Cholesterol Regulates Multiple Forms of Vesicle Endocytosis at a Mammalian Central Synapse

Abstract

Endocytosis in synapses sustains neurotransmission by recycling vesicle membrane and maintaining the homeostasis of synaptic membrane. A role of membrane cholesterol in synaptic endocytosis remains controversial because of conflicting observations, technical limitations in previous studies, and potential interference from nonspecific effects after cholesterol manipulation. Furthermore, it is unclear whether cholesterol participates in distinct forms of endocytosis that function under different activity levels. In this study, applying the whole-cell membrane capacitance measurement to monitor endocytosis in real time at the rat calyx of Held terminals, we found that disrupting cholesterol with dialysis of cholesterol oxidase (COase) or methyl-β-cyclodextrin (MCD) impaired three different forms of endocytosis, i.e., slow endocytosis, rapid endocytosis, and endocytosis of the retrievable membrane that exists at the surface before stimulation. The effects were observed when disruption of cholesterol was mild enough not to change Ca²⁺ channel current or vesicle exocytosis, indicative of stringent cholesterol requirement in synaptic endocytosis. Extracting cholesterol with high concentrations of MCD reduced exocytosis, mainly by decreasing the readily releasable pool (RRP) and the vesicle replenishment after RRP depletion. Our study suggests that cholesterol is an important, universal regulator in multiple forms of vesicle endocytosis at mammalian central synapses.

Introduction

Vesicle endocytosis contributes to synaptic transmission by recycling vesicle membrane and maintaining homeostasis of plasma membrane. A major membrane lipid component, cholesterol has been implicated in regulation of synaptic endocytosis based on observations that cholesterol extraction decreases the depolarization-evoked uptake of the amphiphilic styryl dye FM1-43, horseradish peroxidase (HRP) and vesicular synaptophysin at synapses (Wasser et al., 2007, Dason et al., 2010, Hawes et al., 2010, Rodrigues et al., 2013). However, this notion remains debatable, because cholesterol extraction may also influence action potentials (Zamir and Charlton, 2006, Smith et al., 2010), Ca²⁺ channels (Taverna et al., 2004, Mercer et al., 2012), exocytosis (Zamir and Charlton, 2006, Lang, 2007, Wasser et al., 2007, Dason et al., 2010, Hawes et al., 2010, Linetti et al., 2010, Petrov et al., 2010, Smith et al., 2010), vesicular ATPase (Yoshinaka et al., 2004, Tarasenko et al., 2010), and dispersal of vesicular proteins (Dason et al., 2014). These nonspecific effects, if present in previous studies, can affect endocytosis or confound the interpretation of observations. For example, less uptake of FM dye or HRP after cholesterol extraction may result from decreased vesicle turnover due to inhibited exocytosis, instead of slower endocytosis (Petrov et al., 2010). Endocytosis monitored with pH-sensitive fluorescence-tagged synaptophysin can appear slower when cholesterol extraction inhibits vesicular re-acidification by impairing vesicular ATPase activity. Also, because synaptophysin interacts directly with cholesterol (Thiele et al., 2000), its dynamics upon cholesterol extraction may not represent endocytosis of vesicular membrane. In contrast to these assays, the real time measurement of endocytosis using membrane capacitance does not detect an endocytosis defect in cone ribbon synapses depleted of cholesterol (Mercer et al., 2012). This observation, along with studies reporting normal endocytosis after depleting the plasma membrane cholesterol (Dason et al., 2010, Petrov et al., 2010), casts doubt over a role of cholesterol in synaptic endocytosis.

Depending on levels of synaptic activity, vesicle membrane is retrieved via different molecular pathways of distinct kinetics (Wu et al., 2007, Dittman and Ryan, 2009, Saheki and De Camilli, 2012). As shown by numerous studies, the activity-dependent forms of synaptic endocytosis include clathrin-dependent endocytosis (Jockusch et al., 2005, Granseth et al., 2006, Hosoi et al., 2009, Wu et al., 2009), clathrin-independent endocytosis (Jockusch et al., 2005, Kononenko et al., 2014), actin-dependent ultrafast endocytosis (Watanabe et al., 2013), bulk endocytosis (Koenig and Ikeda, 1989, Holt et al., 2003, Wu and Wu, 2007, Clayton et al., 2010, Gaffield et al., 2011, Nguyen et al., 2012), and kiss-and-run (He et al., 2006, Zhang et al., 2009). Whether cholesterol differentially regulates distinct forms of endocytosis has not been studied. Given the significance of cholesterol in normal functions of synapses and brains (Liu et al., 2010), it is necessary to examine closely the involvement of cholesterol in different endocytosis pathways at synapses.

We addressed the above issues at the rat calyx of Held terminals with whole-cell capacitance measurement. We focused on three widely existing forms of endocytosis, i.e., slow classical endocytosis (Wu et al., 2005, Renden and von Gersdorff, 2007, Hosoi et al., 2009, Yamashita et al., 2010, Yue and Xu, 2014), rapid endocytosis (Wu et al., 2005, Wu et al., 2009), and fast excess endocytosis that retrieves membrane pre-existing at terminal surface (Wu et al., 2009, Xue et al., 2012). We found that disrupting cholesterol with dialysis of COase or MCD impaired all these three forms of endocytosis, which was not necessarily accompanied by a change in Ca²⁺ current or exocytosis. This endocytosis impairment was not mimicked by pharmacological disruption of actin polymerization or prevented by oversupplying phosphatidylinositol 4,5-bisphosphate (PIP₂). A high concentration of MCD inhibited exocytosis mainly by reducing the size and replenishment of the RRP, at least partly as a consequence of endocytosis impairment. Collectively, our results indicate that cholesterol regulates multiple forms of synaptic vesicle endocytosis.

