Molecular Psychiatry (2005) 10, 185–200 & 2005 Nature Publishing Group All rights reserved 1359-4184/05 $30.00 www.nature.com/mp FEATURE REVIEW Active zones for presynaptic plasticity in the brain P Garcı´a-Junco-Clemente1, P Linares-Clemente1 and R Ferna´ndez-Chaco´n1 Departamento de Fisiologı´aMe´dica y Biofı´sica, Facultad de Medicina, Universidad de Sevilla. Avda. Sa´nchez-Pizjua´n4, Sevilla, Spain

Some of the most abundant in the brain such as the synapses formed by the hippocampal mossy fibers, cerebellar parallel fibers and several types of cortical afferents express presynaptic forms of long-term potentiation (LTP), a putative cellular model for spatial, motor and fear learning. Those synapses often display presynaptic mechanisms of LTP induction, which are either NMDA receptor independent of dependent of presynaptic NMDA receptors. Recent investigations on the molecular mechanisms of release modulation in short- and long-term synaptic plasticity in central synapses give a preponderant role to active zone proteins as Munc-13 and RIM1-alpha, and point toward the maturation process of synaptic vesicles prior to Ca2 þ -dependent fusion as a key regulatory step of presynaptic plasticity. Molecular Psychiatry (2005) 10, 185–200. doi:10.1038/sj.mp.4001628 Published online 4 January 2005 Keywords: synaptic plasticity; LTP; active zone; neurotransmitter release; RIM; Munc13

Introduction from hours to weeks. Long-term plasticity changes take place either as an enhancement of synaptic A remarkable property of neuronal circuits is the strength, LTP, or a reduction, long-term depression ability to change the strength of their synaptic (LTD). The long-term synaptic plasticity model more connections in response to prior activity. Synaptic intensively studied, LTP’ was initially described at the plasticity leads to the enhancement or depression of .3 Although LTP has been virtually neural transmission efficacy and influences a number observed in every glutamatergic studied in of important aspects of brain function. Paradigmati- the mammalian brain, the best studied model is the cally, synaptic plasticity is envisioned as the cellular NMDA (N-methyl-D-aspartate) receptor-dependent pillar to explain formation and spatial, motor LTP in the CA1 region of the hippocampus. For many and fear learning. However, changes in synaptic years, the most controversial questions regarding LTP efficacy might also participate in relevant processes have been related to the issue of whether LTP as synapse establishment during brain development induction and/or expression occurs as a pre- or post- or homeostatic maintenance of excitability to avoid synaptic phenomena. Presynaptically expressed plas- neuronal excytotoxicity. ticity generally translates into an increase or a Traditionally, classification of different types of decrease of neurotransmitter release. Postsynaptically synaptic plasticity attends to the time scale at which expressed plasticity involves changes at the number or they occur. Short-term synaptic plasticity changes the properties of postsynaptic receptors with no happen as potentiation (STP) or depression (STD) of changes in the amount of neurotransmitter release. synaptic responses, and they last from hundreds of Although it is generally accepted that the site of miliseconds to a few minutes. According to their induction and expression of NMDA-dependent LTP is duration and kinetics, short-term enhancements are basically postsynaptic, the existence of presynapti- defined as facilitation, augmentation or post-tetanic cally induced LTP based on the activation of potentiation. Short-term plasticity is generally ex- presynaptic NMDA receptors has been recently de- plained by direct presynaptic effects as residual monstrated.4 Additionally, a number of synapses have elevations of intracellular Ca2 þ concentration, for been described during the last decade where LTP enhancing changes, or depletion of some pool of turned out to be NMDA independent, being induced readily releasable vesicles, for depressing changes and expressed at the presynaptic terminals. Indeed, (reviewed Regehr and Stevens,1 and Zucker and presynaptic LTP is expressed in some of the most Regehr2). In contrast, long-term plasticity changes last abundant synapses in the brain such as the hippo- campal mossy fibers, the cerebellar parallel fiber and Correspondence: Dr R Ferna´ndez-Chaco´n, Departamento de corticothalamic terminals. The best studied example Fisiologı´aMe´dica y Biofı´sica, Facultad de Medicina, Universidad of presynaptic LTP occurs at the mossy fibers in the de Sevilla. Avda. Sa´nchez-Pizjua´n 4, Sevilla, Spain. CA3 region of the hippocampus. Excellent reviews E-mail: [email protected] Received 22 October 2004; revised 26 October 2004; accepted 10 have described general mechanisms of synaptic November 2004 plasticity and discussed the controversies of the site Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 186 of expression of LTP.5–9 In this review, we will action potential compared to the first.1 It is accepted describe different synaptic types displaying distinct that synapses with a low probability