1. Neuroscience
Download icon

Spatiotemporal dynamics of multi-vesicular release is determined by heterogeneity of release sites within central synapses

  1. Dario Maschi
  2. Vitaly A Klyachko  Is a corresponding author
  1. Washington University School of Medicine, United States
Research Article
  • Cited 1
  • Views 1,612
  • Annotations
Cite this article as: eLife 2020;9:e55210 doi: 10.7554/eLife.55210

Abstract

A synaptic active zone (AZ) can release multiple vesicles in response to an action potential. This multi-vesicular release (MVR) occurs at most synapses, but its spatiotemporal properties are unknown. Nanoscale-resolution detection of individual release events in hippocampal synapses revealed unprecedented heterogeneity among vesicle release sites within a single AZ, with a gradient of release probability decreasing from AZ center to periphery. Parallel to this organization, MVR events preferentially overlap with uni-vesicular release (UVR) events at sites closer to an AZ center. Pairs of fusion events comprising MVR are also not perfectly synchronized, and the earlier event tends to occur closer to AZ center. The spatial features of release sites and MVR events are similarly tightened by buffering intracellular calcium. These observations revealed a marked heterogeneity of release site properties within individual AZs, which determines the spatiotemporal features of MVR events and is controlled, in part, by non-uniform calcium elevation across the AZ.

Introduction

Information transmission in the brain critically relies on the number of vesicles released in response to each action potential. Not surprisingly, there have been several major efforts to model this process in recent years (Neher, 2010; Pan and Zucker, 2009; Rotman et al., 2011). Although the initial hypothesis was that, at most, only a single vesicle is released by a given synapse in response to an action potential (i.e., uni-vesicular release (UVR)), we now know that two or more vesicles can be released in response to a single action potential, even within the same synaptic bouton (Rudolph et al., 2015). This phenomenon has been named multi-vesicular release (MVR). Indeed, MVR is a ubiquitous release mechanism, occurring at both small and large central synapses throughout the brain (Auger et al., 1998; Christie and Jahr, 2006; Foster et al., 2005; Huang et al., 2010; Leitz and Kavalali, 2011; Malagon et al., 2016; Oertner et al., 2002; Rudolph et al., 2011; Singer et al., 2004; Taschenberger et al., 2002; Tong and Jahr, 1994; Wadiche and Jahr, 2001). Because the vast majority of small central synapses contain only a single active zone (AZ) (Schikorski and Stevens, 1997; Schikorski and Stevens, 1999), it follows that individual AZs must be capable of supporting both UVR and MVR.

MVR has been suggested to serve a wide range of functions, including enhancing synaptic reliability, controlling synaptic integration, enhancing efficient information transmission by complex spikes, and inducing synaptic plasticity (Rudolph et al., 2015). However, despite the prevalence and functional significance of MVR, the regulatory mechanisms and spatiotemporal organization of MVR events within individual synaptic AZs are poorly understood. Indeed, we know relatively little about the functional organization of vesicle release within the AZ in general—be it related to UVR or MVR events—although findings of nanoscale co-clustering of presynaptic docking factors and vesicle release machinery (Bademosi et al., 2016; Glebov et al., 2017; Tang et al., 2016; Weyhersmüller et al., 2011) have underscored the idea that vesicles are released from relatively stable ‘release sites’. Progress toward this end is hampered by the fact that the AZ is extremely small and thus beyond the resolution limits of conventional experimental approaches (Schikorski and Stevens, 1997; Schikorski and Stevens, 1999).

Recently, we were able to overcome this limitation by developing a nanoscale imaging modality that is capable of resolving the locations of individual vesicle release events in active hippocampal synapses in culture with ~27 nm precision (Maschi and Klyachko, 2017). Using these tools, we uncovered the presence of multiple, distinct release sites within individual AZs, at which vesicle fusion occurs repeatedly in response to action potentials. Having demonstrated our ability to identify individual release sites reliably, we now ask a series of questions related to the organization and regulation of these sites within a single AZ: do all sites support vesicle fusion involved in both UVR and MVR events? What controls the relative probability that a site is involved in an MVR event? And, at a more fundamental level: is the probability of release uniform across all sites?

To address these questions, we applied nanoscale imaging tools to detect and study the organizational principles of UVR and MVR events at individual hippocampal synapses in dissociated neuronal cultures. Our results reveal that release site characteristics are highly heterogeneous, even within a single AZ. Specifically, we find that the closer a site is to the center of the AZ, the higher its release probability. Interestingly, this gradient of release site properties also underlies the spatial and temporal organization of vesicle release involved in MVR events. This gradient of release site properties and the spatial features of MVR are also similarly affected by buffering intracellular calcium. Together, our analyses suggest a new level of functional organization of the AZ that is determined by the heterogeneous landscape of release site properties. The spatiotemporal organization of MVR is shaped by the gradient of release site properties across individual AZs and depends, in part, on the non-uniform elevation of calcium across the AZ following an action potential.

Results

Detection of MVR events

To detect MVR events and resolve their locations in hippocampal synapses, we took advantage of a nanoscale imaging approach that we recently developed (Maschi and Klyachko, 2017), combined with the use of a pH-sensitive indicator, pHluorin, which was targeted to the vesicle lumen via vGlut1 (vGlut1-pHluorin) (Balaji and Ryan, 2007; Leitz and Kavalali, 2011; Voglmaier et al., 2006). vGlut1-pHluorin was expressed in cultures of excitatory hippocampal neurons using lentiviral infection at DIV3. Imaging was then performed at DIV16–19 at 37°C. Release events were evoked using 1 Hz stimulation for 120 s (Figure 1A). Robust detection of individual release events was achieved at 40 ms/frame rate throughout the observation time period.

Figure 1 with 1 supplement see all
Non-uniform spatial features of MVR events and release sites within an AZ.

(A) Sample spatial distribution of ten UVR events and a single MVR (arrows) event within a hippocampal bouton evoked by 120 APs at 1 Hz. Release sites were defined using a hierarchical clustering algorithm with a cluster diameter of 50 nm (Maschi and Klyachko, 2017) and are shown by crossed circles. Images (before and after 1 AP stimulation) show a sample MVR event highlighted by arrows. (B) Examples of MVR events in different boutons. (C) Proportion of MVR events as a function of intra-event separation distances. (D, E) Cumulative distributions of AZ area (D) and number of clusters/release sites (E) for all recorded boutons (black) and boutons exhibiting MVR events (red). (F) Spatial overlap of MVR and UVR events. Percentages of MVR events in which none, one or both events in the pair occurred at release sites that also harbored at least one UVR event. (G) Probability of reuse by UVR events of more proximal vs. more distal release sites engaged in MVR event pairs. (H) Analysis of release probability (Pr-site) of more distal (left bars) and more proximal (right bars) release sites engaged in MVR event pairs compared to other release sites equidistant to the AZ center within ±25 nm but not engaged in the MVR event during the observation period (shown schematically on the left). (I) Proportion of release sites that are reused at least once during the observation period as a function of the distance to the AZ center. Numbers shown represent average release site re-use in a given bin. N = 3781 (UVR) and 245 (MVR) events, from 90 dishes from 11 independent cultures; *p<0.05, **p<0.01, ***p<0.001. Two-sample KS-test (D, E); Chi-square test (G); Paired t-test (H).

Using a hierarchical clustering algorithm, as well as simulations and statistical considerations, we have previously determined that release events are not randomly distributed throughout the AZ, but occur in a set of defined and repeatedly reused release sites within the AZs (Maschi and Klyachko, 2017). Our current data are consistent with that finding (Figure 1A and see 'Materials and methods' for details). In addition, visual examination of these recordings revealed a subset of events that involve the simultaneous fusion of two vesicles in the same bouton following a single action potential (Figure 1A,B). To identify these double fusion events (i.e., MVR) automatically and to determine their precise spatial locations, we used a well-established mixture-model fitting approach with two fixed-width Gaussians to approximate the point spread function (PSF)-like images of each vesicle (Jaqaman et al., 2008; Thomann et al., 2003). We previously used a conceptually similar fitting approach to localize individual non-overlapping UVR events, achieving ~27 nm precision (Maschi and Klyachko, 2017).

Here, when studying instances of MVR, we found that although the distance between the two fusion events comprising MVR varied widely, shorter separation distances were more frequently observed (Figure 1C). Over 90% of fusion event pairs that are involved in MVR were separated by less than 600 nm. The chances of misidentifying two events in the neighboring boutons as occurring in the same bouton is negligible, as we have previously used 3D FIB-SEM reconstruction of our neuronal cultures to show that the average bouton-to-bouton distance is an order of magnitude larger than the average event-to-event distance (Maschi and Klyachko, 2017). At the other end of the spectrum, it is important to note that this mixture-model fitting approach does not reliably fit the subset of double fusion events that occur so close as to be nearly overlapping. We examine this subset of unresolved MVR events using additional computational tools that are presented in the subsequent sections.

MVR events preferentially occur at release sites with higher release probability

We first asked how the incidence of MVR is distributed in the synapse population. Previous studies suggest that the synaptic release probability is a strong predictor of a propensity for MVR (Christie and Jahr, 2006; Huang et al., 2010). Moreover, we know that AZ size is a major determinant of synaptic release probability (Holderith et al., 2012; Matz et al., 2010). We thus explored the relationship between AZ size (see 'Materials and methods' for details) and the probability of observing MVR events at individual synaptic boutons. We also used the number of release sites per bouton as another, functional measure of synaptic release probability. Individual release sites within each bouton were defined using a hierarchical clustering algorithm with a cluster diameter of 50 nm (Maschi and Klyachko, 2017). Boutons at which at least one MVR event was observed had a significantly larger AZ (Figure 1D and Table 1; N = 3781 (UVR) and 245 (MVR) events, 90 dishes from 11 independent cultures; p<0.001, two-sample KS-test) and a significantly larger number of release sites than the synapse population overall (Figure 1E; Table 1). These results suggest that MVR events are more likely to occur at boutons that have larger AZs and higher overall release probability.

