Abstract
Whole-terminal capacitance measurements were used to examine membrane retrieval that follows Ca2+-triggered exocytosis in single synaptic terminals. Exocytosis was followed by endocytosis only when the internal solution contained a hydrolyzable analog of ATP. ATP-γ-S, a poorly hydrolyzable ATP analog, did not support endocytosis but instead produced a rapid and profound inhibition of membrane retrieval. Under similar conditions, the GTP analogs GTP-γ-S and GDP-β-S failed to block endocytosis, suggesting that ATP is the preferred substrate. Furthermore, the requirement for ATP was independent of the role of ATP in regulating intraterminal Ca2+, and the role of Ca2+ in endocytosis was different from that of ATP. The results suggest a direct, acute requirement for ATP hydrolysis in compensatory fast endocytosis in synaptic terminals. Given that the capacitance technique detects changes in membrane surface area, ATP must be required for the membrane fission step or at a step that is a prerequisite for membrane fission.
Three distinct patterns of endocytosis have been identified in secretory cells when capacitance measurements are used to monitor membrane retrieval. The first of these, “compensatory endocytosis,” is observed immediately after Ca2+-dependent exocytosis and serves to restore the membrane capacitance precisely back to baseline. Compensatory endocytosis typically has a time constant of a few seconds, although after strong stimuli or elevations in basal Ca2+, the recovery time can be quite prolonged (von Gersdorff and Matthews, 1994a,b). This form of endocytosis has been reported in both synaptic terminals (von Gersdorff and Matthews, 1994a,b) and neuroendocrine cells (Smith and Neher, 1997;Engisch and Nowycky, 1998). In neuroendocrine cells and neuronal somata, but not synaptic terminals, Ca2+-triggered endocytosis may also lead to a large undershoot of the baseline capacitance (Thomas et al., 1994;Smith and Neher, 1997; Engisch and Nowycky, 1998; Heidelberger, 1998). This “excess retrieval” is thought to be related to cellular housekeeping functions rather than neurotransmitter release (Smith and Neher, 1997; Engisch and Nowycky, 1998). A third pattern of endocytosis, “rapid endocytosis,” has been observed after an exocytotic burst in adrenal chromaffin cells (Artalejo et al., 1995). A similar rapid retrieval of membrane has been observed only at extremely high internal Ca2+ in synaptic terminals (Heidelberger, 1998), and its significance to synaptic function is unclear. Rapid endocytosis in chromaffin cells involves the GTPase dynamin and tyrosine phosphorylation but not clathrin (Artalejo et al., 1995; Nucifora and Fox, 1999). In contrast, very little is specifically known about the underlying molecular mechanism of compensatory endocytosis, particularly at synapses. Because compensatory endocytosis is coupled both temporally and in magnitude to the preceding exocytotic response and is the predominant mechanism in synaptic terminals, it is this form of endocytosis that is likely to play a significant role in synaptic function.
Previous morphological studies have suggested that metabolic energy is required for the replenishment of synaptic vesicles at active zones (Atwood et al., 1972; Schaeffer and Raviola, 1978). ATP has also been shown to be required at multiple reactions in receptor-mediated endocytosis, including at the very early steps of coated pit formation (Smythe et al., 1989; Schmid and Smythe, 1991) and at the membrane fission step that forms a coated vesicle from a coated pit (Smythe et al., 1989; Schmid and Carter, 1990; Schmid and Smythe, 1991). However, because more than one pathway of membrane retrieval exists in nerve terminals (Miller and Heuser, 1984; Koenig and Ikeda, 1996), whether ATP was specifically required at an early step in fast compensatory endocytosis remained uncertain. To address this question, changes in membrane surface area associated with synaptic vesicle fusion and retrieval were monitored using time-resolved capacitance measurements in single synaptic terminals of retinal bipolar neurons, along with measurements of intraterminal Ca2+. The validity of the capacitance approach for these glutamatergic neurons is well established (von Gersdorff et al., 1998). The results demonstrate that, not only is ATP specifically required in the fast compensatory endocytic pathway, but that it is necessary for endocytosis to be initiated.
MATERIALS AND METHODS
Preparation of synaptic terminals. Synaptic terminals of Mb1 bipolar neurons were prepared acutely from dark-adapted goldfish retina using a combination of enzymatic and mechanical treatment (Heidelberger and Matthews, 1992). Isolated synaptic terminals were identified on the basis of their characteristic appearance, size (terminals of 8–12 μm in diameter), and electrophysiological profile (Kaneko and Tachibana, 1985; Heidelberger and Matthews, 1992). All experiments were performed at room temperature (21–24°C).