of release (Pr) forms of presynaptic plasticity where an enhancement normally present PPF, while synapses with a high Pr of neurotransmitter release occurs. We will then tend instead to have paired-pulse depression.1 discuss recent advances that point out putative During mossy fiber LTP, Zalutsky and Nicoll molecular targets in the synaptic vesicle cycle as observed a decrease in PPF consistent with an responsible for plastic changes in neurotransmitter elevation in Pr. In contrast, no similar changes were release, which are important for short- and long-term observed during associational-commisural LTP. synaptic potentiation. Although a number of studies supported that hypoth- esis,21–26 there are other reports indicating that the mechanism of mossy fiber LTP induction involves an Presynaptic LTP independent of NMDA receptor initial rise in postsynaptic Ca2 þ -concentration.27–30 Hippocampal mossy fibers However, the presynaptic expression of mossy fiber Based on studies of the possible role of LTP in LTP is unanimously accepted and several groups have learning and memory, it has been stated the so-called provided solid evidences supporting that view: a synaptic plasticity and memory (SPM) hypothesis: decrease in PPF during mossy fiber LTP,20,31–33 an ‘activity-dependent synaptic plasticity is induced at increase in quantal content (the number of vesicles appropriate synapses during memory formation and released during a presynaptic action potential)33 and a is both necessary and sufficient for the information significant reduction in the number of release fail- storage underlying the type of memory mediated by ures.33 Synaptic release probability can be directly the brain area in which that plasticity is observed’.10 measured as the rate of inhibition of synaptic In that sense, hippocampal LTP is probably involved responses using MK-801, an irreversible open-chan- in the initial encoding and storage of spatial memory. nel blocker of the NMDA receptor.34,35 MK-801 only There is general agreement regarding the existence of blocks irreversibly NMDA receptors that have been different types of LTP in the hippocampus.6 Nowa- opened by glutamate and, therefore, synapses that are days, many investigators accept that LTP at the CA1 activated become unavailable to subsequent stimuli. region of the hippocampus is triggered at the post- The rate of decline of the NMDA-dependent EPSCs synaptic terminals and an important mechanism of (excitatory post synaptic currents) depends on the Pr, expression resides in the trafficking and function of because a higher Pr results in a higher number of glutamate receptors of AMPA (alpha-amino-3-hydro- NMDA receptors being opened per stimulus. xy-5-methyl-4-isoxazole propionic acid) type.11,12 Although mossy fiber LTP is NMDA receptor inde- Nevertheless, a number of experiments led to propose pendent, Weisskopf and Nicoll36 showed that NMDA presynaptic mechanisms of LTP expression such as receptors are present at mossy fiber synapses and the existence of retrograde messengers able to modify used the MK-801 method to reveal that upon LTP presynaptic function.13–18 In any case, a form of induction, there is a large increase in the probability hippocampal LTP exists, which is independent of of release,37 emphasizing the presynaptic expression NMDA receptors and genuinely expressed presynap- of that type of LTP. Although most of the studies were tically at the mossy fiber pathway formed between the performed in slices, the presynaptic nature of mossy axons of granule cells of the dentate gyrus and the fiber LTP has been also shown in cultured granule pyramidal cells of area CA3.19,20 Although it is cells.38,39 generally accepted that mossy fiber LTP expression An important issue is to define the synaptic is presynaptic, there are controversies regarding mechanisms that get modified to increase neurotrans- whether the induction is pre- or post-synaptic. mitter release. Tong et al.38 using cultured granule Zalutsky and Nicoll20 took advantage of the particular cells showed that after total blockade of the evoked property of CA3 pyramidal neurons, which displays NMDAR EPSC by MK-801, the application of the LTP- LTP at two types of synaptic inputs from both the inducing tetanic stimulus resulted in the appearance associational–commisural fibers and the mossy fiber of an NMDAR EPSC that remained during the rest of pathway, to study in parallel the induction and the experiment (more than 30 min).38 Those results expression of LTP at those two different synapses. were consistent with an increase in the number of They showed that although the induction of the release sites and they could be explained either by a associational–commisural pathway required the post- postsynaptic insertion on new NMDA receptors or the synaptic activation of NMDA receptors, membrane activation of previous presynaptically silent sites. depolarization and elevation, mossy fiber Tong et al concluded that the effect was, in part, due LTP did not require any of those effects to be to presynaptic appearance of release sites because the induced.20 In addition, they studied paired-pulse full blockade of functional NMDA receptors with MK- facilitation (PPF) in mossy fibers. PPF is a form of 801 (for open channels) and NMDA (for the non- synaptic plasticity observed in many synapses when blocked with MK-801) abolished the appearance of two action potentials arrive at a presynaptic terminal any NMDAR EPSC upon tetanic stimulation. If the separated by a short time interval (ten to hundred insertion of new NMDA receptors would have milliseconds) and characterized by a higher neuro- occurred, they should have detected an NMDA- transmitter release upon the arrival of the second receptor-dependent current. Those results supported

