Fast, temperature-sensitive and clathrin-independent endocytosis at central synapses

Abstract

The fusion of neurotransmitter-filled vesicles during synaptic transmission is balanced by endocytotic membrane retrieval. Despite extensive research, the speed and mechanisms of synaptic vesicle endocytosis have remained controversial. Here, we establish low-noise time-resolved membrane capacitance measurements that allow monitoring changes in surface membrane area elicited by single action potentials and stronger stimuli with high-temporal resolution at physiological temperature in individual bonafide mature central synapses. We show that single action potentials trigger very rapid endocytosis, retrieving presynaptic membrane with a time constant of 470 ms. This fast endocytosis is independent of clathrin, but mediated by dynamin and actin. In contrast, stronger stimuli evoke a slower mode of endocytosis that is clathrin-, dynamin-, and actin-dependent. Furthermore, the speed of endocytosis is highly temperature-dependent with a Q₁₀ of ~3.5. These results demonstrate that distinct molecular modes of endocytosis with markedly different kinetics operate at central synapses.

Introduction

Synaptic communication in the nervous system relies on the release of synaptic vesicles from nerve terminals. Because vesicle fusion increases presynaptic plasma membrane, a balance between endocytosis of fused membrane and vesicle release is required to maintain synaptic function over time (Südhof, 2004). However, the speed and mechanisms of endocytosis remain controversial, mainly due to technical limitations. Separate modes of endocytosis have been distinguished according to their speed or underlying mechanism. The use of electron microscopy to study synaptic vesicle endocytosis (Heuser and Reese, 1973) has recently provided evidence for an ultrafast endocytosis time course (Watanabe et al., 2013), but only allows investigating “snapshots” at a single time point after stimulation in any one cell. Fluorescence imaging of synaptic vesicle endocytosis can observe endocytosis over time in single synapses (Balaji et al., 2008; Hua et al., 2011; Klingauf et al., 1998; Leitz and Kavalali, 2011; Ryan et al., 1996), but has mostly been restricted to neuronal cultures and features relatively low temporal resolution, which is often limited by the speed of vesicle re-acidification. Measurements of membrane capacitance (C_m) on the other hand allow time-resolved investigation of endocytosis in brain slices with high temporal resolution (Lou et al., 2008; Sun et al., 2002; von Gersdorff and Matthews, 1994; Smith et al., 2008), but usually require strong stimuli to achieve sufficient signal-to-noise ratio. Moreover, C_m measurements at physiological temperature are complicated by the heating bath perfusion, which causes large artifacts in C_m measurements due to bath level fluctuations. Most previous C_m measurements were therefore performed at room temperature without continuous bath perfusion. To overcome these limitations we established low-noise whole-cell presynaptic C_m measurements at physiological temperature at both mature cerebellar and hippocampal mossy fiber boutons, two functionally very different central excitatory synaptic terminals allowing direct presynaptic patch-clamp recordings (Delvendahl et al., 2013; Hallermann et al., 2003; Ritzau-Jost et al., 2014; Vyleta and Jonas, 2014). With these technical advancements we were able to investigate the mechanisms and kinetics of endocytosis following a single action potential (AP) or after a short AP train or stronger voltage-clamp depolarizing at physiological temperature in mature central synapses.