Solutions. External bathing solution for all experiments contained (in mm): 115 NaCl, 2.6 KCl, 1.6 MgCl2, 1.0 CaCl2, 10 HEPES, and 11 glucose, pH 7.3 (255–260 mOsm). Standard internal recording solution contained (in mm): 85–100 Cs-gluconate, 10 tetraethylammonium (TEA)-Cl, 1–2 MgCl2, 33–65 mm Cs-HEPES, 0.5 GTP, 2 mm MgATP, 0.5 EGTA, and 0.2 fura-2. In some experiments Cs-gluconate was replaced by Cs-glutamate and TEA-Cl was reduced to 5 mm. For all internal solutions, free Mg2+ was at least 1 mm. Internal solutions in which free Ca2+ was defined were made by replacing 0.5 EGTA in the standard solution with a combination of 5 mm EGTA and 2.5 mmCaCl2 to give a free Ca2+ concentration of ∼150 nm. For ATP-γ-S solutions, ATP-γ-S was substituted for ATP and 1 mm additional MgCl2 was added to give 3 mm MgCl2. GTP-γ-S solutions were based on the standard solution and contained 3.2 mm GTP-γ-S, a total of either 3 or 10 mm MgCl2, and a reduced ATP concentration of 1 mm. For GDP-β-S solutions, GDP-β-S replaced GTP in the standard internal solution. The pH and osmolarity of the final internal solutions were adjusted to 7.25 and 265 mOsm, respectively. Note that ATP-γ-S was used rather than omitting ATP because, like other L-type calcium channels, the bipolar neuron calcium channels require ATP to maintain their activity.
The time course of loading of terminals with internal solution was monitored by following the increase in fura-2 fluorescence. Calculations suggest that loading with nucleotides should be slightly faster than fura-2 (Pusch and Neher, 1988), which is complete within ∼50 sec after break-in in synaptic terminals. Therefore, a minimum of 50 sec was allowed for dialysis before the start of an experiment. The nucleotide content of solutions with high ATP and GTP-γ-S were verified by HPLC using a spheroclone column from Phenomenex (Torrance, CA). Cs+ salts were prepared from CsOH purchased from Aldrich (Steinheim, Germany and Milwaulkee, WI). Fura-2 was obtained from Molecular Probes (Eugene, OR). ATP-γ-S and GTP were obtained from Boehringer Mannheim (Mannheim, Germany and Indianapolis, IN). GTP-γ-S was obtained from Calbiochem (La Jolla, CA) and Boehringer Mannheim (Indianapolis, IN). GDP-β-S was obtained from Calbiochem. All other reagents were from Sigma (Deisenhofen, Germany and St. Louis, MO). Internal solutions were kept on ice and protected from light when not stored in the dark at −20°C.
Electrical measurements. Conventional whole-cell recordings were performed on synaptic terminals using Sylgard-coated pipettes with resistances of 7–10 MΩ. Because hydrostatic pressure can influence the time course of endocytosis (R. Heidelberger and G. Matthews, unpublished observations), the hydrostatic pressure in the pipette was adjusted to a constant value with either a feedback-controlled device from Lorenz (Katlenburg-Lindau, Germany) or manually with a micrometer-controlled syringe and a monometer. In addition, the angle of the pipette was held constant, and the pipettes were backfilled with a fixed volume of fluid.
Electrical recordings were made using the computer-controlled EPC-9 patch-clamp amplifier from Heka Elektronik (Lambrecht, Germany). Capacitance measurements were performed using the software emulation of a two-phase lock-in amplifier that is provided as part of the EPC-9 Pulse software (Heka Elektronik) or with the automatic capacitance compensation of the EPC-9 amplifier (Chow et al., 1992; Heidelberger et al., 1994). With the software lock-in amplifier, a 1600 Hz sinusoidal stimulus, ≤32 mV peak-to-peak, was applied about the DC holding potential. The resulting current was processed using the Lindau–Neher technique (Lindau and Neher, 1988; Gillis, 1995) to give estimates of the equivalent circuit parameters (Cm,Gs, andGm). The reversal potential of the measured DC current was assumed to be 0. For all experiments, the holding potential, Vh, was −60 mV and all depolarizations were to 0 mV. Terminals with leak current greater than −40 pA were excluded from analysis.
Calcium measurements. For measurement of the average intraterminal Ca2+, alternating excitation at 345 and 390 nm was typically provided by a computer-controlled monochrometer-based system from ASI/T.I.L.L. Photonics (Eugene, OR) as described previously (Messler et al., 1996). In some experiments, alternating excitation was provided by a two-flash lamp system from T.I.L.L. Photonics (Graefeling, Germany). Emitted fluorescence was collected from an ∼20-μm-diameter spot in the object plane through a 470 nm long-pass and a 540 nm short-pass filter and, for the monochrometer-based system, detected by a photomultiplier tube (model R928; Hamamatsu, Hamamatsu City, Japan) or photodiode (model S4753–02; Hamamatsu). This fluorescence signal was sampled by the EPC-9 and acquired using the Fura extension of the Pulse software. For the two-flash lamp system, both excitation and emitted light were detected by photodiodes, simultaneously integrated, and digitized (Heinemann et al., 1994). Fura-2 calibrations were performed in terminals using Ca2+–EGTA-buffered solutions (Heidelberger and Matthews, 1992). Because of potential complications of changes in intraterminal Ca2+ altering the rate of endocytosis (von Gersdorff and Matthews, 1994), terminals with resting Ca2+ >300 nm were excluded from analysis.