Methods

Preparation of brainstem slices

We prepared parasagittal brainstem slices containing the medial nucleus of the trapezoid body (MNTB) as previously described (Xu and Wu, 2005, Yue and Xu, 2014). Briefly, in accordance with guidelines of The Institutional Animal Care and Use Committee, Georgia Regents University, 8 – 10 days old Sprague-Dawley rats of either sex, which were purchased from Charles River Laboratories and bred in house, were acutely decapitated to harvest brainstems in ice-cold low-Ca²⁺ artificial cerebrospinal fluid (aCSF). The low-Ca²⁺ aCSF included (in mM): 125 NaCl, 25 NaHCO₃, 2.5 KCl, 1.25 NaH₂PO₄, 3 MgCl₂, 0.5 CaCl₂, 25 glucose, 0.4 Na ascorbate, 3 myo-inositol and 2 Na pyruvate (bubbled with 95% O₂ and 5% CO₂). The brainstems were sectioned at an automated VT1200S slicer (Leica Microsystems, Wetzlar, Germany) into slices of 180 – 200 μm thickness, which were transferred into normal aCSF of 37 °C to recover for 45 min. The normal aCSF was identical with the low-Ca²⁺ aCSF, except it contained 1 mM MgCl₂ and 2 mM CaCl₂. After recovery, the slices were kept under room temperature (22 – 24 °C).

Electrophysiology

For electrophysiological recording, the brainstem slices were perfused constantly with a bath solution containing (in mM): 105 NaCl, 20 tetraethylammonium chloride (TEA-Cl), 2.5 KCl, 1 MgCl₂, 2 CaCl₂, 25 NaHCO₃, 1.25 NaH₂PO₄, 25 glucose, 0.4 Na ascorbate, 3 myo-inositol, 2 Na pyruvate, 0.001 tetrodotoxin (TTX), and 0.1 3,4-diaminopyridine (pH 7.4 when bubbled with 95% O₂ and 5% CO₂). After visually identifying the calyx of Held terminals in slices, we performed the standard whole-cell patch-clamping technique to measure ionic current and membrane capacitance (Xu and Wu, 2005, Xu et al., 2008) with an EPC-9 amplifier controlled by the Pulse program (HEKA, Germany). The real time capacitance was measured using a lock-in function based on Lindau-Neher's method (Lindau and Neher, 1988), with a sinusoidal wave (60 mV peak-to-peak amplitude, 1 kHz) being superimposed on the holding potential of -80 mV. The resulted capacitance and conductance data were sampled at 1 kHz without averaging, while the membrane current was sampled at 20 kHz after an online Bessel filtering of 2.9 kHz. The series resistance (< 15 MΩ) was compensated by 50 – 65% with a lag of 10 μs. Patch pipettes were filled with (in mM): 125 Cs-gluconate, 20 CsCl, 4 MgATP, 10 Na₂-phosphocreatine, 0.3 GTP, 10 HEPES, 0.05 BAPTA, pH adjusted to 7.2 with CsOH. As indicated, the pipette solution was added with chemical(s) in test. The pharmacological effects were measured between 5 – 12 min after the whole-cell configuration was formed. Chemicals were purchased from Sigma-Aldrich (USA) except for COase (EMD Millipore, Germany), DNF (21^st Century Biochemicals, Inc., USA) and Pitstop 1 (Abcam Biochemicals, USA). For simultaneous recording of postsynaptic responses, we performed the whole-cell patch-clamping technique on both the calyx terminal and the soma of the connected principal neuron of the MNTB with an EPC10/2 amplifier controlled by the Patchmaster software (HEKA). The postsynaptic soma was voltage-clamped at -80 mV and dialyzed with the following (mM): 125 K-glutconate, 20 KCl, 4 MgATP, 10 Na₂-phosphocreatine, 0.3 GTP, 10 HEPES, 0.5 EGTA, pH 7.2 (adjusted with KOH). The series resistance (< 12 MΩ) was compensated by 60 – 80% with a lag of 10 s. The bath solution was further added with 100 μM cyclothiazide, 1 μM strychnine chloride, and 10 μM bicuculline methiodide (all from Abcam Biochemcals) to block desensitization of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors as well as postsynaptic currents mediated by glycine and gamma-aminobutyric acid. Recordings were carried out under 22 – 24 °C.

Data analysis

Since membrane capacitance (Cm) is proportional to area of the membrane, we measured exocytosis as the evoked Cm increase (ΔCm) over the pre-stimulus baseline, and initial rate of endocytosis (Rate_endo) as the linear rate of Cm decay within the first 4 s following the 20 ms depolarization pulse or within the first 2 s following the train of depolarization pulses (see below for details about stimulation). We also determined endocytosis efficiency by measuring the net Cm increase at 25 s after the single pulse (ΔCm_25s) or 45 s after the train stimulation (ΔCm_45s). Time constant (τ) is provided to describe endocytosis kinetics in control condition, but not for judging drug effects, because exponential fitting yields highly variable τ values in the presence of endocytic blockers and is less reliable than linear fitting of the initial Cm decay (Wu et al., 2009). To analyze changes in individual exocytosis events, we measured the size and number of asynchronous events within 4 s after the single pulse or 2 s after the train stimulation, and those of miniature postsynaptic events within 4 s before the depolarization (Yue and Xu, 2014). Data are presented as Mean ± S.E.M. Except that paired t-test is used to compare postsynaptic events before and after depolarization (Fig.3B), the statistical test is one-way ANOVA, followed by tukey test to determine the significance level of differences. In either test, p < 0.05 indicates a significant difference.

Figure 3.

COase and MCD at 2.5 mM do not alter asynchronous release following depol_20msX10. A, Sampled postsynaptic responses with presynaptic dialysis with control, COase (10 U/ml), or MCD (2.5 mM) solution. The right two panels are enlarged postsynaptic currents before and after depol_20msX10. B, Comparison between mEPSC events and asynchronous events after depol_20msX10 in control (n = 10), COase (n = 7) and MCD (n = 13).