Table 1
Data values and statistical analyses.

Columns represent (from left to right): figure/panel number; experimental conditions; number of samples (synapses, dishes and cultures); mean values and standard errors of the means (SEM); statistical test used for comparison; and the P-value resulting from the statistical comparison.

Figure numberConditionsNSynNDishesNCulturesMean ± SEMStatistical testPval
Figure 1DAll378190110.1014 ± 0.0009Two-sample KS-test<0.001
MVR24590110.164 ± 0.005
Figure 1EAll378190116.57 ± 0.05Two-sample KS-test<0.001
MVR24590119.8 ± 0.03
Figure 1GProximal24590110.45 ± 0.05Chi-square test0.0386
Distal24590110.29 ± 0.05
Figure 1HDistal MVR/UVR6690111.24 ± 0.09Paired t-test0.006
Proximal MVR/UVR9590111.28 ± 0.08Paired t-test<0.001
Figure 2CUVR15190110.32 ± 0.02Two-sample KS-test<0.001
MVR14490110.47 ± 0.02
Figure 3BProximal MVR/Proximal MVR24590111 ± 0.03Chi-square test0.017
Distal MVR/Proximal MVR24590110.65 ± 0.02
Figure 3CLarger MVR2459011214 ± 7Two-sample t-test0.0058
Smaller MVR2459011242 ± 7
Figure 3DMVR2459011y = 3.108+0.02106 xLinear fit<0.001
Figure 3EMVR <400 nm12990119.8 ± 0.7Two-sample t-test<0.001
MVR >400 nm115901114.3 ± 0.9
Figure 4AEGTA MVR Linear fit2255711y = 8.801+0.00556 ×Linear fit0.264
MVR <400 nm156571110.1 ± 0.8Two-sample t-test0.131
MVR >400 nm69571112.3 ± 1.2
Figure 4BCtrl MVR24590110.021 ± 0.005One-way analysis of covariance (ANOCOVA)0.022
EGTA MVR22557110.006 ± 0.005
Figure 4CCtrl MVR245901152 ± 3Chi-square test<0.001
EGTA MVR225571169 ± 3
Figure 4DLarger MVR2255711178 ± 7Two-sample t-test<0.001
Smaller MVR2255711216 ± 7
Figure 4ECtrl Pr = 0.04224179011104 ± 8Two-sample t-test<0.001
EGTA Pr = 0.0422338571162 ± 5
Ctrl Pr = 0.0332417901193 ± 3Two-sample t-test<0.001
EGTA Pr = 0.0332338571172 ± 2
Ctrl Pr = 0.02524179011107 ± 2Two-sample t-test<0.001
EGTA Pr = 0.0252338571183 ± 1
Ctrl Pr = 0.01724179011124 ± 1Two-sample t-test<0.001
EGTA Pr = 0.01723385711100 ± 1
Ctrl Pr = 0.00824179011154.6 ± 0.7Two-sample t-test<0.001
EGTA Pr = 0.00823385711129.8 ± 0.8
Figure 4FCtrl MVR24179011y=a+bxcFit0.0030
EGTA MVR23385711a = 17.951
b = 1.0049e+09 
c = 5.5575
Figure 4GDistal MVR/UVR5257111.29 ± 0.09Paired t-test0.0020
Proximal MVR/UVR7757111.5 ± 0.1Paired t-test<0.001
Figure 3—figure supplement 1AUVR136901111.5 ± 0.8Two-sample t-test0.1262
MVR,<400 nm12990119.8 ± 0.7
UVR136901111.5 ± 0.8Two-sample t-test0.0232
MVR,>400 nm115901114.3 ± 0.9
Figure 3—figure supplement 1BUVR6659011y = 35.088-0.0064674 xLinear fit0.207
UVR 0–100 nm285901134.9±0.6Two-sample t-test0.1376
UVR 200–300 nm109901133.4 ± 0.9
Figure 3—figure supplement 1DSynaptic vesicle diametersNSyn = 93
NVesic = 806
3y = 48.109+0.0047331xLinear fit0.161
Figure 4—figure supplement 1ALarger Ctrl MVR2459011214 ± 7Two-sample t-test0.0058
Smaller Ctrl MVR2459011242 ± 7
Larger EGTA MVR2255711178 ± 7Two-sample t-test<0.001
Smaller EGTA MVR2255711216 ± 7
Larger Ctrl MVR2459011214 ± 7Two-sample t-test<0.001
Larger EGTA MVR2255711178 ± 7
Smaller Ctrl MVR2459011242 ± 7Two-sample t-test0.0116
Smaller EGTA MVR2255711216 ± 7
Figure 4—figure supplement 1BCtrl Pr = 0.04224179011104 ± 8Two-sample t-test<0.001
EGTA Pr = 0.0422338571162 ± 5
Ctrl Pr = 0.0332417901193 ± 3Two-sample t-test<0.001
EGTA Pr = 0.0332338571172 ± 2
Ctrl Pr = 0.02524179011107 ± 2Two-sample t-test<0.001
EGTA Pr = 0.0252338571183 ± 1
Ctrl Pr = 0.01724179011124 ± 1Two-sample t-test<0.001
EGTA Pr = 0.01723385711100 ± 1
Ctrl Pr = 0.00824179011154.6 ± 0.7Two-sample t-test<0.001
EGTA Pr = 0.00823385711129.8 ± 0.8

In addition to detecting a variable propensity for MVR across the synapse population, we were interested in whether there is similar variability among release sites within the same synaptic bouton. In other words, are all release sites within a single bouton equally likely to support vesicle fusion involved in both UVR and MVR events or are there specific characteristics of individual release sites that make them more or less likely to support one type of fusion event over the other? We looked at the extent of release site overlap between MVR and UVR events, defining it as a ‘full overlap’ when both release sites involved in MVR were also observed as release sites during UVR events, ‘partial overlap’ when only one release site involved in MVR also served as a release site during UVR, and ‘no overlap’ when neither release site involved in an MVR event was observed as a release site for a UVR event during our observation period. We found full overlap with UVR for ~15% of MVR events, whereas ~47% of MVR events showed partial overlap, and ~38% showed no overlap (Figure 1F; Table 1). These results did not depend on the specific definition of release sites because we obtained essentially the same breakdown using proximity analysis of individual events without defining release sites using any clustering algorithms (Figure 1—figure supplement 1; Table 1).

This observation suggests that release sites can be involved in both UVR and MVR events, but that the likelihood that they are involved in one versus the other is not uniform (explored in more detail below). We also note that because we can only observe release events over a relatively short period (limited to 120 s by natural synapse displacement [Maschi and Klyachko, 2017]), our results cannot be interpreted as indicating that there are some specialized release sites that only support MVR or UVR. MVR is a relatively low probability event (~10% of release events are MVR under our experimental conditions), making much longer recordings necessary to determine whether sites are never involved in MVR; indeed, infinitely long recordings are required to answer this question definitively.

To understand better why some sites may be more likely to support MVR events than others, for each of the two release sites involved in an MVR event, we determined the probability that the same release site is also involved in one or more UVR events during our observation period. We found that this probability was location-dependent: of the two release sites involved in a given MVR event, the one more proximal to the AZ center was significantly more likely to be also involved in UVR events than the more distal site (Figure 1G; Table 1). Thus, the probability that a given release site was involved in both UVR and MVR events formed a gradient from AZ center to periphery, supporting the notion of heterogeneity of release site properties across a single AZ.

To further examine this notion, we quantified the release probability at each release site (Pr-site) based on the number of release events detected during the 120 s observation period. We then compared Pr-site separately for each of the two sites involved in the MVR event to other release sites (i.e., those involved only in UVR events) located equidistantly from the AZ center in the same bouton (i.e. within a ± 25 nm band, Figure 1H). We observed that both release sites engaged in an MVR event had a significantly higher release probability than other equidistant, non-MVR-involved release sites in the same bouton (Figure 1H; Table 1).

We also observed more general patterns among release sites within a given AZ, including those involved in UVR, MVR, or both. First, release sites were highly heterogeneous in terms of release probability, which varied ~5 fold among release sites within the same AZ during the observation time (Pr-site range [0.008–0.042]). Second, we observed a spatial gradient of Pr-site, which decayed with distance from the AZ center (Figure 1I).

Taken together, these findings provide evidence of a marked heterogeneity of release site properties within the individual AZs, characterized by a gradient of Pr-site from the AZ center to the periphery. These results further suggest that, in addition to the radial distribution of release probability, release sites that have a higher propensity for MVR are characterized by a higher release probability than other sites equidistant to the AZ center.

The spatiotemporal features of resolved MVR events are generalizable to MVR events that cannot be resolved

The well-separated MVR events analyzed above had sufficient spatial separation to allow each event in the pair to be individually localized (resolved events). However, because the AZ is very small overall, a significant proportion of MVR events involve vesicle release at sites within a sub-diffraction distance from one another. Such unresolved events would not have been captured in our analyses thus far. Therefore, we next asked to what extent the findings relating to resolvable MVR events could be generalized to unresolved MVR events.