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 187 the hypothesis that one of the mechanisms triggered release sites which includes the activation of pre- by LTP at mossy fiber is the organization or activation viously presynaptically silent or ‘mute’ synapses. of, at least, the presynaptic machinery responsible for In dense core vesicles, the release of transmitters the efficient stimulus-secretion coupling and neuro- happens through the opening of a fusion pore, which transmitter release at previously presynaptically opens either reversibly (kiss-and-run) or irreversibly. silent or ‘mute’ synapses.38,40,41 In mast cells, during reversible fusions, there is a Although a recent study has reported direct partial transmitter release, while upon full irreversi- measurements of synaptic vesicle exocytosis from ble fusion the whole vesicle content is released.48 single mossy fiber terminals by means of membrane Smaller vesicles from chromaffin cells are able to capacitance measurements,42 most of the studies release the whole neurotransmitter content during performed to analyze the synaptic plasticity mechan- reversible fusions.49 Recent experiments support the isms at mossy fibers are based on electrophysiological existence of kiss-and-run in synapses formed between analysis of the simultaneous activity of multiple cultured hippocampal neurons.50,51 Interestingly, the synapses. The postsynaptic terminals contacted by rate of fusion pore expansion is regulated by Ca2 þ and mossy fibers are complex dendritic spines known as phorbol ester receptors.52–54 Modulation of the fusion thorny excrescences,43 which are interdigitated with pore opening has been proposed as a presynaptic protrusions of the mossy fiber terminals, presenting mechanism for synapse unsilencing during LTP in discrete postsynaptic densities apposed to multiple CA3-CA1 hippocampal synapses.55 Choi et al ob- (up to 30–35) presynaptic release sites (reviewed in served silent synapses with responses only mediated Henze et al44), which normally present multiquantal by NMDA receptors but not by AMPA receptors. Since release.45 In order to monitor changes of presynaptic NMDA receptors have a higher affinity for glutamate sites at single synapses, Reid et al46 have studied than AMPA receptors, at low glutamate concentra- Ca2 þ - transients at single thorny excrescences.46 Reid tions in the synaptic cleft only NMDA receptors might et al have characterized, in organotypic hippocampal be activated. That would be the case for a so-called cultures, the Ca2 þ transients at single thorny excres- ‘whispering’ synapse,40 a synapse that is releasing low cences elicited by synaptic activity and have shown amounts of . Choi et al reported an that Ca2 þ transients result mainly from AMPA increase of glutamate concentration at the synaptic receptor-gated Ca2 þ influx through voltage-activated cleft during LTP manifested as a conversion of slow- calcium channels on the spine. Those Ca2 þ transients rising EPSCs to fast-rising EPSCs. They proposed that have a very small NMDA-receptor component and do LTP induces a switch in fusion pore operation from a not depend on Ca2 þ release from intracellular nonexpanding mode in basal conditions to a fully stores.47 They established that the probability of expanded state, which leads to the release of full obtaining an excitatory postsynaptic Ca2 þ -transient vesicle contents and increases glutamate concentra- (EPSCaT) reflects the aggregate probability of release tion at the synaptic cleft. It has been proposed that a of the multiple release sites located at the imaged similar mechanism might underlie the conversion of synapse. Interestingly, they observed a clear shift to immature silent synapses into mature functional bigger EPSCaT size upon LTP induction, which they synaptic contacts.56 Although Choi et al.55 favored attributed to an increase in quantal content probably the fusion pore explanation, they did not rule out that due to the recruitment of additional functional release the slow-rising EPSCs could be generated instead by sites and multivesicular release at single sites, with spillover from neighbor synapses rather than from a no changes in quantal amplitude normally produced slow releasing fusion pore. Recently, Nielsen et al.57 by an increased number of postsynaptic receptors. have reported that at cerebellar mossy fiber-granule Remarkably, they monitored synapses with null cell synapses glutamate spillover from other synapses probability