Results

Ultrafast single-AP-evoked endocytosis

We first used presynaptic patch-clamp recordings and C_m measurements in cerebellar mossy fiber boutons (cMFBs) for a time-resolved investigation of the endocytosis time course. With optimized recording conditions using quartz-glass pipettes (Dudel et al., 2000; Hallermann et al., 2005), we resolved C_m changes associated with single action potentials (APs) at mammalian physiological temperature (36° C). A recorded AP waveform voltage command with a half-duration of 129 μs elicited a Ca²⁺ current with half-duration of 112 ± 4 μs (n = 34 cMFBs), consistent with previous results (Ritzau-Jost et al., 2014), and a transient C_m increase that decayed rapidly within the first second (Figure 1A). However, some C_m transients may be unrelated to exo- and endocytosis (Wu et al., 2005; Yamashita et al., 2005). We therefore specifically blocked exocytosis with tetanustoxin light chain (TeNT-LC), which inhibits synaptic vesicle fusion by cleaving synaptobrevin (Figure 1B, Figures S1E–G). For single APs, a small and rapidly decaying C_m component remained with TeNT-LC (Figure 1C). This C_m component did not change over time and after repeated stimulation (Figure S1G), indicating that the transient C_m increase is not caused by exocytosis of vesicles with pre-assembled SNARE-complexes (Hayashi et al., 1994), but reflects non-exocytosis C_m changes (Yamashita et al., 2005). Subtraction of the C_m trace with TeNT-LC from control revealed that a single AP leads to a C_m increase of 1.3 ± 0.2 fF (n = 39 cMFBs), corresponding to exocytosis of 18 ± 2 synaptic vesicles (assuming a single vesicle capacitance of 70 aF; Hallermann et al., 2003). Exocytosis was followed by a rapid C_m decay with time constant of 470 ± 70 ms (n = 39; Figure 1C). Thus, endocytosis evoked by single APs is ultrafast at cMFBs, consistent with recent results at cultured hippocampal synapses using flash-and-freeze electron microscopy methods (Watanabe et al., 2013).

Figure 1. Ultrafast single-AP-evoked endocytosis.

(A) Top: Voltage command (V_command) used for AP-evoked capacitance recordings in cMFBs. The AP waveform was recorded in a previous experiment with axonal stimulation. An example of a resulting Ca²⁺ current (I_Ca) is depicted below. On the right, V_command and I_Ca are shown on an expanded time scale and half-durations are indicated. Middle: Example single-AP-evoked C_m trace (average of 40 consecutive sweeps). Bottom: Corresponding series and membrane resistance (R_s and R_m, respectively).

(B) TeNT-LC effectively blocks synaptic vesicle exocytosis in cMFBs. Top: Voltage protocol with 3-ms depolarization from −80 mV to 0 mV. Middle: Pharmacologically isolated Ca²⁺ current immediately after break-in (control, black) and after 3:00 minutes of whole-cell recording (5 μM TeNT-LC, blue). Bottom: Corresponding C_m traces. Right: Average data of Ca²⁺ current amplitudes and C_m increase (ΔC_m) elicited with 3-ms depolarizations for control (black) and 5 μM TeNT-LC (blue; n represents number of cMFBs). Data are represented as mean ± SEM.

(C) Grand average of AP-evoked C_m responses (black, n represents number of cMFBs). Blocking synaptic vesicle exocytosis with 5 μM TeNT-LC (blue) revealed a transient C_m increase not related to exocytosis. Subtraction (gray) shows the time course of endocytosis following a single AP. See also Figure S1.

Highly temperature-sensitive endocytosis at hippocampal and cerebellar mossy fiber synapses

AP firing in cMFBs can reach very high frequencies (Ritzau-Jost et al., 2014), and very fast endocytosis may constitute an adaptation to this firing behavior. To address whether the observed fast endocytosis is specific to cMFBs or a general property of central synapses, we analyzed C_m changes in hippocampal mossy fiber boutons (hMFBs), which operate at lower frequencies and have very different functional properties, such as a broader AP-half-width, facilitation of excitatory postsynaptic currents, and prominent long-term potentiation (Delvendahl et al., 2013). At hMFBs, a train of ten 1-ms depolarizations delivered at 50 Hz resulted in an increase in C_m, which was followed by a very rapid C_m decay with a fast time constant of 1.2 ± 0.4 s (n = 8 hMFBs, Figure 2A). We next analyzed the temperature dependence of endocytosis in hMFBs and cMFBs, because (1) previous studies suggested that ultrafast endocytosis requires physiological temperatures (Watanabe et al., 2013), and because (2) the comparison of the temperature dependence of endocytosis at hMFBs and cMFBs allows testing if these two central synapses use similar endocytotic mechanisms. The speed of C_m decay following 50 Hz trains was highly temperature dependent at hMFBs, being much faster at 36° C than at 30° C or at room temperature (Figures 2A–C). As in the calyx of Held, the time course of endocytosis follows a double-exponential at high temperatures (Renden and von Gersdorff, 2007). The resulting Q₁₀ temperature coefficient of the amplitude-weighted time constant of the C_m decay in hMFBs was 3.3 ± 1.9 (bootstrap SEM, corresponding to a 16–84% confidence interval; Figure 2D), which is higher than previously reported (Balaji et al., 2008; Fernández-Alfonso and Ryan, 2004; Granseth and Lagnado, 2008). We also analyzed the temperature dependence of endocytosis at cMFBs. Comparison of the C_m decay in response to 3-ms depolarizations at 23° C and 36° C in cMFBs (Figure S2A) revealed a Q₁₀ temperature coefficient of 3.8 ± 1.3 (bootstrap SEM; Figure 2D). The strong overlap of the 16–84% confidence intervals indicates that the Q₁₀ of endocytosis is similar in hMFBs and cMFBs. These results show that endocytosis is very fast at cMFBs and hMFBs with very similar temperature-sensitivity, suggesting that similar mechanisms operate at these central synapses. We therefore focus on cMFBs from now on to gain insights into the molecular machinery mediating fast endocytosis.

