The ubiquitous nature of multivesicularreleaseStephanie Rudolph1, Ming-Chi Tsai1, Henrique von Gersdorff2, andJacques I. Wadiche1

1 Department of Neurobiology and Evelyn McKnight Brain Institute, University of Alabama at Birmingham, Birmingham,

AL 35294, USA2 The Vollum Institute, Oregon Health and Science University, Portland, OR 97239, USA

Feature Review

Glossary

Active zone: ultrastructurally and functionally specialized area at the pre-

synaptic terminal where readily-releasable vesicles fuse (see also Release site).

Coefficient of variation (CV): standard deviation s divided by the mean (e.g., s

of current amplitude fluctuations and mean current). CV is a normalized

measure of relative variability within a distribution (e.g., numerous trials of an

evoked synaptic current).

Depletion: temporal unavailability of releasable vesicles as a result of foregoing

release. Vesicular depletion can cause a decrease in synaptic strength.

Desensitization: receptors bound by transmitter but in a non-conducting state

can contribute to failure of receptor opening following repeated or prolonged

exposure to neurotransmitter. Desensitization can cause a decrease in synaptic

strength.

Desynchronization: the temporal jitter of vesicle fusion between synaptic

release sites (intersite asynchrony) or within a release site (intrasite

asynchrony).

EPSC/EPSP: excitatory postsynaptic current/potential.

Fast-off/low-affinity antagonist: rapidly-dissociating competitive antagonist.

IPSC/IPSP: inhibitory postsynaptic current/potential.

Multivesicular release (MVR): the release of >1 vesicles per active zone in

response to a single presynaptic action potential. Transmitter released by

multiple vesicles interacts with a common population of postsynaptic

receptors.

Quantal theory: postulates that the spontaneously occurring miniature currents/

potentials represent the basic element of the evoked synaptic current/potential.

Receptor occupancy: the percentage of receptors bound by released transmit-

ter. 100% occupancy refers to receptor saturation.

Release site: site within the active zone where a vesicle undergoes fusion. It

remains controversial whether release sites are clearly defined molecularly or

whether release can take place at any location in the active zone. For the

purpose of this review, we define the active zone as a structure that provides

release sites, and a single active zone can have multiple release sites.

Slow-off/high-affinity antagonist: slowly-dissociating competitive antagonist.

Spillover: refers to transmitter escaping the synaptic cleft that activates

extrasynaptic receptors located on the releasing cell or receptors on adjacent

cells.

Synaptic strength: defines the average response amplitude at a synaptic

contact in response to an action potential. The total number of release sites,

readily releasable vesicles, Pr, and postsynaptic factors such as single receptor

conductance and receptor density, determine synaptic strength.

‘Simplicity is prerequisite for reliability’ (E.W. Dijkstra [1])Presynaptic action potentials trigger the fusion of vesiclesto release neurotransmitter onto postsynaptic neurons.Each release site was originally thought to liberate atmost one vesicle per action potential in a probabilisticfashion, rendering synaptic transmission unreliable. How-ever, the simultaneous release of several vesicles, ormultivesicular release (MVR), represents a simple mecha-nism to overcome the intrinsic unreliability of synaptictransmission. MVR was initially identified at specializedsynapses but is now known to be common throughoutthe brain. MVR determines the temporal and spatial dis-persion of transmitter, controls the extent of receptoractivation, and contributes to adapting synaptic strengthduring plasticity and neuromodulation. MVR consequent-ly represents a widespread mechanism that extends thedynamic range of synaptic processing.

MVR occurs throughout the brainFast chemical communication between neurons occurs atultrastructurally defined synaptic junctions through therelease of neurotransmitters. At each presynaptic releasesite, neurotransmitter-filled vesicles are docked on theplasma membrane ready to fuse upon the arrival of anaction potential (Figure 1Ai,Bi). Vesicle fusion and neuro-transmitter release then result in receptor activation. Thestrength of the synaptic signal at the postsynaptic mem-brane is determined by the number of release sites (N; seeGlossary), the probability that a vesicle is released (Pr),and the amplitude of the postsynaptic response elicited bythe content of each synaptic vesicle (q) [2]. Seminal workcorrelating morphology with physiology led to the idea thatan action potential allows the stochastic fusion of, at most,one vesicle at each release site (univesicular release; UVR)[3–8]. Under this model the maximum number of vesiclesreleased corresponds to the number of anatomically de-fined release sites, N [9]. The tenet of UVR was thought toapply to most synapses (see [10] for review); however,several observations are inconsistent with the ‘one site,one vesicle’ hypothesis, including considerable amplitude

0166-2236/

� 2015 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tins.2015.05.008

Corresponding authors: Rudolph, S. (stephanie_rudolph@hms.harvard.edu);Wadiche, J.I. (jwadiche@uab.edu).

428 Trends in Neurosciences, July 2015, Vol. 38, No. 7

fluctuations of evoked synaptic responses [11] and hightransmitter concentration in the synaptic cleft [12]. Toreconcile these findings, a more flexible hypothesis pro-poses that multiple vesicles can be released per active zonewith each action potential, defined as MVR. Indeed, stud-ies have established that MVR occurs at many inhibitoryand excitatory synapses throughout the brain, including

Univesicular release (UVR): the release of �1 vesicle per active zone in

response to a single presynaptic action potential.

Variance-mean analysis (VMA): a statistical method to estimate synaptic

quantal parameters N, Pr, and q.

Vesicular release probability (Pr): the likelihood that a vesicle fuses upon arrival

of an action potential at any release site within an active zone.

(Ai)

(Aii)

(C)

(Bii)

(Bi)

4 mM

1 mM200 µs

5 mM 2 mM

10 ms

UVR MVR dMVR

[transmi�er]

EPSC

10 ms500 pA

TRENDS in Neurosciences

Figure 1. Multiquantal release at single synapses. (Ai,ii) Electron micrograph and corresponding cartoon of a hippocampal synapse imaged following rapid freezing.

Arrows indicate the presence of three docked vesicles at the active zone. (Bi) Electron micrograph of two synaptic vesicles fusing with the within an active zone

simultaneously. (Bii) Cartoon representation of (Bi) illustrating neurotransmitter released from MVR interacts with a common pool of postsynaptic receptors (green and

orange). Adapted from [22]. (C) Overview of different release modalities and their consequences for postsynaptic currents. UVR causes a low-concentration, short-lived

transmitter transient in the synaptic cleft that results in a small EPSC (left). Synchronous fusion of several vesicles, MVR, results in elevated synaptic transmitter

concentration and a large EPSC (middle). Desynchronized MVR prolongs but reduces the transmitter concentration transient, resulting in a smaller and slowed EPSC (right).

Adapted from [39]. Abbreviations: EPSC, excitatory postsynaptic current; MVR, multivesicular release; UVR, univesicular release.

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

hippocampus, cerebral cortex, cerebellum, hypothalamus,and at sensory synapses [12–21]. Due to the sub-millisecondspeed of vesicle fusion direct electron microscopic evidence ofmultiple vesicle fusions (omega figures) at single releasesites has been rare (but see [22,23]) (Figure 1). Nevertheless,functional data from many studies strongly suggests thatMVR is a widespread phenomenon among synapses – moreprevalent than originally assumed.

In this review we delineate the evidence and conse-quences of MVR, including its effects on the time courseand concentration of transmitter, receptor occupancy, anddesensitization. We will examine the experimentalapproaches that have helped to identify MVR, explorehow MVR shapes synapse properties, and argue thatMVR not only promotes synaptic strength and reliabilitybut also enhances the dynamic range of a neuron duringplasticity and neuromodulation [24,25]. Finally, we reviewthe role of MVR in a wide range of physiological contexts,

underlining MVR as an indispensable mechanism for neu-ronal computation.