of obtaining EPSCaT, even after paired- is the main cause for slow-rising EPSCs. Novel pulse stimulation, which upon LTP induction were experimental approaches such as the use of geneti- successfully activated and gave rise to EPSCaTs. cally modified mice encoding reporters to monitor Those cases were interpreted as LTP-mediated sy- neurotransmitter release at single synapses might be napse unsilencing. Although the authors could not very useful to analyze the mechanisms of synapse exclude that the unsilencing reflects insertion of unsilencing during LTP. receptors at a postsynaptically silent synapse, they An enormous amount of data support the role of considered that possibility unlikely because of two Ca2 þ influx through postsynaptic NMDA receptors in reasons: (1) application of 0 Mg2 þ to activate NMDA triggering LTP at the hippocampal CA1 region. Ca2 þ receptors failed to uncover subsequently unsilenced binds to (CaM) to activate calmodulin- synapses and (2) they observed that those spines were kinase II (CaMKII), which gets autophosphorylated. equipped with AMPA receptors as visualized with CaMKII remains activated when Ca2 þ returns to basal immunocytochemistry. In conclusion those studies levels and increases the conductance and trafficking indicate that mossy fiber LTP is expressed as a of postsynaptic AMPA receptors (reviewed in Mal- potentiation of neurotransmitter release at single enka et al,8 Bredt and Nicoll,12 Malenka and Nicoll,58 synapses resulting from an increase of probability of Nicoll59 and Griffith60). A considerable number of release from functionally active sites, an increase in genetic studies in mouse have provided a wealth of multivesicular release and/or the recruitment of new information of the role of alpha-CaMKII in plasticity

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 188 and cognition.61 Recently, presynaptic CaMKII have stimulation in the absence of climbing fiber activa- also been implicated in the control and plasticity of tion.79–85 In a very similar manner as it happens at the neurotransmitter release in hippocampal neurons.62,63 hippocampal mossy fiber, parallel fiber LTP seems to Hinds et al62 have generated region specific knockout be triggered by a presynaptic Ca2 þ influx, and it is mice lacking alpha-CaMKII at the hippocampal CA3 resistant to blockers of glutamate receptors and Ca2 þ area and therefore, in those mice, CA3-CA1 synapses chelation at postsynaptic Purkinje cells.81,86–88 In were deprived of presynaptic alpha-CaMKII. At the addition, cerebellar granule cells in culture stimulated CA3-CA1 synapses and in basal conditions the CA3- to release neurotransmitters were shown to activate alpha-CaMKII KO mice have normal PPF ratio, glutamate-evoked inward currents in adjacent glial normal Pr and no changes in the releasable pool of cells driven by AMPA/kainate receptors. Interestingly, synaptic vesicles. Remarkably, upon repetitive stimu- under those experimental conditions and upon low- lation, the mutant synapses showed significantly frequency stimulation, it was possible to induce LTP higher levels of synaptic facilitation due to an in glial cells with similar characteristics as in parallel increased Pr during maintained stimulation. Hinds fiber–Purkinje cells LTP.86 Although the results in et al then concluded that, in wild-type synapses, cultured neurons were consistent with LTP induced presynaptic alpha-CaMKII functions as a negative and expressed at the presynaptic terminals, the regulator of Pr during high-frequency presynaptic postsynaptic currents in glia and Purkinje cells were stimulation, which could serve as a regulatory mediated by AMPA/kainate receptors and because of mechanism to maintain Pr in an optimal range for that a possible postsynaptic mechanism of LTP normal synaptic operation. expression linked to AMPA/kainate receptors it could Which are the biochemical mechanisms responsi- not be ruled out. A confirmation of the presynaptic ble for induction and expression of presynaptic origin came from a set of experiments where synapti- mossy fiber LTP? The current view is that presynaptic cally evoked glutamate release was monitored as Ca2 þ influx activates a calcium/calmodulin-regulated currents of electrogenic glial glutamate transporters, adenylyl cyclase, which increases cAMP levels and which were activated upon binding of released activates protein-kinase A (PKA).64 Several proteins glutamate. LTP of isolated transport currents was of the synaptic vesicle cycle are substrates for PKA evoked by stimulation at 4 Hz and was blocked by and their relevant significance in long-term plasticity removal of external Ca2 þ during stimulation. LTP will be discussed below. A number of studies have expression was associated with a decrease in the rate evidenced the role of that biochemical cascade in of synaptic failures and a decrease in PPF, which mossy fiber LTP (Huang et al,32 Weisskopf et al,37 and strongly suggested that parallel fiber LTP is expressed Huang and Kandel,65 but see also Kamiya et al,66). presynaptically.87 Similarly as it happens at the Different types of presynaptic glutamate receptors