To identify unresolved MVR events, we took advantage of the quantal nature of vesicular release to distinguish MVR from UVR events based on amplitude (Balaji and Ryan, 2007; Leitz and Kavalali, 2011). At 2 mM extracellular Ca+2 and at 37°C, over 90% of the events in hippocampal neurons are UVR (Leitz and Kavalali, 2011; Maschi and Klyachko, 2017). Thus, we analyzed individually each synaptic bouton with a minimum of five fusion events to determine the mean and intrasynaptic variability (standard deviation) of quantal event amplitude. We then set the threshold for MVR event detection at two standard deviations above the mean quantal event amplitude for that bouton (Figure 2A,B). On the basis of this analysis, we estimated that in our studies, MVR events represent ~9% of all release events. Of these events, we were able to robustly identify ~50% as MVR based solely on their amplitude (Figure 2B). This approach does not rely on spatial information, and thus is complementary to the mixture-model fitting approach that we used above; the MVR event populations that were identified by these two approaches are partially overlapping (~20%; data not shown).

Spatiotemporal features of resolved MVR events generalize to unresolved MVR events.

(A) MVR (red) and UVR (black) events were separated on the basis of the event amplitude. Examples of an identified resolved MVR event (top) and two unresolved MVR events (middle and bottom) are shown with corresponding images. (B) An intensity histogram for all detected events from panel (A) reflects the quantal nature of fusion events. Gaussian fits to the first peak (UVR events, blue) and second peak (MVR events, red) and their relative abundances are shown. (C) Asymmetry analysis of unresolved MVR events vs UVR events. The asymmetry score was calculated using asymmetrical Gaussian fit to the event image to determine maximal (δ1) and minimal (δ2) width (insert). (D) Proportion of unresolved MVR events as a function of asymmetry score, which correlates with intra-event separation distances. (E) Mean distance to the AZ center for unresolved (top) and resolved (bottom) MVR events. Distance was calculated from the peak of the Gaussian fit for unresolved MVR events and from the center of the line connecting two fusions within the resolved MVR events. (F) Probability of overlap of unresolved MVR and UVR events in the same bouton as a function of distance to the AZ center. Event overlap was determined using proximity analysis with a radius of 25 nm. Only more symmetrical MVR events (asymmetry score <0.5) were included in this analysis. Points represent the proportion of MVR events within a given distance band that overlapped with UVR events. Pie chart: proportion of unresolved MVR events that overlap or not with UVR events during the observation period. N = 151 (UVR) and 144 (MVR) events, 90 dishes from 11 independent cultures. *** p<0.001; Two-sample KS-test (C).

The identified MVR events were then analyzed on the basis of asymmetry considerations, using an asymmetric Gaussian model fit that takes into consideration the pixelated nature of the image to determine the width (sigma) of the Gaussian fit in the maximal (longitudinal, δ1) direction and the minimal (transverse, δ2) direction. The ratio δ12–1 (asymmetry score) represents an estimate of asymmetry of the double-event image, which correlates with the distance between the two sub-diffraction events forming the image (DeCenzo et al., 2010). Distributions of asymmetry scores for the single and double events indicate that they represent two distinct populations (Figure 2C; Table 1) and thus validate our approach to robustly distinguish unresolved MVR events from UVR events.

We then examined the spatiotemporal features of the unresolved MVR events. First, we observed that unresolved MVR events preferentially have smaller asymmetry scores (Figure 2D) and thus tend to occur at smaller separation distances, similarly to the resolved MVR events (Figure 1C). Next, we examined the localization of unresolved MVR events relative to the AZ center/periphery. We observed that more asymmetrical (more spatially separated) events occurred closer to the AZ center, whereas symmetrical events tend to be more peripheral (Figure 2E, top). An equivalent calculation for the resolved MVR events (see figure legend for details) showed a very similar relationship (Figure 2E, bottom), supporting the notion that the two subpopulations of MVR events have similar spatial properties. Finally, as described for the resolved MVR events above, we examined the extent to which release sites were used for either MVR or UVR, or both. Only a subpopulation of strongly overlapping MVR events (asymmetry score <0.5) were used in this analysis because these highly symmetrical events could be well-approximated by a single symmetrical Gaussian fit, making this analysis comparable to that of the resolved MVR events. The extent of overlap of MVR and UVR events at individual release sites was comparable between unresolved and resolved MVR events (Figures 2F and 1F), with overlap more likely occuring closer to the AZ center in both cases (Figures 2F and 1G). These data suggest that unresolved and resolved MVR events have comparable spatial features. Thus, our observations can likely be generalized across the entire population of MVR events.

Temporal desynchronization of release events comprising MVR

We next asked whether there is also heterogeneity within the temporal properties of MVR events. Previous studies present evidence of temporal ‘jitter’ on a millisecond timescale (~1–5 ms) within the release pair comprising an MVR event at both excitatory and inhibitory cerebellar synapses (Auger et al., 1998; Auger and Marty, 2000; Crowley et al., 2007; Malagon et al., 2016; Rudolph et al., 2011). Upon initial inspection of resolved MVR events, we noticed that one of the two fusion events comprising an MVR is often noticeably larger in amplitude than the other (Figure 3A). In our experimental approach, action potentials are synchronized with the beginning of the frame acquisition; thus, considering the quantal nature of fusion events, we hypothesized that this difference in amplitude reflects imperfect synchronization between the two release events (if one release event occurs later in the recording frame, we would expect to collect a smaller number of photons for this delayed event, i.e., smaller amplitude) (Figure 3A, top).

Figure 3 with 2 supplements see all
Spatiotemporal organization of release events comprising an MVR.

(A) Sample image (right) and intensity profile (left) of an MVR event with noticeable difference in intra-event amplitude. The top insert shows a cartoon representation of a relationship between a time delay (red arrow) of the second fusion after an action potential and the resulting amplitude difference within an MVR event. (B) Probability that the proximal or distal event within MVR pairs is larger, normalized to that of the proximal event. (C) Distance to the AZ center from the larger and smaller events within MVR pairs. (D, E) Amplitude difference of the two events comprising MVR as a function of intra-event separation. Linear fit (D) and t-test of pooled data (E) are shown. *p<0.05, ***p<0.001; Chi-square test (B); Paired t-test (C); Two-sample t-test (E). N = 245 MVR events, from 90 dishes from 11 independent cultures.

To test this hypothesis, we looked at the relationship between the relative amplitude and the spatial organization of the release pairs comprising each MVR event. We observed that the larger amplitude event (assumed to occur first) was more likely to occur closer to the AZ center than the smaller one in the pair (Figure 3B; Table 1). Accordingly, the average distance to the AZ center was significantly shorter for the larger amplitude event relative to the event with the smaller amplitude within the pair (Figure 3C; Table 1). Most importantly, this amplitude difference was correlated with the distance between the two events within the pair (Figure 3D,E), such that larger amplitude differences were associated with larger separation distances (Figure 3D,E). This spatial organization parallels the gradient of release site release probability from the AZ center to the periphery (Figure 1I, and see below). We note that a component of the amplitude differences can arise from the uncertainty in determining the fusion event amplitude; we estimated this uncertainty to be ~10% (Figure 3—figure supplement 1A). Thus, the uncertainty in our measurements may account for the amplitude differences that we measured for the most closely spaced MVR events, but it does not account for the differences that we measured for MVR events that are further apart (Figure 3—figure supplement 1A). Moreover, the positive correlation between the amplitude difference and the spatial separation of the two events comprising MVR cannot be explained by random noise or by measurement uncertainty. We thus interpret our results as indicating that the amplitude difference between the two fusion events that comprise an MVR, at least in part, reflects imperfect synchronization.

Given the observed amplitude differences within the release pairs comprising an MVR event, and the acquisition duration of 40 ms per frame, we estimated the maximal time difference between the two events comprising MVR to be less than ~4 ms for the majority of MVR events in our recordings. We note that this value overestimates the true extent of desynchronization because, as noted above, a component of the amplitude difference arises from uncertainty in the amplitude measurement itself. Thus, we estimate the maximal time delay to be ~2 ms if the measurement uncertainty is factored in. These values are in a close agreement with previous reports, which found desynchronization within individual MVR events to be in the range of 1–5 ms (Auger et al., 1998; Auger and Marty, 2000; Crowley et al., 2007; Malagon et al., 2016; Rudolph et al., 2011).

We also considered the possibility that factors other than desynchronization, such as differences in vesicle size or cleft pH along the AZ plane, contribute to the difference in amplitude within MVR events and its spatial arrangement relative to the AZ center. We used Large-Area Scanning Electron Microscopy (LaSEM) micrographs of our cultures (Maschi et al., 2018) to determine the relationship between the size of vesicles positioned near the AZ (within 100 nm, defined previously as tethered vesicles; Maschi et al., 2018) and their position relative to the AZ center. Vesicle diameter did not exhibit any measurable change as a function of distance from the AZ center (Figure 3—figure supplement 1C,D), indicating that the amplitude differences that we observed are not due to systematic differences in vesicle size. We next examined cleft pH at different locales within the AZ. The peak amplitude of vGlut1-pHluorin signal during individual fusion events is determined in large part by the cleft pH. Thus, we measured peak pHluorin signal as a function of distance to AZ center. The vGlut1-pHluorin signal amplitude (Figure 3—figure supplement 1B) did not measurably change as a function of distance from the AZ center, suggesting that a gradient of cleft pH is unlikely to explain the differences in MVR event amplitude. Finally, we note that in our imaging experiments, dozens of synapses are positioned in random orientation and are recorded simultaneously. Thus, in any given recording, some boutons have the center of the AZ in focus, whereas in others, only the periphery is in focus, and so on for all possible configurations in-between. Thus, there is no bias towards any one configuration or tilt of the AZ relative to the imaging plane that can lead to systematically larger fusion event amplitude at the AZ center versus that in the periphery. Indeed, this notion is highlighted by the fact that, in our measurements, the average release event amplitude is indistinguishable between events located at the AZ center versus those in the periphery (Figure 3—figure supplement 1B, insert). Thus, we conclude that the observed differences in amplitude within the release pair comprising an MVR event are unlikely to result from a bias in AZ position relative to the imaging plane.