How MVR shapes synaptic transmissionReceptor occupancy and transmitter concentration

Heterogeneity in synapse size, vesicular transmitter con-tent, transmitter clearance, and receptor affinity contrib-ute to the variability of neurotransmitter-receptoroccupancy levels [12,26–30]. MVR can only increasestrength at individual synapses when receptor occupancyby transmitter released from a single vesicle is sufficientlylow. Indeed, evidence from many synapses suggests thattransmitter from one vesicle is unlikely to fully occupypostsynaptic receptors [11,31,32]. For example, at excit-atory hippocampal synapses, transmitter from multiplevesicles can interact with a common population of postsyn-aptic receptors to increase synaptic strength (Figure 1B),and Pr determines the extent of receptor occupancy

429

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

[12]. These observations are consistent with several stud-ies showing increased receptor occupancy under high Pr

conditions [33–35]. Low receptor occupancy is particularlyevident between cholecystokinin (CCK)-positive interneur-ons and CA3 pyramidal cells where MVR enhances theinhibitory postsynaptic current (IPSC) by several fold[30]. Hence, MVR allows synapses with low receptor occu-pancy to enhance their dynamic range. However, the ex-tent of receptor occupancy varies markedly acrosssynapses. For example, the effect of MVR on IPSC ampli-tude is minimal at synapses between cerebellar molecularinterneurons (MLIs), implying near full occupancy of post-synaptic receptors by the contents of a single vesicle[13]. While such high-occupancy synapses forfeit dynamicrange due to insensitivity to changes in Pr, they operatewith greater fidelity. Together, these examples demon-strate that the degree of receptor occupancy is an impor-tant consideration in assessing the consequences of MVR.

The temporal and spatial spread of transmitter

Evoked transmission across synapses is typically tightlytimed to the presynaptic stimulation [36], suggesting tem-poral coordination across release sites (Figure 1C, left).Nevertheless, MVR has been reported to be either highlycoordinated [16,37] (Figure 1C, center) or desynchronous[38,39] in a manner that alters the neurotransmitterconcentration at postsynaptic receptors (Figure 1C). Numer-ical simulations suggest that desynchronization of MVR candifferentially influence glutamatergic AMPA receptor(AMPAR) and NMDA receptor (NMDAR) activation[40]. Near-synchronous MVR (tens of microseconds) increasesAMPAR activation supra-linearly, while more pronouncedtemporal dispersion of MVR (several milliseconds) can resultin AMPAR desensitization. Conversely, vesicle release syn-chrony has little effect on NMDAR activation. Therefore,activity-dependent desynchronization of MVR might not onlyalter the time course of the synaptic currents but also itsrelative receptor contributions. Although the molecular mech-anisms underlying MVR are currently unknown (Boxes 2 and3), it is tempting to speculate how it regulates both synapticintegration and the recruitment of downstream signaling.

The elevated transmitter concentrations that resultfrom MVR can, under some circumstances when transmit-ter clearance is impeded by restricted geometry, promoterebinding of transmitter to synaptic receptors [41]. MVRcan also promote transmitter escape from the synaptic cleftto activate extrasynaptic receptors [17,42–46]. Both re-binding and spillover are expected to prolong the durationof synaptic currents, and indeed there is a correlationbetween MVR and increased decay time [30,47]. Transmit-ter spillover most likely occurs when the transmitter con-centration is high, for example after MVR, when releasesites are closely spaced, or in the absence of glial diffusionbarriers and uptake machinery [48–51]. A clear distinctionbetween spillover from a neighboring release site and MVRis often difficult because both phenomena can result inprolonged presence of transmitter in the synaptic cleft andat extrasynaptic sites [52–54]. However, studies that uselow-affinity antagonists suggest that spillover does notincrease peak transmitter concentration in the cleft,whereas MVR does [14,17], allowing pharmacological

430

distinction when release sites are sufficiently far apart.At many synapses, glial cell processes that enwrap synap-ses establish an anatomical diffusion barrier and providehigh densities of excitatory amino acid transporters(EAATs) to separate release sites. Together with perisy-naptic neuronal EAATs, these transporters rapidly bufferand uptake glutamate to spatially and temporally confinetransmitter spread [55] and maintain extracellular gluta-mate concentrations in the nM range [56].

In accord with experimental findings that only high levelsof MVR can temporarily overwhelm glutamate clearancemechanisms [46,57], and that spillover is more apparent atsynapses lacking glial ensheathment [58], simulationssuggest that synapse crosstalk occurs only when near-synchronous MVR is combined with a lack of astrocyticcoverage [40,59]. Hence, although MVR significantlyincreases transmitter concentration, efficient clearancelargely eliminates spillover, highlighting that independenceof glutamatergic synapses remains preserved even duringMVR. As a result, at excitatory synapses with low MVR suchas Schaffer collateral (SC)–CA1 contacts, spillover is mostevident in response to repetitive stimulation and whenmonitored with high-affinity, slowly desensitizing NMDARsthat are more suitable to detect lower glutamate concentra-tions [44,60]. However, the near-simultaneous stimulationof adjacent fibers that occurs with many experimental para-digms may exaggerate the consequences of spillover[44,61]. A notable exception is glutamate spillover fromthe climbing fiber (CF) to MLIs in the cerebellar cortex.Here, high MVR allows glutamate to activate interneuronAMPARs and NMDARs, generating slow and long-lastingexcitation [62] that increases firing for tens of milliseconds(see below and [63,64]).

Another potential consequence of prolonged transmitterpresence following MVR is receptor desensitization. De-sensitization can diminish receptor availability, contributeto short-term depression, shape the time course of synapticcurrents, and is most common among synapses with clus-tered release sites, high Pr, and slowed transmitter clear-ance [65–68]. Hence, desensitization occurs at some but notall MVR synapses. For example, there is minimal desensi-tization of AMPARs at synapses where glutamate is rap-idly cleared by glial and neuronal EAATs (CF–Purkinjecell (CF–PC) synapses) [69,70], whereas MVR results inprofound desensitization at retinogeniculate contacts[51]. In conclusion, MVR can exacerbate desensitization,but synapse tortuosity, glial coverage, receptor composi-tion, and the spacing of release sites are likely the mainfactors [71,72]. Less is known about how MVR of GABAaffects spillover and desensitization. However, the lack ofefficient clearance mechanisms predicts a greater extent ofGABA escape from the synaptic cleft, perhaps with morepronounced effects on GABAA receptor desensitization andtonic inhibitory currents [73–75].

Vesicle release probability as a predictor of MVR?

What is the principal determinant of MVR? The mostparsimonious hypothesis is that Pr regulates whethermultiple vesicles are released concurrently. If each dockedvesicle can fuse independently in response to an actionpotential, then the simultaneous release of multiple vesicles

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

will occur more readily at synapses with higher Pr. Forexample, if two docked vesicles are each released with aprobability of 0.9, then the simultaneous release of twovesicles will occur with a probability of 0.81. Consistentwith this idea, the exceptionally high Pr at the CF terminal(�0.35–0.9) [76,77] coincides with significant MVR. Yet,some high Pr synapses display exclusively UVR [6,7], where-as various low Pr synapses exhibit MVR [15,17,34,78], dem-onstrating that Pr does not always reliably predict theincidence of MVR.

However, increasing Pr promotes MVR. Many low Pr

synapses generate small UVR postsynaptic currents(PSCs) that are prone to failure, but repetitive stimulationcan facilitate PSC amplitude. Release models predict thatthis facilitation depends on an increase in the number offusion-ready vesicles and/or Pr [79]. Activity-dependentlong- and short-term increases of MVR, for example duringfacilitation, augmentation, and long-term potentiation(LTP), have been observed at numerous synapses withlow initial Pr [11,17,34,80,81], supporting the view thathigh Pr promotes MVR. Considering inter-terminal het-erogeneity in synapse size, Pr, number of docked vesicles,and molecular make-up [82–84], it is also probable that theextent of MVR differs strongly from terminal to terminal[78,85]. It is therefore important to note that althoughmost central synapses have low Pr, mechanisms that ele-vate Pr (such as frequency facilitation, decreased presyn-aptic inhibition, or presynaptically expressed LTP) mayalso increase the probability of MVR [11,14,15,34,80,81].

Functions of MVR throughout the brainHigh-fidelity transmission and adaptation at sensory

synapses: MVR par excellence at synaptic ribbons

Sensory processing requires neuronal circuits to transformgraded information into a temporal code of action poten-tials. To continuously perform these computations synap-ses need to adapt to sustained levels of transmitter release.Unlike conventional synapses that require action poten-tials to release vesicles, the presynaptic cells at manyprimary sensory synapses typically operate via gradedmembrane potential changes that can last for severalseconds. Here, vesicles are often tethered to a specializedprotein structure called a ribbon that is thought to main-tain and synchronize the release of hundreds of vesicles persecond ([16,86–89], but see [90]). For example, at retinalbipolar cell terminals a 200 ms depolarization can release�6000 synaptic vesicles at 50 synaptic ribbon release sites,with each ribbon releasing �120 vesicles, a number similarto the total number of docked vesicles at the ribbon plusthose tethered to the ribbon [91]. In fact, vesicle pooldepletion is thought to be the major mechanism for spikeadaptation during sensory stimuli [92,93]. Release canoccur with rapid and slow kinetics at ribbon synapses,resulting in multiphasic and/or large-amplitude excitatorypostsynaptic currents (EPSCs) [89,94,95]. Several hypoth-eses have been proposed to explain the near-synchronousfusion of several tens of vesicles, including coordination ofrelease sites and tight coupling of calcium channels to therelease machinery [21,89,96], and fusion of several vesiclesto each other before exocytosis (compound fusion [97]).Recently, however, fusion pore regulation of UVR events

has been proposed as an alternative mechanism to explainmultiphasic EPSCs [90].

