Rapid bulk endocytosis and its kinetics of fission pore closure at a central synapse

Abstract

Upon exocytosis, fused vesicles must be retrieved for recycling. One route of retrieval is to generate endosome-like structures, from which small vesicles bud off. Endosome-like structures are widely thought to be generated slowly (≈1 min) from the plasma membrane, a process called bulk endocytosis. Although the concept of bulk endocytosis seems established, the kinetic evidence showing the instant of the bulk membrane fission at synapses is still missing. The present work provides this missing piece of evidence at a calyx-type synapse. We used the capacitance measurement technique, which offers a high time resolution (≈1 ms) to resolve the fission process. The instant of bulk membrane fission was reflected as a brief downward capacitance shift (DCS) of ≈20–500 fF (mean = 123 fF) with 10–90% decay time of ≈30–500 ms. At least 8.6–11.0% of exocytosed vesicles were retrieved by DCSs. During a DCS, the decrease in the fission pore conductance was detected, from which we found that the fission pore diameter decreased at ≈39 nm/s. This provided the measurement of the rate of fission during bulk endocytosis at synapses. The DCS frequency peaked (≈0.021 Hz) in <10 s after stimulation and decayed with a half time <20 s, indicating that the time course of bulk endocytosis is much faster than previously estimated with low time-resolution techniques. Our results also suggest that bulk endocytosis was composed of two kinetically different steps: the DCS that reflected the fission process and the time between stimulation and the DCS, during which membrane invagination led to the fission pore formation.

Endocytosis is essential for the maintenance of synaptic transmission, because it recycles, and thus prevents depletion of, synaptic vesicles in the nerve terminal. Three forms of endocytosis are hypothesized at synapses (1). In the best-characterized form, full collapse of vesicle occurs with the plasma membrane upon exocytosis, followed by clathrin-mediated invagination and fission during endocytosis. “Kiss-and-run” exocytosis and endocytosis involve fusion followed rapidly by fission without full collapse of vesicle membrane. The third form is by forming an intermediate membrane compartment, the endosome-like structure, from which vesicles bud off.

After strong stimulation, endosome-like structures dominate at many different nerve terminals, suggesting this third form represents a major mechanism to retrieve vesicles (2–6). When the endosome-like structure was first observed, it was proposed to result from the fusion of multiple small vesicles newly retrieved from the plasma membrane (4, 7). Recent electron microscopic studies reveal deep invaginations of plasma membrane resembling endosome-like structures in conditions that block or slow down endocytosis (5, 8–10). These observations led to an alternative but more widely accepted hypothesis that the endosome-like structure comes from retrieval of a large piece of membrane directly from the plasma membrane, called bulk endocytosis. However, the possibility that deep membrane invaginations are generated only when endocytosis is blocked or slowed down cannot be ruled out. In line with this possibility, deep-membrane invaginations are not reported or are rarely observed in physiological conditions at synapses such as the frog neuromuscular junction and the calyx-type synapse (2–4, 6). Furthermore, the possibility that deep-membrane invaginations are not internalized as one piece, but serve as the platform from which small vesicles pinch off, has not been ruled out (9, 10). In summary, although many studies support the existence of bulk endocytosis, the kinetic evidence showing the instant of bulk membrane scission at synaptic nerve terminals has remained missing. This missing piece of evidence is provided in the present work.

We studied bulk endocytosis at a live synaptic nerve terminal, the calyx of Held, by the use of the capacitance measurement technique, which offers a time resolution (≈1 ms) much faster than traditionally used electron microscopy and imaging techniques. With this technical improvement, we characterized bulk endocytosis, including the time course, the frequency, and the kinetics of the fission process in various stimulation conditions.

Results

DCSs Reflect Bulk Endocytosis.