Analysis. To control for differences in internal solutions and variability in cell preparations, data were either collected in a pairwise manner or compared with control data obtained under identical experimental conditions. Data were exported into Igor from WaveMetrics Inc. (Lake Oswego, OR) for analyses. More than 75 terminals that had Ca2+ and capacitance responses were examined for this study; however, many terminals were excluded from analyses as a result of the strict criteria of a resting Ca2+ <300 nm and a leak current less than −40 pA. Terminals were also excluded if the first stimulation was given earlier than 50 sec after break-in. ATP hydrolysis is required for refilling of the release-ready pool of synaptic vesicles (Heidelberger, 1998) (R. Heidelberger, P. Sterling, and G. Matthews, unpublished observations); therefore, in experiments comparing the effects of ATP-γ-S with ATP on endocytosis, only the first endocytotic response was analyzed. For all experiments,n represents the number of terminals examined rather than the total number of times the event was witnessed. All pooled data are expressed as mean ± SEM. Comparisons between test groups and controls were analyzed with a two-sample t test or a pairedt test using SAS software (SAS Institute, Cary, NC). Statistically significant differences are indicated in the text and figure legends.
RESULTS
ATP is acutely required for fast endocytosis
Observations from a previous study in which exocytosis was evoked by flash-photolysis of caged Ca2+ provided hints that, in addition to its role in maintaining exocytosis, ATP might be required for fast compensatory endocytosis in synaptic terminals (Heidelberger, 1998). However, this interpretation was tenuous because of the global elevation of Ca2+ and the reported inhibitory effects of high cytosolic Ca2+ on endocytosis (von Gersdorff and Matthews, 1994; Rouze and Schwartz, 1998; Neves and Lagnado, 1999). To address the role of ATP under more controlled conditions, acutely isolated synaptic terminals of retinal bipolar neurons were voltage clamped, and exocytosis followed by compensatory synaptic endocytosis was triggered by stepping the membrane potential from −60 to 0 mV. This protocol activates Ca2+ influx through presynaptic L-type Ca2+ channels (Heidelberger and Matthews, 1992) to trigger the release of the neurotransmitter glutamate (Tachibana et al., 1993; von Gersdorff et al., 1998). Changes in surface area associated with membrane addition and retrieval were monitored using time-resolved membrane capacitance measurements, and intraterminal Ca2+ was ratiometrically calculated with the fluorescent Ca2+indicator dye fura-2 (Heidelberger and Matthews, 1992). As expected (von Gersdorff and Matthews, 1994a), after a 500 msec depolarization, terminals dialyzed with ATP-containing internal solution exhibited exocytosis followed by fast compensatory endocytosis (Fig.1A). In contrast, terminals dialyzed with internal solution containing ATP-γ-S secreted in response to a first stimulus, but they failed to rapidly retrieve membrane (Fig. 1B). The mean time constant of endocytosis was more than four times longer in terminals with ATP-γ-S than in ATP terminals (ATP-γ-S, τ = 26 ± 8 sec,n = 5; ATP, τ = 5.5 ± 1 sec,n = 6), indicating that ATP may be required for fast compensatory endocytosis. This inhibitory action of ATP-γ-S was observed <60 sec after achieving the whole-cell recording configuration, suggesting that ATP is acutely required for compensatory endocytosis. However, internal Ca2+ failed to fully return to resting levels after stimulation without ATP. Even when 1 mm EGTA was added to the ATP-γ-S solution to increase the Ca2+ buffering capacity, the mean resting poststimulus Ca2+ was ∼203% of the prestimulus resting Ca2+ (n = 5) compared with ATP controls, in which the poststimulus Ca2+ returned to within 15% of the basal level (n = 5).
Terminals dialyzed with ATP but not ATP-γ-S exhibit fast compensatory endocytosis. A, The time-resolved capacitance record of an isolated synaptic terminal dialyzed with standard internal solution (2 mm ATP). In response to a 500 msec depolarization from aVh of −60 to 0 mV, the membrane capacitance (Cm) increased by ∼186 fF. The time course of restoration of the membrane capacitance back to baseline could be described by a single exponential with a time constant of ∼2.96 sec. The corresponding membrane conductance (Gm) and series conductance (Gs) are shown in the bottom panels. Lack of correlated changes betweenCm and Gs orGm suggest that changes inCm reflect changes in membrane surface area.B, The time-resolved capacitance record of an isolated synaptic terminal dialyzed with internal solution in which 2 mm ATP was replaced by 2 mm ATP-γ-S. In response to a 500 msec depolarization from aVh of −60 to 0 mV, the membrane capacitance increased by ∼55 fF. Note the slow time course of membrane retrieval.Gm and Gs are shown in the bottom panels. Again, there are no correlated changes in Cm andGm or Gs. For both A and B, timing of the depolarization is given by the arrow. Terminals are from the same trituration.