MVR is prominent throughout the retinal circuit. Rodbipolar cells (RBCs) receive input from light-sensing rodsand relay the visual information to AII amacrine cells. Thisretinal ribbon synapse exhibits highly synchronous MVR[16] that is modulated by the number of open calciumchannels. MVR requires at least two open calcium channelsto trigger the release of multiple vesicles from one ribbonrelease site [98]. Astoundingly, transmitter from the 2–4simultaneously released vesicles onto amacrine cells neitherfully occupies nor desensitizes postsynaptic AMPARs, allow-ing linear summation of each vesicle and a wide range ofsynaptic operation [93,99]. By contrast, MVR at cone bipolarcell–ganglion cell synapses fully occupies AMPARs andactivates perisynaptically located NMDARs in a concentra-tion-dependent manner [45,100,101]. In addition, at somebipolar cell terminals presynaptic activation of GABAC

receptors regulates Pr and MVR, and determines the extentof spillover and NMDAR activation [100,102,103].

In summary, sensory synapses take advantage of MVRto regulate firing duration and frequency, and also toensure reliable and sustained postsynaptic firing at highfrequencies. Moreover, MVR at ribbon synapses results inthe release of thousands of vesicles at single active zonesover prolonged periods of stimulation. The synaptic ribbontherefore appears to be a highly specialized structure thatpromotes continuous MVR at primary sensory synapses.

MVR at excitatory and inhibitory hippocampal synapses

Although initial studies rejected the MVR hypothesis[104,105], considerable variability in postsynaptic cur-rents, as well as in anatomical and molecular properties,led to the speculation that hippocampal synapses use MVRto increase their dynamic range [11,30,83,106] (Box 1 andFigure 2). Indeed, modeling of SC–CA1 EPSCs also favorsthe idea that MVR underlies current amplitude fluctua-tions [29,107]. A pioneering study in cultured hippocampalneurons identified MVR based on observations that multi-ple quanta interact with a common population of postsyn-aptic receptors [12]. Consistent with this notion, NMDAR-mediated calcium influx at single SC–CA1 spines increasesin a stepwise manner during repetitive stimulation, dem-onstrating that MVR underlies enhanced synaptic potencyduring facilitation [11,15,38]. In addition, evidence fromelectrophysiological studies suggests that cleft glutamateconcentration increases during trains of stimuli [17]. Final-ly, presynaptic modulators can control MVR in an input-and state-specific manner. For example, activation of thepresynaptic estrogen receptor b (ERb) induces acute po-tentiation of MVR specifically at low Pr SC–CA1 contacts[108]. Moreover, the transition from UVR to MVR mayunderlie the expression of late LTP at CA3–CA1 synapses[109]. Together these studies converge on the view thatMVR dynamically promotes reliable transmission at typi-cally low Pr hippocampal synapses and contributes toactivity- and context-dependent plasticity.

Increasing reliability of cortical and subcortical synapses

Heterogeneity of release mechanisms prevails in thecerebral cortex. Studies using statistical arguments and

431

Box 1. Detecting multivesicular release

Variance-mean analysis (VMA) determines the quantal parameters q, Pr,

and N by relating synaptic current fluctuations (expressed as CV) to

mean synaptic responses under various Pr conditions [143,144]

(Figure 2Aiii). Together with anatomical reconstruction (Figure 2Ai,ii)

and Gaussian histograms of synaptic response amplitudes, VMA serves

to identify release modality [119,145–149]. Narrow-amplitude histo-

grams with equally spaced peaks are thought to reflect release of several

vesicles. Uniquantal models assume that peaks represent the number of

release sites, each releasing at most one vesicle, whereas multiquantal

models assume that peaks reflect the release of multiple vesicles from a

common active zone [11,20]. Single-peak histograms are thought to

correspond to single-site UVR contacts [4,150], where synaptic varia-

bility (e.g., electronic distance from recording site, vesicle size, trans-

mitter content, and receptor density) generate amplitude fluctuations

([10,26,151], but see [28,29,40,152]). However, amplitude fluctuations

could also be consistent with saturation, spillover, or desensitization

during MVR.

MVR can be measured directly when the presynaptic neuron only

forms a single contact with the postsynaptic neuron [13,127]. Here,

spontaneous PSCs in have a significant fraction of double-peaked

events that occur within short time intervals, and statistical analysis

suggests that double peaks appear more often than expected by

chance. Therefore, double peaks likely emerge from multiple vesicles

released from a common active zone that interact with the same

postsynaptic receptor population.

To detect MVR across a population of synapses, a receptor antago-

nist with a fast unbinding rate (typically a ‘low-affinity antagonist’)

[153] is a powerful but simple tool that takes advantage of the

non-equilibrium interaction of neurotransmitter and antagonist with

receptor binding sites [105]. Unlike a slowly unbinding, high-affinity

antagonist that inhibits the PSC in an antagonist concentration-de-

pendent manner (Figure 2Bi), the extent of inhibition by the fast-off

antagonist also depends on transmitter concentration if the antagonist

dissociation rate is shorter than the presence of synaptically released

transmitter. As a result, both transmitter and antagonist compete for

receptor binding sites until transmitter is cleared [12,154] (Figure 2Bii).

This allows the fast-off antagonist to distinguish between MVR and

UVR because MVR results in higher transmitter concentrations in the

synaptic cleft than UVR (Figure 2Biii). The advantages of the fast-off

antagonist are easy implementation, fewer assumptions about quantal

parameters, the number of release sites and synaptic inputs. However,

it is unsuitable to report release heterogeneity between release sites,

and performs less reliably at closely spaced sites because spillover and

MVR might become indistinguishable [155].

Optical quantal analysis monitors NMDAR-mediated calcium fluxes

in response to vesicle fusion at individual presynaptic boutons, and

requires the loading of the neuron with a calcium-sensitive fluorescent

dye [15,18]. Similar to quantal current, the fusion of a vesicle causes a

quantal fluorescence change (Figure 2Ci). Hence, a stepwise increase

of fluorescence signals reflects an increase in the number of fusing

vesicles, or MVR (Figure 2Cii,iii). Another optical technique uses the

pH-sensitive green fluorescent protein pHluorin coupled to a vesicular

transporter (e.g., vGlut–pHluorin) to monitor vesicle fusion at single

sites and takes advantage of a pH-dependent increase in fluorescence

that occurs when the acidified vesicle is exposed to the basic extra-

cellular milieu [38].

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

anatomical reconstructions find evidence for UVR at intra-cortical excitatory synapses [6,110,111]. Nevertheless,MVR can also enable faithful transmission at intracorticalsynapses in layer 4 (L4) of the somatosensory cortex. HereL4 neuron terminals display high Pr and postsynapticreceptor saturation that reliably initiate action potentials,although MVR is less prevalent at L4 connections in thevisual cortex [19]. It is possible that the differential ex-pression of MVR represents distinct coding of informationdepending on content or valence.

Thalamic input to the cortex recruits robust feed-for-ward inhibition crucial for limiting cortical excitation andrefining sensory processing [112,113]. High Pr thalamocor-tical synapses onto interneurons display significant MVR(up to seven vesicles), driving reliable inhibition ontocortical pyramidal neurons [114]. However, the close spac-ing of release sites hampers a clear distinction betweenMVR and spillover at these contacts. In addition, corti-cothalamic projections from L6 pyramidal neurons showuse-dependent enhancement of synaptic strength mediat-ed by increased Pr, MVR and, consequently, receptor satu-ration [81]. In conclusion, while intracortical processingseems to rely primarily on UVR synapses, the evidencesuggests that MVR is important at the input and outputstages where signal transmission requires greater reliabil-ity.

In the striatum, cholinergic modulation of MVR and lowlevels of receptor saturation enhance the dynamic range ofexcitatory inputs onto medium spiny neurons. MVR isprevalent under basal conditions, but acetylcholine releasefrom interneurons dramatically decreases synaptic poten-cy (and MVR) by activating presynaptic muscarinic acetyl-choline receptors (mAChRs) [18]. The decrease in MVRaccelerates the time course of excitatory postsynapticpotentials (EPSPs), an effect that can be attributed to

432

MVR recruiting local voltage-gated conductances orNMDARs, and shows that MVR can control dendriticnonlinearities. Similarly to mAChRs, presynaptic GABAB

receptors at cortical pyramidal neurons reduce synapticpotency by decreasing Pr and MVR [115]. Therefore, themodulation of the number of vesicles released at singlesynaptic sites appears to control local signal integration inaddition to synaptic strength.

