Loose Coupling Between Ca2+ Channels and Release Sensors at a Plastic Hippocampal Synapse

See allHide authors and affiliations

Science  07 Feb 2014:
Vol. 343, Issue 6171, pp. 665-670
DOI: 10.1126/science.1244811

From Channel to Sensor

The coupling between voltage-activated calcium channels and calcium sensors of exocytosis on synaptic vesicles is a key factor that determines the timing and efficiency of transmitter release. It is still largely unclear how tight this coupling is at mature synapses in the central nervous system. Vyleta and Jonas (p. 665) found that mossy fiber boutons contain high concentrations of endogenous calcium buffers that normally limit the amount of calcium that reaches the calcium sensor responsible for neurotransmitter release. As a consequence, the calcium signal available to trigger release is small for the initial action potential. However, after high-frequency stimulation, the endogenous calcium buffer binds calcium and is less able to buffer calcium entry, which allows more calcium to reach the calcium sensor, increasing neurotransmitter release and synaptic facilitation.


The distance between Ca2+ channels and release sensors determines the speed and efficacy of synaptic transmission. Tight “nanodomain” channel-sensor coupling initiates transmitter release at synapses in the mature brain, whereas loose “microdomain” coupling appears restricted to early developmental stages. To probe the coupling configuration at a plastic synapse in the mature central nervous system, we performed paired recordings between mossy fiber terminals and CA3 pyramidal neurons in rat hippocampus. Millimolar concentrations of both the fast Ca2+ chelator BAPTA [1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid] and the slow chelator EGTA efficiently suppressed transmitter release, indicating loose coupling between Ca2+ channels and release sensors. Loose coupling enabled the control of initial release probability by fast endogenous Ca2+ buffers and the generation of facilitation by buffer saturation. Thus, loose coupling provides the molecular framework for presynaptic plasticity.

The coupling between voltage-activated Ca2+ channels and Ca2+ sensors of exocytosis on synaptic vesicles is a key factor that determines the timing and probability of transmitter release (13). Work at the young calyx of Held in the auditory brainstem suggested that coupling is loose, with source-sensor distances of ~100 nm (2, 4). However, recent evidence indicates that, at several excitatory and inhibitory synapses in the mature central nervous system, coupling is substantially tighter than previously thought, with source-sensor distances of only 10 to 20 nm (59). Tight coupling offers several functional advantages, including speed, temporal precision, and energy efficiency of synaptic transmission (3). Does any synapse in the mature brain make use of loose coupling, and, if so, does this have specific consequences for the functional properties of synaptic transmission? To address these questions, we focused on hippocampal mossy fiber synapses on CA3 pyramidal neurons (1014) (Fig. 1A). These synapses express several presynaptic forms of plasticity (15), which may be linked to a loose coupling configuration (16). Furthermore, their presynaptic terminals are accessible to direct patch-clamp recording. This allowed us to quantitatively probe channel-sensor coupling by using Ca2+ chelators, at a level of rigor previously only achieved at the calyx of Held (4, 17, 18).

Fig. 1 Transmitter release from mossy fiber boutons is highly sensitive to the slow Ca2+ chelator EGTA.

(A) (Top) Schematic showing paired whole-cell recording configuration. (Bottom) Infrared–differential interference contrast image of a paired recording between a mossy fiber bouton and a postsynaptic CA3 pyramidal neuron to which the bouton was attached at the apical dendrite. Photomontage of individual images at slightly different focal planes. Images were captured in the cell-attached mode before reaching the whole-cell configuration. (B) Paired whole-cell recording with 0.1 mM EGTA in the presynaptic recording pipette (control). (C) Paired whole-cell recording with 0.2 mM BAPTA. (D) Paired whole-cell recording with 10 mM EGTA. For (B) to (D), the presynaptic action potential is shown on top, and 10 consecutive EPSCs in the postsynaptic CA3 pyramidal neurons are shown superimposed (gray, blue, and red) at the bottom, overlayed with the average (black in all cases; 20-s repetition interval). (E) Concentration-effect relationship of EPSC peak amplitude against presynaptic BAPTA or EGTA concentration for separate paired recordings. (F to H) Mean synaptic delay (F), EPSC 20 to 80% rise time (G), and EPSC decay time constant (H) against presynaptic BAPTA or EGTA concentration. Error bars indicate SEM.