In summary, our results support the notion that vesicle release associated with MVR is desynchronized and follows a specific spatial organization with respect to the center of the AZ. Specifically, the first of the two events in the MVR pair is preferentially located closer to the AZ center. This spatial organization of MVR events parallels our finding that release probability and the probability that a release site is involved in an MVR also follows a center-to-periphery spatial gradient.

Double events do not result from asynchronous release overlapping temporally with synchronous events

Given the intrinsically limited temporal resolution of our imaging tools, we must consider the possibility that the double events that we observed arose not from MVR, but rather from asynchronous release events generated by preceding stimulation and temporarily overlapping with UVR. We addressed this possibility in five complementary ways.

First, if the double events do in fact arise from overlap of synchronous release with asynchronous events generated by preceding stimulation, the probability of observing such double events should increase over the course of the stimulus train. However, we detected no increase in the double event probability during the stimulus train (Figure 3—figure supplement 2A, Table 1). In fact, the probability of observing double events is slightly higher for the first stimulus, for which there is no preceding stimulation. These results argue against any significant contribution of asynchronous release to the observed double-release events.

Second, synchronous release events, including those associated with MVR, are time-locked with the stimulus and are acquired for the entire duration of the frame. Thus, the amplitude distribution of MVR events should appear as a single Gaussian peak centered at ~2 q value (twice the amplitude of a UVR event), as indeed is the case in our measurements (Figure 2B). By contrast, asynchronous release is, by definition, not time-locked with the stimulus and thus occurs randomly at any time during the acquisition frame. As a result, single asynchronous release events are acquired for a wide range of durations that are less than a full frame, and thus must have a skewed non-Gaussian amplitude distribution shifted towards smaller values than those produced by a full-size synchronous event. We confirmed that this is the case using asynchronous events that we recorded in 3 mM Sr2+ in otherwise identical conditions (Figure 3—figure supplement 2B, Table 1). Thus, we conclude that the double events in our recordings are synchronous events and have properties that are distinct from those of asynchronous release.

Third, we compared the spatiotemporal properties of the double events in our recordings with that of asynchronous release events recorded in 3 mM Sr2+. The asynchronous release events detected in the same frame in the same bouton (Figure 3—figure supplement 2C) did not exhibit the spatiotemporal patterns evident for the double events (Figure 3D). The fact that the spatiotemporal properties of the double events are distinct from those of asynchronous events supports our conclusion that the double events in fact reflect instances of MVR.

Fourth, the synchronicity of double events in our recordings is relatively high, arguing against a significant contribution from asynchronous release. As mentioned above, we estimate that the double events in our recordings are synchronous with the stimulation within a few milliseconds. This result is consistent with previous reports of MVR, and aligns with the accepted definition of MVR events (Auger et al., 1998; Auger and Marty, 2000; Crowley et al., 2007; Malagon et al., 2016; Rudolph et al., 2011).

Finally, we note that previous studies find minimal asynchronous release evoked by 1 Hz stimulation under nearly identical experimental conditions (37°C, 2 mM extracellular Ca2+) (Raingo et al., 2012). Thus, at least five complimentary lines of evidence strongly suggest that double events in our recordings do indeed reflect synchronous MVR events, with no significant contribution of asynchronous release.

Spatiotemporal features of MVR events and release site properties are calcium-dependent

We next set out to explore the mechanistic origin of the spatial organization of MVR events relative to the AZ center. Our previous study suggested a possible role of calcium in the spatial regulation of release site properties because we observed that release site usage is regulated in an activity-dependent manner, such that site usage shifts towards the periphery during trains of activity (Maschi and Klyachko, 2017). Previous studies have also found that the propensity for MVR events increases as extracellular calcium levels increase (Leitz and Kavalali, 2011). However, whether calcium regulates the spatiotemporal organization of MVR events and/or the properties of the release sites remains unknown. To test this possibility, we pre-incubated neurons with a cell-permeable calcium chelator EGTA-AM (30 µm) for 20 min. EGTA does not directly affect vesicle release because it is too slow to buffer rapid calcium rise near voltage-gated calcium channels (VGCCs), but it is effective in buffering the ensuing slower calcium elevation caused by diffusion. We observed several effects of EGTA on the spatiotemporal distribution of MVR events. First, EGTA affected MVR event desynchronization: although the amplitudes of the two fusion events comprising MVR were still different, the difference no longer depended on their separation distance (Figure 4A,B; Table 1; N = 225 synapses, from 57 dishes from 11 independent cultures). Second, in the presence of EGTA, a larger proportion of MVR events occurred at shorter intra-event distances (Figure 4C; Table 1), that is, pairs of release events comprising MVR were more likely to be closer to each other in the presence of EGTA. In addition, the average distances from both events in the MVR pair to the AZ center were significantly shortened in the presence of EGTA (Figure 4—figure supplement 1A, Table 1). Thus, calcium buffering causes MVR events to occur at shorter separation distances and more proximal to the AZ center. However, the preferential localization of the larger (earlier) event in the MVR pair closer to the AZ center was still observed in the presence of EGTA (Figure 4D, Table 1). These effects of buffering calcium diffusion with EGTA suggest that several major spatiotemporal features of MVR events are determined, in part, by calcium diffusion following an action potential.

Figure 4 with 1 supplement see all
Spatiotemporal features of MVR events and release site properties are calcium-dependent.

(A) Effect of EGTA on the correlation between the spatial separation and amplitude difference between two events comprising an MVR. (B) The effect of EGTA was assessed by comparing the slopes of the correlation in panel (A) in control (from Figure 3D) and EGTA (from A) conditions. (C) Proportion of MVR events with intra-event separation <400 nm in EGTA and control conditions. (D) Distance to the AZ center from the larger and smaller events within MVR pairs in the presence of EGTA. (E,F) Average distance to the AZ center (E) and its relative change (F) for individual release sites binned on the basis of their release probability, in EGTA and control conditions. (G) Release probability of more distal (left bars) and more central (right bars) release sites engaged in MVR event pairs compared to all other release sites equidistant to the AZ center within ±25 nm, in the presence of EGTA. ***p<0.001; ns, not significant. Statistical tests were as follows: two-sample t-test (A,E); one-way analysis of covariance (ANOCOVA) (B); Chi-square test (C); and Paired t-test (D,G). Control: N = 245 MVR events from 90 dishes from 11 independent cultures. EGTA: N = 225 MVR events from 57 dishes from 11 independent cultures.

In line with the idea that the spatiotemporal features of MVR events reflect release site heterogeneity, we observed that the preferential utilization of more central release sites was exacerbated in the presence of EGTA (Figure 4E, Figure 4—figure supplement 1B; Table 1). Interestingly, buffering intraterminal calcium had a larger effect on more central release sites with higher Pr-site values than on the more peripheral sites with lower Pr-site (Figure 4F). These results are in line with our earlier findings that the release site utilization shifts in the opposite direction (i.e. from AZ center towards periphery) when intraterminal calcium is elevated during high-frequency stimulation (Maschi and Klyachko, 2017). It is also consistent with the shorter separation distance within MVR events in the presence of EGTA (Figure 4C), and the shorter distances of MVR events to the AZ center in the presence of EGTA (Figure 4—figure supplement 1A). Indeed, the preferential use of sites with higher release probability (as compared to other, equidistant sites) during MVR events persists in the presence of EGTA (Figure 4G; Table 1), while the utilization of these sites shifts closer to the AZ center in the presence of EGTA.

Together, these observations suggest that the gradient of release site properties, as well as the spatiotemporal features of MVR events, are, in part, determined by the calcium concentration landscape across the AZ.

Discussion

Although MVR is well established as a ubiquitous release mechanism that occurs at many types of synapses (Rudolph et al., 2015), its spatiotemporal organization and regulation within the AZ is largely unknown. We took advantage of our ability to detect individual release events at central synapses with nanoscale precision to reveal three major findings: (i) release sites within the same AZ have highly heterogeneous properties, reflecting a gradient of release probability that decreases from the AZ center to periphery; (ii) MVR events exhibit non-uniform patterns of spatial and temporal organization that parallel the center-to-periphery organization of release site release probability; and (iii) both the gradient of release site properties and the spatiotemporal features of MVR are determined, in part, by the intraterminal calcium elevation following an action potential. Together, these results suggest that the non-uniform spatiotemporal dynamics of MVR events arise from the heterogeneity of release site properties within the individual AZs.

Our results suggest a model of MVR events in which the earlier of the two events in the MVR pair is similar to a UVR event, in that it occurs closer to the AZ center because release sites with higher release probability are localized preferentially more proximally to the AZ center. A second release event is then triggered occasionally with a short delay after the same action potential, at a more peripheral release site that is primed for release, in part due to calcium spread from the AZ center. This notion is supported by the effects of calcium buffering with EGTA and by our observations that both events in the MVR pair occur at sites with higher release probability than other equidistant sites in the same bouton. These observations reveal a previously unrecognized complex landscape of release probability within the AZ, which determines the spatial organization of MVR, and arises, in part, from a calcium concentration gradient across individual AZs.