Dynamics of transmitter release in the cerebellum

MVR has been studied in detail at cerebellar CF–PCsynapses. Following an action potential the near-synchro-nous fusion of �3–5 vesicles per release site gives rise tocleft glutamate concentrations of >10 mM that fully occupyAMPARs and generate large EPSCs [14,33] that result inbursts of action potentials termed complex spikes [116–118]. Characteristic of high Pr, repetitive stimuli depletesynaptic vesicles and cause a decrease in CF EPSC ampli-tude [70,119]. Nonetheless, MVR and saturation arethought to help speed recovery from depression, allowingreliable transmission even during sustained CF activity atphysiological frequencies observed in vivo (<30% depres-sion at 2 Hz [39,120,121]). Firing at these frequencies alsocauses a slowing of the EPSC waveform that is consistentwith desynchronization of MVR within a release site,termed intra-site asynchrony [39]. Synaptic depressionand desynchronization of MVR generate complex spikesthat propagate more successfully along the PC axon thanduring near-synchronous MVR, thereby enhancing infor-mation transfer to PC target neurons. Downregulation ofMVR by various presynaptic receptors (including GABAB,adenosine, adrenergic, and metabotropic glutamate recep-tors) [39,122,123] can disrupt associative plasticity of par-allel fiber (PF) inputs [123], suggesting that modulation ofMVR at CF synapses has a role in cerebellar learning.

(Ai) (Aiii)

(Bi) (Bii)

Varia

nce

(pA2 ) 4000

2000

00 250 500

Mean current (pA)

75 mV80 pA

10 ms

(Aii)

50 ms

G/R0.5

50 ms

G/R0.2

G/R0.1

50 ms

ControlSlowFast

10 ms500 pA

200 pA10 ms

(Ci) (Cii)

1 µm

NF = 26q = 24.1 pA

High Pr

Low Pr

(Ciii)

(Biii)

1 2

TRENDS in Neurosciences

Figure 2. Detecting multivesicular release (MVR). (Ai) Reconstruction from paired recordings of a presynaptic basket (soma and dendrites in black, axon in red) and

postsynaptic pyramidal cell (soma and dendrites in blue, axon in gray). Inset, electron microscopy (EM)-identified boutons shown at higher magnification. (Aii) Electron

micrographs show synaptic contacts between the presynaptic basket terminals and the pyramidal cell soma. Arrowheads indicate synaptic junctions. (Aiii) Presynaptic

action potentials (red) and resulting currents (blue) recorded under low release probability (Pr, left) and high Pr (right) conditions. Below, variance versus mean plot from

current responses recorded in (Aiii) was used to derive quantal parameters (bottom). Adapted from [30]. (Bi) (Top) Slow-off, high-affinity antagonist (labeled as ‘slow’)

occupies postsynaptic receptors (in orange). (Bottom) As a result of slow antagonist unbinding, synaptically released neurotransmitter (NT) binds only to the fraction of

receptors that is not bound by the antagonist. (Bii) (Top) Due to its fast off-rate a low-affinity antagonist (labeled as ‘fast’) rapidly unbinds from receptors. (Bottom)

Synaptically released transmitter competes with the antagonist for receptor binding sites. Therefore, inhibition by the antagonist depends on the transmitter concentration

and time course. (Biii) The slow high-affinity antagonist inhibits postsynaptic currents recorded under high and low Pr conditions to the same extent (left). The fast low-

affinity antagonist inhibits the excitatory postsynaptic current (EPSC) recorded under high Pr conditions to a lesser extent than at low Pr conditions (right). Adapted from

[39]. (Ci) Raster image of a dendrite with spines (red fluorescence) and the calcium transient after synaptic stimulation (green fluorescence). White lines and arrows indicate

the position of the line scan and the time of synaptic stimulation, respectively. Synaptic stimulation results in a rapid increase of the green fluorescence (calcium-sensitive)

that does not occur with the red fluorescence (calcium-insensitive). (Cii) The ratio of green/red fluorescence in response to a single (left) or a pair of synaptic stimuli (right)

shown by arrows. Failures of synaptic transmission can be clearly distinguished from successes. (Ciii) Average fluorescence responses to a single successful stimulus

(yellow) and to the second stimulus in a pair when the first stimulation produced a failure (green). The failure to both stimuli is also shown (black). Adapted from [15].

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

Despite the lack of direct synaptic connections betweenCFs and MLIs, recent studies indicate that glutamatespillover from CFs evokes slow AMPAR and NMDAR-mediated EPSCs in MLIs that lead to a prolonged increasein firing rate [62–64]. These findings suggest that MVR canenhance the temporal and spatial spread of CF activation,

and provide the mechanistic details for the long-lastingmodulation of PC firing rate following CF activation in vivo[124,125].

Desynchronization of MVR on a supra-millisecond time-scale is also evident at inhibitory synapses between MLIs[13] where 0–3 vesicles can be released per action potential

433

Box 2. A mechanism for MVR

It is not known how MVR occurs. It appears most likely that many

vesicles fuse independently with the plasma membrane (heterotypic

fusion) at most synapses [13–15,39]. However, alternative explana-

tions have been considered. One example is the fusion of multiple

vesicles with each other before exocytosis (homotypic fusion or

compound fusion). First observed in secretory cells [156], activity-

and calcium-dependent compound fusion that could explain see-

mingly coordinated MVR was also reported at ribbon synapses [97]

and at the calyx of Held [157]. However, the conditions under which

compound fusion could occur at conventional synapses remain

unknown. Several lines of evidence argue against compound fusion

as a general mechanism for MVR: at the ribbon MVR can persist even

in the presence of fast calcium buffers [21,158], MVR exhibits similar

intra- and intersite temporal jitter [39], and proteins of the vesicle

release machinery located on the plasma membrane and on synaptic

vesicles are asymmetrically distributed [159]. These observations

support the notion of multiple independent release events within a

release site [13–15,39]. In addition, fusion pore dynamics were hy-

pothesized to govern peak transmitter concentration in the synaptic

cleft. Two fusion modes were proposed: incomplete fusion (‘kiss and

run’), that frees transmitter slowly and might prolong the transmitter

transient, or rapid full fusion, resulting in high transmitter concen-

tration [160,161]. However, the lack of kinetic changes in miniature

mEPSC (mEPSC) time course at varying Pr argues against such a

mechanism to explain fluctuations in transmitter concentration

[17,39]. In conclusion, the most likely factors in regulating the num-

ber of releasable vesicles, setting Pr, and determining the temporal

pattern of fusion that should be investigated in the context of MVR

are density, recruitment, and type of presynaptic calcium channels,

calcium-buffering proteins, as well as proteins of the presynaptic

vesicle-release machinery [84,98,162–165].

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

at single release sites [126]. In these electrically-compactinterneurons, desynchronized IPSCs occurring within timeintervals of 5–20 ms could provide long-lasting inhibitionthat limits the integration of incoming PF-mediated exci-tation. Consistent with MVR at MLI contacts, MLIs canalso release multiple synchronized vesicles onto PCs via acalcium store-dependent mechanism, resulting in minia-ture IPSCs with exceptionally large amplitude (‘maxi-mini-IPSCs’) [37].

Finally, at least a subset of low Pr PF terminals alsoundergoes MVR during long- and short-term activity-de-pendent potentiation, albeit with considerable intersynap-tic heterogeneity [34,78,80,127,128]. Although the origin ofthis heterogeneity is unknown, differences in PF boutonsize, calcium dynamics, and sensitivity to G-protein-cou-pled receptor activation have been discussed[129,130]. The high glutamate concentration generatedby MVR is thought to activate extrasynaptic NMDARson MLIs [78], and extrasynaptic metabotropic receptorson PCs and MLIs, as well as activating AMPARs onBergmann glia [33,52,131–133].

MVR during synapse maturation and development

During postnatal development synapses undergo exten-sive morphological and physiological changes. One promi-nent and well-studied example is the calyx of Held synapseof the auditory pathway [134]. Morphological refinement(i.e., a transition from a cup-shaped structure to a morefenestrated and claw-like morphology) occurs concomitant-ly with a progressive speeding of the presynaptic actionpotential that decreases Pr and shortens synaptic delay[135]. MVR is less prominent at the mature compared tothe immature calyx [136], allowing faster glutamate clear-ance and less AMPAR desensitization and saturation[66,136]. These coordinated changes enable this auditorysynapse to fire spikes at high frequencies (up to 1 KHz).The expression of MVR at a given synapse may therefore besubject to developmental remodeling and track changes inPr.