hippocampal mossy fibers, LTP in parallel fiber– have an important role in presynaptic mossy fiber Purkinje cell synapses is mediated by presynaptic LTP67–69 and recent excellent reviews have discussed Ca2 þ influx, which activates Ca2 þ sensitive adenylyl that issue in detail.70–73 Although mossy fibers have cyclase to increase cAMP levels and activates probably been the most extensively studied model of PKA.81,86,88,89 The potentiation of glutamate release presynaptic LTP, there are other synapses in cerebel- putatively occurs upon the PKA-mediated phosphor- lum and neocortex that share similar mechanisms. ylation of presynaptic proteins involved in neuro- transmitter release, as it will be discussed below. Cerebellar parallel-Purkinje cell synapses The cerebellum is a very attractive and useful Corticothalamic fibers structure to study cellular and molecular mechanisms The major route for afferents to the neocortex of motor learning and memory. Together with a very originates in the thalamus. Every message that arrives well-known neuronal circuit, in cerebellum, there are to the neocortex goes through the thalamus first. several behavioral tests as eye and eyelids movements Reciprocally, the thalamus receives massive input that are easily quantified to evaluate behavioral from the neocortex via corticothalamic fibers. The conditioning. An important variety of synaptic plas- thalamus is therefore proposed as a crucial structure ticity types, in both excitatory and inhibitory sy- for cortico-cortical communication (reviewed in Sher- napses, have been described in the cerebellum man et al90 and Guillery and Sherman91). Repetitive (reviewed in Hansel et al74). Each Purkinje cell stimulation of corticothalamic fibers produces STP receives B200 000 excitatory synapses from the axons and LTP in vitro an in vivo,92–99 indicating that of granule cells, the parallel fibers. In neural models corticothalamic synapses modify the effectiveness for associative eyelid conditioning, parallel fibers with which the neocortex influences the thalamus. drive the conditioned, and climbing fibers the un- Castro-Alamancos and Calcagnotto96 have studied conditioned stimuli into Purkinje cells. Parallel fiber LTP in mouse brain slices at the glutamatergic LTD has been related to the behavioral acquisition of a synapses formed by corticothalamic fibers contacting conditioned response,75–77 while LTP seems to parti- the ventrobasal thalamus. Several observations led cipate in the extinction of the conditioned response them to conclude that corticothalamic LTP (ct-LTP) is (reviewed in Medina et al78 and Hansel both induced and probably expressed presynapti- et al74). Parallel fiber LTP occurs upon low-frequency cally: (1) ct-LTP was resistant to

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 189 antagonists, which strongly suggested independence et al4; reviewed in Pare110). In mouse brain slices, of postsynaptic Ca2 þ or postsynaptic depolarization; Humeau et al used stimulation patterns containing (2) ct-LTP was not affected by NMDA-receptor Poisson-distributed stimuli to mimick physiological blockade or disinhibition; (3) ct-LTP required extra- activity of converging thalamic and cortical afferents cellular Ca2 þ , suggesting that a rise in presynaptic during fear conditioning. They induced LTP at Ca2 þ is necessary; and (4) during ct-LTP there is a cortical but not at thalamic afferents only when both decrease in PPF, suggesting an increase in the release thalamic and cortical afferents were simultaneously probability. In addition, as it has been observed in stimulated (Figure 1a); the independent stimulation hippocampal mossy fibers and parallel fibers–Pur- of either thalamic or cortical inputs failed to induce kinje cell synapses, ct-LTP is abolished by PKA LTP at cortical afferents, indicating that it is associa- blockers. Furthermore, the forskolin-mediated activa- tive LTP. Cortical LTP induction was blocked in the tion of adenylyl cyclase causes a long-lasting poten- presence of a competitive NMDA receptor antagonist tiation with a characteristic decrease in PPF. and evoked upon a puff-application of NMDA coin- The synapses described above are independent of cident with train stimulations of cortical afferents, NMDA receptors. Nevertheless, the existence of indicating an NMDA-receptor-dependent LTP. Sur- excitatory and inhibitory synapses where neurotrans- prisingly, chelation of postsynaptic Ca2 þ and com- mitter release is potentiated through the activation of plete blockade of postsynaptic NMDA receptors, by presynaptically located NMDA receptors has been intracellular application of the NMDA-receptor open recently described. channel blocker MK-801, did not interfere with the induction of cortical LTP. In contrast, the application of a membrane permeant Ca2 þ chelator acting on the Presynaptic LTP dependent of presynaptic NMDA presynaptic terminals abolished LTP. Those results receptors suggested that LTP induction is presynaptic. Further- A number of papers have described the existence of more, they observed that upon LTP induction the presynaptic NMDA receptors and their