To probe the distance between Ca2+ channels and release sensors, we measured the effects of exogenous Ca2+ chelators with different binding rates (Fig. 1) (3, 19). Paired recordings from mossy fiber terminals and postsynaptic CA3 pyramidal neurons were obtained by using presynaptic solutions with minimal Ca2+ buffer capacity (0.1 mM EGTA, control), or different concentrations of either the fast Ca2+ chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA) or the slow Ca2+ chelator EGTA (12, 10, and 13 pairs total, respectively; Fig. 1, B to D). Mossy fiber terminals were stimulated in the whole-cell current-clamp configuration with brief current pulses, whereas excitatory postsynaptic currents (EPSCs) were recorded in postsynaptic CA3 pyramidal neurons under voltage-clamp conditions. On average, 1 mM BAPTA and 10 mM EGTA reduced EPSC peak amplitudes to 10 ± 7% and 8 ± 2% of the control value (mean ± SEM, 4 and 10 pairs, P = 0.004 and 0.0002, respectively; Fig. 1E). Both Ca2+ chelators had no detectable effects on the synaptic delay (Fig. 1F; P = 0.8 and 0.1 for 1 mM BAPTA and 10 mM EGTA). However, 1 mM BAPTA and 10 mM EGTA significantly reduced the 20 to 80% rise time (to 53 and 61% of control; Fig. 1G; P = 0.0002 and 0.0001) and the decay time constant of the average EPSC (to 42 and 57% of control; Fig. 1H; P = 0.001 and 0.009).

To estimate the physical distance between Ca2+ channels and release sensors, we analyzed the concentration dependence of the BAPTA and EGTA effects (Fig. 2). To minimize intersynapse variability, EPSC amplitudes in the presence of chelators were normalized by the EPSC amplitudes previously measured by using noninvasive stimulation (materials and methods). BAPTA and EGTA suppressed transmitter release in a concentration-dependent manner, with half-maximal inhibitory concentrations of 0.094 ± 0.011 mM and 1.22 ± 0.46 mM, respectively (Fig. 2A) (59). Next, the concentration-effect data for BAPTA and EGTA were fit by a model of Ca2+ buffering and diffusion based on linear approximations (Fig. 2B) (1, 6). The main free parameter was the coupling distance between Ca2+ channels and Ca2+ sensors, whereas several other parameters (e.g., the physicochemical properties of the buffers and the cooperativity of transmitter release) were well constrained (fig. S1). An analysis of the entire data set revealed a coupling distance of ~75 nm (distance of 73 ± 15 nm for a single-channel model; mean distance of 65 nm and cluster width of 80 nm for a multichannel model; Fig. 2, C and D). Variations in resting Ca2+ concentration and buffer product over a wide range gave distance values of 70 to 88 nm (Fig. 2E). Thus, the mean source-sensor distance at hippocampal mossy fiber synapses was markedly longer than at several other central synapses (59). However, the estimated distance was substantially shorter than the distance between neighboring release sites at mossy fiber synapses (400 to 500 nm) (20), suggesting that local rather than global presynaptic Ca2+ elevations drive transmitter release.

Fig. 2 Loose source-sensor coupling at mossy fiber synapses.