Results

Oxidation or extraction of cholesterol inhibits slow Cm decay

Among the different activity-dependent forms of endocytosis in the calyx (Wu et al., 2005, Wu et al., 2009), we first explored the cholesterol involvement in slow endocytosis evoked by a prolonged pulse such as depol_20ms, i.e., a depolarization pulse from -80 mV to 0 mV for 20 ms (Wu et al., 2005, Renden and von Gersdorff, 2007, Hosoi et al., 2009, Yamashita et al., 2010, Yue and Xu, 2014). In control terminals, depol_20ms evoked inward Ca²⁺ current with a peak amplitude (ICa_peak) of 2.1 ± 0.1 nA (n = 8) and a current charge (QCa) of 34.4 ± 2.3 pC (Fig.1A, C), followed by an abrupt increase in membrane capacitance (ΔCm) of 454 ± 22 fF (n = 8) (Fig. 1B, C), which resulted from vesicle fusion. Following ΔCm, Cm decayed at an initial rate (Rate_endo) of 32.2 ± 3.3 fF/s and efficiently recovered towards the baseline following a mono-exponential kinetics (τ = 15.8 ± 2.4 s). The net Cm increase at 25 s after depol_20ms (ΔCm_25s) was 66 ± 19 fF, indicating that 86% of the added terminal area has been recovered. The Cm decay induced by depol_20ms represents the classical clathrin-, dynamin-dependent endocytosis because it can be blocked by inhibiting functions of clathrin and dynamin (Yamashita et al., 2005, Lou et al., 2008, Xu et al., 2008, Hosoi et al., 2009, Wu et al., 2009, Yue and Xu, 2014). To determine whether this endocytosis requires cholesterol, we dialyzed the calyx terminals with either 10 U/ml COase or 2.5 mM MCD. COase can oxidize cholesterol into 4-cholesten-3-one as the first step of cholesterol degradation, and reduce membrane cholesterol (Mercer et al., 2012) probably because cholestenone desorbs from the membrane more readily than cholesterol (Neuvonen et al., 2014). Similar to MCD that extracts cholesterol, COase has been found to impair cholesterol-dependent functions such as Ca²⁺ channel mobility on retina cone synapses (Mercer et al., 2012), neurotransmission mediated by brain-derived neurotropic factor in cortical neurons (Assaife-Lopes et al., 2010), and opening of N-methyl-D-aspartate receptors in cerebellar granule cells (Korinek et al., 2015). Dialysis of COase for 5 – 12 min did not change Ca²⁺ influx (QCa) or ΔCm, but decreased the rate of initial Cm decay (Rate_endo = 15 ± 4.8 fF/s, n = 6, p < 0.01) and increased the amount of membrane stranded at 25 s after depol_20ms (ΔCm_25s = 230 ± 43 fF, p = 0.04; Fig.1B, C). As a control of COase activity, we tested effects of COase after heat inactivation with boiling water for 10 min (Torabi et al., 2007). The boiled COase did not affect QCa, ΔCm, Rate_endo or ΔCm_25s (Fig.1B, C). MCD, which has been widely used to deplete cholesterol (Wasser et al., 2007, Dason et al., 2010, Mercer et al., 2012, Yao et al., 2013), similarly slowed Cm decay after depol_20ms without changing QCa or ΔCm (Fig.1B, C). These effects were similar for depol_20ms repeated every 80 s during 5 – 12 min of dialysis, and for a first depol_20ms applied at either 5 – 6 min or 11 – 12 min (supplementary figure). Dialysis with MCD did not alter cell morphology, resting membrane conductance or resting capacitance. The slowing of Cm decay by MCD was caused by intracellular extraction of cholesterol, because Cm decay was normal when the pipette solution was filled with cholesterol-saturated MCD (2.5 mM MCD : 0.42 mM cholesterol). Consistent with previous studies (Hosoi et al., 2009, Wu et al., 2009), Cm decay following depol_20ms was inhibited by DNF, a 12-mer peptide that can disrupt formation of clathrin-coated pits by specifically interfering with functions of amphiphysin in the clathrin-mediated endocytosis (Jockusch et al., 2005). Pitstop 1, a recently developed inhibitor of clathrin-coat maturation (von Kleist et al., 2011), also impaired Cm decay induced by depol_20ms, resulting in slower Rate_endo (12.0 ± 2.9 fF/s, n = 7, p < 0.01) and larger ΔCm_25s (271 ± 58 fF, p < 0.01, Fig.1B, C). Taken together, both oxidation and extraction of cholesterol within the calyx terminals did not change Ca²⁺ current or exocytosis induced by depol_20ms, but inhibited the following clathrin-dependent Cm decay, suggestive of cholesterol requirement for the slow classical endocytosis.

Figure 1.

Dialysis of COase or MCD inhibits slow endocytosis. A, Sampled ICa induced under depol_20ms from a calyx dialyzed with the control pipette solution. B, Sampled Cm traces showing exocytosis and endocytosis induced by depol_20ms from terminals dialyzed with the control pipette solution or a solution containing DNF (100 μM), Pitstop 1 (80 μM), boiled COase (10 U/ml), COase (10 U/ml), MCD (2.5 mM), or MCD supplemented with cholesterol (2.5 mM : 0.42 mM). C, QCa, ΔCm, Rate_endo and ΔCm_25s induced by depol_20ms from control (n = 8), DNF (n = 6), Pitstop 1 (n = 7), boiled COase (n = 6), COase (n = 6), MCD (n = 7), and MCD supplemented with cholesterol (n = 7). * p < 0.05, ** p < 0.01.

Oxidation or extraction of cholesterol inhibits rapid endocytosis

In addition to the slow classical endocytosis, rapid endocytosis with a putatively different molecular mechanism occurs under intense stimulation to accelerate membrane recovery following a large amount of vesicle exocytosis (Wu et al., 2005). We next studied whether this rapid endocytosis involves cholesterol. In calyces dialyzed with the control pipette solution, a train of ten depol_20ms delivered at 10 Hz (referred to as depol_20msX10) triggered Ca²⁺ influx with a total QCa of 222 ± 12 pC (n = 9), and, ΔCm of 1213 ± 77 fF, which was followed by bi-exponential Cm decay with τ of 2.2 ± 0.3 s and 18 ± 2 s, respectively (Fig.2). The Rate_endo within 2 s after depol_20msX10 was 194.2 ± 16.9 fF/s, which reflects primarily the contribution from rapid endocytosis owing to its seven times faster kinetics in membrane retrieval (Wu et al., 2005, Wu et al., 2009, Xu et al., 2013). In calyces dialyzed with 10 U/ml COase, Rate_endo following depol_20msX10 decreased to 122.8 ± 21.1 fF/s (n = 7, p = 0.04 vs control), suggestive of impaired rapid endocytosis. COase also slowed the full course of endocytosis due to inhibition of both rapid and slow endocytosis, leaving more membrane stranded on the surface at 45 s after depol_20msX10 (ΔCm_45s = 374 ± 104 fF vs. 103 ± 33 fF in control, p = 0.03). Similar effects were observed for dialysis with 2.5 mM MCD, but not the boiled COase or MCD supplemented with cholesterol (2.5 mM : 0.42 mM; Fig.2B, C). Neither COase nor MCD influenced the sums of QCa and ΔCm induced by depol_20msX10 (Fig.2C). Effects of COase and MCD on the slow Cm decay (Fig.1) and the rapid Cm decay suggest that cholesterol regulates both slow and rapid forms of endocytosis.