What are the molecular underpinnings of release site heterogeneity? Recent nanoscopy studies indicate that release sites colocalize with nanoclusters of presynaptic docking factors, such as RIM1/2 (Tang et al., 2016), which have been suggested to control the recruitment and clustering of VGCCs at the AZ via the RIM binding protein RIM-BP (Davydova et al., 2014; Hibino et al., 2002; Wang et al., 2000). RIM1/2 nanoclusters are more likely to be located near the center of the AZ than in the periphery, suggesting a possible structural basis for the gradient of release site properties that we observed in this work (Tang et al., 2016). Moreover, the enrichment of many scaffold/docking proteins, including RIM, Bassoon, and Munc13, within their clusters is dynamic (Bademosi et al., 2016; Glebov et al., 2017; Smyth et al., 2013; Tang et al., 2016; Weyhersmüller et al., 2011). Thus, the heterogeneity of release site properties may arise, in part, from variability in cluster architecture, such as cluster size or the relative enrichment of tethering/docking/priming factors.

An additional source of heterogeneity could be the fact that variable fractions of many critical components of release machinery, including VGCCs, Syntaxin-1, and Munc18, are mobile within the AZ (Glebov et al., 2017; Schneider et al., 2015; Smyth et al., 2013). For example, a large proportion (>50%) of VGCCs are mobile in the AZ plane and this mobility is calcium-dependent (Schneider et al., 2015). How VGCC mobility is spatially controlled has not been explored, but differential VGCC stability at the AZ center versus the periphery could account, in part, for the heterogeneity of the calcium rise across the bouton. VGCC mobility could also affect coupling between the channels and the vesicles (Eggermann et al., 2012; Miki et al., 2017). This could explain the differential effect of EGTA on peripheral versus more central release sites.

Another possibility is that release site properties are determined, in part, by extrinsic factors. For example, release site refilling and vesicle retention at release sites depends on actin and myosins (Maschi et al., 2018; Miki et al., 2016). Thus, a non-homogeneous spatial distribution of actin cytoskeleton could contribute to differential release site usage across a bouton. In addition, calcium influx at least through some subtypes of VGCCs is also modulated by the balance of phosphorylation/dephosphorylation by CDK5 and calcineurin (Kim and Ryan, 2013), and is directly regulated by a number of presynaptic proteins, such as ELKS (Liu et al., 2014) and Munc13 (Calloway et al., 2015). Although the precise factors that drive release site heterogeneity remain to be elucidated, our results reveal a previously unknown level of structural and functional organization among vesicle release sites within individual AZs, with specific implications for the spatiotemporal dynamics of MVR events.

Materials and methods

Neuronal cell cultures

Request a detailed protocol

Neuronal cultures were produced from rat hippocampus as previously described (Peng et al., 2012). Briefly, hippocampi were dissected from E16 pups, dissociated by papain digestion, and plated on coated glass coverslips. Neurons were cultured in Neurobasal media supplemented with B27. All animal procedures conformed to the guidelines approved by the Washington University Animal Studies Committee.

Experimental design

Request a detailed protocol

All live imaging measurements were replicated in more than 100 boutons derived from 57 to 90 coverslips from 11 independent cultures (see Table 1 for individual experiments). Most experiments were carried out in an unblended manner and no specific randomization strategy was used. Statistical computations were not performed to determine the optimal sample size for experiments.

Lentiviral infection

Request a detailed protocol

VGlut1-pHluorin was generously provided by Drs Robert Edwards and Susan Voglmaier (University of California San Francisco) (Voglmaier et al., 2006). Lentiviral vectors were generated by the Viral Vectors Core at Washington University. Hippocampal neuronal cultures were infected at DIV3 as previously described (Maschi and Klyachko, 2017).

Fluorescence microscopy

Request a detailed protocol

All experiments were conducted at 37°C within a whole-microscope incubator (In Vivo Scientific) at DIV16–19. Neurons were perfused with bath solution (125 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM HEPES, 15 mM glucose, 50 μM DL-AP5, 10 μM CNQX adjusted to pH 7.4). Asynchronous release events were recorded using the same solutions, except that 3 mM Sr2+ and 0 mM CaCl2 were used in the bath. Fluorescence was excited with a Lambda XL lamp (Sutter Instrument) through a 100 × 1.45 NA oil-immersion objective and captured with a cooled CMOS camera (Hamamatsu). With this configuration, the effective pixel size was 65 nm. The focal plane was continuously monitored, and focal drift was automatically adjusted with ~10 nm accuracy by an automated feedback focus control system (Ludl Electronics). Field stimulation was performed by using a pair of platinum electrodes and controlled by the software via Master-9 stimulus generator (AMPI). Images were acquired using an acquisition time of 40 ms, one 45 ms before stimulation and one coincidently with stimulation (0 ms delay).

Large-area scanning electron microscopy (LaSEM)

Request a detailed protocol

The LASEM methods and data used were published previously (Maschi et al., 2018). Briefly, cells were grown on 12 mm glass coverslips, were aspirated of media and were fixed in a solution containing 2.5% glutaraldehyde and 2% paraformaldehyde in 0.15 M cacodylate buffer with 2 mM CaCl2 (pH 7.4) that had been warmed to 37 °C for one hour. The samples were then stained according the methods described by Deerinck et al. (2010). Large areas (~330 × 330 µm) were then imaged at high resolution in a FE-SEM (Zeiss Merlin, Oberkochen, Germany) using the ATLAS (Fibics, Ottowa, Canada) scan engine to tile large regions of interest. High-resolution tiles were captured at 16,384 × 16,384 pixels at 5 nm/pixel with a 5 µs dwell time and line average of 2. The SEM was operated at 8 KeV and 900 pA using the solid-state backscatter detector. Tiles were aligned and exported using ATLAS 5.

Image and data analysis

Event detection and localization

Request a detailed protocol

The fusion event localization at subpixel resolution was performed as previously described (Maschi and Klyachko, 2017) using Matlab and the uTrack software package, which was kindly made available by Dr Gaudenz Danuser's lab (Aguet et al., 2013; Jaqaman et al., 2008). The input parameters for the PSF were determined using stationary green fluorescent 40 nm beads.

Localization precision was determined directly from the least-squares Gaussian fits of individual events as described in Thomann et al. (2002) and Thomann et al. (2003) using in-built functions in uTrack software (Aguet et al., 2013; Jaqaman et al., 2008). Spatial constraints of the vesicle lumen imply that only a few VGlut1-pHluorin molecules can be located within individual vesicles. Given our observations that the fluorescence signal evoked by vesicle fusion did not disperse significantly during our acquisition time of 40 ms, the small number of VGlut1-pHluorin molecules per vesicle and their lateral movement upon fusion, if present, do not strongly affect localization precision at the time at which our measurements are made.

Localization of resolved MVR events (Figures 1, 3 and 4) was performed using a mixture-model multi-Gaussian fit as described in Thomann et al. (2002) and Thomann et al. (2003) using in-built functions in uTrack software (Aguet et al., 2013; Jaqaman et al., 2008).

Unresolved MVR events (Figure 2) were identified on the basis of the event amplitude. The single event amplitude and its variability were determined for each bouton individually. Photobleaching was accounted for by fitting the event intensity changes over time. The threshold for MVR event detection was set at two standard deviations above the mean single event amplitude determined individually for each bouton. Localization of unresolved MVR events was determined using an asymmetrical Gaussian model fit that was based on the minimization of the residuals.

Definition of release sites

Request a detailed protocol

Release sites were defined using a hierarchical clustering algorithm based on built-in functions in Matlab as described previously (Maschi and Klyachko, 2017; Maschi et al., 2018; Wang et al., 2016). We have previously compared the results of this clustering analysis, obtained with the experimentally observed distribution of fusion events, with the the same number of simulated events distributed randomly across the same AZs (Maschi and Klyachko, 2017). We found that randomly distributed release events result in a very different pattern of clustering than the experimentally observed events, and do not reproduce the observed features of real release event clusters. The observed clusters thus do not arise from a random distribution of release events, but rather represent a set of defined and repeatedly reused release sites within the AZs.

Release site release probability

Request a detailed protocol

The release probability of individual release sites was calculated based on the number of release events detected per release site, divided by the duration of the observation period. UVR and MVR events were counted equivalently in this analysis, with each of the two release events comprising an MVR counted independently as a single release event at the two release sites that harbored them.

AZ dimensions and center

Request a detailed protocol

The AZ size was approximated on the basis of the convex hull encompassing all vesicle fusion events in a given bouton. This measurement is in a close agreement with the ultrastructural measurements of AZ dimensions (Maschi and Klyachko, 2017). AZ center was defined as the mean position of all fusion events in a given bouton.

Event proximity analysis

Request a detailed protocol

To determine the spatial overlap of MVR and UVR events, a proximity analysis was performed in which overlap was defined as having at least one UVR event occurring within 25 nm of an MVR event during the observation period.

Synapse identification and analysis of vesicle diameter in LaSEM data

Request a detailed protocol

Three characteristic features were used for synapse identification: the presence of a synaptic vesicle cluster, the postsynaptic density, and the uniform gap between pre- and postsynaptic membranes. Distance to the AZ center was determined from the projection of the vesicle center position to the AZ plane.

Fit regression models

Request a detailed protocol

Nonlinear and linear fit regression models were generated using built-in functions in Matlab.

Data inclusion and exclusion criteria

Request a detailed protocol

A minimum of five detected release events per bouton was required for all analyses.