(A) Concentration-effect curves for BAPTA and EGTA. EPSC peak amplitudes were normalized by the EPSC amplitude in preceding bouton-attached configuration to reduce intersynapse variability. Data points were fit with a Hill equation (Hill coefficient of 1.77, maximal amplitude of 5.08). Error bars, SEM. (B) Structure of the model of Ca2+ diffusion and buffering based on the analytical steady-state solution to the linearized reaction-diffusion equations. (Left) Single-channel model; (right) multichannel model. Red dots, Ca2+ channels; green dots, synaptic vesicles with Ca2+ sensors. (C) Plot of ratio of EPSC peak amplitude in the presence of chelator to that in control conditions against the concentration of BAPTA (blue squares) and EGTA (red circles). Curves represent the predictions of the models fit to the entire data set (continuous curves, single-channel model; dashed curves, multichannel model). The best fit was obtained with r = 73 nm (single-channel model) or d = 65 nm and σ = 80 nm (multichannel model). Error bars, SEM. (D) Statistical errors. Histogram of estimated coupling distance in 1000 bootstrap replications fit by using the single-channel model. (E) Systematic errors. Plot of coupling distance against resting Ca2+ concentration and endogenous buffer product. Red point indicates default parameter values.

A hallmark property of mossy fiber synapses is the pronounced facilitation of transmitter release (15). However, the mechanisms of facilitation at central synapses remain controversial (2124). We therefore examined the dynamics of transmission and tested the effects of Ca2+ chelators (fig. S2). Presynaptic terminals were stimulated by 50-Hz trains of brief pulses with either control or BAPTA- or EGTA-containing solutions loaded into presynaptic terminals. Mossy fiber EPSCs showed depression in control solutions (fig. S2A) but marked facilitation after loading with exogenous Ca2+ chelators (fig. S2, B and C). The facilitation induced by BAPTA resembled the “pseudofacilitation” in cortical pyramidal neuron–interneuron synapses, although the extent was larger than previously reported (25).

Why do unitary mossy fiber EPSCs under control conditions show large initial amplitude (Fig. 1B) and marked depression (fig. S2A), whereas several previous studies reported low release probability and facilitation (11, 15, 26, 27)? To resolve this apparent contradiction, we stimulated mossy fiber boutons noninvasively in the tight-seal, bouton-attached voltage-clamp configuration with brief voltage pulses (Fig. 3, A to G; fig. S3; and table S1). Suprathreshold stimuli reliably evoked action currents, which consecutively triggered EPSCs in postsynaptic CA3 pyramidal neurons (Fig. 3, B to D). Average unitary EPSCs evoked by bouton-attached stimulation had mean synaptic delays of 1.43 ± 0.07 ms, 20 to 80% rise times of 1.33 ± 0.12 ms, and peak amplitudes of 260 ± 28 pA (27 pairs). Moreover, transmitter release from intact boutons showed a marked facilitation (Fig. 3, E and F). On average, EPSC2/EPSC1 was 3.34 ± 0.52, whereas EPSC3/EPSC1 was 3.19 ± 0.63 (21 pairs). In parallel to the onset of facilitation, the 20 to 80% rise time of average EPSCs increased gradually during train stimulation (Fig. 3G). A comparable degree of facilitation was observed for bouton-attached stimulation at near-physiological temperature (fig. S4 and table S1), consistent with properties of mossy fiber transmission in vivo (28). Thus, noninvasive stimulation allowed us to fully replicate the hallmark properties of mossy fiber synaptic transmission, low release probability and facilitation, at the level of unitary EPSCs (11, 15, 26, 27).

Fig. 3 Endogenous Ca2+ buffers control initial release probability and facilitation in mossy fiber boutons.