If a large piece of membrane is pinched off from the plasma membrane, it causes a large and rapid decrease in the area and thus the capacitance (proportional to the area) of the plasma membrane. Using Lindau–Neher's technique (11) to measure capacitance at the calyx of Held, we found large downward capacitance shifts (DCSs) after a strong stimulus, a train of 10 pulses of 20-ms depolarization (from −80 to +10 mV, if not mentioned) at 10 Hz. As shown in our recent study (12), this stimulus induced a capacitance jump (1,287 ± 102 fF, n = 11) that decayed biexponentially with time constants of 1.4 ± 0.4 s (n = 11, amplitude = 407 ± 48 fF) and 18.4 ± 1.7 s (n = 11, amplitude = 880 ± 58 fF), respectively (e.g., Fig. 1A). In this recent study (12), however, we did not characterize DCSs, because they occurred infrequently (see Fig. 2B) that failed to attract our attention, and the data were far from enough for quantification. By accumulating a much larger amount of new data, we found that DCSs were superimposed on the smooth biexponential decay (Fig. 1A, large arrow), and thereby we characterized their properties in the present work. A DCS was not an instantaneous drop of capacitance. Instead, its 10–90% decay time was ≈30–500 ms (e.g., Fig. 1 B–E). Its size ranged from our detection limit of 20 fF up to ≈500 fF (e.g., Fig. 1, but see Fig. 2 A and D for summary).

Fig. 1.

DCS reflects fission during bulk endocytosis. (A) Ten depolarizing pulses of 20 ms from −80 to +10 mV at 10 Hz (vertical arrow) induced a capacitance jump (Cm), which decayed smoothly most of time, except for an abrupt DCS (arrow). The DCS was readily identified when the capacitance decay was low-pass-filtered at 30 Hz (gray) and differentiated (ΔCm/Δt; see SI Text I). Measured G_m and G_s are also shown. (B Left) Plot of the DCS in A in larger scales (same order as in A). The dashed line is a linear regression fit of the baseline before the DCS. (Center) A baseline-corrected, filtered Cm trace (Upper) was obtained by subtracting the dashed line from the low-pass-filtered capacitance trace (gray, Left). As described in SI Text I, this trace (Top) was used to predict changes in G_m (Middle, red) and G_s (Bottom, red). (Right) The measured (black) and the predicted (red) G_m and G_s shown in Left and Middle are superimposed. (C) Cm, G_m, and G_s traces averaged from 62 DCSs, each of which was >100 fF. The Cm trace was baseline-corrected (black), which was further low-pass-filtered at 30 Hz (gray) and used to predict changes in G_m and G_s (red, applies to C–E). (Scale bars apply to C–E.) (D) Similar to C, but a single DCS <100 fF. (E) Similar to C, but an average from 79 DCSs, each of which was <100 fF. (F) Equivalent circuit of the calyx without (Left) and with (Center and Right) bulk endocytosis. Labels are explained in the text.

Fig. 2.

The timing and the size of DCS. (A) DCS size plotted vs. the onset of DCSs (n = 141 DCSs). The stimulus (10 pulses of 20-ms depolarization at 10 Hz) was applied at time 0 (applies to Figs. 2–4). The stimulus was repeated one to four times in a calyx. A total of 154 stimuli were applied to 64 calyces. All data were superimposed. Analysis of these data gave rise to B–D. (B and C) The frequency (B) or accumulated amplitude of DCSs (ΣDCS amp, C) binned every 20 s per stimulus, is plotted vs. the time at which DCS occurred. Data are plotted as mean ± SE (applies to data with error bars in all figures). B Inset shows a DCS (large arrow) occurring at ≈3 s after the stimulus (vertical arrow). (D) DCS size distribution.

DCSs were not accompanied by and thus not caused by changes in the membrane conductance (G_m, e.g., Fig. 1B). DCSs were not paralleled with a downward or an upward shift of the series conductance (G_s, e.g., Fig. 1B), indicating they were not caused by the G_s change. However, DCSs larger than 100 fF were consistently accompanied by a transient decrease in G_s (Fig. 1 B and C). Instead of paralleling the time course of the DCS, the G_s decreased to the maximum at the middle of the DCS and recovered at the end of the DCS (Fig. 1 B and C). Individual DCSs smaller than 100 fF were not often accompanied by a clear G_s change (e.g., Fig. 1D). However, when 79 DCSs (from 64 calyces) with amplitudes of 20–100 fF were averaged, the G_s change became evident (Fig. 1E).

To determine whether the G_s change is caused by the DCS, we derived equations that made use of the measured capacitance change during a DCS to predict changes in G_s and G_m, then compared the predicted with the measured G_s and G_m. The results are summarized below, and described in detail in supporting information (SI) Text I.

