Dependence of paired-pulse ratio on inter-stimulus interval

A: Experimental protocol. To obtain simple synapse recordings, individual granule cell axons were stimulated with an extracellular pipette located in the granule cell layer, and EPSCs were recorded in a postsynaptic MLI. Presynaptic stimulations involved trains of 4 action potentials with various inter-stimulus intervals (ISIs), and with 10 s inter-train intervals. B: Example recording showing responses to trains using 10 ms ISIs (left) and 1 600 ms ISIs (right). Stimulation times indicated by dotted vertical lines. C, Top: Average EPSC in response to AP #1 (1st trace, labelled ‘average s1’) followed by average responses to AP #2 for ISIs varying from 10 ms to 3200 ms, from the same experiment (20 repetitions for each ISI). C, Bottom: A measure of the paired-pulse ratio, PPR, calculated from the ratio of the mean numbers of released SVs for AP #2 over the average release for AP #1 from the same data. Numbers of SVs released by individual presynaptic stimulations were determined by deconvolution, using the mean quantal EPSC as kernel. D: Plot of PPR as a function of ISI (same analysis as in C, bottom: grey dots, individual cells (means from > 20 repetitions for each cell); blue dots and associated error bars, means ± sem from n = 8 cells). The red curve is a double exponential fit to the data with Afast = 0.94, τfast = 230 ms, Aslow = 0.30, τslow = 2100 ms. Inset: Blow-up of results for ISIs of 200 ms or less. * and ** indicate data points that differ from PPR = 1 with p < 0.05 and p < 0.01 respectively.

LFD reflects a decrease in docking site occupancy.

A: Sequential 2-state docking model. SVs coming from the intermediate pool (IP) transit through a replacement site (RS, with occupancy ρ) and an associated docking site (DS, with occupancy δ) before release. Taken together, one RS and one DS constitute one docking unit (DU). sf, sb, rf and rb represent transition rates as indicated. pr represents the release probability of a docked SV following an AP. B: Experimental protocol. To probe the state of SV pools in the synapse, trains of 8 APs at 100 Hz were applied either in isolation (control train), or after a depressing train of 8 APs at low stimulation frequency (recovery train). C, Top: Plots (means ± sem from n = 7 cells; > 20 repetitions for each cell) of si for the control train (black), for the low frequency train (2 Hz, blue), and for the recovery train (yellow). Dotted line: Average s1 value from control and low frequency trains. C, Bottom: Same, after normalization with respect to s1 value. Exponential fit with y0 = 0.71 ± 0.03 and τ = 0.5 ± 0.4 AP # (= 0.25 ± 0.2 s; red curve). D: Superimposed si plots for control (black) and recovery (yellow). Comparison of the SV release between control and recovery trains allows us to evaluate the occupancy of the RRP and of the IP at the end of the low frequency train. SV numbers in response to the 1st (s1), 2nd (s2), and last (s5-8) APs respectively report changes in δ (red), in ρ (blue), and in the IP size (purple). E: Ratios of si values between recovery 8-AP train and control train. s1 is significantly reduced, indicating a decrease in δ, but neither s2 nor s5-8 are changed, indicating no change in ρ or in IP size.

Simulations of s, δ and ρ curves after an AP stimulation

A, up: Model depiction of an AZ with 3 DUs at various time periods before and after an AP stimulation. After release, SVs transition from RS (blue; some of these SVs are placed sidewise as the 2-step model cannot specify the exact location of the RS) to DS (white) within tens of ms (transient docking), then after hundreds of ms they undock, before eventually returning to their basal state. A, bottom: Simulated time course of δ (red curve) and ρ (blue curve) before and after a presynaptic AP (at time 0). Resting δ and ρ values indicated by dotted lines. During transient docking, δ is high and ρ is low, so that RRP replenishment is allowed (RS gate open, green shade). During undocking, δ is low and ρ is high, so that RRP replenishment is blocked (RS gate closed, red shade). B: PPR curve as a function of ISI (blue dots, from Fig. 1) together with simulation results (red curve).

