was further indicated by strong staining of the synaptic terminal region with FITC-CTX_B (Fig. 1B), which selectively labels lipid raft ganglioside GM1 subunits (Obrocki and Borroni 1988; Yamamoto et al. 2005). Untreated cones showed only autofluorescence in the inner segment and no fluorescent signal at the synaptic terminal. Dendrites of isolated bipolar cells did not show FITC-CTX_B staining (data not shown), consistent with an absence of lipid rafts on postsynaptic membranes in the outer plexiform layer (OPL) (Cooper and McLaughlin 1984). These data show that FITC-CTX_B staining of the OPL observed in salamander retina (Mercer et al. 2011b) was primarily due to staining of the synaptic plasma membrane of rods and cones and confirm that cholesterol is enriched in lipid rafts at the cone ribbon synapse (Cooper and McLaughlin 1984).

Fig. 1.

Cholesterol localization at the cone synapse and effects of pharmacological treatments on retinal cholesterol levels. A: to visualize membrane cholesterol, an enzymatically dissociated double cone was fixed and stained with filipin III (300 μg/ml for 90 min). When visualized under epifluorescence with 380-nm excitation/525-nm emission, cones exhibited generalized cholesterol staining throughout the inner segment (IS), cell body (CB), and terminal, but cholesterol appeared more concentrated at the synapse (arrows). Control cones visualized without filipin III binding showed autofluorescence in the IS but no synaptic fluorescence (B; arrows). C: to specifically target cholesterol-rich lipid rafts, we stained cones with 2 μg/ml FITC-CTX_B for 45 min and visualized them using 488-nm excitation/525-nm emission on a spinning disk confocal microscope. Cones labeled with FITC-CTX_B showed concentrated staining at the cone pedicle (arrow). D: cones without FITC-CTX_B showed no staining at the synaptic terminal (arrow). E: we quantified cholesterol levels in the retina using the Amplex Red cholesterol assay kit. Control retinas exhibited a cholesterol level of 12.6 ± 0.69 ng/μg protein (control), which dropped to 4.94 ± 0.52 ng/μg protein (39% of control; ***P < 0.0001 vs. control) and 7.18 ± 0.66 ng/μg protein (58% of control, ***P < 0.0001) after methyl-β-cyclodextrin (MβCD) or cholesterol oxidase (COase) treatment, respectively. Membrane cholesterol was restored to a level of 11.3 ± 0.41 ng/μg protein (repletion) by repleting tissue with a complex of cholesterol and MβCD (91% of control; P = 0.19 vs. control). Scale bar for immunohistochemistry, 10 μm.

Cholesterol depletion caused a reduction in staining by FITC-CTX_B in the OPL, suggesting a reduction in cholesterol-rich lipid rafts (Mercer et al. 2011b). We quantified changes in whole retina membrane cholesterol content using the Amplex Red cholesterol assay kit. In control retinas, cholesterol averaged 12.6 ± 0.69 ng cholesterol/μg protein. This level dropped to 4.94 ± 0.52 ng cholesterol/μg protein (40.9% of control; P < 0.0001) after cholesterol depletion with MβCD. COase also caused a decrease in total retina cholesterol, diminishing the level to 7.18 ± 0.66 ng cholesterol/μg protein (57.6% of control; P < 0.0001). Repleting MβCD-treated tissue with an MβCD-cholesterol complex restored cholesterol levels to 11.3 ± 0.41 ng cholesterol/μg protein (91.2% of control; P = 0.19). Although we could not measure membrane cholesterol levels specifically in cones, these results show that pharmacological treatment with MβCD or COase produced significant changes in whole retina membrane cholesterol (Fig. 1E).

Ca_V channel mobility vs. synaptic cholesterol depletion.