Figure 2.

Dialysis of COase or MCD inhibits rapid endocytosis. A, Sampled ICa induced under depol_20msX10 from a control calyx. B, Sampled Cm traces showing exocytosis and endocytosis induced by depol_20msX10 from terminals dialyzed with the control pipette solution or solution containing boiled COase (10 U/ml), COase (10 U/ml), MCD (2.5 mM), or MCD supplemented with cholesterol (2.5 mM : 0.42 mM). C, QCa, ΔCm, Rate_endo and ΔCm_45s induced by depol_20msX10 from control (n = 8), boiled COase (n = 6), COase (n = 7), MCD (n = 8), and MCD supplemented with cholesterol (n = 6).

Oxidation or extraction of cholesterol does not enhance asynchronous release

We have interpreted the slower kinetics of Cm decay after disrupting cholesterol as the result of inhibition of slow and rapid endocytosis. However, because Cm is proportional to the surface membrane area that increases upon exocytosis and decreases during endocytosis, Cm decay can slow down if the number and size of vesicles being fused in the asynchronous release drastically increase following depolarization. This concern is reasonable because MCD is known to enhance spontaneous exocytosis (Zamir and Charlton, 2006, Wasser et al., 2007, Petrov et al., 2014). We addressed this concern by monitoring excitatory postsynaptic current (EPSC) from the MNTB principal neurons during presynaptic dialysis with the control pipette solution or a pipette solution containing COase or 2.5 mM MCD. Consistent with our recent study (Yue and Xu, 2014), miniature EPSCs (mEPSCs) in control synapses (n = 10) occurred at 6.1 ± 1.1 Hz and had an average amplitude of 30.4 ± 3 pA, while asynchronous events after depol_20ms became 19 ± 3.3 Hz (p < 0.01, paired t-test) and 35.9 ± 2.7 pA (p = 0.01, paired t-test) (Fig.3A, B). Dialysis with COase (n = 7) or MCD (n = 13) did not change the increases of event frequency and amplitude after depol_20msX10 (p = 0.31 – 0.99, Fig.3B), indicating that they do not alter the asynchronous release after depol_20msX10. Similarly, COase or MCD did not change the frequency and amplitude of mEPSC before depol_20ms and asynchronous release after depol_20ms (data not shown). The slowing of Cm decay by COase and MCD indeed results from inhibition of endocytosis, but not from facilitation of asynchronous release following depolarization. Furthermore, dialysis with COase or 2.5 mM MCD did not change the amplitude of EPSC upon depol_20ms (COase: 8.9 ± 1.1 nA, MCD: 8.6 ± 0.6 nA; p = 0.70 and 0.95, respectively, vs. 9 ± 0.6 nA in control; n = 7 – 13 synapses), which agrees with the similar amount of exocytosis measured as ΔCm (Fig.1C).

Cholesterol disruption inhibits endocytosis of the retrievable membrane

The slow and rapid forms of endocytosis are usually compensatory, retrieving the same amount of membrane as that newly fused so that the increased Cm upon depolarization recovers to the baseline. In addition, excess endocytosis is frequently observed under intense stimulation at the calyx terminals (Wu et al., 2009, Xue et al., 2012) and in endocrine cells (Smith and Neher, 1997, Engisch and Nowycky, 1998, Lee and Tse, 2001), which includes contribution from the retrievable membrane pre-existing at the surface. Whether the retrieval of pre-existing membrane requires cholesterol has never been studied. We solved this issue by investigating the excess endocytosis following stimulation with ten depolarization pulses from -80 mV to 0 mV for 50 ms at an interval of 50 ms (referred to as depol_50msX10) from calyces bathed in 6 mM Ca²⁺ (Fig.4A, B). Consistent with previous studies (Wu et al., 2009, Xue et al., 2012), depol_50msX10 evoked a ΔCm of 1298 ± 37 fF (n = 7) in control terminals, followed by Cm decay generating a negative overshoot (ΔCm_25s = - 693 ± 124 fF) below the pre-stimulus level, which indicates retrieval of membrane pre-existing at the surface. The subsequent depol_50msX10 induced less or even none overshoot because of depletion of the pre-existing retrievable membrane (Xue et al., 2012). Indeed, contribution of pre-existing membrane to the endocytosis evoked by the first depol_50msX10 can be calculated as the difference of ΔCm_25s between the first and the third traces (C1 - C3 in the right panel of Fig.4B), which was - 763 ± 136 fF (Fig.4C) and very similar to the published result (Xue et al., 2012). In terminals dialyzed with COase, the first depol_50X10 induced ΔCm of 1194 ± 129 fF (n = 5) and a small overshoot (ΔCm_25s = - 87 ± 66 fF; Fig.4C, D). The difference between responses to the first and third depol_50msX10 was - 255 ± 42 fF, indicative of 1/3 less retrieval from pre-existing membrane than the control (p = 0.02). Instead, the boiled COase did not affect exocytosis or endocytosis induced by depol_50msX10.

Figure 4.