Statistical analysis

Request a detailed protocol

Statistical analyses were performed in Matlab. Statistical significance was determined using two-sample two-tailed t-test, paired t-test, Kolmogorov-Smirnov (K-S) test, one-way analysis of covariance (ANOCOVA) or chi-square test, where appropriate. The number of experiments reported reflects the number of different cell cultures tested. The value of N is provided in the corresponding figure legends and in Table 1. The statistical tests used to measure significance are indicated in each figure legend along with the corresponding significance level (p value). Data are reported as mean ± SEM and p<0.05 was considered statistically significant. Analysis of the samples was not blinded to condition. Randomization and sample size determination strategies are not applicable to this study and were not performed.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
    NCMIR methods for 3D EM: a new protocol for preparation of biological specimens for serial block-face SEM
    1. TJ Deerinck
    2. E Bushong
    3. A Thor
    4. MH Ellisman
    (2010)
    Microscopy.
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50

Decision letter

  1. Karel Svoboda
    Reviewing Editor; Janelia Research Campus, Howard Hughes Medical Institute, United States
  2. Ronald L Calabrese
    Senior Editor; Emory University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This paper presents a clever high-resolution imaging approach to measure spatial aspects of vesicles release at central synapses. Using a deconvolution approach the authors can distinguish single-vesicle and multi-vesicle release events. The key finding is that release events are inhomogeneously distributed across the active zone. Multi-vesicle events share release sites with single-vesicle events, and are more frequent at the center of the AZ, where the release probability is highest. This study provides an unprecedented view of spatial aspects of synaptic vesicles release.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Spatiotemporal dynamics of multi-vesicular release is determined by heterogeneity of release sites in central synapses" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions, we regret to inform you that your work will not be considered further for publication in eLife.

The reviewers and editor were excited by the approach and some of the data. However, the study is at a relatively preliminary stage and it is unlikely that the additional data and analysis required can be accomplished in two months. The eLife policy is to reject a paper under these conditions.

The mechanisms of release have often been studied with electrophysiological methods. However, these methods average over synapses, washing out interesting fluctuations that are informative about the function of single synapses. Over the last two decades a variety of postsynaptic and presynaptic imaging methods have been developed to look at synaptic transmission at single synapses. Some methods have high SNR and can be used for studying synapses in intact tissue (i.e. calcium imaging), whereas others (e.g. synaptopHluorin) promise better time resolution and spatial resolution, but are typically used in reduced preparations. The current study uses imaging synaptopHluorin to track single release events at individual synapses. The paper reports multi-vesicular release, with the dominant release site closer to the center of the active zone (AZ). Release probability was highest at the center of the AZ and dropped off towards the edge. This gradient, and spatial features of MVR, were similarly tightened by buffering intracellular calcium.

This is a nice imaging paper on vesicle release at single synapses. All reviewers agreed that the study is potentially very interesting, but don't find the data set and quantification convincing in their current state.

Essential revisions:

1) The paper has to be reframed. The Introduction might have been appropriate in the late 1990's, when UVR ("one bouton, one spike, one quantum") was a reasonable hypothesis based on prior work. Obviously, UVR was interesting in principle because it would pose serious biophysical riddles. If one vesicle releases, what signal tells other docked vesicles not to release? This signal would have to travel over microsecond timescales. The UVR came from work on the NMJ (misunderstood in this context) and, famously, the Mauthner cell. The original data has been reanalyzed and the statistical analysis has been questioned (some would say debunked, see Ninio, 2007, J. Neurophysiol.).

At central synapses support for UVR was always very spotty. In contrast, imaging studies with single synapse resolution tended to show MVR, as did many excellent electrophysiological studies using more indirect analyses. Yes, individual synapses often release single quanta, but this is because their overall release probability is often low. Under conditions of elevated release probability quanta are released based on the binomial model (e.g. Oertner et al., 2002; reviewed in Rudolph et al., 2015). It seems there is no evidence for UVR left and the whole thing is best forgotten except as a lesson in science sociology.

This means that MVR and UVR are likely the same process (operating independently at release sites/docked vesicles with different individual release probs). The authors have demonstrated that sites that host MVR could also harbor a UVR event later. This interchangeability actually suggests that they may be similarly regulated. In fact, none of the reviewers see support for a difference in spatial or temporal profile between MVR and UVR. Please analyze the distributions of time intervals for a UVR occurring at sites of an MVR, an MVR occurring at sites of an MVR, a UVR occurring at sites of a UVR, and a UVR occurring at sites of a UVR to see any difference. In Figure 1I the authors did not classify UVR into distal UVR and proximal UVR, as they did for MVR. This may have biased the comparison.

2) The authors concluded that they observed desynchronization of MVR events based on the ~8-10% difference in fluorescence intensity profiles of doublet events. However, from Figure 3A, the difference in the peak fluorescence intensities of two events is surely larger than 25%. Given 40 ms exposure time, it is hard to believe that ~4 ms of desynchronization generated such a large difference in fluorescence intensities. Moreover, if it is truly due to desynchronization, we would expect that the fluorescence intensity profiles of two events reversed over the next time point. Please explain.

3) Related to above. The second main finding is a desynchronization of MVR, with the more central release site leading and the peripheral site lagging by up to 4 ms. This evidence is indirect (via amplitude) and somewhat unconvincing (see above). How do we know that the amplitude difference is due to timing and not due to other factors (cleft pH, out-of-focus release, vesicle size etc.)?

As discussed in the 2017 Neuron paper, AZs may be tilted relative to the imaging plane, which could also affect the measured amplitude of distal events. Given these alternative explanations, it would be reassuring to have confirmatory evidence for delayed release events. A 4 ms delay is huge, twice the typical synaptic delay. Such delayed events, even if rare, should be readily visible in voltage clamp recordings. This is all a bit puzzling and needs to be explained better.

4) The authors argue that EGTA "affected the MVR event de-synchronization" and abolishes "the preferential localization of the earlier event in the MVR pair closer to the AZ center". In both cases, a significant difference under control conditions (Figure 3E, Figure 3C) was found 'non-significant' in EGTA (Figure 4B, Figure 4D). However, the number of analyzed synapses was more than twice as high under control conditions, which could account for the higher significance found in the t-tests. To compare t-tests of identical power, it may be necessary to conduct additional EGTA experiments. More importantly, the correct comparison is testing for changes in the difference between smaller and larger with and without EGTA. As is, the analysis is statistically flawed. This likely requires more experiments and analysis.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for choosing to send your work entitled "Spatiotemporal dynamics of multi-vesicular release is determined by heterogeneity of release sites in central synapses" for consideration at eLife.

Specifically, while the reviewers and editors found the manuscript much improved, they felt that a complete rewrite would still be necessary, and therefore this version still falls short of what is needed. One reviewer stated:

“In my opinion, the results strongly suggest that active zones close to the center of a synapse have a higher release probability than more distal AZs. This is an important and interesting finding that should be simple to understand. Unfortunately, it is (still) obfuscated by their MVR/UVR site classification, which they now partially take back ("our results cannot be interpreted to indicate that there are some specialized release sites that only support MVR or UVR"). But the confusion already starts at Figure 1A, where a release site is labeled in red (for MVR) that has in fact released only a single vesicle. So while it makes sense to talk about UVR and MVR events, it is very confusing to talk about UVR and MVR sites, only to state at the end that they are probably not fundamentally different. (Since typically only a single MVR event was observed per bouton, meaningful single bouton statistics is not possible). What is needed is a complete rewrite of the manuscript, including a change in nomenclature and analysis strategy, not just the addition of some text blocks.”

https://doi.org/10.7554/eLife.55210.sa1

Author response

Essential revisions:

1) The paper has to be reframed. The Introduction might have been appropriate in the late 1990's, when UVR ("one bouton, one spike, one quantum") was a reasonable hypothesis based on prior work. Obviously, UVR was interesting in principle because it would pose serious biophysical riddles. If one vesicle releases, what signal tells other docked vesicles not to release? This signal would have to travel over microsecond timescales. The UVR came from work on the NMJ (misunderstood in this context) and, famously, the Mauthner cell. The original data has been reanalyzed and the statistical analysis has been questioned (some would say debunked, see Ninio, 2007, J. Neurophysiol.).

At central synapses support for UVR was always very spotty. In contrast, imaging studies with single synapse resolution tended to show MVR, as did many excellent electrophysiological studies using more indirect analyses. Yes, individual synapses often release single quanta, but this is because their overall release probability is often low. Under conditions of elevated release probability quanta are released based on the binomial model (e.g. Oertner et al., 2002; reviewed in Rudolph et al., 2015). It seems there is no evidence for UVR left and the whole thing is best forgotten except as a lesson in science sociology.

We completely agree, and have extensively reframed the manuscript to reflect this point. We now state in the Introduction that:

“Although initially it was hypothesized that, at most, only a single vesicle can be released from a given synapse with each action potential (i.e. uni-vesicular release (UVR)), it is now widely accepted that two or more vesicles can fuse simultaneously in response to a single action potential in the same synaptic bouton, leading to the notion of multi-vesicular release (MVR) (Rudolph et al., 2015). […] Despite its prevalence, spatiotemporal organization of MVR within the synaptic active zone (AZ) and its regulation are poorly understood (Rudolph et al., 2015)”.

This means that MVR and UVR are likely the same process (operating independently at release sites/docked vesicles with different individual release probs). The authors have demonstrated that sites that host MVR could also harbor a UVR event later. This interchangeability actually suggests that they may be similarly regulated. In fact, none of the reviewers see support for a difference in spatial or temporal profile between MVR and UVR. Please analyze the distributions of time intervals for a UVR occurring at sites of an MVR, an MVR occurring at sites of an MVR, a UVR occurring at sites of a UVR, and a UVR occurring at sites of a UVR to see any difference. In Figure 1I the authors did not classify UVR into distal UVR and proximal UVR, as they did for MVR. This may have biased the comparison.