Experiments using single-particle tracking techniques to visualize individual QD-labeled Ca_V channels at cone synapses in the retinal slice preparation have shown that channels normally remain close to vesicle release sites, but depleting membrane cholesterol with MβCD enlarges the surface area traversed by an individual channel (Mercer et al. 2011b). To compare release efficiency with changes in Ca_V channel domain size, we measured changes in the size of the confinement domain in isolated cones using a high-NA (1.45) objective to provide high spatial resolution. Consistent with data obtained from retinal slices using a 1.0-NA objective (Mercer et al. 2011b), single-particle tracking experiments on Ca_V channels at the synapses of dissociated control cones (n = 15) showed that Ca_V channels are normally confined to a limited surface area of 0.18 ± 0.000005 μm², but depletion of membrane cholesterol with MβCD (n = 10) expanded the confinement area to 0.88 ± 0.00014 μm² (P = 0.0032 vs. control). Expansion of the confinement area is evident in individual examples of QD trajectory plots and was quantified from the asymptotic level of MSD (Fig. 2). The average confinement areas (L²) determined using Eq. 2 are plotted in the bar graphs in Fig. 2B. To test whether cholesterol depletion also increased the mobility of the presynaptic ribbon complex, we labeled ribbons by patch clamping isolated cones and introducing a Hylite 488-labeled peptide that binds to the ribbon protein, ribeye (Mercer et al. 2011b; Zenisek 2008). The small movements of the ribbon did not increase significantly in cholesterol-depleted cells (control: L² = 0.048 ± 0.0004 μm², n = 10; MβCD: L² = 0.054 ± 0.0004 μm², n = 10; P = 0.67), indicating that increases in synaptic Ca_V channel mobility are independent of the ribbon complex.

Fig. 2.

Depletion of synaptic cholesterol increased the confinement domains of individual voltage-gated Ca²⁺ (Ca_V) channels. A: mean squared displacement (MSD) calculated using Eq. 1 plotted against time interval. Data were fit with Eq. 2 to determine the size of the confinement area (L²) from the asymptotes of the curves. B: bar graph showing the average confinement areas. Ca_V channels at the synapses of enzymatically dissociated cones (n = 15; black squares in A) were confined to a membrane surface area of 0.18 μm² (control). Depletion of membrane cholesterol with MβCD (n = 10; white triangles in A) or COase (n = 11; black diamonds in A) significantly increased the confinement area to 0.88 μm² (**P = 0.003 vs. control) and 0.37 μm² (*P = 0.014 vs. control), respectively. Repleting membrane cholesterol with the MβCD-cholesterol complex (n = 9; white circles in A) after an initial MβCD treatment reduced the confinement area to 0.26 μm² (P = 0.17 vs. control). C–F: examples of trajectory plots of quantum dot (QD)-labeled Ca_V channels in isolated cones from different conditions: control (C), MβCD treated (D), COase treated (E), and cholesterol repleted (F).

MβCD can have other effects in addition to cholesterol depletion (Zidovetzki and Levitan 2007). Consistent with expansion of the confinement domain being primarily due to cholesterol depletion, repleting membrane cholesterol by applying a complex of MβCD and cholesterol after initially depleting cholesterol with MβCD restored the confinement area to a smaller size of 0.26 ± 0.000015 μm², not significantly different from control (P = 0.17; Fig. 2). We also depleted cholesterol using the enzyme COase. Consistent with a smaller effect on retinal cholesterol levels than MβCD, COase caused a smaller expansion of the confinement area to 0.37 ± 0.000020 μm² (n = 11, P = 0.014; Fig. 2). These results show that depletion of synaptic cholesterol increased the confinement area for Ca_V channels and that the normal confinement area can be restored by replacing membrane cholesterol.

Effects of membrane cholesterol on neurotransmission.