Figure 2. Highly temperature-sensitive endocytosis at hippocampal and cerebellar mossy fiber synapses.

(A) Example traces of C_m recordings in hMFBs evoked by a train of ten stimuli (1-ms depolarizations to +20 mV) delivered at 50 Hz (arrow) at 36° C (red), 30° C (gray), and 24° C (blue). Lower traces represent corresponding membrane and series resistance (R_m and R_s, respectively).

(B) Grand average C_m traces recorded at 36° C, 30° C, and 24° C (color code as in A; n represents number of hMFBs). The decay of the grand average traces was best fit with the sum of two exponentials with time constants of 1.1 s (67%) and 11.3 s for 36° C, the sum of two exponentials with time constants of 2.3 s (43%) and 7.4 s for 30° C, and a single exponential function with time constant of 10.7 s for 24° C.

(C) Average amplitude-weighted time constant (τ_w) of the exponential fits to the C_m decay for the three temperatures. Data are represented as mean ± SEM, n is given in B.

(D) Left: Histogram of Q₁₀ values by bootstrap analysis of endocytosis rates obtained in hMFBs based on τ_w data shown in panel C. Right: Corresponding histogram of Q₁₀ values by bootstrap analysis of endocytosis rates obtained in cMFBs using 3-ms depolarizations at 23° C and 36° C (cf. Figure S2A). See also Figure S2.

Clathrin-independent single-AP-evoked endocytosis

We applied inhibitors of proteins involved in important molecular pathways mediating endocytosis and studied endocytosis following single APs in cMFBs. AP-evoked fast endocytosis was not affected by the clathrin inhibitor pitstop 2 (control: endocytosis τ = 470 ± 70 ms, n = 38; pitstop 2: τ = 490 ± 80 ms, n = 30; p = 0.92; Figures 3A and 3B), indicating that fast endocytosis is clathrin-independent. In contrast, the dynamin-inhibitor dynasore and the actin polymerization inhibitor latrunculin A blocked endocytosis evoked by single APs almost completely (both p < 0.0001, Figures 3A and 3B). The endocytosis inhibitors had no effect on amplitude or kinetics of presynaptic Ca²⁺ currents (Figure S3). These data suggest that fast AP-evoked endocytosis in cMFBs is clathrin-independent, but mediated by dynamin and actin, consistent with findings at calf chromaffin cells and cultured mouse hippocampal synapses (Artalejo et al., 1995; Watanabe et al., 2013; Watanabe et al., 2014).

Figure 3. Clathrin-independent single-AP-evoked endocytosis.

(A) Top: Voltage command used for AP-evoked C_m recordings in cMFBs. Bottom: Grand averages of AP-evoked C_m responses (n represents number of cMFBs) for control (black), application of a clathrin-inhibitor (pitstop 2, 25 μM; orange), a dynamin-inhibitor (dynasore, 100 μM; red), and an actin-inhibitor (latrunculin A, 25 μM; green). Gray solid lines are exponential fits to the C_m decay.

