Figures and data
SF-iGluSnFR imaging of quantal glutamate release in individual presynaptic boutons of the axonal arbour.
(a) Experimental design. Schematic of the imaging configuration. Whole-cell voltage-clamp recordings were used to evoke action potentials (APs) in SF-iGluSnFR-transfected pyramidal neurons while monitoring glutamate release from their axonal boutons. (b) Identification of active boutons. Image analysis from an axonal fragment of a representative wild type (WT) neuron. Top: stimulation paradigm showing the consecutive paired-pulse (20 Hz) and train (50 APs at 20 Hz) protocols. Example raw fluorescence images acquired after AP stimulation during paired-pulse trials and the train. Bottom: corresponding images after application of a spatiotemporal filter, illustrating vesicular release successes and failures in active boutons within the selected region of interest. Bottom right: maximal projection of the filtered stack, visualising all boutons that released at least one vesicle during the stimulation protocol (five active boutons in this example). (c) Automatic event detection. Top: raw fluorescence trace from an example bouton (b1, WT) indicated by an arrowhead in (b). Bottom: corresponding deconvolved trace showing detected quantal release events. Vertical dotted lines mark AP timings; gaps between traces correspond to 10 s intersweep intervals. Events were identified as local maxima exceeding the detection threshold θ (horizontal dashed line), defined as 4 standard deviations (σ) of baseline fluorescence noise (see Methods). Blue and red circles denote synchronous and asynchronous events (<10 ms and >10 ms after the preceding AP, respectively). (d) Quantal analysis. Amplitudes of all detected events (top raster) were used to generate a quasi-continuous amplitude distribution by bootstrap resampling with bouton specific noise, which improved convergence and stability of the Gaussian mixture fit (see Methods). The resulting distribution was fitted with a sum of Gaussian functions to estimate the quantal amplitude (q, green line on the deconvolved traces in (c)), corresponding to the mean single vesicle SF iGluSnFR response. Peaks at q and 2q indicate one and two vesicle release events, respectively. See additional examples in Supplementary Figs. 1 and 2. (e) Functional bouton parameters for the example bouton b1 (WT). nT, nS, and nA represent total, synchronous, and asynchronous release efficacies (average number of quanta released per action potential); nA/nT denotes the asynchronous release fraction, and FDIis Facilitation-Depression Index. Parameters were calculated for the 1st (AP1) and 2nd (AP2) stimuli of the paired-pulse protocol and for the steady-state phase of the train (Train; APs 11– 50). See Methods for exact definitions.
Cell level analysis of glutamate release in Syt7-/- and wild type neurons during paired-pulse protocol.
(a, b) Quantification of vesicular release at the first (a) and second (b) action potentials. Left panels: average time courses of evoked glutamate release in wild type (WT) and Syt7-/- neurons (quanta per frame). For each bouton, the number of detected quanta in individual frames following the first or second action potential was averaged across ten paired-pulse trials (frames 0–11; 4 ms per frame). Bouton responses were then averaged within each neuron and subsequently across neurons (shaded areas indicate SEM across cells). A 10 ms threshold after the somatic action potential (between frames 2 and 3, i.e. at 10 ms) was used to separate synchronous and asynchronous events (see Methods). Right panels: cell-averaged total, synchronous, asynchronous release, and the asynchronous fraction. (c) Paired-pulse ratios for total, synchronous, and asynchronous components, calculated as the ratio of cell-averaged release at the second versus the first action potential. One Syt7-/- neuron lacked asynchronous release on the first pulse in all boutons but exhibited asynchronous events on the second; because the corresponding PPR is undefined, this datapoint is not plotted but was treated as maximally facilitated when computing the median across cells. Boxplots show the median and interquartile range with whiskers extending to 1.5x the interquartile range and individual datapoints overlaid. Wild type, n = 15 cells; Syt7-/-, n = 19 cells. * p < 0.05, ** p < 0.01, *** p < 0.001, NS not significant; two-tailed Mann–Whitney U tests. Exact p values and additional statistical details are provided in Supplementary Table 1 and SourceData.xlsx.
Cell level analysis of glutamate release in Syt7-/- and wild type neurons during the steady-state of the 20 Hz train.
(a) Comparison of total, synchronous, and asynchronous vesicular release per action potential in wild type (WT) and Syt7-/- neurons during 20 Hz trains of 50 stimuli. Traces represent cell-averaged vesicular release per bouton for each action potential, shown as the mean across neurons (shaded areas indicate SEM across cells). Synchronous and asynchronous events were separated using a 10 ms threshold after the somatic action potential, as in Fig. 2. (b) Average time courses of evoked glutamate release during the steady-state portion of the train (APs 11–50). Left: classification of synchronous and asynchronous components using a fixed 10 ms threshold. Right: corrected classification in which the portion of ongoing asynchronous release falling within the 10 ms window was reassigned from synchronous to asynchronous bins (see Methods). Shaded areas indicate SEM across cells. (c) Cell-averaged adjusted total, synchronous, and asynchronous release during the steady-state after correction for ongoing asynchronous release, and the corresponding asynchronous fraction. (d) Paired-pulse ratios for total, synchronous and asynchronous release during the train steady-state, calculated as the ratio of the cell-averaged release at APs 11–50 to the release at the first action potential. One Syt7-/- neuron showed no detectable asynchronous release on the first pulse and therefore did not yield a defined asynchronous PPR; this datapoint is not shown and was treated as maximally facilitated when computing the median. Boxplots show the median and interquartile range, with whiskers extending to 1.5× the interquartile range and individual datapoints overlaid. Wild type, n = 15 cells; Syt7-/-, n = 19 cells. * p < 0.05, ** p < 0.01, NS not significant; two-tailed Mann–Whitney U tests. Exact p values and additional statistical details are provided in Supplementary Table 1 and SourceData.xlsx.