A calyx was electrically equivalent to a single compartment circuit (Fig. 1F Left) composed of three parameters, an access resistor (R_a) that links to the voltage clamp and a membrane resistance (Rm_c) in parallel with a membrane capacitance of the calyx (Cm_c). If a DCS reflected fission of the bulk membrane with a capacitance of Cm_b, Cm_c (or Rm_c) is in parallel with a circuit composed of the Cm_b in series with the fission pore resistance (R_p, Fig. 1F Center). Because we did not detect any change in G_m during a DCS (e.g., Fig. 1A), the bulk membrane resistance was too large and thus could be neglected in the circuit (Fig. 1F). When a sine-wave command is applied, the circuit during bulk membrane fission (Fig. 1F Center) is equivalent to a single compartment circuit (Fig. 1F Right) composed of R_a, Cm′, and Rm′, where

In these two equations, ω = 2 π f, where f, the sine wave frequency, was set as 1,000 Hz in our experiments. From this equivalent single-compartment circuit (Fig. 1F Right), Lindau–Neher's technique measures the membrane resistance (Rm _LN′) and access resistance (R _{a_LN′}) based on

where

and

In Eqs. 5–7,

From Eqs. 3 and 4, Rm R_LN′ and R R_aLN′, and thus Gm (= 1/Rm_LN′) and Gs (= 1/R_aLN′) can be predicted if R_a, Rm′, Cm′, and Rm_c in Eqs. 5–9 are known. R_a and Rm_c can be obtained from the measured G_s and G_m after a DCS (e.g., Fig. 1A); Cm′ is the measured capacitance at each time point during a DCS; and Rm′ can be obtained from Eq. 2. For Eq. 2, Cm_b is measured as the amplitude of the DCS (e.g., Fig. 1A), and R_p can be measured from Eq. 1 as

where Cm_c is the measured capacitance after a DCS (e.g., Fig. 1A). With these equations, the measured capacitance during and after a DCS and the measured G_m and G_s after a DCS can be used to predict changes in G_m and G_s during a DCS.

The predicted G_m showed little change, which was similar to the observed one (e.g., Fig. 1B). The predicted G_s matched well with the observed G_s change during individual DCSs larger than 100 fF (n = 62 DCSs from 64 calyces, e.g., Fig. 1B). The match was further confirmed when all DCSs larger than 100 fF were averaged (n = 62 DCSs, Fig. 1C). The predicted change in G_s for individual DCSs smaller than 100 fF was small and closed to the noise level (e.g., Fig. 1D), explaining why a G_s change was not observed consistently for smaller DCSs. However, when the noise was reduced by averaging 79 DCSs below 100 fF, the measured mean G_s change matched well with the predicted change (Fig. 1E). In summary, the close match between the predicted and the measured G_s change confirmed that a DCS reflects membrane fission, the final step of bulk endocytosis. These results also ruled out the possibility that a DCS is a sequential fission of many small vesicles, because the latter would not cause a detectable G_s change.

The Frequency and Size of DCS.

To characterize bulk endocytosis, we determined the frequency, size, contribution, and onset of DCSs. The number of DCSs ranged from 0 to 6 in 80 s after the stimulus. Because the frequency was low, we repeatedly applied the stimulus at an interval of ≈100 s (20-s baseline plus 80 s after the stimulus) for 154 times in 64 calyces. We observed a total of 141 DCSs. DCSs occurred during baseline but at a low frequency of 0.003 ± 0.001 Hz (n = 64 calyces, Fig. 2 A–C; see SI Text II for additional controls and discussion). The frequency increased to 0.021 ± 0.002 Hz (n = 64 calyces, P < 0.001) within 10 s after the stimulus, followed by a decline to the baseline within 80 s with a half-decay time <20 s (Fig. 2 A and B). Some DCSs occurred within ≈0.4–3.0 s after the stimulus (Figs. 2B and 3 A and D). A similar time course was observed when the amplitude of DCS was summed every 20 s (Fig. 2C). These results showed a rapid time course of bulk endocytosis.

Fig. 3.