Ultrastructural and electrophysiological data indicate that Ca_V channels normally remain within 50–100 nm of the release machinery at cone synapses (Mercer et al. 2011a; tom Dieck et al. 2005). By allowing Ca_V channels to move further away from release sites, we predicted that an expanded confinement area would reduce the efficiency with which Ca_V channel openings stimulate vesicle release. We calculated the change in efficiency between Ca_V channel opening and fusion by determining the number of individual Ca_V channel openings and synaptic release events during simultaneous recordings of cone I_Ca and horizontal cell EPSCs (Fig. 3). Cones were voltage-clamped at −70 mV and depolarized to −10 mV for 100 ms to activate I_Ca and stimulate exocytosis of the entire immediately releasable pool of ∼20 vesicles per ribbon (Bartoletti et al. 2010). As shown in Fig. 3A, the peak of the horizontal cell EPSC was attained quickly, often before the I_Ca attained its peak amplitude. As described in methods, we determined the rate of Ca_V channel openings by linear deconvolution of cone I_Ca, assuming a single-channel current of 0.3 pA with a mean open time of 1.1 ms and 13 ribbons per cone. We determined the number of quantal release events by linear deconvolution of the horizontal cell EPSC, using an average mEPSC waveform and scaling for the number of ribbon contacts estimated from the size of the EPSC (Bartoletti et al. 2010, 2011). As illustrated in Fig. 3B, release efficiency (i.e., the number of vesicle release events associated with a single Ca_V channel opening) rose and declined quickly as the releasable pool of vesicles was emptied. With depolarizing test steps to −10 mV, release from cones attained a maximal efficiency of 0.32 ± 0.029 vesicle fusion events per Ca_V channel opening per ribbon (n = 20; ∼3 channel openings per fusion event). In retinal slices treated with MβCD, peak fusion efficiency declined significantly to 0.22 ± 0.026 fusion events per channel opening (n = 13, P = 0.028; Fig. 3, D and E; ∼4.5 channel openings per fusion event). Depletion of cholesterol with COase also caused a significant reduction in fusion efficiency to 0.21 ± 0.033 fusion events per channel opening (n = 9, P = 0.035; Fig. 3E). Repletion of cholesterol restored release efficiency to 0.26 ± 0.030 fusion events per channel opening (n = 12, P = 0.21 compared with control; Fig. 3E). There were no statistically significant differences in cone I_Ca amplitude, EPSC amplitude, EPSC half-width, or EPSC time to peak in any of the treatment conditions (Table 1).

Fig. 3.

Depletion of membrane cholesterol reduced the efficiency of vesicle release. To determine effects of cholesterol manipulation on the ability of Ca_V channel openings to trigger vesicle release, we analyzed synaptic release using simultaneous recordings of I_Ca in cones and postsynaptic currents (PSCs) in horizontal cells. From these paired recordings, we deconvolved the number of presynaptic Ca_V channel openings and vesicle release events. A: a control example of I_Ca from a cone (thick gray trace) overlaid on the simultaneously recorded PSC from a horizontal cell (thin black trace). The rate of Ca_V channel openings per ms per ribbon was obtained by deconvolution of I_Ca. Deconvolution of the PSC using the average mEPSC waveform yielded the release rate. In the example in A, the release rate was normalized for 5 ribbon contacts. Dividing the release rate per ribbon by the rate of Ca_V channel openings per ribbon yielded the number of release events per opening (B). C: PSC from a horizontal cell (thin black trace) and I_Ca from a cone (thick gray trace) in a retinal slice treated with MβCD. D: the number of release events per Ca_V channel opening. In this example, the release rate was normalized for 4 ribbon contacts. E: bar graph showing peak release efficiency (i.e., the number of vesicle fusion events triggered by a Ca_V channel opening at the peak of release) in control (0.32 ± 0.029, n = 20), MβCD (0.22 ± 0.026, n = 13; *P = 0.028 vs. control), COase (0.21 ± 0.033, n = 9; *P = 0.028 vs. control), and cholesterol repletion (0.26 ± 0.030, n = 12; P = 0.21) conditions. F: schematic diagram of the anatomic arrangements used to model release at the cone ribbon synapse as described in the text.

Table 1.

Properties of I_Ca and PSCs after cholesterol manipulation

	Control	MβCD	COase	Repletion
Cone ICa amplitude, pA	141 ± 10.2	145 ± 10.0 (P = 0.80)	134 ± 20.7 (P = 0.87)	145 ± 19.8 (P = 0.83)
EPSC amplitude, pA	111 ± 17.2	74.5 ± 10.5 (P = 0.11)	100.0 ± 19.3 (P = 0.72)	94.6 ± 11.3 (P = 0.51)
EPSC half-width, ms	12.6 ± 1.8	10.2 ± 1.4 (P = 0.33)	15.2 ± 2.8 (P = 0.09)	14.6 ± 3.3 (P = 0.20)
EPSC time to peak, ms	8.8 ± 0.8	8.0 ± 0.6 (P = 0.46)	11.0 ± 2.7 (P = 0.11)	7.8 ± 0.6 (P = 0.39)
n	20	15	9	12

Data are means ± SE, and measurements were compared with control conditions using unpaired Student's t-test (P values). To test whether cholesterol depletion by methyl-β-cyclodextrin (MβCD) affected other properties of release besides the latency of postsynaptic current (PSC) responses, we measured the peak amplitude of cone voltage-gated Ca²⁺ current (I_Ca) and a number of PSC parameters: peak amplitude, half-width of the PSC waveform, and excitatory PSC (EPSC) time to peak. COase, cholesterol oxidase.