Dialysis of COase or MCD inhibits endocytosis of the pre-existing retrievable membrane at the surface. A, Sampled ICa induced under depol_50msX10 from a control calyx bathed in 6 mM Ca²⁺. B and C, left, Sampled Cm traces showing excess endocytosis following 3 consecutive depol_50msX10 from terminals dialyzed with the control pipette solution or solution containing DNF (100 μM), Pitstop1 (80 μM), boiled COase (10 U/ml), COase (10 U/ml), MCD (2.5 mM), and MCD (10 mM). middle, Responses to the first and the third depol_50msX10 (C1 and C3, respectively) are superimposed to show disappearance of the negative overshoot due to depletion of pre-existing membrane over repetitive depol_50msX10. right, Subtracting the third response (C3) from the first (C1) gives rise to estimate of the endocytosed amount of the membrane pre-existing at the terminal surface following the first depol_50msX10. D, QCa, ΔCm, and the amount of pre-existing membrane (C1 - C3) retrieved by the first depol_50msX10 from control (n = 8), DNF (n = 10), Pitstop 1 (n = 6), boiled COase (n = 6), COase (n = 7), 2.5 mM MCD (n = 7), and 10 mM MCD (n = 6).

Dialysis of MCD inhibited the retrieval of pre-existing membrane in a dose-dependent manner. Pre-existing membrane retrieved by the first depol_50msX10 was calculated as - 184 ± 48 fF in 2.5 mM MCD (n = 5, p < 0.01), and - 39 ± 19 fF in 10 mM MCD (n = 5, p < 0.01). In contrast, DNF or Pitstop 1 did not affect the endocytosis amount of the pre-existing membrane (- 655 ± 95 fF, n = 8, p = 0.62; - 494 ± 117 fF, n = 6, p = 0.35; Fig.4B, D), implicating its independence of acute formation of the clathrin-coated pits. Therefore, cholesterol regulates not only the slow clathrin-dependent endocytosis (Fig.1) and rapid endocytosis (Fig.2), but also the putatively clathrin-independent endocytosis of the retrievable terminal membrane (Fig.4).

Oversupply of PIP₂ does not prevent endocytosis impairment by cholesterol extraction

How does cholesterol regulate endocytosis? Cholesterol itself can control the physical properties of membrane and regulate curvature formation (Chen and Rand, 1997). Cholesterol extraction has also been reported to decrease levels of PIP₂ in plasma membrane and interfere with actin organization (Kwik et al., 2003, Hao and Bogan, 2009). As both PIP₂ and actin have been indicated important for endocytosis (Shupliakov et al., 2002, Bourne et al., 2006, Di Paolo and De Camilli, 2006, Yao et al., 2013), could they mediate cholesterol regulation of endocytosis?

We first tested whether increase of PIP₂ prevents the endocytosis impairment induced by MCD. Dialysis of 5 μM PIP₂ into the terminals did not change exocytosis or endocytosis induced by depolarization (Fig.5). Co-dialysis of 5 μM PIP₂ and 2.5 mM MCD reduced Rate_endo after depol_20ms to 15.9 ± 3.2 fF/s (n = 6, p < 0.01 vs. 32.2 ± 3.3 fF/s in control) and increased ΔCm_25s to 172 ± 16 fF (p = 0.04 vs. 66 ± 19 fF in control). The co-dialysis also reduced Rate_endo after depol_20msX10 to 87.7 ± 20.3 fF/s (n = 7, p < 0.01 vs. 194.2 ± 16.9 fF/s in control) and increased ΔCm_45s to 380 ± 79 fF (p = 0.03 vs. 103 ± 33 fF in control). Endocytosis inhibition by co-dialysis of PIP₂ and MCD was similar to that for dialysis of MCD alone (p = 0.76 – 0.87, Fig.5). Oversupplying PIP₂ did not effectively rescue endocytosis inhibition induced by cholesterol depletion with MCD, suggesting that normal cholesterol regulation of endocytosis in our experimental conditions does not involve PIP₂ production or increase.

Figure 5.

Oversupply of exogenous PIP₂ does not prevent MCD from inhibiting endocytosis. Ai, Sampled Cm traces showing endocytosis induced by depol_20ms from terminals dialyzed with the pipette solution containing PIP₂ (5 μM) with or without MCD (2.5 mM). Aii, Averaged Cm traces after being normalized by exocytosis, with S.E.M. as the error bars. Data of control and MCD are from the same recordings analyzed in Fig.1. B, QCa, ΔCm, Rate_endo and ΔCm_25s induced by depol_20ms from control (from Fig.1), MCD (from Fig.1), PIP₂ (n = 7) and PIP₂ plus MCD (n = 6). Ci, Sampled Cm traces showing endocytosis induced by depol_20msX10 from terminals dialyzed with the pipette solution containing PIP₂ with or without MCD. Cii Averaged Cm traces after being normalized by exocytosis, with S.E.M. as the error bars. Data of control and MCD are from the same recordings analyzed in Fig.2. D, QCa, ΔCm, Rate_endo and ΔCm_45s induced by depol_20msX10 from control (from Fig.2), MCD (from Fig.2), PIP₂ (n = 6) and PIP₂ plus MCD (n = 7). Note that co-application of PIP₂ and MCD inhibited endocytosis similarly as MCD alone.

Disruption of actin polymerization leads to mild inhibition of endocytosis

Next, we examined the involvement of actin by testing effects of latrunculin A, which disrupts actin polymerization. The solvent, dimethyl sulfoxide (DMSO, 0.1%), did not change the evoked QCa, exocytosis or endocytosis (Fig.6). In terminals dialyzed with latrunculin A (20 μM), Rate_endo following depol_20ms decreased to 21.5 ± 2.8 fF/s (n = 10, p = 0.09 vs. 29.7 ± 3.2 fF/s in DMSO, n = 10) and ΔCm_25s increased to 186 ± 20 fF (p = 0.03 vs. 73 ± 22 fF in DMSO), indicative of impairment of membrane endocytosis after depol_20ms. Latrunculin A was less efficacious than 2.5 mM MCD (Fig.6Aii, B), and appeared to inhibit endocytosis at the later phase after depol_20ms. Similarly, latrunculin A did not change the rate of rapid endocytosis immediately following depol_20msX10 (Rate_endo = 213.8 ± 31.8 fF/s, n = 12 vs. 186.3 ± 25.1 fF/s in DMSO, n = 10), but increased ΔCm_45s to 293 ± 46 fF (p = 0.04 vs. 143 ± 33 fF in DMSO), which reflects impairment of slow endocytosis following depol_20msX10. Furthermore, compared to dialysis of MCD (2.5 mM) alone, co-dialysis of latrunculin A and MCD (2.5 mM) did not produce additive inhibition of slow endocytosis or rapid endocytosis (p = 0.37 – 0.99, Fig.6). These observations suggest that at most actin polymerization contributes partially to cholesterol regulation of slow endocytosis.