DCS frequency depends on the stimulus intensity and exocytosis. (A) A DCS occurred soon after a 20-ms depolarization. Inset enlarges the DCS. (B and C) The frequency (B) and the accumulated amplitude of DCSs (ΣDCS amp, C), per stimulus, binned every 20 s, are plotted vs. the DCS onset time after a 20-ms depolarization (triangles, n = 34 calyces). For comparison, the data after 10 pulses of 20-ms depolarization at 10 Hz are also plotted (gray circles, same as Fig. 2 B and C). (D) Sampled capacitance changes induced by 10 depolarizing pulses at 10 Hz at 3 min (Left) and 13 min (Right) after break in with a pipette containing BoNT/C (1 μM). D Inset shows DCSs in larger scales. (E) DCS frequency (per stimulus) plotted vs. the time after 10 pulses of 20-ms depolarization at 10 Hz. Data were obtained at 1–3 min (Left) and 10–15 min (Right) after break in with a pipette containing BoNT/C (1 μM) (n = 30 calyces).

The amplitude of each DCS ranged from ≈20 to 500 fF with a mean of 131 ± 11 fF (n = 141 DCSs from 64 calyces). DCS amplitude distribution was eccentric with a peak near the threshold of detection, 20–60 fF (Fig. 2D). The eccentric distribution implies that we may have significantly underestimated DCSs because of events below the threshold of detection, 20 fF. The total amount of membrane retrieved by DCSs within 80 s after a stimulus was 99 ± 6 fF (n = 154 stimuli from 64 calyces), which was 8.6 ± 0.7% of the stimulus-induced capacitance jump, i.e., the total amount of endocytosis. This percentage represents the minimum contribution of bulk endocytosis, because we may have significantly underestimated DCSs below 20 fF.

DCS Depends on Stimulation Intensity and Exocytosis.

The results shown thus far were obtained with one strong stimulus, 10 pulses of 20-ms depolarization at 10 Hz. Next, we determined whether DCS occurs after milder stimulus, requires exocytosis to be triggered, and occurs after action-potential trains.

When a single pulse of 20-ms depolarization was applied, the DCS frequency increased from a baseline of 0.004 ± 0.001 to 0.008 ± 0.002 Hz (48 calyces, Student's t test, P < 0.05) in the first 20 s after stimulation (e.g., Fig. 3A, summarized in Fig. 3 B and C, triangles). The DCS frequency (0.008 ± 0.002 Hz) within 20 s after stimulation was significantly less than that (0.021 ± 0.002 Hz, n = 64 calyces; Fig. 3B, circles) after 10 depolarizing pulses at 10 Hz (Student's t test, P < 0.01). Similarly, the summed amplitude of all DCSs within 20 s after a 20-ms depolarization (Fig. 3C, triangles) was significantly less than that after 10 depolarizing pulses at 10 Hz (Fig. 3C, circles). These results suggest that the DCS frequency and summed amplitude depended on the stimulation intensity. However, the summed amplitude of DCS within 60 s after a 20-ms depolarization was 11.0 ± 1.6% (n = 77 stimuli from 48 calyces) of the capacitance jump. This percentage was similar to that (8.6 ± 0.7%) after 10 depolarizing pulses at 10 Hz, suggesting that the relative contribution of bulk endocytosis to total endocytosis is similar within the two stimulation patterns tested.

To determine whether the stimulation triggers DCS by exocytosis, we added 1 μM botulinum neurotoxin C (BoNT/C, Wako Chemicals USA, Inc., Richmond, VA) into the pipette solution to block exocytosis but not calcium currents (12). The bath was heated to 33–34°C to facilitate the block (13). At 1–3 min after break in, the capacitance jump induced by 10 depolarizing pulses at 10 Hz was 1,347 ± 63 fF (n = 30). The DCS frequency within 20 s after stimulation was 0.023 ± 0.004 Hz (n = 30, Fig. 3 D and E Left). The 20–80% rise time for DCSs larger than 100 fF was 103 ± 13 ms (n = 20 DCSs), which was significantly faster than that (163 ± 15 ms, n = 62 DCSs, P < 0.01) in control, suggesting that temperature increase may speed up the fission time course.