In the developing cerebellum, PCs receive multiple CFinputs that undergo activity-dependent pruning. Strongevidence suggests that the CFs with the greatest extent ofMVR evoke the largest depolarization-dependent calciumsignals in PC dendrites, and consequently evade synapseelimination [137,138]. The mechanisms that establishMVR before pruning remain incompletely understood,but these results suggest that MVR may contribute tosynapse refinement.

Sensitization of homeostatic circuits

Behavioral responses to external stimuli often intensifyupon stimulus re-occurrence. Several lines of evidencesuggest that MVR may be a synaptic correlate of suchbehavioral sensitization. One example is the stress re-sponse of the hypothalamus–pituitary–adrenal (HPA) axisthat generates a long-lasting sensitivity to repetitivestressors [139] that involves enhanced activity of parvo-cellular neurosecretory cells (PNCs) in the paraventricularnucleus of the hypothalamus. A recent study found poten-tiated excitatory synaptic transmission onto PNCs in ani-mals pre-exposed to a single stressor compared to naive

434

animals. The mechanism of enhancement is consistentwith an activity-dependent depression of NMDAR currentthat diminishes calcium influx and the release of a tonicretrograde inhibitor from PNCs. Disinhibition of transmit-ter release in turn increases MVR. Remarkably, bothbehavioral sensitization to the stressor and enhancedMVR last several days [20]. Therefore, potentiation ofMVR may be the synaptic correlate of enhanced HPA axissensitivity.

Noradrenergic modulation of MVR may also regulatehormone secretion from magnocellular neurosecretorycells (MNCs) in the hypothalamus. Noradrenaline aug-ments presynaptic calcium via a1 receptor activationand calcium release from ryanodine-sensitive stores, tran-siently enhancing synchronous MVR onto MNCs [140]. Inelectrically-compact cells such as MNCs [141] large-ampli-tude MVR-unitary EPSCs are sufficient to trigger actionpotential firing and, hence, release of vasopressin andoxytocin, hormones crucial for the regulation of hydration,stress, hemorrhage, lactation, and parturition. Together,these results demonstrate that neuroendocrine output inthe hypothalamus partially underlies the short and long-term regulation of MVR [142].

Concluding remarksWe present abundant evidence that MVR is a commonmechanism for enhancing synaptic reliability and over-coming the stochastic variability of synaptic transmission.Although the precise molecular mechanisms underlyingMVR remain unknown (Box 2), MVR determines synapticbehavior during repetitive and sustained activation,affects the temporal and spatial activation profile of neu-

Box 3. Outstanding questions

� The molecular determinants of UVR and MVR are unknown; the

composition of active zones that support MVR need further eluci-

dation. Important factors may include calcium channel density,

subtypes and adapter proteins, calcium sensors of the release

machinery, calcium buffering, availability of release sites, and

trans-synaptically acting proteins.

� The detailed mechanisms that govern the temporal pattern of MVR

(synchronized vs desynchronized) are unresolved.

� Presynaptic signaling pathways that underlie activity-dependent

regulation of MVR, especially during intermediate and short-term

plasticity, need additional exploration.

� The consequences of MVR versus UVR for synaptic integration at

single spines and dendritic branches are not well studied. This

includes understanding how MVR contributes to the duration of

the postsynaptic potential and to the recruitment of voltage-gated

and calcium-dependent conductances.

� Consequences of MVR for network function (spillover-mediated

recruitment of inhibition, action potential propagation) have been

described in detail in the cerebellum. Whether similar phenomena

apply to other circuits is unknown.

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

rotransmitter receptors and voltage-gated conductances,and therefore controls synaptic integration, downstreamintracellular signaling, and the induction of plasticity. Thenumerous examples of MVR discussed here illustrate thatsynapses differentially exploit MVR depending on theirfunctional demands. Nevertheless, several outstandingquestions remain (Box 3). In conclusion, we suggest thatMVR is a fundamental and simple mechanism for increas-ing the dynamic range of synaptic transmission, and wechallenge the long-held hypothesis that MVR is a rarephenomenon manifested exclusively at high Pr contacts.

AcknowledgmentsThis work was supported by National Institutes of Health grants to J.I.W.(R01NS065920) and H.v.G. (R01DC012938 and R01DC04274). We thankDrs Monica Thanawala, Pascal Kaeser, Jessica Hauser, AnastasiosTzingounis, Jada Vaden, Linda Overstreet-Wadiche, and members ofthe laboratory of J.I.W. for reading the manuscript.

References1 Dijkstra, E.W. (1982) Selected Writings on Computing: A Personal

Perspective, Springer-Verlag2 Del Castillo, J. and Katz, B. (1954) Quantal components of the end-

plate potential. J. Physiol. 124, 560–5733 Redman, S. and Walmsley, B. (1983) Amplitude fluctuations in

synaptic potentials evoked in cat spinal motoneurones at identifiedgroup Ia synapses. J. Physiol. 343, 135–145

4 Gulyas, A.I. et al. (1993) Hippocampal pyramidal cells exciteinhibitory neurons through a single release site. Nature 366, 683–687

5 Lawrence, J.J. and McBain, C.J. (2003) Interneuron diversity series:containing the detonation–feedforward inhibition in the CA3hippocampus. Trends Neurosci. 26, 631–640

6 Silver, R.A. et al. (2003) High-probability uniquantal transmission atexcitatory synapses in barrel cortex. Science 302, 1981–1984

7 Murphy, G.J. et al. (2004) Sensory neuron signaling to the brain:properties of transmitter release from olfactory nerve terminals. J.Neurosci. 24, 3023–3030

8 Biro, A.A. et al. (2005) Quantal size is independent of the releaseprobability at hippocampal excitatory synapses. J. Neurosci. 25, 223–232

9 Korn, H. et al. (1981) Fluctuating responses at a central synapse: n ofbinomial fit predicts number of stained presynaptic boutons. Science213, 898–901

10 Redman, S. (1990) Quantal analysis of synaptic potentials in neuronsof the central nervous system. Physiol. Rev. 70, 165–198

11 Conti, R. and Lisman, J. (2003) The high variance of AMPA receptor-and NMDA receptor-mediated responses at single hippocampalsynapses: evidence for multiquantal release. Proc. Natl. Acad. Sci.U.S.A. 100, 4885–4890

12 Tong, G. and Jahr, C.E. (1994) Multivesicular release from excitatorysynapses of cultured hippocampal neurons. Neuron 12, 51–59

13 Auger, C. et al. (1998) Multivesicular release at single functionalsynaptic sites in cerebellar stellate and basket cells. J. Neurosci.18, 4532–4547

14 Wadiche, J.I. and Jahr, C.E. (2001) Multivesicular release at climbingfiber–Purkinje cell synapses. Neuron 32, 301–313

15 Oertner, T.G. et al. (2002) Facilitation at single synapses probed withoptical quantal analysis. Nat. Neurosci. 5, 657–664

16 Singer, J.H. et al. (2004) Coordinated multivesicular release at amammalian ribbon synapse. Nat. Neurosci. 7, 826–833

17 Christie, J.M. and Jahr, C.E. (2006) Multivesicular release at Schaffercollateral–CA1 hippocampal synapses. J. Neurosci. 26, 210–216

18 Higley, M.J. et al. (2009) Cholinergic modulation of multivesicularrelease regulates striatal synaptic potency and integration. Nat.Neurosci.