Bouton-level dependence of the asynchronous release fraction on release efficacy.
Boutons within each neuron were sorted by total release efficacy at the first action potential nT (1) and divided into three equal-sized groups (terciles). For each neuron, nT (1) and the asynchronous release fraction nA (1)/nT (1) were averaged within each tercile. Left and middle panels: per-cell, per-tercile values for wild type and Syt7-/-, respectively, showing the dependence on release efficacy within each genotype. Dashed lines connect terciles from the same neuron; bold points and solid lines indicate population medians. Right panel: comparison of tercile medians between genotypes, illustrating the main effect of genotype across efficacy classes. Statistical significance was assessed using linear mixed-effects models, with Model 1 testing efficacy dependence within genotype (left and middle panels) and Model 2 testing genotype effects across efficacy classes (right panel); the genotype-by-efficacy interaction was not significant. Wild type, n = 15 cells; Syt7-/-, n = 19 cells. Exact statistics are provided in SourceData.xlsx. Significance bars indicate p values (*** p < 0.001; * p < 0.05).
Effect of Syt7 knockout on short-term plasticity across boutons with different release efficacy.
Boutons within each neuron were sorted by total release efficacy and divided into three terciles, as in Fig. 4. To avoid bias introduced by sorting on the first-pulse response, grouping was based on the mean total release across the two compared responses. For each neuron and tercile, short-term plasticity was quantified using the Facilitation-Depression Index and plotted against the corresponding tercile mean. (a) Paired-pulse stimulation, FDIT (2,1). (b) Steady-state phase of the train, FDIT (Tr,1). Left and middle panels: per-cell, per-tercile FDI values for wild type and Syt7-/-, respectively. Dashed lines connect terciles from the same neuron; bold points and solid lines indicate population medians. Right panels: comparison of tercile medians between genotypes, illustrating loss of facilitation in low-efficacy Syt7-/- boutons and enhanced depression at medium and high efficacy. The dashed horizontal line at FDI = 1 indicates no change between responses. Statistical significance was assessed using linear mixed-effects models, with Model 1 testing efficacy dependence within genotype (left and middle panels) and Model 2 testing genotype effects across efficacy classes (right panels); the genotype-by-efficacy interaction was not significant (see Methods). Wild type, n = 15 cells; Syt7-/-, n = 19 cells. Exact statistics are provided in SourceData.xlsx. Significance bars indicate p values (*** p < 0.001; ** p < 0.01; * p < 0.05).
Failure-based analysis reveals functional segregation of release modes and Syt7-dependent facilitation.
(a) Schematic of failure-based analysis. Boutons from wild-type (WT) and Syt7-/- neurons were stratified into efficacy bins based on mean first-pulse release nT (1). Within each bin, paired-pulse ratios were calculated for: all trials 
Model for differential regulation of synchronous and asynchronous release by calcium sensors.
Vesicles progress from empty release sites through loosely-docked to tightly-docked states in Ca2+-dependent manner supported by either Syt7 or Syt3. Tightly-docked vesicles are competent for both synchronous and asynchronous fusion, with the mode determined by Ca2+ dynamics and sensor complement. During action potentials, high local [Ca2+]local activates low affinity sensors (Syt1/Syt2) for synchronous release, with Syt7 enhancing release probability to enable facilitation. Residual [Ca2+]res drives asynchronous release through high-affinity sensors (Syt7, Syt3, Doc2α/β). Syt7 deletion abolishes synchronous facilitation while partially reducing asynchronous release, but asynchronous facilitation persists through remaining high-affinity sensors. This segregation allows independent regulation of synchronous and asynchronous components of synaptic transmission.
Median cell-level release parameters in wild type and Syt7-/- neurons during paired-pulse and train stimulation.
Quantal analysis in representative wild type boutons (related to Fig. 1).