Simulation of SV movements during HFD and LFD

A, upper panel: si curve during control runs with 8 APs at 100 Hz (black dots: data from Fig. 2; red curve: simulated results; model parameters in Table 1). A, lower panel: simulated values of δ (red) and ρ (blue) observed just before AP stimulations, as a function of AP #. B: Same as A, but during 2 Hz 8-AP stimulations (data from Fig. 2). Note the behavior of δ and ρ compared to A, here ρ remains stable throughout the train, while δ undergoes a sharp initial decrease. At the end of the train there is a low δ value and a high ρ value for 2 Hz trains, while at the end of a 100 Hz 8-AP train (A) there is a relatively high δ and a low ρ. C: Same as in A, but during recovery (100 Hz; data from Fig. 2). The initial δ and ρ values for the recovery train simulation are the end-of-train values from the 2 Hz simulation. D: Proposed SV movements during high-frequency depression (HFD) and LFD. Central sequence: Proposed timeline of SV movements inside a DU following an AP stimulation. HFD loop: Proposed sequence of SV movements during HFD. The DU exits the main timeline at 10 ms after the AP to enter a loop with repetitive high frequency AP stimulation, which results in high rates of exocytosis and RRP replenishment, as the RS gate is open (green shade). Eventually, SV depletion leads to depression. LFD loop: proposed sequence of SV movements during LFD. Here the DU exits the main timeline as the RS gate is closed, and it undergoes an idle docking-undocking cycle at each ISI (blue shade). HF recovery after LFD: proposed sequence of SV movements during high frequency recovery train after LFD (yellow shade). This sequence of events happens if high frequency is applied at any point during an LFD train and transfers the DU to the HFD loop (green shade) after the 2nd AP.

RRP depletion does not cause LFD.

A: Exemplar traces from an LFD experiment, illustrating responses to 1st and 2nd AP (AP times indicated by vertical grey lines) during low-frequency trains. Black traces show EPSCs including a success in response to the 1st AP, while purple traces (marked with a star) show EPSCs when there was a failure in response to the 1st AP. B, left: Group data showing a similar extent of LFD (mean ± sem of LFD across cells; n = 7 cells for 500 ms; 8 cells for 800 and 1600 ms) when there was a failure in response to the 1st AP (purple bar; m. ± sem) compared to the corresponding data taken from all traces (red bar). The LFD value was obtained by calculating the ratio of the mean numbers of released SVs for the second AP over the average release for the first AP. B, right: Same analysis, except that now LFD values are calculated separately for each s1 value. Dots indicate predictions from the model of Fig. 3.

Prolonged low frequency trains produce a gradual decrease in docking site occupancy.

A: A long AP train at low frequency (200 APs @ 2 Hz) was followed by a long recovery train (50-100 APs @ 100 Hz; experimental protocol in insert). Normalized plot of si during long low frequency train, showing an initial rapid depression (as in Fig. 2, notice the reduction in release for the first binned data point) followed by a gradual depression during the entire train duration (light blue: mean ± sem of individual trials; dark blue dots and associated error bars: binned data for 10 consecutive APs; red: linear fit to the data; n = 16 trials, 8 cells). B: Time course of synaptic depression during the recovery train, with exponential fit (red; time constant: 41.2 ISIs, or 412 ms). n = 6 trials, 3 cells for 50 APs, 6 trials, 4 cells for 100 APs. C: A series of control trains (8-APs @ 100 Hz, separated by 10 s-long inter-train intervals) were recorded before the long AP train to establish the characteristics of the synapse. Normalized si plots for this control and the first 8-APs of the recovery 100 Hz train are compared here. s1, s2 and s5-8 are indicated in the plot. D: The s1 ratio (red) between recovery and control trains is < 1, indicating a decrease in δ at the end of the low frequency train. By contrast, neither s2 (blue) nor s5-8 (purple) are changed, suggesting no change in ρ or in IP size. E: Simulation of control trains shown in C (black dots: average normalized data; red curve: simulation). F: Simulation of recovery trains shown in C (yellow dots: average normalized data). Decreasing the release probability from pr = 0.55 to pr = 0.2 without changing the other simulation parameters fails to provide a satisfactory fit of recovery data (black curve), while decreasing the DS occupancy from δ = 0.48 to δ = 0.18 without changing the other simulation parameters provides a satisfactory fit of the recovery data (red curve).