functional cortical terminals underwent a persistent decrease in influence of neurotransmitter release.100,101 In addi- PPF, which is consistent with an increase in Pr. tion, it has been demonstrated that presynaptic Indeed, they demonstrated that increase of Pr with NMDA receptors are also implicated in several types external high Ca2 þ concentration occluded LTP of plasticity. Two recent studies have indeed shown induction. Since homosynaptic LTP at cortical affer- that presynaptic NMDA receptors enhance neuro- ents has been previously shown to be mediated by an transmitter release in both excitatory glutamatergic increase in Pr, Humeau et al hypothesized that both synapses in the lateral amygdala4 and in GABAergic homosynaptic and heterosynaptic LTP might share a inhibitory synapses formed between interneurons and common mechanism of presynaptic expression. In Purkinje cell in the cerebellum.102 support of that hypothesis, they found that Poisson- induced heterosynaptic LTP could be suppressed Heterosynaptic associative plasticity in cortical upon induction of homosynaptic LTP by pairing afferents in the lateral amygdala afferent stimulation with postsynaptic stimulation. The lateral amygdala is considered a region where Humeau et al proposed that glutamate released by synaptic plasticity takes place for the acquisition of thalamic afferents might directly activate NMDA conditioned fear responses.103 In vivo studies in receptors located on presynaptic terminals of cortical amygdala probably provide the strongest evidences afferents, which is consistent with previous electron of a relation between LTP and learning (reviewed in microscopy studies reporting presynaptic NMDA Stevens104). Several groups have demonstrated that receptors at the cortical afferents in the lateral LTP of cortical or thalamic afferents in the amygdala amygdala.111 An important question arising was why is postsynaptically induced through NMDA receptors it was not possible to induce cortical LTP stimulating and it is expressed presynaptically as an increase of only the cortical afferents? Humeau et al thought that Pr upon presynaptic and postsynaptic paired activa- glutamate released by cortical afferents might be tion.105–108 Classically, the induction of associative rapidly cleared from the synaptic cleft, and indeed, synaptic plasticity requires the coincident activation their hypothesis was supported by an experiment of both presynaptic and postsynaptic terminals. In showing that cortical LTP could be elicited stimulat- homosynaptic plasticity, the intrinsic activity of the ing only cortical terminals in the presence of a low synapse itself modifies its own functional state. On concentration of a glutamate uptake blocker (Figure the other hand, in heterosynaptic plasticity, the 1b). That study is a clear demonstration of LTP synaptic activity between two neurons is modified mediated by NMDA presynaptic receptors. by the extrinsic action of a third neuron. Hetero- synaptic plasticity has been generally considered as a GABA release enhancement at cerebellar nonassociative process and independent of synaptic interneuron-Purkinje cell synapses activity.109 A recent study of LTP in lateral amygdala One of the most powerful synaptic contacts in the has uncovered a novel mechanism of heterosynaptic brain happens at the synapses formed by climbing and associative plasticity genuinely induced and fibers on Purkinje cells. A single climbing fiber axon expressed at the presynaptic terminals (Humeau forms B1400 synapses with each Purkinje cell.

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 190 Excitatory activity driven by climbing fibers is allows Purkinje cells to downregulate GABA release counterbalanced by inhibitory GABAergic interneur- through the retrograde release of endocannabinoids ons that project to the axo-somatic regions of Purkinje that activate presynaptic cannabinoid receptors (CB1) cells. An interesting mechanism, known as depolar- located in GABAergic interneurons. After the initial ization-induced suppression of inhibition (DSI), stimulation, CB1 receptor activation reduces both probability of release and the overall synaptic efficacy for tens of seconds112–116 (Figure 1c, d). Duguid and Smart102 have recently identified a novel long-lasting mechanism (several minutes) termed depolarization- induced potentiation of inhibition (DPI), which follows the termination of DSI and implies the potentiation of GABA release, reviewed in Tzingou- nis and Nicoll117 (Figure 1c). Interneuron-Purkinje cell synapses in the basal state present paired-pulse depression as expected in a synapse with a high release probability.118 Duguid and Smart observed PPF during DSI, presumably due to a decrease in release probability. In contrast, during DPI, they reported an increase in paired-pulse depression consistent with an increased release probability, therefore indicating a presynaptic mechanism. Inter- estingly, DPI turned out to be blocked by the NMDA- receptor antagonist APV and promoted by glutamate- uptake blockade. In addition, the authors observed that Ca2 þ chelation in Purkinje cells abolished DPI.