To examine the effects of Ca_V channel domain size on release further, we simulated release at the cone synapse using a model that incorporates known physiological and anatomic properties of the synapse (Bartoletti et al., 2011). As illustrated in Fig. 3F, the ribbon synapse was modeled as an evaginating arc with 100-nm radius centered at the ribbon apex, giving way above the arciform density to invaginating arcs of 200-nm radius along the flanks of the ribbon. Fifty-six Ca_V channels were distributed randomly on the plasma membrane beneath the immediately releasable pool of 20 vesicles positioned in two rows on either side of a 250-nm-long ribbon. To replicate limited effects of diffusible chelators on release efficiency, the model assumed a diffusion barrier that slows the diffusion of Ca²⁺ and Ca²⁺ chelators by a factor of 1:200. In control conditions with 5 mM EGTA as the diffusible buffer, the model predicted a peak efficiency of 0.28 fusion events per Ca_V channel opening, similar to the measured efficiency of 0.32 fusion events per channel opening. To mimic the expansion of Ca_V channel domain size caused by COase treatment, we expanded the region within which Ca_V channels are distributed by 1.5-fold (from 100 to 150 nm). This reduced the predicted peak efficiency to 0.18 fusion events per channel opening, similar to the experimentally observed value of 0.21. Increasing the range over which channels are distributed by 2.2-fold (to 220 nm), to mimic treatment with MβCD, reduced the predicted peak efficiency to 0.12 fusion events per channel opening. However, this version of the model assumed that the diffusion barrier extended unchanged into the cytoplasm beyond the arciform density. Incorporating the plausible assumption that the influence of the diffusion barrier diminished as channels strayed beyond the arciform density (by reducing the diffusion barrier from 1:200 to 1:100 in the light gray region above the dashed line in Fig. 3F) improved model predictions to 0.21 and 0.18 fusion events per channel opening with channel confinement regions of 150 and 220 nm, respectively.

Consistent with earlier results (Bartoletti et al. 2011), we found that replacing 5 mM EGTA with 1 mM BAPTA in the cone pipette solution did not significantly alter peak release efficiency (0.30 ± 0.036 fusion events per channel opening per ribbon, n = 16). With BAPTA as the diffusible chelator, cholesterol depletion with MβCD diminished release efficiency to 0.22 ± 0.033 fusion events per channel opening (n = 10). With 1 mM BAPTA, the model predicted a peak efficiency of 0.31 fusion events per channel opening. Expanding the Ca_V channel confinement region to 220 nm to simulate MβCD treatment reduced efficiency to 0.18 fusion events per channel opening. Incorporating a reduction in the strength of the diffusion barrier from 1:200 to 1:100 beyond the arciform density increased this predicted efficiency from 0.18 to 0.22 fusion events per channel opening.

In addition to studying release in paired recordings, we also examined exocytosis from cones using capacitance techniques to measure the increase in surface area accompanying vesicle fusion (Bartoletti et al. 2010; Rabl et al. 2005). The time constant for release of the rapidly releasable pool of vesicles is ∼3–5 ms (Rabl et al. 2005; Thoreson et al. 2004), so we predicted that MβCD should have a bigger effect on exocytosis evoked by short 5-ms test steps than longer 8-ms test steps. Consistent with this prediction, significantly less vesicle release was evoked by a short 5-ms test pulse after membrane cholesterol depletion (Fig. 4A; control: 143 ± 19.6 fF, n = 6; MβCD: 48.4 ± 18.5 fF, n = 5; P = 0.0072), whereas exocytotic capacitance jumps evoked by an 8-ms test pulse were similar in both conditions (Fig. 4B; control: 130 ± 14.8 fF, n = 10; MβCD: 100 ± 23.5 fF, n = 10; P = 0.27).

Fig. 4.