Presynaptic calcium transients in single PF varicosities

A1, left: Diagram of the experimental recording. A1, center: Image of an OGB-6F loaded PF varicosity before stimulation. A1, right: Same, at the peak of the response to a single presynaptic AP. A2: Average Ca2+-dependent fluorescence transients (ΔF/F0 traces from n = 5 trials: see Methods) registered in response to the 5 first APs (left trace) and to the 5 last APs (right trace) in response to a train of 50 APs at 1 Hz, from the same varicosity as in A1. Imaging was interrupted between the two traces to avoid photodamage and photobleaching. B: Average traces from a group of 10 varicosities from 8 cells at 1 Hz. C: Average traces from another group of 9 varicosities from 6 cells, in response to the first and last 5 APs from a train of 200 APs at 2 Hz. D: Mean Ca2+-dependent fluorescence transients (closed symbols: means across varicosities, with attached bars representing ± the sem; open symbols: values for individual varicosities) are the same for the 1st, 2nd, and last AP in a train (left: 1 Hz stimulation; black; right: 2 Hz stimulation; blue). E1-E2: Evolution of the basal fluorescence level during and after a long AP train. E1: Initial and final basal fluorescence levels are not statistically different for long trains at 1 Hz (left, black), while there is a significant increase at the end of trains at 2 Hz compared to its initial value (right, blue). E2: Return of the ΔF/F0 fluorescence trace to its initial baseline (mean from 5 varicosities) after the end of long 2 Hz trains indicates a shift of the basal fluorescence level by 11 % near the end of the train (yellow curve: exponential fit to the decay, with time constant 750 ms).

LFD during doublet vs. singlet train stimulations

A: Stimulation protocols. Either an 80-AP train of single stimulations (‘singlets’) or a sequence of 80 double stimulations 10 ms apart (‘doublets’) were applied at 0.2 Hz. Having 2 stimuli provides an evaluation of ρ and δ during LFD by comparing release during control and after AP# 1 or AP# 2 during doublets (sid1 and sid2, respectively). B: Example traces of a doublet experiment. C: During the first 20 doublet stimulations, the ratio between sid1 and ctrl s1 is < 1, indicating a decrease in δ; however, sid2 is not significantly different to ctrl s2 indicating that the occupancy of the replacement sites (ρ) remains the same as control during LFD. D: Normalized plot of si during singlet trains (sis) showing a 2-component depression (compare to Fig. 3). The initial component is followed by a gradual steep depression during the entire train duration (mean ± sem of individual trials, normalized with respect to the mean s1 value obtained during preliminary control trains; blue dots and associated error bars: binned data for 5 consecutive APs; red, linear fit to the data; 8 trials from 5 cells). E: Normalized plot of si during doublet trains (mean ± sem of individual trials, normalized with respect to the mean value obtained during preliminary control trains; dots and associated error bars: binned data for 5 consecutive APs; red, linear fit to the data; 9 trials from 6 cells). Top: SV release for the first AP of each doublet (sid1) displays a rapid LFD component but no significant slow component. Bottom: SV release for the second AP of each doublet (sid2), showing no significant deviation from the control value.

Trains of 4 successive EPSCs with different ISIs

Each panel shows numbers of released SVs in response to 4 consecutive APs at different ISI values. These numbers are normalized with respect to the mean s1 value, calculated across all ISI values (dotted lines). Data (dots: mean values; shaded areas: ± sem) are in blue for ISI values of 20 to 3200 ms, and in black for control 8-AP trains with an ISI of 10 ms. Red curves show simulations based on the model of Fig. 3.

LFD at various ISI and external calcium concentration values

Same experiments as in Fig. 2 at varying stimulation frequencies during low frequency trains (A: 1 Hz, n = 5; B: 2 Hz, n = 6; C: 0.5 Hz, n = 5) and varying external calcium concentrations (A: 3 mM; B and C: 1.5 mM). Red curves show exponential fits (A: asymptote: 0.78; τ = 2.3 ISIs, or 2.3 s; C: asymptote: 0.65; τ = 3.1 ISIs, or 6.2 s).