(B) Average time constants of endocytosis and amplitudes of AP-evoked ΔC_m with endocytosis inhibitors (color code as in A). The speed of endocytosis was unaltered by pitstop 2 (p = 0.92), but significantly slowed by dynasore and latrunculin A (p < 0.001). Data are represented as mean ± SEM. See also Figure S3.

Distinct molecular modes of endocytosis

Our approach allows comparing the speed and molecular mechanism of endocytosis evoked by single APs and by stronger stimuli in a single bouton. Trains of 20 APs at a frequency of 300 Hz elicited presynaptic Ca²⁺ currents with amplitude facilitation characteristic for Ca_V2.1 channels at cMFBs (Figure 4A; Ritzau-Jost et al., 2014). Such AP trains evoked a C_m increase that was blocked by TeNT-LC (Figure 4B). The C_m decay following AP train stimulation was not affected by the presence of pitstop 2, but strongly prolonged by application of dynasore or latrunculin A (Figure 4B). The resulting time constants of endocytosis were 920 ± 190 ms for control, 1000 ± 160 ms for pitstop 2 (p = 0.92), and ~6 s for dynasore and latrunculin A, with comparable amplitude of the C_m increase under all conditions (Figures 4E and 4F). Thus, trains of APs at high-frequency result in endocytosis that is slower than that for single APs (cf. Figure 1), which is consistent with previous findings (Sun et al., 2002), but both stimuli trigger endocytosis operating with similar molecular mechanisms.

Figure 4. Distinct molecular modes of endocytosis.

(A) Top: Voltage command of 20 APs at a frequency of 300 Hz for C_m recordings in cMFBs. Bottom: Corresponding Ca²⁺ currents with slight amplitude facilitation.

(B) Example C_m traces following 20 APs at 300 Hz for control (black), application of a clathrin-inhibitor (pitstop 2, 25 μM; orange), a dynamin-inhibitor (dynasore, 100 μM; red), an actin-inhibitor (latrunculin A, 25 μM; green), and with TeNT-LC (5 μM, blue).

(C) Top: 30-ms voltage step from −80 mV to 0 mV. Bottom: Corresponding Ca²⁺ current.

(D) Example C_m traces following a 30-ms depolarization as shown in C.

(E) Summary of the effect of endocytosis inhibitors on exocytosis evoked by different stimuli recorded in cMFBs (color code as in panel B and D). The endocytosis inhibitors dynasore, latrunculin A, and pitstop 2 had no effect on exocytosis evoked by weaker stimuli (1 AP, 20 APs, 1 ms, and 3 ms), whereas TeNT-LC completely blocked synaptic vesicle exocytosis. For 30-ms step pulses, exocytosis was reduced by latrunculin A. Inset: Enlargement of the first three stimulus types on a logarithmic scale.

(F) Summary of the effect of endocytosis inhibitors on time constants of endocytosis evoked by different stimuli. The clathrin-inhibitor (pitstop 2, orange) had no effect on fast endocytosis evoked by single APs and AP trains (arrows), but stronger impact on endocytosis evoked by depolarizations of 1, 3, or 30 ms duration. In contrast, a dynamin-inhibitor (dynasore, red) and an actin-inhibitor (latrunculin A, green) reduced the speed of endocytosis evoked by all tested stimuli. Throughout the figure, data are represented as mean ± SEM and asterisks indicate significance level with Kruskal-Wallis tests and post hoc Mann-Whitney-U tests both with Bonferroni-Holm correction. See also Figures S3 and S4.