The requirement for ATP is not mediated by Ca2+
To distinguish between a direct requirement of ATP in fast endocytosis from an indirect requirement mediated by elevation of cytosolic Ca2+, Ca2+ was defined in the internal solutions to be 150 nm, a value close to resting internal Ca2+ in undialyzed synaptic terminals (Heidelberger and Matthews 1992). Figure2 shows representative capacitance responses from a pair of terminals dialyzed with Ca2+–EGTA-buffered internal solutions containing either ATP (Fig. 2A) or ATP-γ-S (Fig.2B). In both terminals, a 500 msec depolarizing voltage step from −60 to 0 mV evoked a transient increase in intraterminal Ca2+ that triggered the addition of membrane. Whereas the terminal with ATP exhibited a rapid return of membrane capacitance to baseline (τ = 3.2 sec), endocytosis was completely blocked in the terminal with ATP-γ-S. The mean time constant of endocytosis in terminals dialyzed with Ca2+-buffered internal solution containing ATP was 2.58 ± 0.32 sec (n = 8) (Fig.3A), in good agreement with values reported previously for the time course of fast endocytosis in synaptic terminals (von Gersdorff and Matthews, 1994a,b). In contrast, the mean time constant of endocytosis was 67.0 ± 30 sec (n = 6) in ATP-γ-S terminals (Fig. 3A), excluding four terminals in which there was no return of the membrane capacitance during the observed time interval (>40 sec after stimulation). The inhibition of fast endocytosis by ATP-γ-S, introduced into the terminals as a Li+salt, could not be attributed to the presence of Li+. At concentrations of up to 20 mm, Li+ did not significantly effect the time course of fast endocytosis (τ = 2.69 ± 0.85 sec, n = 5). In addition, the Li+ salt of another nucleotide, GTP-γ-S, did not block fast endocytosis (Fig.4).
ATP-γ-S does not support rapid endocytosis, even when differences in internal Ca2+ are minimized.A, B, Top panels, The capacitance records from isolated synaptic terminals dialyzed with Ca2+–EGTA-buffered internal solutions containing either ATP (A) or ATP-γ-S (B). Exocytosis was evoked by a 500 msec depolarization from −60 to 0 mV. Fast endocytosis followed exocytosis with ATP but not ATP-γ-S. Timing of the depolarizations are indicated by the arrows. Bottom panels, The corresponding Ca2+ records, ratiometrically calculated from changes in fura-2 fluorescence.
The ATP requirement for endocytosis is independent of intraterminal Ca2+. A, Exocytosis followed by endocytosis was evoked by a 500 msec depolarization from −60 to 0 mV. The mean endocytosis time constant is shown for terminals dialyzed with Ca2+–EGTA-buffered internal solutions containing either ATP (n = 8; black bar) or ATP-γ-S (n = 6; gray bar). Four terminals dialyzed with ATP-γ-S-containing internal solution failed to endocytose in the >40 sec after exocytosis and are not included in this figure. B, There are no statistical differences in the mean Ca2+parameters between terminals with ATP versus ATP-γ-S when Ca2+-buffered internal solutions are used. Same group of terminals as in A. Data are expressed as mean ± SEM.
GTP-γ-S does not block endocytosis, but it does decrease the amplitude of exocytosis. A, Capacitance record from the synaptic terminal of a bipolar neuron that was dialyzed with internal solution containing 3.2 mm GTP-γ-S and 2 mm Mg2+. Two 2 sec depolarizations from −60 to 0 mV were given at the times indicated by thearrows to evoke exocytosis. Note the reproducibility of the response. The time constant of endocytosis was ∼4 sec.B, Pooled data showing the average time constant of endocytosis from five cells (9 responses) dialyzed with GTP-γ-S compared with ATP terminals (2 or 10 mm;n = 8). In GTP-γ-S cells, free Mg2+ was either 2 mm(n = 3) or 10 mm (n= 2). Voltage steps were from −60 to 0 mV and were 0.25–1 sec in duration. Data expressed as mean ± SEM. C, The amplitude of the capacitance response to a 500 msec depolarization was diminished in terminals with GTP-γ-S (n = 5;light gray bar) relative to terminals with Ca–EGTA-buffered internal solutions containing either ATP (n = 13; black bar) or ATP-γ-S (n = 15; dark gray bar) or literature estimates (dotted line) of the average size of the release-ready pool of synaptic vesicles in bipolar neuron synaptic terminals (Heidelberger, 2001). Increasing the duration of Ca2+ influx to 2 sec increased the amplitude of the capacitance response in GTP-γ-S terminals (n = 6 terminals, 12 responses).