At 10–15 min after break in with a pipette containing BoNT/C (Fig. 3 D and E Right), the capacitance jump induced by 10 depolarizing pulses at 10 Hz was reduced to 125 ± 28 fF (n = 30), or 11 ± 2% of that at 1–3 min after break in. The decrease was not because of run down, because in the absence of BoNT/C, the capacitance jump decreased by only ≈16% at 10–15 min after break in (12). The DCS frequency within 20 s after stimulation was 0.002 ± 0.001 Hz (n = 30, Fig. 3 D and E Right), which was significantly less than that (Fig. 3 D and E Left) at 1–3 min after break in (Student's t test, P < 0.01). The frequency decrease was not because of run down of endocytosis, because the DCS number (within 80 s after 10 depolarizing pulses at 10 Hz) was not significantly different at 1–5 min (0.80 ± 0.07 DCS per trace, n = 103 traces) and 10–15 min (0.67 ± 0.16 DCS per trace, n = 16 traces, P > 0.7) after break in with a control pipette solution. Thus, exocytosis was required for the stimulation to increase the DCS frequency. This requirement also ruled out the possibility that DCS is caused by a stimulation-dependent conductance change in the axon connected to the calyx.

To determine whether DCS could also be triggered by action-potential trains, we mimicked an action potential with a 1-ms depolarization from −80 to 7 mV [action potential-equivalent stimulus (AP-e)] (12, 14). In physiological conditions, neurons that give rise to calyces may fire from a few to hundreds of hertz for a long time (15, 16). Thus, we used 50–500 AP-e at 30–100 Hz to mimic physiological action-potential trains. After 500 AP-e at 100 Hz, the DCS frequency (Fig. 4 A and B) or the summed amplitude (Fig. 4 A and C) increased and peaked within 20 s, then decayed with a half decay time <20 s (n = 46 calyces). After 50 AP-e at 30 Hz, the DCS frequency (0.008 ± 0.002 Hz, n = 52 calyces) or the summed amplitude (32 ± 8 fF/20 s, n = 52 calyces) within the first 20 s after stimulation was significantly increased from the baseline (Fig. 4 D–F, P < 0.01), but significantly less than the frequency (0.022 ± 0.003 Hz, n = 46 calyces, P < 0.01) or the summed amplitude (65 ± 11 fF/20 s, n = 46 calyces, P < 0.01) after 500 AP-e at 100 Hz. Thus, DCS after AP-e trains also depended on the stimulation intensity. DCS size distribution obtained after AP-e trains was eccentric with a peak at 20–60 fF (not shown). In summary, the results obtained with AP-e trains (Fig. 4) were similar to those obtained with trains of 20-ms depolarization (Figs. 2 and 3).

Fig. 4.

Bulk endocytosis occurs after trains of AP-e. (A) Sampled capacitance changes and a DCS (arrow) induced by 500 AP-e at 100 Hz (vertical arrow). Inset shows the DCS in larger scales. (B and C) The frequency (B) and the accumulated amplitude (ΣDCS amp, C) of DCSs per stimulus, binned every 20 s, are plotted vs. the time after 500 AP-e at 100 Hz. The stimulus was applied one to four times at each of 46 calyces recorded. (D–F) Similar to A and B, but with a different stimulus (50 AP-e at 30 Hz, n = 52 calyces).

Rate of Fission Pore Closure.

Eq. 10 (see also SI Text I, section 2) allowed us to measure the fission pore conductance, G_p (= 1/R_p). G_p ranged from ≈0.5 to 18 nS at the onset of DCSs, which decreased to ≈0.1–0.3 nS at the end of DCSs (e.g., Fig. 5 A–C). The detectable initial G_p (Fig. 5D) and final G_p (not shown) increased as the DCS size increased, consistent with the prediction from Eq. 10 (see SI Text III for details). The largest detectable initial G_p (18 nS) was likely to reflect the initial G_p value during fission.

Fig. 5.

Fission G_p measurements. (A–C) Baseline-corrected Cm, G_p, and D_p for three different sizes of DCSs (black). The 20–80% decrease of D_p was fit with a linear regression line (red) with a slope (red) of −40 nm/s (A), −28 nm/s (B), and −31 nm/s (C), respectively. Cm trace was low-pass-filtered at 30 Hz (gray), from which G_p and D_p were calculated (Eqs. 10 and 11). (D) The detectable initial G_p increased as the DCS size increased (circles). Data were obtained after 10 depolarizing pulses at 10 Hz (also applies to E). (E) The rate of D_p closure during fission (D_p rate) plotted vs. the DCS size (black circles). The data were also sorted based on the DCS size and averaged for every 10 data points. The resulting mean D_p rate (±SE) was plotted vs. the corresponding mean DCS size (red). The red line is a linear regression fit of the mean data (correlation coefficient: −0.01).