We next examined if even stronger stimuli such as prolonged depolarizations (Figures 4C and 4D) entail endocytosis with time course similar to that for single APs or trains of APs. The amount of exocytosis was not reduced by endocytosis inhibitors for short durations of depolarizations, but significantly reduced by latrunculin A for 30 ms pulses (Figure 4E), consistent with previous findings (Hosoi et al., 2009; Lee et al., 2012; Rizzoli and Betz, 2005; Sankaranarayanan et al., 2003). TeNT-LC blocked synaptic vesicle exocytosis for all types of stimuli as expected (Figures 4E and S1E–G; Sakaba et al., 2005). To directly compare the kinetics of endocytosis, we focused on the initial C_m decay within 2 s after the stimulus. The biexponential C_m decay of 1–30 ms depolarizations is analyzed in Figure S4. Pitstop 2, dynasore, and latrunculin A markedly slowed endocytosis following depolarizing pulses, demonstrating that slow endocytosis is clathrin-, dynamin-, and actin-dependent (Figure 4D). These findings demonstrate that two distinct modes of endocytosis occur at presynaptic terminals that are likely to be triggered depending on the type of stimulus. Indeed, endocytosis was still faster for AP train stimuli (50 APs) than for short depolarizations when both stimuli produced similar ΔC_m (Figures S4C–D). These data indicate that actin and dynamin are required for all modes of endocytosis, whereas clathrin is not required for a rapid mode of endocytosis elicited by single APs or short AP trains (Figure 4F).

Discussion

Using presynaptic C_m measurements we here demonstrate that endocytosis following single APs at physiological temperature is very fast and clathrin-independent. Endocytosis evoked by strong stimuli, however, is slower and requires clathrin. Our study extends previous findings by (i) determining the speed and mechanisms of endocytosis for single APs in mature synapses in brain slices, thoroughly addressing potential C_m artifacts, by (ii) quantifying the temperature coefficient of endocytosis in two separate central synapses, and by (iii) dissecting distinct temporal and molecular modes of endocytosis. Our data provide a potentially unifying explanation for previous discrepancies regarding kinetics and modes of endocytosis.

The time course and mechanisms of endocytosis we obtained with time-resolved methods are in good agreement with recent flash-and-freeze findings, where endocytotic structures were observed from 50 ms until ~1 second after the stimulus (Watanabe et al., 2013; Watanabe et al., 2014). The 470-ms time constant of AP-evoked endocytosis is consistent with these recent electron microscopic findings, because initial endocytotic structures (i.e., shallow and deep pits) are not detected with C_m measurements. Previous studies of fast endocytosis using C_m measurements have been confounded by C_m artifacts (Wu et al., 2005; Yamashita et al., 2005). Here, we were able to distinguish exo- and endocytosis from transient C_m artifacts (Figure 1; Yamashita et al., 2005). The slowing of endocytosis kinetics with increasing AP number (cf. Figures 1, 4, and S4) is consistent with previous findings (Sun et al., 2002). The observed fast endocytosis following single APs is also comparable to the speed of endocytosis in cone photoreceptor cells (Van Hook and Thoreson, 2012).

In contrast to the fast endocytosis observed here with presynaptic C_m measurements (see also Sun et al., 2002; Van Hook and Thoreson, 2012; Watanabe et al., 2013), fluorescence imaging techniques usually reveal slower endocytosis rates (e.g., Balaji et al., 2008; Kononenko et al., 2013; but see Leitz and Kavalali, 2014). One possible explanation is provided by the strong impact of both temperature (Figure 2) and stimulation strength (Figure 4F) on endocytosis kinetics. We determined a very high temperature coefficient of endocytosis that might be caused by the strong temperature dependence of dynamin GTPase activity (Q₁₀ of 5.7; Leonard et al., 2005). Taking the high temperature dependence into account, our results using square depolarizing pulses are compatible with endocytosis time constants of 10–40 s obtained with fluorescence imaging at room temperature (Balaji et al., 2008; Granseth et al., 2006; Hua et al., 2011; Kononenko et al., 2013; Ryan et al., 1996). Furthermore, C_m measurements report endocytosis of presynaptic membrane in contrast to some fluorescence imaging techniques based on tagged vesicular proteins that measure protein endocytosis. It is thus possible that the endocytosis of some synaptic vesicle proteins has a slower speed and relies on different mechanisms than membrane retrieval (Kononenko and Haucke, 2015). However, we also note that whole-cell recordings may disrupt normal intracellular function via the dialysis of mobile proteins and changes in membrane tension (Heidelberger et al., 2002; Hull and von Gersdorff, 2004).