Cholesterol depletion diminished vesicle release evoked by 5-ms steps but not by 8-ms steps. Vesicle exocytosis from cones was measured using capacitance techniques. Voltage-clamped cones were stepped from −70 to −10 mV for short pulses to elicit a burst of vesicle fusion. Capacitance measurements were suspended during the test step and for 20 ms afterwards to avoid gating charges and allow time for the phase angle feedback circuitry to settle. A: examples of exocytotic capacitance responses evoked by 5-ms test steps in a control and MβCD-treated cone. B: examples of exocytotic capacitance responses evoked by 8-ms test steps in the same control and MβCD-treated cones. C: using a brief pulse of 5 ms (just faster than the time constant for release of ∼3 ms; Rabl et al. 2005), MβCD treatment caused a significant reduction in capacitance responses (48.4 ± 18.5 fF, n = 5) compared with control cells (143 ± 19.6 fF, n = 6; **P = 0.0072). Release evoked from cones stepped to −10 mV for 8 ms exhibited no significant differences in the 2 conditions (control amplitude: 130 ± 14.8 fF, n = 10; MβCD amplitude: 100 ± 23.5 fF, n = 10; P = 0.27).

It has been reported that MβCD treatment can impair clathrin-mediated endocytosis (Rodal et al. 1999; Subtil et al. 1999). To measure endocytosis, we analyzed the time required for recovery of membrane capacitance following release. In control cones, the time required for the depolarization-evoked capacitance increases to decay by 50% averaged 195 ± 62.1 ms (n = 5) following 5-ms steps and 215 ± 34.5 ms (n = 6) following 8-ms steps. The time required for a 50% decline in the capacitance increase was similar after MβCD treatment, averaging 145 ± 34.6 ms (n = 8, P = 0.46 vs. control) for 5-ms steps and 240 ± 44.8 ms (n = 9, P = 0.70 vs. control) following 8-ms steps. These results suggest that the rapid endocytosis observed in cones was not significantly affected by synaptic cholesterol depletion and is consistent with other evidence that this mechanism is probably not clathrin-dependent at ribbon-style synapses (LoGiudice and Matthews 2007; Wu et al. 2007).

Cholesterol depletion can alter AMPA receptor recycling (Glodowski et al. 2007; Hering et al. 2003), so we tested whether glutamate receptor currents in horizontal cells were altered by cholesterol depletion with MβCD. To elevate glutamate rapidly and stimulate inward AMPA receptor currents, we photolytically liberated glutamate from MNI-glutamate (0.5 mM) by applying a bright UV flash to the OPL while recording from horizontal cells (Rabl et al. 2006). Glutamatergic currents in control neurons reached an average amplitude of 187.1 ± 72.1 pA (n = 5), whereas MβCD-treated horizontal cells exhibited currents with an average amplitude of 216.5 ± 69.3 pA (n = 5; P = 0.999; Table 2). In addition, the time constants for current decay did not differ significantly between control (τ_decay = 36.8 ± 11.8, n = 5) and MβCD-treated neurons (τ_decay = 21.0 ± 6.02, n = 5; P = 0.27; Table 2). These findings suggest that MβCD treatment did not significantly alter the number or single-channel conductance of glutamate receptors in the postsynaptic membrane, induce glutamate receptor desensitization, or disrupt glutamate reuptake.

Table 2.

Effects of cholesterol depletion on glutamatergic currents

	Control	MβCD
Flash current amplitude, pA	187 ± 72.1 (n =5)	217 ± 69.3 (n = 5) (P = 0.999)
Flash decay time constant, ms	36.8 ± 11.8 (n = 5)	21.0 ± 6.02 (n = 5) (P = 0.27)
mEPSC amplitude,pA	4.40 ± 0.47 (n = 6 cells, 75.2 ± 24.0 mEPSCs/cell)	5.51 ± 1.86 (n = 4 cells, 89.0 ± 18.6 mEPSCs/cell) (P = 0.50)
mEPSC frequency, Hz	119 ± 11.9	125 ± 18.9 (P = 0.81)
mEPSC half-width, ms	1.79 ± 0.20	1.47 ± 0.26 (P = 0.35)

Data are means ± SE. P values for comparisons with control conditions were determined using unpaired Student's t-test. To examine potential postsynaptic effects of cholesterol depletion, we recorded currents evoked by photolysed 4-methoxy-7-nitroindolinyl-l-glutamate (MNI-glutamate) and miniature EPSCs (mEPSCs) from horizontal cells voltage-clamped at a holding potential of −60 mV. We found no significant differences in the amplitude or decay of glutamatergic currents evoked by UV photolysis of caged glutamate. We also found no differences in the amplitude, frequency, or half-width of mEPSCs in control vs. MβCD-treated conditions (P > 0.05 for all parameters).