Figure 1 Presynaptic NMDA receptors mediates potentia- tion of neurotransmitter release in excitatory and inhibitory neurons. (a) Presynaptic induction of heterosynaptic asso- ciative plasticity revealed as changes in EPSP after simultaneous stimulation (arrow) of thalamic (open blue circles) and cortical (filled red circles) afferents at lateral amygdala (scale bars, 2 mV, 50 ms) (modified from Humeau et al4 with permission; copyrightr 2003 Nature Publishing Group). (b) Glutamate released by thalamic afferents activate NMDA receptors localized in presynaptic terminals of cortical afferents. It is still unknown how glutamate spillover or axo-axonic synapses might contribute to this process. NMDA receptor activation increases intracellular Ca2 þ concentration, which leads to an enhancement in Pr at cortical synapses only when cortical and thalamic afferents are simultaneously activated. (c) Changes in mIPSC fre- quencies after a train of conditioning stimuli (arrow) in inhibitory synapses between interneurons and Purkinje cells. The stimulus produces first a depolarization-induced suppression of inhibition (DSI) followed by a depolariza- tion-induced potentiation of inhibition (DPI) (modified from Duguid and Smart102 with permission; copyrightr 2004 Nature Publishing Group). (d) The stimulation of climbing fibers terminals activates voltage-gated calcium channels (VGCC) at postsynaptic Purkinje cells terminals, which increases intracellular Ca2 þ concentration. This enhancement in Ca2 þ concentration produces two different effects in a time-dependent manner. During the first 10 s, Ca2 þ triggers endocannabinoid synthesis, which inhibits GABA release (dotted red arrow) through the activation of presynaptic CB1 receptors (1, red arrows). Next, Purkinje cells release glutamate by unknown mechanisms and presynaptic activated NMDA receptors mediate an enhance- ment of GABA release through Ca2 þ -induced calcium release from intracellular stores (modified from Tzingounis and Nicoll117 with permission; copyrightr 2004 Nature Publishing Group).

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 191 The authors proposed that DPI occurs through a new probably includes several stages within the priming retrograde signaling pathway, which involves gluta- process123 where the active zone proteins Munc-13- mate release from Purkinje cells and activation of 1124,125 and RIM1-alpha play fundamental roles.126 NMDA receptors in the presynaptic terminal of 102 GABAergic interneurons (Figure 1d). In this study Molecular targets at the synaptic vesicle cycle they did not investigate how glutamate might be released from Purkinje cell dendrites, which consti- Munc13 proteins tutes an important question to be answered. They Munc13 proteins (Munc13-1, Munc13-2, Munc13-3) nevertheless, emphasize the importance of maintain- are active zone proteins encoded by three different ing the glutamate concentration tightly controlled at genes that present two or three C2 domains and one the synaptic cleft and they remark on the putative role C1 domain.127 A number of genetic studies in the of Bergman glia in that process. Bergman glia express nematode,128 Drosophila129 and mice have established high levels of glutamate transporters, which rapidly that Munc13 proteins are essential for priming and sequesters any released glutamate, encapsulating the generation of the readily releasable pool (RRP) of Purkinje cell dendritic arbor limiting glutamate synaptic vesicles.130,124 Priming is a biochemical diffusion out of the synaptic cleft toward presynaptic event necessary for the vesicles to become fusion NMDA receptors. Therefore, DPI would be activated competent and it is a step required subsequent to the only in situations of high activity when glutamate ultrastructural appearance of docked vesicles. The transporters may became transiently saturated. Ca2 þ RRP include the synaptic vesicles competent for influx through presynaptic NMDA receptors seems to fusion upon the arrival of an action potential to the trigger Ca2 þ -induced Ca2 þ release (CICR) from pre- presynaptic terminal and the RRP size directly reports synaptic ryanodine-sensitive stores, which have pre- the efficiency of priming.131 Munc13s unfold and viously been implicated in spontaneous119 and activate the SNARE protein syntaxin and promote the evoked GABA release.120 The authors have shown formation of the SNARE complex.131–133 In addition, that DPI can be elicited either by direct voltage-clamp recent studies have shown that Munc13 proteins play depolarization of Purkinje cells or through a physio- an important role in potentiation of neurotransmitter logical stimulation of climbing fiber synaptic inputs. release in short-term synaptic plasticity.125 Rosen- A surprising aspect of the novel mechanism is that mund et al125 using knockout mice for Munc13-1 and NMDA receptors get activated in the presence of Munc13-2, have shown that in cultured hippocampal Mg2 þ concentrations (1 mM) prone to voltage-depen- neurons there are two functional classes of synapses dent block of NMDA receptors, and without any depending on the differential localization of Munc13- previous depolarization. As possible reasons, the 1 and Munc13-2 in the same neuron along