The Ca2+ records in Figure 2 indicate that intraterminal Ca2+ returned to baseline after closure of Ca2+ channels with either ATP or ATP-γ-S when using a Ca2+-buffered internal solution. Analysis of the Ca2+ records indicated that there were no significant differences in the resting Ca2+ before and after a stimulus between terminals with ATP or ATP-γ-S (Fig. 3B). Furthermore, there was no statistically significant difference in the peak Ca2+ after membrane depolarization or in the time course or recovery of Ca2+ to basal levels (Fig. 3B). Thus, intraterminal Ca2+ was well controlled in these experiments and comparable between groups. In addition, little rundown of ICa was observed with successive depolarizations in terminals with either ATP or ATP-γ-S, consistent with the suggestion that Ca2+ did not remain high under the membrane (data not shown). Therefore, the inability of terminals with ATP-γ-S to rapidly retrieve membrane after exocytosis cannot be ascribed to differences in Ca2+ between terminals with ATP or ATP-γ-S, indicating that the role of ATP in fast endocytosis is independent of its role in regulating intraterminal Ca2+.
Neither GTP-γ-S nor GDP-β-S inhibit fast endocytosis
To determine whether the observed requirement for ATP was direct or mediated via the intraterminal conversion of ATP-γ-S to GTP-γ-S, terminals were dialyzed with internal solution in which (1) GTP was omitted and the nonhydrolyzable GTP analog GTP-γ-S was added at a concentration of 3.2 mm, (2) the ATP concentration was reduced from 2 to 1 mm, and (3) free Mg2+ concentration was either 2 or 10 mm to facilitate nucleotide exchange. Figure4A shows a representative recording. Each exocytotic response was followed by the rapid retrieval of membrane and the restoration of membrane capacitance to baseline. That terminals can exocytose in the presence of GTP-γ-S is not surprising given that the inhibition of exocytosis by GTP-γ-S is a very slow process (Hess et al., 1993; Artalejo et al., 1995). Interestingly, the mean time constant of endocytosis in GTP-γ-S terminals (τ = 3.19 ± 0.72 sec, n = 3 terminals, 7 responses) was virtually identical to controls (τ = 3.20 ± 0.56 sec,n = 8) for depolarizations <1 sec in duration (Fig.4B) and quite distinct from the prolonged time course of endocytosis observed with ATP-γ-S (Figs. 1-3). Longer depolarizations, which give rise to slower rates of endocytosis (von Gersdorff and Matthews, 1994a,b), failed to consistently reveal an inhibitory effect of GTP-γ-S on endocytosis. Of the six terminals with GTP-γ-S that met the criteria for analysis, only one did not exhibit compensatory endocytosis, and there was no statistically significant difference between the mean time constant for endocytosis after a 2 sec depolarization in terminals with GTP-γ-S compared with controls (GTP-γ-S, τ = 13.66 ± 4.65 sec,n = 6 terminals, 12 responses; controls, τ = 5.46 ± 1.93 sec, n = 5 terminals, 8 responses;p = 0.13). GTP-γ-S did appear to have a suppressive effect on the mean amplitude of the membrane addition triggered by a 500 msec depolarization (Fig. 4C). This could be overcome by increasing the duration of the membrane depolarization (Fig.4C), consistent with a smaller average peakICa in GTP-γ-S terminals (GTP-γ-S, −66 ± 5 pA, n = 7; GTP, −173 ± 18 pA,n = 12). GTP-γ-S has been reported to decreaseICa in adrenal chromaffin cells (Artalejo et al., 1995) and in neurons (Scott et al., 1991). G-proteins may also inhibit exocytosis downstream of Ca2+ entry (Blackmer et al., 2001).
Next, the nonhydrolyzable guanosine diphosphate GDP-β-S was substituted for GTP in the internal solution. GDP-β-S also failed to block fast endocytosis (Fig.5A). The mean time constant of endocytosis in terminals dialyzed with GDP-β-S was 3.06 ± 0.86 sec (n = 4) (Fig. 5B). Longer depolarizations (2–5 sec in duration) also failed to block fast endocytosis (τ = 4.96 ± 0.45 sec, n = 5). As with GTP-γ-S, the average amplitude of the capacitance response with GDP-β-S after a 500 msec depolarization was smaller (58 ± 18 fF, n = 4) than the average size of the response with ATP (Fig. 4C), and ICawas smaller, on average, in amplitude (−49 ± 7 pA,n = 4). The lack of a dramatic, quick inhibition of endocytosis by either GTP-γ-S or GDP-β-S suggests that the rapid inhibition of fast endocytosis by ATP-γ-S reflects an acute and specific requirement for ATP over GTP.