Time dependent transition rates during LFD

A: Time dependent transition rates (calculated per 10 ms time bins) are shown for the 4 SV transitions of the model in Fig. 2A. Note that the first 10 ms period is excluded from this plot (see the corresponding transition probabilities in Table 1). sf is stable as a function of time, while rf, rb and sb follow exponential functions of time (x, in ms). In the case of rb, the exponential starts with a delay of 150 ms. B: Following an AP, SV movements were described using 4 different DU states: “full” when both RS and DS were occupied; “up” when RS was occupied while DS was empty; “down” when RS was empty while DS was occupied; “empty” when both RS and DS were empty. Release resulted in an increase of % up and a simultaneous fall of % full (gray shade). The next 10 ms following stimulation were dominated by the calcium-dependent movement of SVs from the RS to the DS (rf, ‘transient docking’) resulting in an increase in % down and a decrease in % up. In subsequent 10 ms time segments, SV movements were calculated based on the time-dependent changes in rf, rb, sf and sb shown in A. Between 10 and 100 ms after the AP, up and down state proportions were stable, with a low % empty and up DUs and a high % down and full DUs (green shade, RS gate open). Next, SVs moved from DS to RS (undocking, with rate rb), resulting in a gradual increase in % up and a decrease in % down (red shade, RS gate closed). C: DU states before individual APs for 8-AP trains. Left: Control. Center: LFD 2 Hz. Right: Recovery.

Lack of recovery of LFD over tens of s after AP train

The time interval between the last AP in the LFD train and the 1st AP in the awakening train varied in the experiments presented in Fig. 7. These differences were used to analyze the time it takes for synapses to recover after a long low-frequency train. This figure shows the normalized s1 value for recovery trains following long LFD trains (200 APs @ 2 Hz), as a function of the time duration intervening between the end of the LFD train and the onset of the recovery train. Failures were observed in a majority of the trials. The 0-delay point corresponds to the mean s200 value at the end of the LFD train. The red line is a linear fit of the data that is constrained to pass through the 0-delay point. The data indicate no recovery for up to 60 s.

Binomial analysis of released SV numbers during prolonged LFD

A: Exemplar experiments. In each case, the distribution of released SV numbers (k) in response to the 1st AP during control runs (8 APs @ 100 Hz; black curves) is shown together with the distribution of released SV numbers during LFD (red curves; obtained from two of the experiments shown in Fig. 7). Next, we calculated the binomial distributions in two different models of reduced responsiveness. Both models assumed unchanged N values. In one model (blue dotted curves), the reduced response was assumed to reflect global failure at the AZ level for some stimulations, while other stimulations were assumed to be as effective as in the control. This model mimicks situations such as unwanted stimulation failures during LFD. In the other model, a homogeneous reduction of P (= δ* pr; the release probability per DU) was assumed (purple dotted curves). Both models were constrained to account for the failure rate observed during LFD. In both experiments, the lower P curve is much closer to the experimental results (red curves for k = 1 to 3) than the failure increase curve. B: Group data analysis (n = 16 trials from 8 cells). Left: Residuals (dotted lines: individual LFD trials; dots and associated SEM, means across trials) calculated from the comparison of the two models with LFD data (top: increased failure model; bottom: decreased P model). Right: Superimposition of mean residual curves for the two models, showing that the lower P model is a much better representation of the data than the failure model.

Prolonged low frequency trains with postsynaptic BAPTA

BAPTA (10 mM) was added to the internal solution used for MLI recording, keeping all other experimental conditions as before. A: A long AP train at low frequency (100 APs @ 2 Hz) was followed by a short 8-AP recovery train (experimental protocol in insert). Normalized plot of si during long low frequency train (red curve showing binned data for 10 consecutive APs), showing a two-component depression with an initial rapid depression (65 ± 3 % of the control) followed by a gradual depression during the entire train duration (slope = -0.2 ± 0.05 % per AP; red: linear fit to the data ; n = 12 trials, 3 cells). The results are similar to those with 1 mM EGTA shown in Fig. 6 (reproduced here as a dark blue curve). B: A short (8-AP) 100 Hz recovery train was applied immediately at the end of the low-frequency train. Normalized si plots for control and the recovery 100 Hz trains are compared here. C: The s1 ratio between recovery and control trains is < 1, indicating a decrease in δ at the end of the low frequency train while neither s2 (blue) nor s5-8 (purple) are changed, suggesting no change in ρ or in IP size. The similarity of results with and without postsynaptic BAPTA supports a presynaptic mechanism for LFD.