Two additional representative wild type boutons (WT b2 and WT b3) analysed using the same quantal imaging and analysis pipeline as in Fig. 1. Left: raw fluorescence traces (top) and corresponding deconvolved signals (bottom). Vertical dotted lines indicate action potential (AP) timings; gaps correspond to 10 s intersweep intervals. The detection threshold (θ) is shown as a horizontal dashed line. Blue and red circles mark synchronous and asynchronous release events (<10 ms and >10 ms after the preceding AP, respectively). Right: amplitude distributions of deconvolved events were converted into quasi-continuous histograms using a bootstrap procedure incorporating bouton-specific noise and fitted with Gaussian mixtures to estimate the quantal amplitude (q). Peaks at q, 2q, and 3q correspond to the release of one, two, or three vesicles. Gaussian fitting was restricted to three components to ensure stable estimation of q (see Methods). Rare higher-amplitude events consistent with the release of four or more vesicles were occasionally observed (for example, WT b3) but were not included in the fitting procedure. Tables summarise bouton-specific quantal parameters for the first (AP1) and second (AP2) pulses of the paired-pulse protocol and for the steady-state phase of the train (Train; APs 11– 50). Partial overlap of successive events during 20 Hz trains occasionally reduced apparent event amplitudes, leading to minor under-detection (∼10–15%) of small asynchronous events (see Methods).
Quantal analysis in representative Syt7-/- boutons (related to Fig. 1)
Three representative boutons from Syt7⁻/⁻ neurons (Syt7⁻/⁻ b1–b3) analysed using the same quantal imaging and analysis pipeline as in Fig. 1. Left: raw fluorescence traces (top) and corresponding deconvolved signals (bottom). Vertical dotted lines indicate action potential (AP) timings; gaps correspond to 10 s intersweep intervals. The detection threshold (θ) is shown as a horizontal dashed line. Blue and red circles mark synchronous and asynchronous release events (<10 ms and >10 ms after the preceding AP, respectively). Right: amplitude distributions of deconvolved events were converted into quasi-continuous histograms using a bootstrap procedure incorporating bouton-specific noise and fitted with Gaussian mixtures to estimate the quantal amplitude (q). Peaks at q, 2q, and 3q correspond to the release of one, two, or three vesicles. Gaussian fitting was restricted to three components to ensure stable estimation of q (see Methods); in these examples, no events exceeded 3q. Tables summarise bouton-specific quantal parameters for the first (AP1) and second (AP2) pulses of the paired-pulse protocol and for the steady-state phase of the train (Train; APs 11– 50). Partial overlap of successive events during 20 Hz trains occasionally reduced apparent event amplitudes, leading to minor under-detection (∼10–15%) of small asynchronous events (see Methods).
Comparison of conventional paired-pulse ratio (PPR) and Facilitation-Depression Index (FDI) definitions.
Relationship between PPRT (2,1) = nT (2)/nT (1) (red line) and FDIT (2,1) = nT (2)/mean [nT (1), nT (2)] (blue line). The two measures are related by 
Bouton-to-bouton heterogeneity of release properties within single neurons.
Representative examples from one wild type neuron (n = 101 boutons) and one Syt7-/- neuron (n = 48 boutons) illustrating variability of release properties across boutons supplied by the same axon. Boutons were ranked by total release efficacy and divided into three equal-sized groups (terciles), as in Figs. 4 and 5. Shown are relationships between (a) the asynchronous release fraction at the first action potential nA (1)/nT (1) and total release efficacy nT (1); (b) short-term plasticity during paired-pulse stimulation, quantified using the Facilitation-Depression Index FDIT (2,1), plotted against the mean total release across the first and second action potentials; and (c) short-term plasticity during the steady-state phase of the stimulus train, quantified as FDIT (Tr,1), plotted against the mean total release across the first and steady-state responses. Vertical dashed lines indicate the division of boutons into low-(T1), medium-(T2), and high-efficacy (T3) terciles. Dark-coloured points connected by lines show tercile means ± SEM within each neuron. Dashed horizontal lines in (a) indicate the cell-averaged asynchronous release fraction, and dashed horizontal lines in (b, c) indicate FDI = 1, corresponding to no facilitation or depression. The number of boutons contributing to each panel differs due to stimulus-specific inclusion criteria (see Methods, Data Inclusion and Exclusion Criteria).
Paired-pulse ratios for total and synchronous, but not asynchronous, release depend on first-pulse outcome.
(a–c) Paired comparisons from the failure-based analysis in Fig. 6. For each genotype, paired-pulse ratios were computed separately for three trial subsets: all trials, trials in which the first action potential evoked release (success), and trials in which it did not (failure). Within each first-pulse efficacy bin, these three values form a matched set, and statistical significance was assessed using the Friedman test. Dashed lines connect matched data points from the same efficacy bin across the three conditions. Left column, wild type; right column, Syt7⁻/⁻. (a) Total release (PPRT). (b) Synchronous release (PPRS). (c) Asynchronous release (PPRA). For total and synchronous release, the distribution of paired-pulse ratios differed significantly across failure, all-trial, and success conditions in both genotypes (Friedman test; *p < 0.05, ***p < 0.001). In contrast, asynchronous release showed no significant difference across conditions in either genotype (NS). These results indicate that the paired-pulse ratio of the asynchronous component, unlike that of total and synchronous release, is unaffected by whether the first stimulus evoked release.