GDP-β-S does not inhibit endocytosis.A, Capacitance record from an isolated synaptic terminal that was dialyzed with an internal solution containing GDP-β-S in place of GTP. Exocytosis was evoked by a 500 msec depolarization from −60 to 0 mV. Timing of depolarization is indicated by thearrow. The time constant of membrane recovery was 3.5 sec. B, Pooled data show that GDP-β-S terminals (n = 4; gray bar) endocytosed with a nearly identical time course as ATP terminals (2 or 10 mm;n = 8; black bar) after 500 msec depolarizations (GDP-β-S) or 0.250–1 sec depolarizations (ATP).
Depletion of ATP does not mediate the reported calcium-dependent inhibition of fast endocytosis
Endocytosis is Ca2+ sensitive in nerve terminals, with higher internal Ca2+concentrations slowing the rate of membrane retrieval (von Gersdorff and Matthews, 1994; Hsu and Jackson, 1996; Neves and Lagnado, 1999). However, reports from other cells have indicated that elevated internal Ca2+ may stimulate endocytosis (Engisch and Nowycky, 1998; Klingauf et al., 1998; Sankaranarayanan and Ryan, 2001). Because intraterminal Ca2+ is regulated primarily by plasma membrane Ca2+ ATPases (Zenisek and Matthews, 2000), it is conceivable that depletion of ATP attributable to high Ca2+ might underlie the reported Ca2+-dependent inhibition of endocytosis in synaptic terminals. To test this hypothesis, the relationship between ATP, Ca2+, and the rate of compensatory membrane retrieval was examined. First, ATP in the internal solution was increased from 2 to 10 mm (Fig.6). After membrane depolarization, terminals with high ATP exhibited fast endocytosis after Ca2+-triggered exocytosis with time constants indistinguishable from controls (2 mm ATP, τ = 2.85 ± 0.70 sec, n = 18; 10 mm ATP, τ = 3.2 ± 0.58 sec,n = 4 terminals, 8 responses), indicating that 2 mm ATP is sufficient to support fast endocytosis in response to moderate stimulation. When the duration of depolarization was increased to 2 sec, the time constant of endocytosis increased as expected (von Gersdorff and Matthews, 1994a,b), but the mean time constants of endocytosis were not statistically different between the two conditions (10 mm ATP, 4.82 ± 0.87 sec, n = 3 terminals, 6 responses; 2 mm ATP, 5.46 ± 1.93 sec, n= 4 terminals, 8 responses). These results suggest that depletion of cytosolic ATP does not underlie the prolongation of the time course of compensatory endocytosis observed with a long depolarization.
The time course of Ca2+recovery predicts the time course of endocytosis. A,B, Top panels show the capacitance records from synaptic terminals of bipolar neurons dialyzed with internal solution containing 10 mm ATP and stimulated to exocytose with either a 500 msec (A) or 1 sec (B) depolarization from −60 to 0 mV.Bottom panels, The ratiometrically calculated intraterminal Ca2+ determined with fura-2.Insets show the corresponding Ca2+current. The thickness of the trace that precedes and follows the inward current carried by Ca2+ is attributable to the superimposed sine wave stimulus used to monitor membrane capacitance. Curves in the top panels are the time course of membrane recovery predicted from the recovery time course of intraterminal Ca2+. Curves were drawn according to the following:Cm(t) =a + be−αt, where α(t) = 0.54(1 − 1/(1 + (0.460/[Ca2+])4)), from von Gersdorff and Matthews (1994b). C, Black bars show the mean rate of endocytosis experimentally observed in terminals with 10 mm ATP after depolarizations of the indicated durations from −60 to 0 mV. Gray bars show the rate of endocytosis predicted from the recovery time course of intraterminal Ca2+. For mild to moderate duration depolarizations, the time course of endocytosis matched the time course of endocytosis predicted by the time course of recovery of intraterminal Ca2+ in terminals with 10 ATP. However, after a 2 sec depolarization, there was a statistically significant difference between the measured and the predicted rates of endocytosis (p < 0.005). After a 5 sec depolarization, the measured and predicted rates were also significantly different (p < 0.07). For ≤500 msec,n = 4 responses, 2 terminals. For 1 sec,n = 2 responses, 1 terminal. For 2 sec,n = 4 responses, 3 terminals. For 5 sec,n = 5 responses, 3 terminals. Data are expressed as mean ± SEM.
The question of whether the reported relationship between Ca2+ and the time course of endocytosis (von Gersdorff and Matthews, 1994b) still held when ATP was high was addressed next. If this relationship did not hold, this would also lend support to the hypothesis that inhibition of endocytosis by elevated intraterminal Ca2+ was not directly mediated by Ca2+ but rather by ATP depletion. After the closure of voltage-gated Ca2+ channels, intraterminal Ca2+ typically recovered with a time course described by a single exponential function (Fig. 6). The predicted time course of endocytosis was calculated from this using the published relationship between intraterminal Ca2+ and the rate of membrane retrieval (von Gersdorff and Matthews, 1994b). The time course of endocytosis predicted by the intraterminal Ca2+ (Fig.6A, curve) is superimposed on the record of membrane capacitance (Fig. 6A, dots). There is good agreement between the time course of endocytosis that was predicted by recovery of internal Ca2+ and the observed time course of endocytosis. Figure 6Bshows a similar analysis of a 1 sec depolarization. As has been noted previously for long depolarizations (von Gersdorff and Matthews, 1994;Neves and Lagnado, 1999), the capacitance record exhibited a brief delay between the end of membrane addition and the start of endocytosis, and this delay was mimicked by the time course of endocytosis predicted from the time course of Ca2+ recovery (Fig. 5B).