(A) Schematic illustration of the tight-seal, bouton-attached voltage-clamp stimulation and recording configuration. (B) Unitary EPSCs evoked by presynaptic stimulation in the bouton-attached configuration. Top traces, voltage steps (50 to 600 mV, 0.1 ms) applied to the presynaptic terminal. Middle traces, currents recorded in the presynaptic terminal. Bottom traces, EPSCs in the CA3 pyramidal neuron. Red traces show successful stimulation. (C and D) Unitary EPSCs evoked by presynaptic 50-Hz bouton-attached stimulation [600 mV, 0.1 ms; same recording as in (B)]. Four consecutive sweeps are shown superimposed (gray), overlayed with the average (black). The action current is the average of four individual sweeps. (E to G) Mean EPSC amplitude (21 pairs), normalized EPSC amplitude (21 pairs), and 20 to 80% rise time (19 to 21 pairs) against stimulus number for 50-Hz bouton-attached stimulation. Data points are connected by lines for clarity; horizontal dashed line in (F) indicates unity. Error bars, SEM. (H) Schematic illustration of the washout experiment. (I) Paired recording in which the mossy fiber bouton was first stimulated in the bouton-attached voltage-clamp configuration (left) and subsequently in the whole-bouton current-clamp configuration (~1 min post break-in, right). EPSCs were recorded in the CA3 pyramidal neuron under whole-cell voltage-clamp conditions. Three and five consecutive sweeps are shown for bouton-attached and whole-bouton stimulation, respectively. (J) Summary of results from experiments like those shown in (I). Dialysis of the nerve terminal with control pipette solution markedly increased peak EPSC1 amplitude (five pairs). Bars indicate mean ± SEM; circles represent data from individual experiments. Lines connect data points from the same experiment. (K) Short-term dynamics of EPSCs with bouton-attached (open circles) and subsequent whole-bouton stimulation (solid circles). Data from the same paired recording as shown in (I). Data points are connected by lines.

Our results demonstrate that exogenous Ca2+ buffers markedly affect both release probability and short-term dynamics at the hippocampal mossy fiber synapse. To examine whether endogenous Ca2+ buffers in presynaptic terminals have similar effects, we compared transmitter release in the noninvasive recording configuration with that after washout of endogenous buffers in the same synapse (Fig. 3, H to K). Dialysis of the bouton with control pipette solution produced a marked increase in the peak amplitude of EPSC1 (from 304 ± 59 to 1470 ± 242 pA; five pairs; P = 0.004; Fig. 3, I and J). Thus, endogenous Ca2+ buffers may control initial release probability. Furthermore, transition from bouton-attached to whole-bouton configuration with control pipette solution led to a switch from facilitation to depression (Fig. 3K). Different mechanisms may contribute to this change. First, the effects of endogenous buffers on facilitation could be indirect, mediated by changes in release probability. Alternatively, the endogenous Ca2+ buffers may have direct effects on facilitation. To distinguish between these possibilities, we adopted a method to correct for the overlaying effects of pool depletion (fig. S5 and materials and methods) (17, 29). The extent of corrected facilitation was markedly higher in the noninvasive recording configuration than after washout of the endogenous buffers (fig. S5B). Thus, endogenous Ca2+ buffers have direct effects on facilitation, independently of their effects on pool depletion.

If the endogenous buffers act directly, as our results suggest, facilitation should be rescued by exogenous Ca2+ buffers. We perfused the presynaptic terminal with different concentrations of EGTA and BAPTA (Fig. 4, A to E). Dialysis of mossy fiber boutons with 1 mM EGTA resulted in a pronounced increase in EPSC1 amplitude (from 456 ± 32 for bouton-attached to 1292 ± 266 pA for whole-bouton stimulation; three pairs; P = 0.11) and reduction of paired-pulse facilitation (EPSC2/EPSC1 = 2.2 ± 0.6 versus 1.0 ± 0.4; three pairs; P = 0.03; fig. S6, B and C). Thus, the properties of endogenous buffers were not rescued. In contrast, dialysis of mossy fiber boutons with submillimolar concentrations of BAPTA preserved EPSC1 amplitude (from 317 ± 127 pA for bouton-attached to 321 ± 138 pA for whole-bouton stimulation with 0.3 mM BAPTA; three pairs; P = 0.97; Fig. 4B). Moreover, BAPTA largely rescued the short-term dynamics phenotype, with facilitation followed by depression during a train of 10 stimuli (Fig. 4F). However, 0.3 mM BAPTA did not exactly mimic the onset of facilitation during repetitive stimulation. This may suggest that either the native buffers have different binding properties and concentration or that additional factors (adenosine triphosphate or guanosine triphosphate) are involved.

Fig. 4 A fast Ca2+ buffer rescues both low release probability and facilitation in experiments and transmitter release models.