We thank the reviewers for this important point and agree. We note that this concern is caused in a large part by a misunderstanding due to lack of clarity in the original version of our manuscript. We did indeed classify UVR into distal and proximal in Figure 1I, as we did for MVR, but we failed to note this clearly in the figure. We now clarify that this analysis was performed on MVR and UVR events equidistant to the AZ center (Figure 1H). In other words, for each MVR event pair, the proximal MVR site was compared only to the proximal UVR sites, and the distal MVR site was compared only to the distal UVR sites. These results show that the release sites that are more likely to harbour both UVR and MVR are characterized by higher release probability than other equidistant sites.

We also completely agree with the reviewers that our results do not indicate that there are specialized sites of UVR vs. MVR, but rather the same sites can harbor both types of events. In fact, our finding that MVR sites have higher release probability is based on observation that UVR events occur more frequently at the release sites at which MVR events are observed. We have now expanded the description of the results to clarify this point. We also stated in the Results that “our results cannot be interpreted to indicate that there are some specialized release sites that only support MVR or UVR”.

Following the reviewers’ suggestion, we also attempted to compare the distributions of time intervals of UVR and MVR events occurring at sites of UVR or MVR and vice versa. However, we came to realization that this analysis is not possible within our current approach because we cannot determine the distribution of time intervals between MVR events at any given release site. This is because the very low probability of observing MVR events, which is typically only 1-2 per bouton within the 120 sec observation window. As a result, MVR events were almost never observed at the same release site more than once in the same bouton during this time period, thus making analysis of inter-event intervals not feasible. We previously showed that intrinsic slow displacement of synaptic boutons precludes precise analysis of release event localization for much longer time periods, which is what would be required for such analysis.

In summary, we have revised the manuscript to include these points and to clarify that our data does not suggest that there are specialized sites of MVR vs UVR, but rather the same sites support both types of release. We further emphasize that it was not the intent of our study to contrast UVR vs MVR. Rather we believe that the critical finding is the highly heterogeneous properties of release sites within the individual AZs. The sites with higher release probability are located closer to the AZ center and are more likely to harbor both UVR and MVR events. The spatiotemporal features of MVR reflect this heterogeneity of release sites across individual AZs and depend, in part, on the non-uniform landscape of calcium elevation across the AZ following an action potential. Our analyses thus suggest a new level of structural and functional organization of release sites within individual AZs that determines the spatiotemporal dynamics of MVR. We have also revised the Introduction and Discussion of the manuscript to clarify these points.

2) The authors concluded that they observed desynchronization of MVR events based on the ~8-10% difference in fluorescence intensity profiles of doublet events. However, from Figure 3A, the difference in the peak fluorescence intensities of two events is surely larger than 25%. Given 40 ms exposure time, it is hard to believe that ~4 ms of desynchronization generated such a large difference in fluorescence intensities. Moreover, if it is truly due to desynchronization, we would expect that the fluorescence intensity profiles of two events reversed over the next time point. Please explain.

We thank the reviewers for these points and performed extensive additional experiments and analyses to determine origins of amplitude differences within MVR events, and to solidify our interpretation of these differences as evidence for desynchronization of MVR. Specifically:

a) Following the reviewer’s suggestions, we perform additional studies to demonstrate that amplitude differences within MVR events are not due to differences in vesicle size (Figure 3—figure supplement 1D) or changes in cleft pH (Figure 3—figure supplement 1B). We also clarified in the text that due to the configuration of our experiments, the observed differences cannot arise simply from the tilt of AZs relative to the imaging plane. This is also evident in the observation that the average amplitude of UVR events is indistinguishable at the AZ center vs periphery (Figure 3—figure supplement 1B, insert). These results are discussed in more detail in the response to point #3 below.

b) We performed extensive additional experiments to confirm that the observed scaling of the amplitude difference within the MVR events as a function of the intra-event separation is eliminated by EGTA-AM (and see response to point #4 below for more details).

c) We performed additional analyses to examine how much the uncertainty in determination of fusion event amplitude contributes to the observed amplitude differences within the MVR event pairs. We found that this uncertainty is comparable to the amplitude differences of the closely-spaced MVR events (~10%), but cannot account for the amplitude differences within MVR events with large separation (Figure 3—figure supplement 1A). Moreover, the positive correlation between the spatial separation and amplitude difference of the two events comprising MVR cannot be explained by measurement uncertainty.

d) We clarified in the text that desynchronization of MVR events is a well established phenomenon, with several studies reporting MVR event desynchronization in the range of ~1-5ms in both excitatory and inhibitory central synapses (Auger et al., 1998; Auger and Marty, 2000; Crowley et al., 2007; Malagon et al., 2016; Rudolph et al., 2011). We also emphasized that our initial estimate for the maximal time difference within our double events of ~4ms is an overestimate because a component of the amplitude differences arise from measurement uncertainty. We provided a corrected upper bound estimate on the MVR desynchronization to be ~2ms, if the measurement uncertainty is factored in. These values are well within the range of 1-5ms reported previously for MVR event desynchronization in the above studies.

e) The reviewers pointed out that the fluorescence intensity profiles of two events are expected to reverse over the next time point if the differences are truly due to desynchronization. Indeed, we agree that this would be the case, but only if the dwell time of the fluorescent signal generated by the individual events is exactly the same for both events. However, unlike the precise initiation of fusion, the dwell time of individual events is highly variable, spanning from ultrafast (~150-250 ms), to fast (~5-12 s) to ultraslow (>20 s) (Chanaday and Kavalali, 2018). A few initial milliseconds of desynchronization do not have any significant impact over such variability. Therefore there is no correspondence between the timing of the event start in one frame and the duration of observation of the same event in any subsequent frame.

f) The example shown in Figure 3A has a ~25% difference in amplitude within the MVR event and was chosen to highlight the differences in amplitude, which can be visually obscured by pixelation and noise. This example is reasonably representative of the differences observed for the far apart MVR events, but we can replace it with a different example if reviewers suggest it.

3) Related to above. The second main finding is a desynchronization of MVR, with the more central release site leading and the peripheral site lagging by up to 4 ms. This evidence is indirect (via amplitude) and somewhat unconvincing (see above). How do we know that the amplitude difference is due to timing and not due to other factors (cleft pH, out-of-focus release, vesicle size etc.)?

As discussed in the 2017 Neuron paper, AZs may be tilted relative to the imaging plane, which could also affect the measured amplitude of distal events. Given these alternative explanations, it would be reassuring to have confirmatory evidence for delayed release events. A 4 ms delay is huge, twice the typical synaptic delay.

We thank the reviewers for this important point. As mentioned in the response to point #2 above, we performed extensive additional experiments to address the reviewer’s concerns. We provide more details of these experiments here:

a) Following the reviewers’ suggestions, we examined if amplitude differences within MVR events can arise from differences in vesicle size. We determine the vesicle size from the EM microscopy of our synapses in culture as a function of distance to the AZ center (Figure 3—figure supplement 1C, D). Vesicle size did not exhibit significant changes with distance to the AZ center (Figure 3—figure supplement 1D). Thus the amplitude differences within MVR events are not due to differences in vesicle size.

b) We also determined if amplitude differences within MVR events can arise from differences in the cleft pH at different locales of the AZ. If this is the case, changes in the cleft pH should similarly affect amplitudes of UVR and MVR events. Yet we did not detect any significant changes in the intensity of UVR events throughout the whole AZ (Figure 3—figure supplement 1B). Thus changes in cleft pH are unlikely to explain the differences in MVR event amplitude.

c) We clarified in the text that in our recordings dozens of synapses are positioned in random orientation and are recorded simultaneously. Some boutons have the center of the AZ in focus and others have the periphery, and yet others are positioned in all other possible configurations inbetween (schematic in Author response image 1). Thus there is no any systematic bias towards any one possible configuration or tilt of the AZ relative to the imaging plane that can lead to systematically larger fusion event amplitude at the AZ center vs. periphery. Indeed, this notion is highlighted by the fact that the average UVR event amplitude is invariable in our measurements, independently of the distance to the AZ center (Figure 3—figure supplement 1B, insert). Thus the observed differences in MVR event amplitude are unlikely to result from tilt of AZ relative to the imaging plane.

Author response image 1
Schematic of the wide range of possible AZ positions relative to the focal plane.

Such delayed events, even if rare, should be readily visible in voltage clamp recordings. This is all a bit puzzling and needs to be explained better.

We completely agree and would like to emphasize that desynchronization of MVR events have indeed been reported previously in a number of voltage clamp studies in both excitatory and inhibitory simple synapses:

a) Widespread desynchronization of MVR with delays in the range of 1-5ms within the event pairs was found at single synapses formed by parallel fiber to interneuron connections in cerebellar slices (Figure 6B in Malagon et al., 2016), with examples of up to 5-8ms delays within MVR events (Figure 2B, 3A, 5A in Malagon et al., 2016).

b) Desynchronization of MVR events in the range of 1-10ms was reported at single granule cell to stellate cell synapses in rat cerebellar slices (Figure 5B,E in Crowley et al., 2007).

c) Jitter in the timing within MVR events on the millisecond timescale was found at climbing fiber to Purkinje cell synapses under similar stimulation conditions to ours (Rudolph et al., 2011).

d) Widespread desynchronization of MVR events in the range of 1.5-3.5ms was reported at single synaptic connections between cerebellar stellate and basket cells (Auger et al., 1998; Auger and Marty, 2000).

To address the reviewer’s concern, we have cited these studies and clarified this point in the Results of the revised version. We further emphasize that the novel observation in our study is not in the fact that MVR can be desynchronized, but in revealing a pattern of spatial organization within these MVR events relative to the AZ center, that is correlated with the gradient of release site properties and is regulated by calcium elevation.