To better compare the predicted time course of endocytosis with the observed time course, both time courses after the delays were fitted with single exponential functions, and the resultant rate constants were compared (Fig. 6C). For the terminals stimulated with a 250–500 msec depolarization, the predicted rate constant of endocytosis and the observed rate constant for endocytosis were indistinguishable (predicted rate constant, 0.576 ± 0.044 sec−1; observed rate constant, 0.562 ± 0.080 sec−1; n = 4). Similarly, the predicted rate constant of endocytosis and the observed rate constant of endocytosis evoked by 1 sec depolarizations were in good agreement (predicted, 0.694 ± 0.02 sec−1; observed, 0.625 ± 0.165 sec−1; n = 2). These findings suggest that, for mild to moderate stimulation protocols, the rate of fast endocytosis does follow a fourth-order dependence on the average cytosolic Ca2+ (von Gersdorff and Matthews, 1994), provided that the requirement for ATP is met. For depolarizations 2–5 sec in duration, the observed time course of endocytosis was significantly slower than that predicted by the rate of recovery of intraterminal calcium (observed rate constant, 0.254 ± 0.068 sec−1; predicted rate constant, 0.613 ± 0.039 sec−1;n = 10), consistent with previous reports (von Gersdorff and Matthews, 1994). Although the difference between the observed rate of endocytosis and the rate predicted by the cytosolic calcium achieved statistical significance only for depolarizations ≥2 sec in duration, there was a trend for the disparity between the observed rate and the predicted rate of endocytosis to increase with increasing stimulus duration. This is consistent with an activity-dependent suppression in the apparent rate of fast endocytosis. The mechanism(s) underlying this apparent slowing is unknown, but the present experiments with high intraterminal ATP suggest that local depletion of ATP is not likely to be the major mechanism.
DISCUSSION
Although there have been indications that ATP may be required for membrane retrieval, the present study is the first to specifically examine the role of ATP in fast compensatory endocytosis in a synaptic terminal. The data indicate that ATP, apart from its role in regulating intraterminal Ca2+, is required for fast compensatory endocytosis that follows glutamate exocytosis in retinal bipolar neurons. Interestingly, the inhibition of endocytosis by ATP-γ-S was evident after the very first round of exocytosis and as early as 1 min after achieving the whole-cell recording configuration. This suggests that fast compensatory endocytosis has an acute and dynamic requirement for ATP.
The inability of GTP-γ-S and GDP-β-S to affect fast compensatory endocytosis in synaptic terminals indicates that hydrolysis of ATP, rather than GTP, may be acutely required. This is an important observation because it implies that the mechanism of fast compensatory endocytosis is different from that of the rapid endocytosis observed in calf adrenal chromaffin cells after an exocytotic burst, which is inhibited by both GTP-γ-S and GDP-β-S (Artalejo et al., 1995;Nucifora and Fox, 1999). Furthermore, these results suggest that dynamin, a GTPase with a well documented role in membrane fission in receptor-mediated endocytosis (Cremona and De Camilli, 1997; Schmid et al., 1998), may play little role in compensatory synaptic endocytosis, although this remains to be directly tested. It is also conceivable that a large Ca2+ signal, such as that produced by a combination of flash-photolysis of caged calcium and membrane depolarization (Heidelberger, 1998), may reveal a form of rapid endocytosis that is sensitive to GTP in synaptic terminals. However, the present study focused on the regulation of compensatory endocytosis that follows exocytosis triggered by Ca2+ influx through voltage-gated channels because of the physiological relevance of this type of stimulation. As such, the data provide a clear indication that ATP, but not GTP, is acutely required for the fast compensatory endocytosis that follows exocytosis triggered by Ca2+ influx in synaptic terminals.