19 Huang, C.H. et al. (2010) Multivesicular release differentiates thereliability of synaptic transmission between the visual cortex and thesomatosensory cortex. J. Neurosci. 30, 11994–12004

20 Kuzmiski, J.B. et al. (2010) Stress-induced priming of glutamatesynapses unmasks associative short-term plasticity. Nat. Neurosci.13, 1257–1264

21 Graydon, C.W. et al. (2011) Sharp Ca2+ nanodomains beneath theribbon promote highly synchronous multivesicular release at hair cellsynapses. J. Neurosci. 31, 16637–16650

22 Abenavoli, A. et al. (2002) Multimodal quantal release at individualhippocampal synapses: evidence for no lateral inhibition. J. Neurosci.22, 6336–6346

23 Watanabe, S. et al. (2013) Ultrafast endocytosis at mousehippocampal synapses. Nature 504, 242–247

24 Tsodyks, M.V. and Markram, H. (1997) The neural code betweenneocortical pyramidal neurons depends on neurotransmitter releaseprobability. Proc. Natl. Acad. Sci. U.S.A. 94, 719–723

25 Abbott, L.F. and Regehr, W.G. (2004) Synaptic computation. Nature431, 796–803

26 Bekkers, J.M. et al. (1990) Origin of variability in quantal size incultured hippocampal neurons and hippocampal slices. Proc. Natl.Acad. Sci. U.S.A. 87, 5359–5362

27 Frerking, M. et al. (1995) Variation in GABA mini amplitude is theconsequence of variation in transmitter concentration. Neuron 15,885–895

28 Ishikawa, T. et al. (2002) A single packet of transmitter does notsaturate postsynaptic glutamate receptors. Neuron 34, 613–621

29 Raghavachari, S. and Lisman, J.E. (2004) Properties of quantaltransmission at CA1 synapses. J. Neurophysiol. 92, 2456–2467

30 Biro, A.A. et al. (2006) Release probability-dependent scaling of thepostsynaptic responses at single hippocampal GABAergic synapses.J. Neurosci. 26, 12487–12496

31 Mainen, Z.F. et al. (1999) Synaptic calcium transients in single spinesindicate that NMDA receptors are not saturated. Nature 399, 151–155

32 McAllister, A.K. and Stevens, C.F. (2000) Nonsaturation of AMPA andNMDA receptors at hippocampal synapses. Proc. Natl. Acad. Sci.U.S.A. 97, 6173–6178

33 Harrison, J. and Jahr, C.E. (2003) Receptor occupancy limits synapticdepression at climbing fiber synapses. J. Neurosci. 23, 377–383

34 Foster, K.A. et al. (2005) The influence of multivesicular release andpostsynaptic receptor saturation on transmission at granule cell toPurkinje cell synapses. J. Neurosci. 25, 11655–11665

35 Chanda, S. and Xu-Friedman, M.A. (2010) A low-affinity antagonistreveals saturation and desensitization in mature synapses in theauditory brain stem. J. Neurophysiol. 103, 1915–1926

36 Katz, B. and Miledi, R. (1965) The effect of temperature on thesynaptic delay at the neuromuscular junction. J. Physiol. 181, 656–670

37 Llano, I. et al. (2000) Presynaptic calcium stores underlie large-amplitude miniature IPSCs and spontaneous calcium transients.Nat. Neurosci. 3, 1256–1265

38 Leitz, J. and Kavalali, E.T. (2011) Ca2+ influx slows single synapticvesicle endocytosis. J. Neurosci. 31, 16318–16326

435

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

39 Rudolph, S. et al. (2011) Desynchronization of multivesicular releaseenhances Purkinje cell output. Neuron 70, 991–1004

40 Boucher, J. et al. (2010) Realistic modelling of receptor activation inhippocampal excitatory synapses: analysis of multivesicular release,release location, temperature and synaptic cross-talk. Brain Struct.Funct. 215, 49–65

41 Otis, T.S. et al. (1996) Delayed clearance of transmitter and the role ofglutamate transporters at synapses with multiple release sites. J.Neurosci. 16, 1634–1644

42 Dzubay, J.A. and Jahr, C.E. (1999) The concentration of synapticallyreleased glutamate outside of the climbing fiber–Purkinje cellsynaptic cleft. J. Neurosci. 19, 5265–5274

43 Brasnjo, G. and Otis, T.S. (2001) Neuronal glutamate transporterscontrol activation of postsynaptic metabotropic glutamate receptorsand influence cerebellar long-term depression. Neuron 31, 607–616

44 Arnth-Jensen, N. et al. (2002) Cooperation between independenthippocampal synapses is controlled by glutamate uptake. Nat.Neurosci. 5, 325–331

45 Chen, S. and Diamond, J.S. (2002) Synaptically released glutamateactivates extrasynaptic NMDA receptors on cells in the ganglion celllayer of rat retina. J. Neurosci. 22, 2165–2173

46 Wadiche, J.I. and Jahr, C.E. (2005) Patterned expression of Purkinjecell glutamate transporters controls synaptic plasticity. Nat.Neurosci. 8, 1329–1334

47 Kondo, S. and Marty, A. (1998) Synaptic currents at individualconnections among stellate cells in rat cerebellar slices. J. Physiol.509, 221–232

48 Scanziani, M. et al. (1996) Role of intercellular interactions inheterosynaptic long-term depression. Nature 380, 446–450

49 DiGregorio, D.A. et al. (2002) Spillover of glutamate onto synapticAMPA receptors enhances fast transmission at a cerebellar synapse.Neuron 35, 521–533

50 Takayasu, Y. et al. (2006) Glial glutamate transporters maintain one-to-one relationship at the climbing fiber–Purkinje cell synapse bypreventing glutamate spillover. J. Neurosci. 26, 6563–6572

51 Budisantoso, T. et al. (2012) Mechanisms underlying signal filteringat a multisynapse contact. J. Neurosci. 32, 2357–2376

52 Bergles, D.E. et al. (1997) Glutamate transporter currents inBergmann glial cells follow the time course of extrasynapticglutamate. Proc. Natl. Acad. Sci. U.S.A. 94, 14821–14825

53 Bergles, D.E. and Jahr, C.E. (1997) Synaptic activation of glutamatetransporters in hippocampal astrocytes. Neuron 19, 1297–1308

54 Otis, T.S. et al. (1997) Postsynaptic glutamate transport at theclimbing fiber–Purkinje cell synapse. Science 277, 1515–1518

55 Tzingounis, A.V. and Wadiche, J.I. (2007) Glutamate transporters:confining runaway excitation by shaping synaptic transmission. Nat.Rev. Neurosci. 8, 935–947

56 Herman, M.A. and Jahr, C.E. (2007) Extracellular glutamateconcentration in hippocampal slice. J. Neurosci. 27, 9736–9741

57 Diamond, J.S. and Jahr, C.E. (2000) Synaptically released glutamatedoes not overwhelm transporters on hippocampal astrocytes duringhigh-frequency stimulation. J. Neurophysiol. 83, 2835–2843

58 Carter, A.G. and Regehr, W.G. (2000) Prolonged synaptic currentsand glutamate spillover at the parallel fiber to stellate cell synapse. J.Neurosci. 20, 4423–4434

59 Greget, R. et al. (2011) Simulation of postsynaptic glutamate receptorsreveals critical features of glutamatergic transmission. PLoS ONE 6,e28380

60 Asztely, F. et al. (1997) Extrasynaptic glutamate spillover in thehippocampus: dependence on temperature and the role of activeglutamate uptake. Neuron 18, 281–293

61 Marcaggi, P. and Attwell, D. (2005) Endocannabinoid signalingdepends on the spatial pattern of synapse activation. Nat.Neurosci. 8, 776–781

62 Szapiro, G. and Barbour, B. (2007) Multiple climbing fibers signal tomolecular layer interneurons exclusively via glutamate spillover. Nat.Neurosci. 10, 735–742

63 Mathews, P.J. et al. (2012) Effects of climbing fiber driven inhibitionon Purkinje neuron spiking. J. Neurosci. 32, 17988–17997

64 Coddington, L.T. et al. (2013) Spillover-mediated feedforwardinhibition functionally segregates interneuron activity. Neuron 78,1050–1062

436

65 Trussell, L.O. and Fischbach, G.D. (1989) Glutamate receptordesensitization and its role in synaptic transmission. Neuron 3,209–218

66 Trussell, L.O. et al. (1993) Desensitization of AMPA receptors uponmultiquantal neurotransmitter release. Neuron 10, 1185–1196

67 Kinney, G.A. et al. (1997) Prolonged physiological entrapment ofglutamate in the synaptic cleft of cerebellar unipolar brush cells. J.Neurophysiol. 78, 1320–1333

68 Koike-Tani, M. et al. (2005) Mechanisms underlying developmentalspeeding in AMPA-EPSC decay time at the calyx of Held. J. Neurosci.25, 199–207

69 Dittman, J.S. and Regehr, W.G. (1998) Calcium dependence andrecovery kinetics of presynaptic depression at the climbing fiber toPurkinje cell synapse. J. Neurosci. 18, 6147–6162

70 Hashimoto, K. and Kano, M. (1998) Presynaptic origin of paired-pulsedepression at climbing fibre–Purkinje cell synapses in the ratcerebellum. J. Physiol. 506, 391–405