4) The authors argue that EGTA "affected the MVR event de-synchronization" and abolishes "the preferential localization of the earlier event in the MVR pair closer to the AZ center". In both cases, a significant difference under control conditions (Figure 3E, Figure 3C) was found 'non-significant' in EGTA (Figure 4B, Figure 4D). However, the number of analyzed synapses was more than twice as high under control conditions, which could account for the higher significance found in the t-tests. To compare t-tests of identical power, it may be necessary to conduct additional EGTA experiments. More importantly, the correct comparison is testing for changes in the difference between smaller and larger with and without EGTA. As is, the analysis is statistically flawed. This likely requires more experiments and analysis.

We thank the reviewers for this important point and absolutely agree. We performed the suggested additional experiments and have revised our statistical analyses as suggested. Specifically, the number of analyzed synapses is now 225 and 245 from 11 independent cultures each, for measurements with and without EGTA, respectively. We also corrected our initial statistical analyses and tested for changes in the slope of the dependence of the amplitude difference between smaller and larger events with and without EGTA, as reviewers suggested. We found that EGTA indeed significantly reduced it (Figure 4B). Similarly, we still observed that proportion of MVR events that had shorter intra-event separation increased significantly in the presence of EGTA (Figure 4C). Accordingly, the average distances from both events in the MVR pair to the AZ center were significantly shortened in the presence of EGTA (Figure 4—figure supplement 1A). Thus calcium buffering causes MVR events to occur at shorter separation distances and more proximal to the AZ center. However, the larger event in the MVR pair was still localized closer to the AZ center in the presence of EGTA than the smaller one in the pair (Figure 4D). This is consistent with the observation that release sites with higher release probability are still localized closer to the AZ center in the presence of EGTA (Figure 4E) and that preferential utilization of more central release sites was exacerbated in the presence of EGTA (Figure 4E). We have revised Figure 4 and the description of these results accordingly.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Specifically, while the reviewers and editors found the manuscript much improved, they felt that a complete rewrite would still be necessary, and therefore this version still falls short of what is needed. One reviewer stated:

“In my opinion, the results strongly suggest that active zones close to the center of a synapse have a higher release probability than more distal AZs. This is an important and interesting finding that should be simple to understand. Unfortunately, it is (still) obfuscated by their MVR/UVR site classification, which they now partially take back ("our results cannot be interpreted to indicate that there are some specialized release sites that only support MVR or UVR"). But the confusion already starts at Figure 1A, where a release site is labeled in red (for MVR) that has in fact released only a single vesicle. So while it makes sense to talk about UVR and MVR events, it is very confusing to talk about UVR and MVR sites, only to state at the end that they are probably not fundamentally different. (Since typically only a single MVR event was observed per bouton, meaningful single bouton statistics is not possible). What is needed is a complete rewrite of the manuscript, including a change in nomenclature and analysis strategy, not just the addition of some text blocks.”

We thank the reviewer for this important point and fully agree. We made major revisions to the text and figures of the manuscript to address the reviewer’s criticism as follows:

1) We would like to clarify that the reviewer’s concern was in a large part caused by the lack of clarity in our description of the results, but not in the actual analyses themselves. Specifically, we did not classify sites into UVR or MVR sites in any of the analyses presented in the manuscript, but our description made an unintended impression that this was the case. The vast majority of analyses, except for Figures 1G-I and Figure 4E-G, examined individual UVR and MVR release events only. In the remaining analyses, we compared release probability of release sites that did or did not have an MVR event during the observation period. These analyses were based on counting all release events at each release site and did not distinguish UVR vs. MVR events in the calculation of Pr-site (see below for more details). We have improved the schematics of these measurements in the figures and the presentation of these results in the text to clarify this point.

2) We changed the nomenclature used in the manuscript as the reviewer suggested. Only the terms “UVR events” and “MVR events” are now used throughout the manuscript. We completely remove any mention of “UVR release sites” or “MVR release sites”. We now use the same generic term “release site” for both forms of release throughout the manuscript.

3) We improved the presentation of UVR and MVR event measurements in Figure 1A to clarify that all release sites were considered in the same way, no matter whether the release was a UVR event or an MVR event, as the reviewer suggested.

4) We improved the schematics of the Pr-site measurements in Figure 1H and 4G, and the corresponding legends and the bar graphs in the same panels, to clarify that release sites with or without an MVR event were compared. We further clarify in the Materials and methods that calculation of Pr-site did not distinguish UVR vs MVR events. Each of the two release events comprising an MVR pair was counted independently as a single release event at the two release sites that harbored them.

5) Following the reviewer’s suggestion we extensively revised the Summary, Introduction, Results and Discussion of the manuscript to highlight the critical finding of a gradient of release site release probability from the AZ center to periphery.

6) We extensively revised the manuscript to improve clarity of result presentation and discussion. In summary, following the reviewer’s criticism, we have extensively revised the manuscript to clarify that our study distinguishes only UVR events and MVR events, but we do not analyze or distinguish release sites as “UVR sites” or “MVR sites”.

References:

Auger, C., Kondo, S., and Marty, A. (1998). Multivesicular release at single functional synaptic sites in cerebellar stellate and basket cells. J Neurosci 18, 4532-4547.

Auger, C., and Marty, A. (2000). Quantal currents at single-site central synapses. J Physiol 526 Pt 1, 3- 11.

Chanaday, N.L., and Kavalali, E.T. (2018). Optical detection of three modes of endocytosis at hippocampal synapses. eLife 7.

Crowley, J.J., Carter, A.G., and Regehr, W.G. (2007). Fast vesicle replenishment and rapid recovery from desensitization at a single synaptic release site. J Neurosci 27, 5448-5460.

Foster, K.A., Crowley, J.J., and Regehr, W.G. (2005). The influence of multivesicular release and postsynaptic receptor saturation on transmission at granule cell to Purkinje cell synapses. J Neurosci 25, 11655-11665.

Malagon, G., Miki, T., Llano, I., Neher, E., and Marty, A. (2016). Counting Vesicular Release Events Reveals Binomial Release Statistics at Single Glutamatergic Synapses. J Neurosci 36, 4010-4025.

Oertner, T.G., Sabatini, B.L., Nimchinsky, E.A., and Svoboda, K. (2002). Facilitation at single synapses probed with optical quantal analysis. Nat Neurosci 5, 657-664.

Rudolph, S., Overstreet-Wadiche, L., and Wadiche, J.I. (2011). Desynchronization of multivesicular release enhances Purkinje cell output. Neuron 70, 991-1004.

Wadiche, J.I., and Jahr, C.E. (2001). Multivesicular release at climbing fiber-Purkinje cell synapses. Neuron 32, 301-313.

https://doi.org/10.7554/eLife.55210.sa2

Article and author information

Author details

  1. Dario Maschi

    Department of Cell Biology and Physiology, Washington University School of Medicine, St. Louis, United States
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  2. Vitaly A Klyachko

    Department of Cell Biology and Physiology, Washington University School of Medicine, St. Louis, United States
    Contribution
    Conceptualization, Resources, Software, Supervision, Funding acquisition, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    klyachko@wustl.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3449-243X

Funding

National Institute of Neurological Disorders and Stroke (NS111596)

  • Vitaly Klyachko

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported in part by the R35 NS111596 grant to VAK from NINDS. We thank Drs Meyer Jackson and Gabrielle Edgerton for helpful comments on the manuscript.

Ethics

Animal experimentation: All animal procedures conformed to the guidelines approved by the Washington University Animal Studies Committee (protocol approval # 20170233).

Senior Editor

  1. Ronald L Calabrese, Emory University, United States

Reviewing Editor

  1. Karel Svoboda, Janelia Research Campus, Howard Hughes Medical Institute, United States

Publication history

  1. Received: January 16, 2020
  2. Accepted: January 28, 2020
  3. Accepted Manuscript published: February 6, 2020 (version 1)
  4. Version of Record published: March 6, 2020 (version 2)

Copyright

© 2020, Maschi and Klyachko

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,612
    Page views
  • 284
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Computational and Systems Biology
    2. Neuroscience
    Chen Chen et al.
    Research Article

    While animals track or search for targets, sensory organs make small unexplained movements on top of the primary task-related motions. While multiple theories for these movements exist—in that they support infotaxis, gain adaptation, spectral whitening, and high-pass filtering—predicted trajectories show poor fit to measured trajectories. We propose a new theory for these movements called energy-constrained proportional betting, where the probability of moving to a location is proportional to an expectation of how informative it will be balanced against the movement’s predicted energetic cost. Trajectories generated in this way show good agreement with measured trajectories of fish tracking an object using electrosense, a mammal and an insect localizing an odor source, and a moth tracking a flower using vision. Our theory unifies the metabolic cost of motion with information theory. It predicts sense organ movements in animals and can prescribe sensor motion for robots to enhance performance.

    1. Ecology
    2. Neuroscience
    Felix JH Hol et al.
    Tools and Resources

    Female mosquitoes need a blood meal to reproduce, and in obtaining this essential nutrient they transmit deadly pathogens. Although crucial for the spread of mosquito-borne diseases, blood feeding remains poorly understood due to technological limitations. Indeed, studies often expose human subjects to assess biting behavior. Here, we present the biteOscope, a device that attracts mosquitoes to a host mimic which they bite to obtain an artificial blood meal. The host mimic is transparent, allowing high-resolution imaging of the feeding mosquito. Using machine learning we extract detailed behavioral statistics describing the locomotion, pose, biting, and feeding dynamics of Aedes aegypti, Aedes albopictus, Anopheles stephensi, and Anopheles coluzzii. In addition to characterizing behavioral patterns, we discover that the common insect repellent DEET repels Anopheles coluzzii upon contact with their legs. The biteOscope provides a new perspective on mosquito blood feeding, enabling the high-throughput quantitative characterization of this lethal behavior.