G_p can be used to estimate the fission pore diameter (D_p) (17, 18). The simplest model, as widely assumed, is a cylindrical pore with

where ρ, the bath solution resistivity, is a constant (100 Ω cm), and L, the length of the pore, is commonly assumed to be a constant, equivalent to the thickness of two membrane bilayers (15 nm) (17, 18). Calculated from Eq. 11 with the measured G_p, the initial D_p ranged from ≈3–19 nm (Fig. 5 A–C), which decreased to an undetectable level within 500 ms (Fig. 5 A–C). The slope of the 20–80% decrease in D_p, measured from a linear regression fit, was 39 ± 2 nm/s (n = 141 DCSs from 64 calyces, e.g., Fig. 5 A–C, red lines). This slope reflected the fission pore closure rate. It was similar despite >15-fold differences in the DCS size and thus was not correlated with the DCS size (Fig. 5E, red, linear regression fit, correlation coefficient = −0.01; see Fig. 5 legend for calculation of red symbols).

The accuracy of D_p calculation depends on the estimate of ρ and L (Eq. 11). The calculation assumed a constant L. In theory, the change in G_p during a DCS could be caused solely by an elongation of L. In this condition, a decrease of G_p from 10 to 0.3 nS, as we observed during some DCSs (Fig. 5 A–C), would increase L by 33.3 times. This means an increase of L from an initial value of 15 to 500 nm in ≈100–200 ms. We consider this possibility unlikely. However, we could not rule out the possibility that L changes slightly during a DCS.

Endocytosis Is Composed of Fission Pore Formation and Fission.

The onset time of DCSs (Fig. 2A) reflects the time to form a Ω membrane profile, provided that the Ω profile is not preformed. Preformed Ω profiles are unlikely for three reasons. First, large membrane invagination connected to the calyx membrane is not seen in electron micrographs under resting conditions (3). Second, DCS frequency depended on the stimulation intensity and exocytosis (Figs. 3 and 4), suggesting that DCS reflects a process altered by conditions at synapses but not something inherently constant or static. Third, if DCS is because of fission of a preformed Ω profile, repeated stimulation should exhaust DCS. In contrast, the DCS frequency within the first 20 s (Fig. 6A) or 80 s (not shown) after the stimulus was similar to that after the same stimulus repeated 100 s later at the same calyx (n = 15 calyces, Student's t test, P > 0.8). These results suggest that during the time between stimulation and DCS, membrane invagination takes place and eventually forms a fission pore.

Fig. 6.

Endocytosis is composed of two steps: membrane invagination and fission. (A) The number of DCS in the first 20 s after stimulation was not exhausted by repeated stimulation. The interval between the first and second stimulus (10 depolarizing pulses at 10 Hz) was 100 s (n = 15 calyces). Before these two stimuli, no stimulation was applied. (B) DCS onset time is independent of DCS size. DCS onset time (after 10 depolarizing pulses at 10 Hz) plotted vs. DCS size (open). The data were also sorted based on the DCS size and averaged for every 10 data points. The resulting mean onset time (±SE) was also plotted vs. the corresponding mean DCS size (red). The line is a linear regression fit of the mean data (correlation coefficient: −0.03). Data were taken from Fig. 2A. (C) A schematic drawing showing that endocytosis is composed of two steps, membrane invagination and fission.