Provided that at a minimum of 1 mm ATP was supplied internally, the rate of synaptic endocytosis was independent of ATP concentration and followed the previously described fourth-order relationship between the rate of synaptic endocytosis and intraterminal Ca2+ quite well (von Gersdorff and Matthews, 1994). This is an important conformation because, in addition to reports that high intraterminal Ca2+can suppress endocytosis in nerve terminals (von Gersdorff and Matthews, 1994; Hsu and Jackson, 1996; Rouze and Schwartz, 1998; Neves and Lagnado, 1999; Cousin and Robinson, 2000), intracellular Ca2+ has also been suggested to favor fast endocytosis (Engisch and Nowycky, 1998; Klingauf et al., 1998;Sankaranarayanan and Ryan, 2001) or have little effect on membrane retrieval (Ramaswami et al., 1994; Wu and Betz, 1996). Interestingly, the discord between the rate of endocytosis predicted by the Ca2+ time course and the experimentally observed rate after strong stimulation was not ameliorated by high ATP. This suggests that local depletion of ATP, in addition to not mediating the Ca2+ dependence of endocytosis under standard conditions, does not underlie the profound slowing of the time course of endocytosis that is observed after strong stimulation protocols (von Gersdorff and Matthews, 1994a,b; Neves and Lagnado, 1999; present study). Together, these observations establish that ATP plays a different role in regulating compensatory endocytosis in synaptic terminals than Ca2+.
So where might the ATP-dependent step in compensatory synaptic endocytosis lie? The capacitance approach, because it monitors membrane addition and retrieval, indicates that ATP is required for membrane fission. Therefore, the observed acute requirement for ATP in compensatory synaptic endocytosis must reflect either a requirement at the membrane fission step (Fig.7A) or a step that is absolutely a prerequisite for and temporally associated with membrane fission (Fig. 7B). Precedence for the first role comes from receptor-mediated endocytosis, in which ATP has been suggested to be necessary for the creation of a coated vesicle from a coated pit (Smythe et al., 1989; Schmid and Carter, 1990; Schmid and Smythe, 1991). This interaction could be mediated by actin, an ATPase that may act as a force generator during fission (Lamaze et al., 1997; Qualmann et al., 2000). A prefission role for ATP may be subserved by ATPases such as N-ethylmaleimide-sensitive factor (NSF) or Hrs-2, both of which are implicated in early steps of membrane retrieval by virtue of their roles in dissembling and/or sorting soluble NSF attachment protein receptor (SNARE) proteins (Hay and Scheller, 1997; Kawasaki et al., 1998; Bean et al., 2000). In particular, ATP-γ-S-bound NSF is known to inhibit disassembly of the 20S SNARE complex, whereas without ATP, this stable complex is not formed (Hanson et al., 1997). Thus, by locking the 20S complex, ATP-γ-S may uniquely inhibit protein sorting with the consequence that newly added membrane is unable to be retrieved by the fast compensatory pathway. Alternative explanations are also possible. ATP could regulate an early step in compensatory endocytosis via a phosphorylation reaction that is poorly supported by ATP-γ-S. In addition, ATP-γ-S could conceivably inhibit endocytosis by thiophosphorylating a protein important for triggering endocytosis when in its dephosphorylated state. This would be consistent with the observation that the rate of endocytosis is enhanced when protein phosphorylation is inhibited (Henkel and Almers, 1996; Kavalali et al., 1999). Unfortunately, the ATP requirement of the calcium channels in synaptic terminals does not allow this hypothesis to be readily examined. However, in experiments in which intraterminal Ca2+ was elevated via flash-photolysis of caged Ca2+, compensatory endocytosis was lost when ATP was not included in the internal solution (Heidelberger et al., 1994; Heidelberger, 1998), favoring the hypothesis that ATP hydrolysis, rather than dephosphorylation of a protein, may be required for endocytosis. However, as stated previously, the global elevation of cytosolic Ca2+ in photolysis experiments makes a mechanistic interpretation of these data difficult. Clearly, it will be of great interest to determine the underlying mechanism of ATP action in the fast compensatory endocytosis pathway. The present results lead the way by suggesting that, in a glutamatergic synaptic terminal, ATP is involved in a step or steps necessary for membrane fission after the Ca2+-dependent vesicular release of neurotransmitter.
Possible locations of an early requirement for ATP in the fast compensatory endocytosis pathway in synaptic terminals.A, ATP may be required for the membrane fission step of the fast compensatory pathway, perhaps by acting as an energy source.B, ATP may be required for a postfusion step that is prerequisite for membrane retrieval via the fast compensatory pathway but not be needed for fission. This ATP-dependent step might reflect the disassembly and sorting of SNARE complexes. In addition, the possibility that ATP may both play a prefission role and act at the fission step is not excluded.
Footnotes
This work was supported by National Institutes of Health Grant EY12128, the Esther A. and Joseph Klingenstein Fund, and the Alfred P. Sloan Foundation. I thank Erwin Neher and Gary Matthews for their contributions to the early phases of this project. I thank Andrew Bean and Neal Waxham for stimulating discussions and for performing the HPLC analysis of nucleotide solutions (N. Waxham). I thank Alice Chuang and the Core Grant for Vision Research (National Eye Institute Grant EY10608) for providing statistical support and Kate Pearson for her excellent technical assistance.
Correspondence should be addressed to Ruth Heidelberger, Department of Neurobiology and Anatomy, University of Texas Houston Health Science Center, Houston, TX 77025. E-mail: ruth.heidelberger{at}uth.tmc.edu.