71 Xu-Friedman, M.A. and Regehr, W.G. (2003) Ultrastructuralcontributions to desensitization at cerebellar mossy fiber to granulecell synapses. J. Neurosci. 23, 2182–2192

72 Mugnaini, E. et al. (2011) The unipolar brush cell: a remarkableneuron finally receiving deserved attention. Brain Res. Rev. 66,220–245

73 Rossi, D.J. and Hamann, M. (1998) Spillover-mediated transmissionat inhibitory synapses promoted by high affinity alpha6 subunitGABA(A) receptors and glomerular geometry. Neuron 20, 783–795

74 Overstreet, L.S. et al. (2002) Measuring and modeling thespatiotemporal profile of GABA at the synapse. In TransmembraneTransporters (Quick, M., ed.), pp. 259–275, John Wiley & Sons

75 Bright, D.P. et al. (2011) Profound desensitization by ambient GABAlimits activation of delta-containing GABAA receptors duringspillover. J. Neurosci. 31, 753–763

76 Silver, R.A. et al. (1998) Locus of frequency-dependent depressionidentified with multiple-probability fluctuation analysis at ratclimbing fibre–Purkinje cell synapses. J. Physiol. 510, 881–902

77 Dittman, J.S. et al. (2000) Interplay between facilitation, depression,and residual calcium at three presynaptic terminals. J. Neurosci. 20,1374–1385

78 Nahir, B. and Jahr, C.E. (2013) Activation of extrasynaptic NMDARsat individual parallel fiber-molecular layer interneuron synapses incerebellum. J. Neurosci. 33, 16323–16333

79 Zucker, R.S. and Regehr, W.G. (2002) Short-term synaptic plasticity.Annu. Rev. Physiol. 64, 355–405

80 Bender, V.A. et al. (2009) Presynaptically expressed long-termpotentiation increases multivesicular release at parallel fibersynapses. J. Neurosci. 29, 10974–10978

81 Sun, Y.G. and Beierlein, M. (2011) Receptor saturation controls short-term synaptic plasticity at corticothalamic synapses. J. Neurophysiol.105, 2319–2329

82 Dobrunz, L.E. et al. (1997) Very short-term plasticity in hippocampalsynapses. Proc. Natl. Acad. Sci. U.S.A. 94, 14843–14847

83 Schikorski, T. and Stevens, C.F. (2001) Morphological correlates offunctionally defined synaptic vesicle populations. Nat. Neurosci. 4,391–395

84 Holderith, N. et al. (2012) Release probability of hippocampalglutamatergic terminals scales with the size of the active zone.Nat. Neurosci. 15, 988–997

85 Rosenmund, C. et al. (1993) Nonuniform probability of glutamaterelease at a hippocampal synapse. Science 262, 754–757

86 Parsons, T.D. et al. (1994) Calcium-triggered exocytosis andendocytosis in an isolated presynaptic cell: capacitancemeasurements in saccular hair cells. Neuron 13, 875–883

87 von Gersdorff, H. and Matthews, G. (1994) Dynamics of synapticvesicle fusion and membrane retrieval in synaptic terminals.Nature 367, 735–739

88 Rieke, F. and Schwartz, E.A. (1996) Asynchronous transmitterrelease: control of exocytosis and endocytosis at the salamanderrod synapse. J. Physiol. 493, 1–8

89 Glowatzki, E. and Fuchs, P.A. (2002) Transmitter release at the haircell ribbon synapse. Nat. Neurosci. 5, 147–154

90 Chapochnikov, N.M. et al. (2014) Uniquantal release through adynamic fusion pore is a candidate mechanism of hair cellexocytosis. Neuron 83, 1389–1403

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

91 von Gersdorff, H. et al. (1996) Evidence that vesicles on the synapticribbon of retinal bipolar neurons can be rapidly released. Neuron 16,1221–1227

92 Moser, T. and Beutner, D. (2000) Kinetics of exocytosis andendocytosis at the cochlear inner hair cell afferent synapse of themouse. Proc. Natl. Acad. Sci. U.S.A. 97, 883–888

93 Singer, J.H. and Diamond, J.S. (2006) Vesicle depletion and synapticdepression at a mammalian ribbon synapse. J. Neurophysiol. 95,3191–3198

94 Zenisek, D. et al. (2000) Transport, capture and exocytosis of singlesynaptic vesicles at active zones. Nature 406, 849–854

95 Li, G.L. et al. (2009) The unitary event underlying multiquantalEPSCs at a hair cell’s ribbon synapse. J. Neurosci. 29, 7558–7568

96 Grant, L. et al. (2010) Two modes of release shape the postsynapticresponse at the inner hair cell ribbon synapse. J. Neurosci. 30, 4210–4220

97 Matthews, G. and Sterling, P. (2008) Evidence that vesicles undergocompound fusion on the synaptic ribbon. J. Neurosci. 28, 5403–5411

98 Jarsky, T. et al. (2010) Nanodomain control of exocytosis is responsiblefor the signaling capability of a retinal ribbon synapse. J. Neurosci. 30,11885–11895

99 Singer, J.H. (2007) Multivesicular release and saturation ofglutamatergic signalling at retinal ribbon synapses. J. Physiol. 580,23–29

100 Sagdullaev, B.T. et al. (2006) Presynaptic inhibition modulatesspillover, creating distinct dynamic response ranges of sensoryoutput. Neuron 50, 923–935

101 Zhang, J. and Diamond, J.S. (2006) Distinct perisynaptic and synapticlocalization of NMDA and AMPA receptors on ganglion cells in ratretina. J. Comp. Neurol. 498, 810–820

102 Matsui, K. et al. (2001) Modulation of excitatory synaptictransmission by GABA(C) receptor-mediated feedback in themouse inner retina. J. Neurophysiol. 86, 2285–2298

103 Vigh, J. and von Gersdorff, H. (2005) Prolonged reciprocal signalingvia NMDA and GABA receptors at a retinal ribbon synapse. J.Neurosci. 25, 11412–11423

104 Stevens, C.F. and Wang, Y. (1995) Facilitation and depression atsingle central synapses. Neuron 14, 795–802

105 Clements, J.D. (1996) Transmitter timecourse in the synaptic cleft: itsrole in central synaptic function. Trends Neurosci. 19, 163–171

106 Rosenmund, C. et al. (1998) The tetrameric structure of a glutamatereceptor channel. Science 280, 1596–1599

107 Ricci-Tersenghi, F. et al. (2006) Multivesicular release at developingSchaffer collateral–CA1 synapses: an analytic approach to describeexperimental data. J. Neurophysiol. 96, 15–26

108 Smejkalova, T. and Woolley, C.S. (2010) Estradiol acutely potentiateshippocampal excitatory synaptic transmission through a presynapticmechanism. J. Neurosci. 30, 16137–16148

109 Bolshakov, V.Y. et al. (1997) Recruitment of new sites of synaptictransmission during the cAMP-dependent late phase of LTP at CA3-CA1 synapses in the hippocampus. Neuron 19, 635–651

110 Buhl, E.H. et al. (1997) Effect, number and location of synapses madeby single pyramidal cells onto aspiny interneurones of cat visualcortex. J. Physiol. 500, 689–713

111 Egger, V. et al. (1999) Coincidence detection and changes of synapticefficacy in spiny stellate neurons in rat barrel cortex. Nat. Neurosci. 2,1098–1105

112 Gabernet, L. et al. (2005) Somatosensory integration controlled bydynamic thalamocortical feed-forward inhibition. Neuron 48, 315–327

113 Bruno, R.M. and Sakmann, B. (2006) Cortex is driven by weak butsynchronously active thalamocortical synapses. Science 312, 1622–1627

114 Bagnall, M.W. et al. (2011) Multiple clusters of release sites formed byindividual thalamic afferents onto cortical interneurons ensurereliable transmission. Neuron 71, 180–194

115 Chalifoux, J.R. and Carter, A.G. (2010) GABAB receptors modulateNMDA receptor calcium signals in dendritic spines. Neuron 66, 101–113

116 Eccles, J.C. et al. (1966) The excitatory synaptic action of climbingfibres on the Purkinje cells of the cerebellum. J. Physiol. 182, 268–296

117 Konnerth, A. et al. (1990) Synaptic currents in cerebellar Purkinjecells. Proc. Natl. Acad. Sci. U.S.A. 87, 2662–2665

118 Perkel, D.J. et al. (1990) Excitatory synaptic currents in Purkinjecells. Proc. Biol. Sci. 241, 116–121

119 Foster, K.A. and Regehr, W.G. (2004) Variance-mean analysis in thepresence of a rapid antagonist indicates vesicle depletion underliesdepression at the climbing fiber synapse. Neuron 43, 119–131