Our results suggest that fission pore forms between stimulation and the onset of DCS and closes during DCS. The following two pieces of evidence suggest that fission pore formation and closure are mediated by different kinetic processes. First, if the pore closure rate is the same before and during the DCS, similar fission pore closure rates for the different sizes of DCSs (Fig. 5E) predicts an earlier onset for smaller DCSs. However, the time between stimulation and the DCS onset was independent of the DCS size (Fig. 6B, linear regression fit, correlation coefficient = −0.03; see also figure legends). Second, for DCSs occurring in the first 20 s after 10 depolarizing pulses at 10 Hz, the mean DCS size was 123 ± 11 fF (n = 64 DCSs, Fig. 2A), and the mean onset was 7.9 ± 0.8 s after stimulation (Fig. 2 B and C). Considering that the specific membrane capacitance is 9 fF/μm² (19), a DCS of 123 fF may correspond to a round membrane sheet with a diameter of ≈4,200 nm (Fig. 6C). To convert this round membrane sheet to a sphere within 7.9 s, the diameter must decreased at a rate of 532 nm/s (= 4,200 nm/7.9 s). This is much faster than the rate of fission pore closure (39 nm/s, Fig. 4E), suggesting that membrane invagination is mediated by a mechanism different from fission. The estimate of the rate of membrane invagination (532 nm/s) assumed that membrane invagination occurs at the end of stimulation and finishes at the onset of the DCS. This represented the lower limit, because membrane invagination could be triggered at certain time after stimulation, and finish before the onset of a DCS. We concluded that bulk endocytosis was composed of two kinetically different steps, membrane invagination that forms a fission pore and fission (Fig. 6C).

Discussion

Observation of the Fission Process at Synapses.

By using the capacitance measurement technique that provides a time resolution at ≈1 ms, we revealed the kinetics of bulk membrane fission at the calyx-type synapse. Bulk membrane fission was reflected as a DCS ranging from <20 to 500 fF (Fig. 2D) in a brief time window of ≈30–500 ms. At the onset of a DCS, the fission D_p could be as large as 19 nm. As the DCS proceeds, the fission D_p decreased at ≈39 nm/s (Fig. 5E). The DCS frequency increased as the stimulation intensity increased (Figs. 3 and 4), which is analogous to the electron microscopic observation that raising the stimulation intensity increases the number of endosome-like structures (2–4, 6). Considering that the specific membrane capacitance is ≈9 fF/μm² (19), a DCS ranging from <20 to 500 fF corresponded to a sphere membrane structure with a diameter ranging from <0.8 to 4.1 μm. This estimate is in the same order as the size of the vacuole-like structures (up to 2.5 μm in diameter) observed at goldfish retinal bipolar synapses (20). Although assuming a sphere membrane structure for calculation, because sphere endosome-like structures are found at the calyx (3), we could not rule out the possibility that bulk endocytosis also results in other shapes of membrane structures, like an elongated one observed at frog neuromuscular junctions (2, 4). In summary, these results provide kinetic evidence revealing the instant of bulk membrane fission, which leads to the formation of large endosome-like structures at synapses.

Bulk Endocytosis Is Much Faster Than Generally Assumed.

We found that the frequency of DCSs peaked at ≈7.9–9.6 s after stimulation and decayed with a half-decay time <20 s (Figs. 2–4). Some bulk endocytosis events occurred within a few seconds after stimulus (Fig. 2A). Such a rapid time course is in sharp contrast to the current view that endosome-like structures are generated on a time scale of minutes (2–4, 21). For example, generation of endosome-like structures at frog neuromuscular junctions are estimated to take 5–15 min after a prolonged train of stimulation (2, 4). This discrepancy is likely because of methodological differences. Electron microscopy offers a low time resolution and measures the lifetime of endosome-like structures, whereas the capacitance measurement technique offers a much higher time resolution and measures the time course of generating endosome-like structures from the plasma membrane. Consistent with this explanation, when the optical imaging was used to study bulk endocytosis with an improved time resolution of ≈1 min, a large number of bulk membrane structures were observed at 1 min after high potassium application at goldfish retinal bipolar nerve terminals (20). Similarly, endosomes were observed with electron microscopy when the fixative was applied as early as 15–45 s after high potassium application (6). We suggest modifying the current view to a rapid generation of endosome-like structures, followed by slow bud off of small vesicles from endosome-like structures. Rapid generation of endosome-like structures provides an explanation for why large and deep membrane invaginations are not reported in physiological conditions (2–4) but are observed when endocytosis is blocked or slowed down (5, 8–10).

Subkinetic Steps of Endocytosis: Fission Pore Formation and Closure.