As a further test of possible effects of MβCD treatment on postsynaptic responses, we analyzed individual mEPSCs and found no significant differences in the amplitude (pA), half-width (ms), or frequency of mEPSCs between control and MβCD-treated retinas (Table 2). Together, these data suggest that cholesterol perturbation alters EPSCs principally by presynaptic actions at cone terminals.

DISCUSSION

The present results show that cholesterol depletion increases the mobility of Ca_V channels and that this increase in mobility is accompanied by an increase in the number of Ca_V channel openings associated with release of a single synaptic vesicle at the cone ribbon synapse. Under control conditions, fusion of a single synaptic vesicle at the peak of release is accompanied by the opening of ∼3 Ca_V channels (Bartoletti et al. 2011). After expansion of the Ca_V channel confinement domain by cholesterol depletion, ∼4.5 channel openings were associated with the fusion of each vesicle. These data suggest that when Ca_V channels spend more time far from ribbon release sites, a greater number of channel openings are needed to guarantee that [Ca²⁺] rises high enough at release sites to trigger vesicle fusion. In addition to less efficient coupling between channel opening and fusion, we consider below four other possible effects of cholesterol depletion that might cause an apparent reduction in release efficiency.

1) The first effect is changes in quantal amplitude. Treatment with cholesterol-depleting agents could potentially reduce mEPSCs, and thus alter the results of deconvolution, by inhibiting vesicular glutamate transport (Tarasenko et al. 2010) or by allowing postsynaptic receptors to diffuse away from the active zone (Fernandes et al. 2010; Hering et al. 2003; Pinaud et al. 2009; Renner et al. 2009). However, we found no significant change in the amplitude or kinetics of mEPSCs. We also found no changes in the amplitude or kinetics of glutamatergic currents evoked by flash photolysis of caged glutamate. These data suggest that changes in quantal amplitude were not responsible for the apparent change in release efficiency at cone synapses. The finding that exocytotic capacitance increases stimulated by brief 5-ms steps were inhibited by cholesterol depletion also support a presynaptic site of action. The absence of significant postsynaptic effects of cholesterol depletion is consistent with the observation that cholesterol is not a major component of postsynaptic membranes in horizontal cell dendrites (Cooper and McLaughlin 1984).

At nonribbon synapses, depletion of cholesterol has been shown to enhance Ca²⁺-independent spontaneous release but to diminish evoked release of vesicles (Linetti et al. 2010; Wasser et al. 2007; Zamir and Charlton 2006). At the cone ribbon synapse, we observed no significant change in the amplitude of EPSCs following cholesterol depletion with MβCD or COase. There was also no significant increase in the frequency of mEPSCs in horizontal cells. However, it is worth noting that the vesicle pools that participate in spontaneous and evoked release at cone synapses have not been distinguished experimentally, and many of the spontaneously occurring mEPSCs in horizontal cells are likely to reflect Ca²⁺-triggered release from neighboring photoreceptor cells with relatively depolarized resting membrane potentials.

2) Changes in single Ca_V channel properties could potentially produce an apparent change in efficiency by altering the number of channel openings calculated by deconvolution of I_Ca. However, we did not observe a significant change in L-type I_Ca at cone synapses following cholesterol depletion. This contrasts with evidence that cholesterol depletion alters Ca²⁺ influx through Ca_V1.3 and Ca_V2.1 channels (Purcell et al. 2011; Taverna et al. 2004), perhaps because of differences in the type of Ca_V channels expressed at cone synapses (primarily Ca_V1.4; Zanazzi and Matthews 2009).

3) A large reduction in the affinity of Ca²⁺ for the release sensor could potentially diminish release efficiency, but a large reduction in Ca²⁺ sensitivity would also significantly slow the EPSC time to peak (Duncan et al. 2010), and this was not observed. Small changes in Ca²⁺ sensitivity would be unlikely to cause significant changes in release efficiency, because Ca²⁺ levels attained at the release site are thought to exceed 20 μM, nearly two orders of magnitude above release threshold (Thoreson et al. 2004).