(A) EPSCs evoked by bouton-attached stimulation (left, six sweeps) and subsequent whole-bouton stimulation (right, five sweeps) with 0.3 mM BAPTA in the presynaptic recording pipette (50-Hz stimulation). Dialysis of the mossy fiber bouton with BAPTA resulted in EPSC1 peak amplitude similar to that recorded with bouton-attached stimulation and synaptic facilitation followed by depression. (B and C) Summary graphs showing the effects of 0.3 mM BAPTA on EPSC1 amplitude (B) and short-term dynamics (C). (D) Effects of different concentrations of chelators on EPSC1 peak amplitude, compared to release from intact terminals (dashed line). For each experiment, EPSC1 was normalized to the amplitude in preceding bouton-attached measurements. (E) Effects of different concentrations of chelators on facilitation compared with release from intact terminals (dashed lines indicate mean EPSC2/EPSC1 and EPSC3/EPSC1 values from data in Fig. 3F). Red shaded regions in (D) and (E) indicate the BAPTA concentration range that mimics both release probability and synaptic facilitation. (F) Structure of the transmitter release model. A single release site of a mossy fiber bouton was modeled as a spherical cone. Ca2+ channels were placed in a disc with 50-nm diameter on the surface of the sphere (red bar). The Ca2+ sensor was placed on the synaptic vesicle 80 nm from the border of the disk (red line). Ca2+ transients were calculated as the numerical solution to the set of partial differential reaction-diffusion equations. Black circles, endogenous buffer; red triangles, Ca2+ ions; gray hemisphere, Ca2+ domain. (G) Free internal Ca2+ concentration (top) and release rate (bottom) during two action potentials separated by 20-ms intervals, with the release sensor placed at 80 nm from the edge of the source. A fast endogenous Ca2+ buffer (0.3 mM, Ca2+-binding and unbinding rates two times faster than BAPTA) was included to approximate the endogenous situation. Horizontal dashed lines indicate the maximal Ca2+ concentration and release rate during the second action potential. (H) Plot of paired-pulse facilitation (EPSC2/EPSC1) against coupling distance. Red curve, facilitation in the standard model including all facilitation mechanisms; green, reset of Ca2+ sensor; orange, reset of residual Ca2+; blue, reset of Ca2+ buffer at t = 20 ms. Error bars, SEM.

To further constrain both binding properties and concentration of the endogenous buffers, we compared the EPSC peak amplitude and the EPSC2/EPSC1 and EPSC3/EPSC1 ratios before (bouton-attached configuration) and after dialysis (whole-bouton configuration) over a wide range of buffer concentrations (Fig. 4, D and E). For EGTA, high concentrations of 10 mM were required to mimic the EPSC peak amplitude (Fig. 4D, red symbols). However, these concentrations failed to replicate the facilitation of transmitter release from intact boutons (Fig. 4C). In contrast, BAPTA at a concentration of 0.2 to 0.3 mM reproduced both the EPSC peak amplitude and the extent of facilitation (Fig. 4, D and E, blue symbols). These results suggest that the endogenous buffers have fast, BAPTA-like Ca2+-binding properties and are present in mossy fiber terminals at submillimolar concentration (~0.2 to 0.3 mM). On the basis of washout kinetics, we were able to constrain the diffusion coefficient of the endogenous buffers to >10 μm2 s−1 (materials and methods) (30).

To better understand the interrelation between loose channel-sensor coupling and the action of endogenous buffers, we developed a quantitative model of transmitter release, comprising action potential–induced Ca2+ inflow, buffered diffusion, and release sensor activation (Fig. 4, F to H) (6, 31). When a fast Ca2+ buffer was included, the model adequately predicted both low initial release probability and facilitation of transmitter release (Fig. 4G). Several candidate mechanisms of facilitation have been proposed, including residual Ca2+ (21), Ca2+ bound to the Ca2+ sensor, and saturation of endogenous Ca2+ buffers (2325). We tested the contribution to facilitation by resetting each individual factor to its initial value during the time interval between two subsequent action potentials (Fig. 4H). Resetting the occupancy of the Ca2+ sensor or the residual Ca2+ concentration had only minimal effects on facilitation (Fig. 4H, top right and bottom left, respectively). In contrast, resetting the occupancy of the Ca2+ buffer largely abolished facilitation (Fig. 4H, bottom right). Facilitation was minimal for coupling distances <25 nm but was markedly enhanced as the coupling distance was increased (Fig. 4H). Thus, loose coupling was essential for the generation of facilitation via buffer saturation. After action potential activity, the concentration of the free buffer was reduced relatively uniformly (fig. S6A), suggesting that buffer saturation was global rather than local (32).