Functional block of different proteins involved in clathrin-mediated endocytosis revealed Ω membrane profiles with wide or narrow necks under the electron microscope (22). These observations led to the hypothesis that endocytosis is composed of two steps, membrane invagination and fission (22). However, the kinetics of these two steps was unclear at synapses. The present work attempted to measure the kinetics of these two steps during bulk endocytosis. We found that bulk endocytosis was composed of two steps, membrane invagination that forms a fission pore and the fission pore closure. During the fission pore closure, the D_p decreased at a rate of ≈39 nm/s. This rate was in the same order as that during bulk endocytosis in pituitary nerve terminals, which do not form synapses but secrete peptide hormones (17). During formation of the fission pore, membrane invagination must reduce the neck of the Ω membrane profile at a rate much faster than the rate of fission pore closure. We estimated that the neck diameter could decrease at a rate of 532 nm/s or higher for DCSs that occurred within the first 20 s after stimulation.

Method Used to Study Bulk Endocytosis and Fission Pore Kinetics.

The present work resolved bulk endocytosis at live synapses by capacitance measurements. Compared with electron microscopic and imaging methods, the capacitance measurement offers a much higher time resolution. More importantly, it allows for measuring the fission pore closure. However, it could not resolve bulk endocytosis of small pieces of membrane (<20 fF), which almost certainly have led to a significant underestimate of bulk endocytosis. Furthermore, it requires whole-cell recordings of large nerve terminals, which are not commonly found in the nervous system.

The fusion and fission G_p were previously measured with Neher–Marty's piecewise-linear capacitance measurement technique at nonsynaptic secretory vesicles (17, 23). This technique requires manual readjustment of the phase setting in the hardware lock-in amplifier whenever necessary (11). Automatic phase setting has been achieved with a software lock in amplifier (PULSE program; HEKA, Lambrecht, Germany) that uses Lindau–Neher's technique, making Lindau–Neher's technique a more popular technique for capacitance measurements (11). However, the PULSE program has not been used to analyze the fission G_p. The equations derived in the present work made it possible to study fusion and fission G_p using the commercially available PULSE program.

Experimental Procedures

Methods for preparing brainstem slices from 7- to 10-day-old Wistar rats and measurements of presynaptic Ca²⁺ currents and capacitance in the medial nucleus of the trapezoid body have been described (14, 24). Recordings were made in a bath solution that pharmacologically isolated Ca²⁺ currents (25). This solution contained 105 mM NaCl, 20 mM tetraethylammonium·Cl, 2.5 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 25 mM NaHCO₃, 1.25 mM NaH₂PO₄, 25 mM dextrose, 0.4 mM ascorbic acid, 3 mM myo-inositol, 2 mM sodium pyruvate, 0.001 mM tetrodotoxin, and 0.1 mM 3,4-diaminopyridine (pH 7.4) when bubbled with 95% O₂ and 5% CO₂. The presynaptic pipette (2.5- to 4.5-MΩ) solution contained 125 mM Cs-gluconate, 20 mM CsCl, 4 mM MgATP, 10 mM Na₂-phosphocreatine, 0.3 mM GTP, 10 mM Hepes, 0.05 mM 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetate or -tetraacetic acid, pH 7.2, adjusted with CsOH. The holding potential was −80 mV. Data were expressed as mean ± SEM. Statistical tests were t or Student's t test.

We measured capacitance using the EPC-9 amplifier together with the PULSE program (HEKA) that applies Lindau–Neher's technique (11). The sinusoidal stimulus was 1,000 Hz with a peak-to-peak voltage of 60 mV. We previously selected calyces showing a single exponential decay of their passive current transients (14, 24). In this study, however, we did not perform any preselection, because the capacitance jump and endocytosis rate are similar for calyces showing a mono- or biexponential decay in their passive current transients (26, 27). The methods for detection and measurement of the DCS are described in SI Text I, section 4. For averaging (Fig. 1 C and E), DCSs were first low-pass-filtered at 30 Hz and differentiated (Fig. 1A). The peak of the differentiated trace was then used to align and thus to average DCSs.

Abbreviations

DCS

downward capacitance shift

D_p

pore diameter

G_p

pore conductance

AP-e

action potential-equivalent stimulus

G_m

membrane conductance

G_s

series conductance

BoNT/C

botulinum neurotoxin C

R_a

access resistor

Rm_c

membrane resistance

Cm_c

capacitance of the calyx

R_p

fission pore resistance.