4) Cholesterol depletion might influence release efficiency by directly altering SNARE protein function. Although we cannot exclude this possibility, the changes in release efficiency predicted by the model from an expansion of Ca_V channel domains were sufficient to explain observed reductions in synaptic efficiency. The predictions of the model were especially good if we assumed that the impact of the diffusion barrier declined as channels moved beyond the arciform density. This is consistent with the idea that properties of Ca²⁺ diffusion within the synaptic evagination differ from those of the bulk cytoplasm (Bartoletti et al. 2011). Taken together, the most parsimonious conclusion of these different experiments is that domain size expansion is the main cause of changes in synaptic efficiency produced by cholesterol depletion.

Ca_V channel mobility was studied by tracking the movements of individual QDs. Although it is possible for individual QDs and antibodies to cross-link with multiple Ca_V channels, we used very low dilutions of all three reagents to produce sparse labeling and minimize this possibility (Bannai et al. 2006). We chose a primary antibody that labels the large extracellular domain of the α₂δ₄ accessory Ca_V channel subunit. This subunit is required for proper trafficking of the channel to the membrane and appears to remain associated with the pore-forming subunit (Bauer et al. 2010). Trajectory values showed a unimodal distribution (Mercer et al. 2011b), consistent with the idea that individual QDs are bound to single channels and that α₂δ₄-subunits rarely, if ever, dissociate from the channel. Ca_V channels can be endocytosed under certain conditions (Cristofanilli et al. 2007), but the amplitude of I_Ca did not differ significantly in control vs. treated conditions, suggesting that the number of endocytosed channels was not significantly altered by cholesterol depletion.

To examine the effects of cholesterol depletion on neurotransmission, many studies employ drugs from the cyclodextrin class to chelate and deplete membrane cholesterol. Cyclodextrins are cyclic oligosaccharides composed of six to eight dextrose molecules that complex with cholesterol and remove it from a lipid bilayer (Irie and Uekama 1997; Stella and He 2008) but also can have other effects (Zidovetzki and Levitan 2007). Enzymatically oxidizing membrane cholesterol (Churchward and Coorssen 2009; Puglielli et al. 2005) can target cholesterol in a more specific manner. Cholesterol depletion with both MβCD and cholesterol oxidase reduced retinal cholesterol levels (Fig. 1E), increased Ca_V channel confinement area (Fig. 2), and reduced synaptic efficiency (Fig. 3). Furthermore, effects of MβCD could be reversed by bathing MβCD-treated tissue with a complex of cholesterol and MβCD to restore cholesterol to the membrane (Abulrob et al. 2005). Together, these results indicate that the observed changes in Ca_V channel mobility and synaptic efficiency were mainly due to cholesterol depletion and not other effects of MβCD.

Cholesterol-rich lipid rafts are enriched around the margins of the active zone at photoreceptor ribbon synapses (Cooper and McLaughlin 1984). The present results provide further evidence that cholesterol is localized to cone photoreceptor terminals and that depletion of membrane cholesterol enhances the lateral mobility of synaptic L-type Ca_V channels (Mercer et al. 2011b). P/Q-type (Ca_V2.1) channels but not L-type Ca_V1.2 channels appear to accumulate in lipid rafts at central nervous system (CNS) synapses (Davies et al. 2006; Taverna et al. 2004). L-type channels in photoreceptor terminals (Zanazzi and Matthews 2009) may also therefore be unlikely to accumulate in lipid rafts. The presence of membrane cholesterol tends to promote overall membrane fluidity, but removal of cholesterol lipid rafts can nevertheless enhance the mobility of integral membrane proteins by removing barriers to free diffusion (Renner et al. 2009). A removal of lipid rafts surrounding the active zone may explain the expansion of Ca_V channel confinement areas at photoreceptor synapses caused by cholesterol depletion.