Facilitation by buffer saturation was observed over a wide range of functional properties of endogenous buffers and structural properties of mossy fiber synapses (figs. S7, B and C, and S8). Simultaneous analysis of the dependence of release probability and facilitation on Ca2+-binding rate (kon) and buffer concentration (cB), followed by comparison with our experimental observations in the bouton-attached configuration, further suggested that the endogenous buffers have a kon > ~2 times faster than BAPTA and a cB of ~100 to 150 μM. Among three naturally expressed Ca2+-binding proteins, calbindin, calretinin, and calmodulin (3336), none mimicked all experimentally observed release properties (figs. S7D and S9). These results are consistent with the observation that genetic elimination of calbindin has only small effects on facilitation (22).

The present results provide a demonstration of loose coupling between Ca2+ channels and release sensors at a mature cortical synapse. Our findings challenge the prevailing view that loose coupling is a purely developmental phenomenon (5) and instead demonstrate that coupling is regulated in a synapse-specific manner. Together with previous observations of tight coupling at fast-signaling synapses (59), our results suggest that loose coupling is preferentially expressed in dynamic and plastic synapses (16). Our results help to understand several previously elusive functional properties of hippocampal mossy fiber synapses. First, they explain the low release probability after single presynaptic action potentials (11, 27) (fig. S10). This low initial release probability has been highly puzzling, given the large size of the terminals, the large number of active zones, and the huge vesicular pool (20, 37, 38). We demonstrate that loose coupling enables access of endogenous Ca2+ buffers acting as “brakes” on release probability. Second, our results shed light on the mechanism of presynaptic facilitation, a hallmark property of mossy fiber synapses (13, 15, 26). Loose coupling permits facilitation via saturation of endogenous Ca2+ buffers (25) (fig. S10). Last, our results have major implications for the mechanisms of long-term potentiation (LTP) at hippocampal mossy fiber synapses. Previous studies have shown that expression of mossy fiber LTP has a presynaptic locus but is not associated with a rise in the amplitude of global presynaptic Ca2+ transients (39). These results could be explained by a change in the coupling distance, a local change in the concentration of endogenous Ca2+ buffers, or a combination of both. This may allow faster onset of regulation of synaptic strength than a mechanism based on changes in the number of Ca2+ channels (40).

The large size of the presynaptic terminals and the proximal location of the synapse previously led to the idea that the mossy fiber synapse acts as an efficient “teacher” synapse in the network, driving the induction of heterosynaptic plasticity in synapses between CA3 cells (41, 42). However, it is also widely accepted that mossy fiber synapses operate as conditional “detonators” (28). A single action potential in a presynaptic granule cell is not sufficient to trigger a spike in a postsynaptic CA3 pyramidal cell but needs to act in combination with presynaptic short-term plasticity. These properties are critical for network function, making the impact of the mossy fiber synapse dependent on burst activity in granule cells (28, 43). Our results reveal that loose channel-sensor coupling and the presence of fast endogenous Ca2+ buffers are the key properties underlying the conditional detonator function of mossy fiber synapses. The switchlike nonlinearity conveyed by this specific design may assist in the separation of storage and retrieval modes in hippocampal memory circuits (41).

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Table S1

References (4459)

References and Notes

  1. Acknowledgments: We thank E. Neher and R. Shigemoto for critically reading the manuscript, F. Marr and M. Duggan for technical assistance, A. Solymosi for manuscript editing, and the scientific service units of IST Austria for efficient help. Supported by the Fond zur Förderung der Wissenschaftlichen Forschung (P 24909-B24) and the European Union (European Research Council Advanced Grant 268548 to P.J.).
View Abstract

Navigate This Article