Figure 6.

Disruption of actin polymerization leads to mild inhibition of endocytosis. Ai, Sampled Cm traces showing endocytosis induced by depol_20ms from terminals dialyzed with the pipette solution containing latrunculin A (20 μM) with or without MCD (2.5 mM). Aii, Averaged Cm traces after being normalized by exocytosis, with S.E.M. as the error bars. Data of control and MCD are from the same recordings analyzed in Fig.1. B, QCa, ΔCm, Rate_endo and ΔCm_25s induced by depol_20ms from control (from Fig.1), MCD (from Fig.1), latrunculin A (n = 10) and latrunculin A plus MCD (n = 7). Rate_endo for latrunculin A is faster than Rate_endo for MCD (p = 0.04) and for latrunculin plus MCD (p = 0.03). Ci, Sampled Cm traces showing endocytosis induced by depol_20msX10 from terminals dialyzed with latrunculin A or latrunculin A plus MCD. Cii, Averaged Cm traces after being normalized by exocytosis, with S.E.M. as the error bars. Data of control and MCD are from the same recordings analyzed in Fig.2. D, QCa, ΔCm, Rate_endo and ΔCm_45s induced by depol_20msX10 from control (from Fig.2), MCD (from Fig.2), latrunculin A (n = 12), and latrunculin A plus MCD (n = 9). Rate_endo for latrunculin A is faster than Rate_endo for MCD and for latrunculin plus MCD (p < 0.01). Cm_45s for latrunculin A is smaller than that for latrunculin plus MCD (p < 0.01) but not for MCD (p = 0.01). Note that in Aii and Cii, the averaged traces after normalization with ΔCm, which is relatively larger in the presence of latrunculin A, have smaller differences between latrunculin A and DMSO, and, between MCD and latrunculin A plus MCD.

Actin polymerization is known to regulate ultrafast endocytosis and bulk membrane retrieval at synapses (Holt et al., 2003, Richards et al., 2004, Nguyen et al., 2012, Watanabe et al., 2013), and excess endocytosis in insulin-secreting INS-1 cells (He et al., 2008). Could actin participate in cholesterol regulation of the rapid retrieval of pre-existing membrane in excess endocytosis? To this aim, we investigated effects of dialysis with latrunculin A (20 μM) or DMSO (0.1%) into the calyx terminals, which were bathed in 6 Ca²⁺. Cm increase and decay induced by repetitive stimulation with depol_50msX10 were similar in both conditions (Fig.7). Specifically, the pre-existing membrane retrieved by the first depol_50msX10 was -704 ± 110 fF in DMSO (n = 8) and -595 ± 111 fF in latrunculin A (n = 10), showing no significant difference (p = 0.50). This result suggests that unlike cholesterol, actin does not play a significant role in endocytosis of pre-existing membrane.

Figure 7.

Latrunculin does not affect endocytosis of the retrievable membrane pre-existing at the surface. A, Sampled Cm traces showing excess endocytosis following 3 consecutive depol_50msX10 from terminals which were bathed in 6 mM Ca²⁺ and dialyzed with a pipette solution containing either DMSO (0.1%) or latrunculin A (20 μM). The display of traces follows that in Fig.4B and C. B, QCa, ΔCm, and the amount of the pre-existing membrane (C1 - C3) retrieved by the first depol_50msX10 from DMSO (n = 8) and latrunculin (n = 10).

Higher doses of MCD inhibit exocytosis mainly by reducing vesicle supply

Endocytosis impairment is expected to decrease exocytosis by hindering membrane clearance from fusion sites (Hosoi et al., 2009, Wu et al., 2009, Hua et al., 2013) and by reducing the reuse of fused vesicles. However, neither COase nor MCD (tested at 2.5 mM) altered the stimulated exocytosis in measurements of Cm and EPSC (Figs. 1 – 4). This probably happened because the endocytosis inhibition by COase and MCD (2.5 mM) was not severe enough to influence exocytosis immediately. We thus tested whether higher doses of MCD induce more severe inhibition of endocytosis and decrease of exocytosis. Indeed, dialysis with 5 – 10 mM MCD led to stronger inhibition of exocytosis and endocytosis following depol_20msX10 (Fig.8A, B). MCD at 10 mM reduced ΔCm by 58% (507 ± 98 fF, n = 7; p < 0.01 vs. 1213 ± 77 fF in control), slowed Rate_endo (4 ± 3 fF/s, p < 0.01 vs. 194 ± 17 fF/s in control), and increased ΔCm_45s (440 ± 106 fF, or 87% of ΔCm; p = 0.03 vs. 103 ± 33 fF in control), reflecting significant reduction of exocytosis and nearly complete inhibition of both rapid and slow endocytosis following depol_20msX10. There was also a 24.3% decrease of QCa summed over depol_20msX10 (168 ± 16 pC vs. 222 ± 12 pC in control, p = 0.04) but no significant change of QCa upon depol_20ms (30.6 ± 1.3 pC vs. 34.4 ± 2.3 pC in control; 11% reduction). MCD at 5 mM inhibited ΔCm by 20% (970 ± 115 fF, n = 7, p = 0.19) and Rate_endo by 80% (39 ± 9 fF/s, p < 0.01) without affecting QCa (Fig. 8B). In summary, effects of MCD (5 – 10 mM) on exocytosis are inhibitory, which is consistent with current literature.

Figure 8.

High doses of MCD decrease exocytosis by reducing supply of the releasable vesicles. A, Sampled Cm recordings showing exocytosis and endocytosis induced by depol_20msX10 from the terminals dialyzed with a pipette solution containing 5 or 10 mM MCD. B, Summary of QCa, ΔCm, Rate_endo and ΔCm_45s induced by depol_20msX10. C, Sampled Cm jumps induced by depolarization pulses from -80 mV to 0 mV for 1 – 30 ms. D, Summarized ΔCm following pulses of different durations (left). Note that ΔCm induced by the 20 ms pulses nearly saturates in both control (n = 5 – 8) and 10 mM MCD (n = 6 – 9), and thus reflects the RRP size. Right: After being normalized with the RRP size, ΔCm is very similar between control and MCD for a given depolarization duration, indicative of similar release probabilities in both conditions.