Although the CNS accounts for only 2% of the body's mass, it contains ∼25% of the body's total unesterified cholesterol (Chattopadhyay and Paila 2007; Dietschy and Turley 2001). Much of the cholesterol in the CNS is found in glial cells and myelin (Krämer et al. 2001; Saher et al. 2005), but synapses and synaptic vesicles also contain cholesterol (Churchward and Coorssen 2009). The present results provide further evidence that cholesterol-rich lipid rafts are important for clustering Ca_V channels close to synaptic proteins at the presynaptic active zone (Davies et al. 2006; Gil et al. 2004; Lang et al. 2001; Salaün et al. 2005; Taverna et al. 2004, 2007; Xia et al. 2007). Changes in presynaptic Ca_V channel mobility might also contribute to cholesterol-dependent modulation of release at other synapses.

Cholesterol biosynthesis and trafficking defects contribute to various forms of neurodegeneration, including retinal degeneration (Liu et al. 2010). Smith-Lemli-Opitz syndrome (SLOS) is the most common pathology of cholesterol biosynthesis (Waterham 2006) and involves a defect in 7-dehydrocholesterol reductase activity (Fliesler 2010; Kretzer et al. 1981). Human patients with SLOS exhibit deficits in rod but not cone pathways (Elias et al. 2003; Garry et al. 2010). A rat model of SLOS created by disrupting cholesterol biosynthesis showed electrophysiological defects in both rod and cone pathways that could be partially ameliorated by a high-cholesterol diet (Fliesler et al. 2004, 2007). Lipid metabolism defects have also been reported to affect electroretinogram (ERG) b-waves in a rare form of retinitis pigmentosa (Aguirre et al. 1997; Nevet et al. 2010; Zangerl et al. 2006), and impaired cholesterol efflux in the lysosomal storage disease Niemann-Pick type C results in retinal degeneration and reduced ERG a- and b-waves (Claudepierre et al. 2010; Phillips et al. 2008). In addition to its location at the ribbon synapse, cholesterol also helps to partition proteins in photoreceptor outer segments (Albert and Boesze-Battaglia 2005; Andrews and Cohen 1983; Niu et al. 2002; Senin et al. 2004; Wang et al. 2008). Changes in the ERG b-wave caused by disorders of cholesterol metabolism appear to be largely due to changes in photoreceptor light responses (Fliesler and Bretillon 2010), but the present results suggest that changes in synaptic release from cones due to changes in Ca_V channel mobility might also contribute to visual deficits. Changes in the membrane mobility of Ca_V channels or other synaptic proteins could also contribute to neurological disorders in other tissues accompanying genetic or metabolic changes that diminish cholesterol levels.

The present results indicate that increases in presynaptic Ca_V channel mobility caused by cholesterol depletion can alter synaptic strength by increasing the average distance between release sites and Ca_V channels. Changes in Ca_V channel mobility caused by other mechanisms, such as disruption of the actin cytoskeleton (Mercer et al. 2011b), might also change synaptic strength. Interactions between transient receptor potential (TRP) channels and the cytoskeleton can be dynamically regulated (Liu et al. 2011), raising the possibility that perhaps cytoskeletal interactions with Ca_V channels (Cristofanilli et al. 2007; Gregory et al. 2011; Zhang et al. 2005) might likewise be dynamically regulated. The fusion of synaptic vesicles causes a brief expansion of Ca_V channel confinement area (Mercer et al. 2011b), suggesting that the size of the confinement area, and thus release efficiency, might also be affected by changes in the balance between exocytosis and endocytosis. Changes in presynaptic Ca_V channel mobility may thus provide a general mechanism for regulating synaptic strength.

GRANTS

This work was supported by Research to Prevent Blindness, National Eye Institute Grants EY10542 (to W. B. Thoreson) and EY10542-14S1 (to W. B. Thoreson), and a University of Nebraska Medical Center Graduate Student Fellowship (to A. J. Mercer).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: A.J.M., S.L.J., and W.B.T. conception and design of research; A.J.M., R.J.S., M.J.V.H., and W.B.T. performed experiments; A.J.M., R.J.S., S.L.J., M.J.V.H., and W.B.T. analyzed data; A.J.M., S.L.J., and W.B.T. interpreted results of experiments; A.J.M., S.L.J., and W.B.T. prepared figures; A.J.M. drafted manuscript; A.J.M., S.L.J., M.J.V.H., and W.B.T. edited and revised manuscript; A.J.M., R.J.S., S.L.J., M.J.V.H., and W.B.T. approved final version of manuscript.