We found that two modes of endocytosis with distinct underlying mechanisms can be evoked in central synapses in acute brain slices, similar to findings in cultured hippocampal neurons (Kononenko et al., 2014). Our data indicate that fast endocytosis following single APs or short AP trains is independent of clathrin, consistent with previous work (Artalejo et al., 1995; Kononenko et al., 2014; Van Hook and Thoreson, 2012; Watanabe et al., 2014). Moreover, actin and dynamin inhibitors suggest that the molecular mechanism of fast endocytosis involves actin as well as dynamin (Artalejo et al., 1995; Watanabe et al., 2013). The slower endocytosis evoked by voltage steps, on the other hand, requires clathrin in addition to actin and dynamin. Why is endocytosis slower for AP trains or brief depolarizations than for single APs? The membrane area that can be retrieved in an ultrafast manner may be limited according to the concept of a ‘readily retrievable pool’ of vesicles (Hua et al., 2011) and stronger stimuli might exhaust this fast endocytosis capacity (Thomas et al., 1994). However, the gradual slowing of a monophasic decay observed here and in previous studies using C_m measurements (Renden and von Gersdorff, 2007; Sun et al., 2002; Yamashita et al., 2005) is inconsistent with a fast exhaustible mechanism, which would predict varying contributions of two components with constant kinetics. Additional mechanisms must therefore operate at central synapses (Alabi and Tsien, 2013; Cousin, 2015). Indeed, Ca²⁺ was shown to trigger both exo- and endocytosis (Wu et al., 2009; Yamashita et al., 2010), and the amount of Ca²⁺ influx seems to influence the time course of endocytosis (Balaji et al., 2008; Leitz and Kavalali, 2011; Sankaranarayanan and Ryan, 2001; von Gersdorff and Matthews, 1994). Furthermore, trains of 50 APs mimicking the ΔC_m of 1-ms pulse depolarizations resulted in faster endocytosis than 1-ms pulse depolarizations (Figures S4C–D). This finding indicates that the spatiotemporal Ca²⁺ signal also regulates the speed of endocytosis independent of the amount of exocytosis (Hosoi et al., 2009; Kononenko et al., 2014; Midorikawa et al., 2014; Yamashita et al., 2010).

Our mechanistic conclusions are based on pharmacological inhibitors and are thus limited by potential off-target effects (Park et al., 2013; Willox et al., 2014). We also note that we tested only a single pharmacological inhibitor per endocytosis protein. However, the mechanistically important finding that fast endocytosis is clathrin-independent is unlikely to be confounded by off-target effects because pitstop 2 had no effect on rapid endocytosis. If anything, off-target effects would erroneously increase the impact of pitstop 2. In addition, pitstop 2 did not impact presynaptic Ca²⁺ currents (Figure S3). The similar effects of latrunculin A and dynasore argue for a dynamin- and actin-dependence, although we cannot exclude that other forms of endocytosis are also inhibited by these drugs (Park et al., 2013).

In summary, our data demonstrate that the speed of endocytotic membrane retrieval following physiological stimuli is very fast in central synapses. The direct comparison of single APs and stronger stimuli revealed two molecularly distinct modes of endocytosis with markedly different speed. In combination with our findings of a strong temperature dependence of endocytosis kinetics, our results explain why previous studies performed at room temperature, with stronger stimuli, or with lower temporal resolution were unable to detect fast endocytosis. Thus, our data indicate that fast, clathrin-independent endocytosis is an essential feature underlying the function of mature central synapses.

Experimental Procedures

Methods are described in detail in the Supplemental Experimental Procedures. Presynaptic C_m measurements were performed essentially as described previously at cMFBs of 35- to 55-day-old C57BL/6 mice (Ritzau-Jost et al., 2014) and at hMFBs of 22- to 30-day-old rats (Hallermann et al., 2003). All experiments were performed at 35–37° C unless otherwise stated. For low-noise C_m measurements, thick-walled quartz glass pipettes were used (Dudel et al., 2000).