120 Armstrong, D.M. and Rawson, J.A. (1979) Activity patterns ofcerebellar cortical neurones and climbing fibre afferents in theawake cat. J. Physiol. 289, 425–448

121 Foster, K.A. et al. (2002) Interaction of postsynaptic receptorsaturation with presynaptic mechanisms produces a reliablesynapse. Neuron 36, 1115–1126

122 Takahashi, M. et al. (1995) Pre- and postsynaptic determinants ofEPSC waveform at cerebellar climbing fiber and parallel fiber toPurkinje cell synapses. J. Neurosci. 15, 5693–5702

123 Carey, M.R. and Regehr, W.G. (2009) Noradrenergic control ofassociative synaptic plasticity by selective modulation ofinstructive signals. Neuron 62, 112–122

124 Bloedel, J.R. et al. (1983) Increased responsiveness of Purkinje cellsassociated with climbing fiber inputs to neighboring neurons. J.Neurophysiol. 50, 220–239

125 Jorntell, H. and Ekerot, C.F. (2003) Receptive field plasticityprofoundly alters the cutaneous parallel fiber synaptic input tocerebellar interneurons in vivo. J. Neurosci. 23, 9620–9631

126 Trigo, F.F. et al. (2012) Readily releasable pool of synaptic vesiclesmeasured at single synaptic contacts. Proc. Natl. Acad. Sci. U.S.A.109, 18138–18143

127 Crowley, J.J. et al. (2007) Fast vesicle replenishment and rapidrecovery from desensitization at a single synaptic release site. J.Neurosci. 27, 5448–5460

128 Satake, S. et al. (2012) Paired-pulse facilitation of multivesicularrelease and intersynaptic spillover of glutamate at rat cerebellargranule cell–interneurone synapses. J. Physiol. 590, 5653–5675

129 Zhang, W. and Linden, D.J. (2009) Neuromodulation at singlepresynaptic boutons of cerebellar parallel fibers is determined bybouton size and basal action potential-evoked Ca transientamplitude. J. Neurosci. 29, 15586–15594

130 Zhang, W. and Linden, D.J. (2012) Calcium influx measured at singlepresynaptic boutons of cerebellar granule cell ascending axons andparallel fibers. Cerebellum 11, 121–131

131 Karakossian, M.H. and Otis, T.S. (2004) Excitation of cerebellarinterneurons by group I metabotropic glutamate receptors. J.Neurophysiol. 92, 1558–1565

132 Beierlein, M. and Regehr, W.G. (2006) Brief bursts of parallel fiberactivity trigger calcium signals in Bergmann glia. J. Neurosci. 26,6958–6967

133 Tsai, M.C. et al. (2012) Neuronal glutamate transporters regulateglial excitatory transmission. J. Neurosci. 32, 1528–1535

134 Borst, J.G. and Soria van Hoeve, J. (2012) The calyx of held synapse:from model synapse to auditory relay. Annu. Rev. Physiol. 74, 199–224

135 Taschenberger, H. and von Gersdorff, H. (2000) Fine-tuning anauditory synapse for speed and fidelity: developmental changes inpresynaptic waveform, EPSC kinetics, and synaptic plasticity. J.Neurosci. 20, 9162–9173

136 Taschenberger, H. et al. (2002) Optimizing synaptic architecture andefficiency for high-frequency transmission. Neuron 36, 1127–1143

137 Hashimoto, K. and Kano, M. (2003) Functional differentiation ofmultiple climbing fiber inputs during synapse elimination in thedeveloping cerebellum. Neuron 38, 785–796

138 Hashimoto, K. et al. (2009) Translocation of a ‘winner’ climbing fiber tothe Purkinje cell dendrite and subsequent elimination of ‘losers’ fromthe soma in developing cerebellum. Neuron 63, 106–118

139 Armario, A. et al. (2008) Long-term neuroendocrine and behaviouraleffects of a single exposure to stress in adult animals. Neurosci.Biobehav. Rev. 32, 1121–1135

140 Gordon, G.R. and Bains, J.S. (2005) Noradrenaline triggersmultivesicular release at glutamatergic synapses in thehypothalamus. J. Neurosci. 25, 11385–11395

141 Bains, J.S. and Ferguson, A.V. (1997) Nitric oxide regulates NMDA-driven GABAergic inputs to type I neurones of the rat paraventricularnucleus. J. Physiol. 499, 733–746

142 Quinlan, M.E. and Hirasawa, M. (2013) Multivesicular releaseunderlies short term synaptic potentiation independent of releaseprobability change in the supraoptic nucleus. PLoS ONE 8, e77402

143 Clements, J.D. and Silver, R.A. (2000) Unveiling synaptic plasticity: anew graphical and analytical approach. Trends Neurosci. 23, 105–113

437

Feature Review Trends in Neurosciences July 2015, Vol. 38, No. 7

144 Clements, J.D. (2003) Variance-mean analysis: a simple and reliableapproach for investigating synaptic transmission and modulation. J.Neurosci. Methods 130, 115–125

145 Reid, C.A. and Clements, J.D. (1999) Postsynaptic expression of long-term potentiation in the rat dentate gyrus demonstrated by variance-mean analysis. J. Physiol. 518, 121–130

146 Meyer, A.C. et al. (2001) Estimation of quantal size and number offunctional active zones at the calyx of held synapse by nonstationaryEPSC variance analysis. J. Neurosci. 21, 7889–7900

147 Lawrence, J.J. et al. (2004) Quantal transmission at mossy fibre targetsin the CA3 region of the rat hippocampus. J. Physiol. 554, 175–193

148 Saviane, C. and Silver, R.A. (2006) Errors in the estimation of thevariance: implications for multiple-probability fluctuation analysis. J.Neurosci. Methods 153, 250–260

149 Ikeda, K. et al. (2008) Distinctive quantal properties ofneurotransmission at excitatory and inhibitory autapses revealedusing variance-mean analysis. J. Neurosci. 28, 13563–13573

150 Arancio, O. et al. (1994) Excitatory synaptic connections onto rathippocampal inhibitory cells may involve a single transmitterrelease site. J. Physiol. 481, 395–405

151 Wall, M.J. and Usowicz, M.M. (1998) Development of the quantalproperties of evoked and spontaneous synaptic currents at a brainsynapse. Nat. Neurosci. 1, 675–682

152 Franks, K.M. et al. (2003) Independent sources of quantal variabilityat single glutamatergic synapses. J. Neurosci. 23, 3186–3195

153 Laake, J.H. et al. (1995) Glutamine from glial cells is essential for themaintenance of the nerve terminal pool of glutamate: immunogoldevidence from hippocampal slice cultures. J. Neurochem. 65, 871–881

154 Clements, J.D. et al. (1992) The time course of glutamate in thesynaptic cleft. Science 258, 1498–1501

438

155 Kullmann, D.M. and Asztely, F. (1998) Extrasynaptic glutamatespillover in the hippocampus: evidence and implications. TrendsNeurosci. 21, 8–14

156 Pickett, J.A. and Edwardson, J.M. (2006) Compound exocytosis:mechanisms and functional significance. Traffic 7, 109–116

157 He, L. et al. (2009) Compound vesicle fusion increases quantal sizeand potentiates synaptic transmission. Nature 459, 93–97

158 Brandt, A. et al. (2005) Few CaV1.3 channels regulate the exocytosisof a synaptic vesicle at the hair cell ribbon synapse. J. Neurosci. 25,11577–11585

159 Takamori, S. et al. (2006) Molecular anatomy of a traffickingorganelle. Cell 127, 831–846

160 Choi, S. et al. (2000) Postfusional regulation of cleft glutamateconcentration during LTP at ‘silent synapses’. Nat. Neurosci. 3,330–336

161 Renger, J.J. et al. (2001) A developmental switch in neurotransmitterflux enhances synaptic efficacy by affecting AMPA receptoractivation. Neuron 29, 469–484

162 Schoch, S. et al. (2002) RIM1alpha forms a protein scaffold forregulating neurotransmitter release at the active zone. Nature 415,321–326

163 Kaeser, P.S. et al. (2011) RIM proteins tether Ca2+ channels topresynaptic active zones via a direct PDZ-domain interaction. Cell144, 282–295

164 Hoppa, M.B. et al. (2012) Multivesicular exocytosis in rat pancreaticbeta cells. Diabetologia 55, 1001–1012

165 Sheng, J. et al. (2012) Calcium-channel number critically influencessynaptic strength and plasticity at the active zone. Nat. Neurosci. 15,998–1006