What pathways did MCD target to cause inhibition of exocytosis? Since depol_20msX10 depletes both vesicles initially residing in the RRP and vesicles subsequently replenished into the RRP during the train (Wu et al., 2005), reduction of ΔCm could be caused by decrease of release probability, reduction of the RRP, and/or impairment in vesicle replenishment after RRP depletion. To find out the cause(s), we measured ΔCm evoked by depolarization pulses stepping from -80 mV to 0 mV for 1 – 30 ms. Pulses of 20 and 30 ms were tested at last so that exocytosis evoked by shorter pulses was not affected by vesicle depletion. Consistent with the literature (Sun et al., 2002), ΔCm increased with longer pulses until it largely saturated at the 20 ms pulse for both control (n = 5 – 8) and 10 mM MCD (n = 6 – 9, Fig.8D), which is attributed to depletion of the RRP. ΔCm upon depol_20ms is therefore taken as the RRP size in both cases. The RRP size in terminals dialyzed with 10 mM MCD was 295 ± 22 fF (n = 9), about 36% smaller than that in control (461 ± 28 fF, n = 6, p < 0.01). Normalizing the ΔCm induced by different pulses to the RRP size, we calculated the proportion of the RRP vesicles released by each pulse. The normalized numbers are similar between control and MCD (the right panel, Fig.8D), indicating that MCD did not affect the release probability upon brief depolarization pulses (e.g., 1 – 2 ms). Taken together, 10 mM MCD inhibited exocytosis mainly by reducing vesicles residing in the RRP and vesicles being replenished after depletion of the RRP, likely as an immediate consequence of endocytosis impairment (Hosoi et al., 2009, Wu et al., 2009, Hua et al., 2013).

Discussion

Taking advantage of whole-cell patch-clamp measurements from the calyx terminals, the current study investigated the requirement of cholesterol for presynaptic functions with an emphasis on vesicle endocytosis. Our major novel finding is that cholesterol regulates multiple forms of endocytosis, which function under different activity levels and with different molecular mechanisms.

As discussed in the introduction, involvement of cholesterol in synaptic endocytosis has been suggested, but with great controversy because of contradictory observations, limits of previous assays, and complication from non-specific effects. Furthermore, assays using FM dye or HRP uptake have used minutes long potassium depolarization to induce detectable endocytosis, which lacks the temporal resolution to separate classical endocytosis and bulk endocytosis (Gaffield et al., 2011) or to identify the activity-dependent switching of endocytosis pathways as shown in pHluorin imaging of hippocampal boutons (Kononenko et al., 2014). Previous studies do not inform which endocytosis pathway in synapses requires cholesterol. Using the whole-cell membrane capacitance measurement, we found that three common forms of endocytosis, i.e., slow endocytosis, rapid endocytosis, and endocytosis of the retrievable membrane pre-existing at the terminal surface, were all impaired after cholesterol oxidation or extraction. Endocytosis of the retrievable terminal membrane was resistant to the DNF peptide and Pitstop 1 (Fig.4), which disrupt formation or maturation of the clathrin-coated pits (Jockusch et al., 2005, von Kleist et al., 2011), indicating that this endocytosis either retrieves pre-formed clathrin-coated pits or does not involve clathrin at all. On the other hand, the slow endocytosis is inhibited by DNF and Pitstop 1 (Fig.1), and thus depends on formation of the clathrin-coated pits (Hosoi et al., 2009, Wu et al., 2009). Therefore, our results suggest that cholesterol participates in multiple forms of endocytosis regardless of their different molecular requirements, for example, for clathrin.

The preferential presynaptic target of cholesterol disruption is endocytosis, not Ca²⁺ channel current or exocytosis, because dialysis of COase or 2.5 mM MCD into the calyx terminals inhibited endocytosis without affecting QCa or exocytosis (Figs.1 and 2). This dialysis did not cause any noticeable change to cell morphology, resting membrane conductance or capacitance either. Furthermore, endocytosis was induced with depolarization pulses and measured as Cm decay, which does not involve action potentials or vesicle reacidification. Therefore, the observed inhibition of endocytosis by COase and MCD (2.5 mM) should not be a consequence of global disturbance in terminals or off-target effects on action potentials, Ca²⁺ channels, exocytosis, or vesicle re-acidification, suggesting that cholesterol directly regulates synaptic vesicle endocytosis. Because endocytosis contributes to exocytosis by recycling vesicle membrane and by clearing the newly fused membrane to facilitate vesicle priming (Hosoi et al., 2009, Wu et al., 2009, Hua et al., 2013), the requirement in endocytosis confers to cholesterol the importance in exocytosis. Indeed, we observed significant reduction in exocytosis along with nearly complete block of endocytosis at the terminals dialyzed with 10 mM MCD (Fig.8A, B). The reduction of exocytosis resulted from decreases of the RRP size and vesicle replenishment after RRP depletion, consistent with expectation for endocytosis block (Hosoi et al., 2009, Wu et al., 2009, Hua et al., 2013). MCD at 10 mM also inhibited QCa during depol_20ms by 11% and QCa over depol_20msX10 by 24.3% (Fig.8B). In comparison to depol_20ms, depol_10ms resulted in 55% reduction in QCa (from 37.3 ± 1.4 pC to 20.6 ± 1.3 pC, n = 6) but only 27.1% reduction in ΔCm (from 461 ± 28 fF to 336 ± 38 fF, Fig.8D). Inhibition of QCa by MCD could thus contribute a minor part to the observed decreases in the measured RRP and vesicle replenishment. Consistent with the finding in hippocampal neurons (Wasser et al., 2007), MCD did not alter the release probability under brief depolarization (Fig.8D). Taken together, our results have revealed the stringent requirement for cholesterol in endocytosis and its importance in supplying the releasable vesicles.

Two pieces of evidence indicate cholesterol itself is essential for synaptic vesicle endocytosis. First, inhibition of endocytosis after cholesterol extraction does not depend on reduction of membrane PIP₂ or block of actin polymerization. Oversupplying PIP₂ via patch-clamp pipettes, that prevents a Rho kinase inhibitor from blocking endocytosis in the calyx terminals (Taoufiq et al., 2013), did not alleviate the impairment of endocytosis by 2.5 mM MCD (Fig.5). This result does not necessarily exclude importance of PIP₂ in endocytosis. Rather it is likely that MCD at this concentration did not reduce PIP₂ at the related sites. In this study, blocking actin polymerization with latrunculin A did not affect the putatively clathrin-independent retrieval of pre-existing membrane (Fig.7), and caused mild inhibition of slow endocytosis, which was less efficacious than 2.5 mM MCD (Fig.6). Co-application of latrunculin and MCD did not produce additive inhibition either. While disruption of presynaptic actin polymerization has been reported to impair clathrin-mediated endocytosis (Shupliakov et al., 2002), clathrin-independent endocytosis (Kononenko et al., 2014) and bulk membrane retrieval (Holt et al., 2003, Richards et al., 2004, Nguyen et al., 2012), it has little effect in some other studies (Betz and Henkel, 1994, Kuromi and Kidokoro, 1998, Sankaranarayanan et al., 2003, Gaffield et al., 2011, Bleckert et al., 2012). The causes underlying such controversy are not well understood yet, but could be related to activity- and temperature-dependent switch of actin functions in endocytosis. For example, treating hippocampal boutons with latrunculin does not affect endocytosis induced by 20 Hz stimulation (Sankaranarayanan et al., 2003), but inhibits endocytosis following stimulation at 40 Hz (Kononenko et al., 2014) or ultrafast endocytosis found under near physiology temperature (Watanabe et al., 2013). Actin in nerve terminals probably serves a molecular scaffolding but not propulsive role (Sankaranarayanan et al., 2003), which depends on activity and temperature. It remains interesting to investigate at the calyx whether actin contributes to acceleration of endocytosis under physiology temperature (Renden and von Gersdorff, 2007). Nevertheless, under the current experimental conditions, PIP₂ production and actin polymerization have limited contribution to the cholesterol regulation of vesicle endocytosis at synapses. Second, disruption of cholesterol impaired three distinct forms of endocytosis in the calyx, regardless of their underlying molecular mechanisms, stimulation, and quantity of the preceding vesicle fusion. The common requirement for cholesterol in synaptic endocytosis is reminiscent of previous studies on non-neuron cells where cholesterol participates in various forms of endocytosis, including clathrin-dependent endocytosis (Rodal et al., 1999, Subtil et al., 1999), macropinocytosis (Grimmer et al., 2002, Vidricaire and Tremblay, 2007), and clathrin-independent endocytosis (Kirkham and Parton, 2005, Lariccia et al., 2011). As a lipid regulating membrane rigidity and fluidity, cholesterol is known to control the bending and curvature formation of lipid membranes (Chen and Rand, 1997), which is critical for further membrane invagination (Subtil et al., 1999) and progress towards membrane scission (Yao et al., 2013). Such biophysical property of cholesterol may account for the importance of cholesterol in the three forms of endocytosis we studied.

Vesicular sterol has been indicated important for synaptic endocytosis (Dason et al., 2010), while the role of plasma membrane cholesterol remains controversial. Bath incubation with MCD reduces HRP uptake or synaptophysin endocytosis in cultured hippocampal synapses (Wasser et al., 2007, Hawes et al., 2010) but does not affect FM dye uptake in Drosophila neuromuscular junctions (Dason et al., 2010). It is also unclear whether MCD can get into the cytoplasm to target vesicular cholesterol after incubation for tens of minutes. In our tests, dialysis of MCD significantly inhibited endocytosis of the retrievable membrane on the terminal surface (Fig.4). Because the terminals were voltage-clamped at -80 mV and TTX remained present to block action potential firings, there should be little vesicle fusion and insertion of vesicular membrane into the surface before test. Therefore, depleting cholesterol from the pre-existing, retrievable surface membrane can effectively block its own endocytosis. This unique finding suggests that cholesterol at plasma membrane participates in regulation of synaptic endocytosis.

To date, mounting evidence has highlighted the importance of cholesterol in the nervous system by showing that impairment of cholesterol homeostasis is associated with neurodegenerative disorders such as Niemann-Pick Type C (NPC) disease, Alzheimer disease, Huntington disease, and Parkinson's disease (Liu et al., 2010). NPC disease is well known to result from impaired homeostasis of membrane cholesterol when mutations in NPC1 or NPC2 lead to disrupted cholesterol trafficking from lysosomes to plasma membrane. α -Synuclein, a presynaptic protein linked to Parkinson's disease, can interact with cholesterol-enriched lipid rafts (Fortin et al., 2004), and deplete membrane cholesterol when applied externally in excess (Ronzitti et al., 2014). Interestingly, endogenous synuclein proteins are also reported to regulate synaptic vesicle endocytosis (Vargas et al., 2014). In our study, acute extraction of cholesterol with MCD inhibited endocytosis starting at 2.5 mM, and further reduced exocytosis and Ca²⁺ channel current at 10 mM. The inhibition of Ca²⁺ current at the calyx is opposite to enhancement of L-type Ca²⁺ channel currents by MCD in recombinant cells (Davies et al., 2006), chick cochlear hair cells (Purcell et al., 2011), and dopaminergic neurons (Ronzitti et al., 2014). Since the Ca²⁺ channels at the calyx terminals of P8 – P10 rats are mainly N-type and P/Q-type, it is likely that regulation of Ca²⁺ channels by cholesterol depends on the specific types of Ca²⁺ channels. It remains interesting to determine how acute effects of cholesterol disruption are related to the above disease related scenarios. To summarize, the current study has demonstrated a universal cholesterol requirement in multiple forms of synaptic vesicle endocytosis and its significance in vesicle supply, providing new insights into the synaptic functions of cholesterol. Our findings can potentially promote the mechanistic understanding of the neurodegeneration associated with the impairment of cholesterol homeostasis.

List of Abbreviations

COase

cholesterol oxidase

MCD

methyl-β-cyclodextrin

PIP₂

phosphatidylinositol 4,5-bisphosphate

ATP

adenosine triphosphate

HRP

horseradish peroxidase

Cm

membrane capacitance

EPSC

excitatory postsynaptic current

RRP

readily releasable pool