a single authors argue that the subunit composition of those axon. Synapses equipped with Munc13-1 exhibit NMDA receptors might make them less sensitive to depression during a train of action potentials (10 Hz) Mg2 þ blockade or that the input resistance and small and show very small or no augmentation upon high- volume of interneuronal axons allow small glutamate- frequency stimulation. In contrast, Munc13-2 depen- activated currents to depolarize the terminal and dent synapses respond with facilitation and augmen- relieve the Mg2 þ block. tation under the same stimulation conditions. DPI constitutes a novel mechanism for Purkinje Rosenmund et al remarked that those properties of cells to regulate their excitability modifying the different Munc13 isoforms in the same axon, might strength of inhibitory synaptic inputs during repeti- dynamically change the short-term plasticity proper- tive climbing fiber stimulation. ties in different synapses by switching between The examples described above illustrate the im- Munc13-1 and Munc13-2. Those findings mean that portance and the different repertoire of synaptic types an augmenting synapse could be transformed into a where the potentiation of neurotransmitter release depressing one or vice versa and suggest that such a occurs as a crucial phenomenon for LTP. An attractive mechanism could participate in the establishment of hypothesis is that presynaptic proteins controlling long-lasting plasticity changes. the synaptic vesicle cycle are key molecular targets for Which signaling pathways act on Munc13s? those mechanisms. During the synaptic vesicle cycle, Munc13 proteins are multimodular proteins targeted synaptic vesicles go through different steps.121,122 by different signaling cascades. Munc13 proteins bind Vesicles filled with neurotransmitters form a reserve diacyl-glicerol (DAG) and beta-phorbol-esters (beta- vesicle pool. Some of them get docked to the plasma PE)134 and calmodulin.135 Phorbol ester binding membrane and maturate in a priming reaction to translocates Munc13 proteins to the plasma mem- become fusion competent. Upon the arrival of an brane.134,136 Interestingly, the high-frequency action potential, Ca2 þ influx through calcium depen- mediated augmentation attributed to Munc13-2 is dent voltage channels triggers the opening of the mirrored by application of phorbol esters.125 fusion pore to release neurotransmitters. Vesicles are Rhee et al studied synaptic transmission in knockin recycled by reversible closure of the fusion pore (‘kiss mice with a punctual mutation in the Munc13-1 C1 and run’) or membrane retrieval through clathrin- domain, which abolishes DAG and beta-PE binding.137 mediated endocytosis. The synaptic vesicle matura- The mutation leads to neonatal death indicating that a tion to reach Ca2 þ -dependent fusion competence functional C1 domain is essential for life. Normally

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 192 wild-type synapses present augmentation in response versions elicited a Ca2 þ -dependent increase in RRP to phorbol-esters applications, in contrast to the size and Pvr upon 10-Hz stimulation. In contrast, the knockin mice that did not have any increase in rescue with the mutant version revealed a normal neurotransmitter release induced by phorbol-esters. increase in Pvr but a decrease in RRP size. Junge et al Rhee et al demonstrated that, instead of PKC, Munc13 concluded that the Munc13/CaM complex is neces- proteins are the main presynaptic DAG/beta PE sary for Ca2 þ -dependent refilling of RRP, while the receptors mediating augmentation in hippocampal increase in Pvr happens as a separate mechanism that neurons. In cultured hippocampal neurons, the might be attributed to a low-affinity secretory Ca2 þ release of the RRP of synaptic vesicles is elicited sensor. In a model where the size of the RRP is upon stimulation with hypertonic sucrose extracellu- determined by the equilibrium between the priming lar solutions. The ratio of the release evoked by an and the unpriming rates, the increase of RRP size will action potential vs the release evoked by sucrose gives occur through the acceleration of priming rate, which the so-called vesicular release probability (Pvr).138 might be mediated by the Munc13/CaM complex. According to their sensitivity to DAG/beta-PE binding These findings constitute an important step toward to the C1 domain, Munc13-1 determines two different the understanding of the molecular mechanisms of synaptic vesicle pools. One pool is a low Pvr pool, short-term synaptic plasticity. Several studies have which increases its size upon DAG/beta-PE stimula- reported synaptic activity-induced refilling in cere- tion or by ongoing synaptic activity, replenishes bellar climbing fibers, hippocampal neurons and the quickly and contributes very little during low-fre- calyx of Held,140–142 which, according to the findings quency stimulation. That pool depends on Munc13-1 of Junge et al, is most likely triggered by the Ca2 þ - soluble molecules, which are not associate with other dependent association between Munc13s and CaM. active zone regulatory components. The other pool is a Sakaba and Neher143 have described, in the calyx of high Pvr pool independent of DAG/beta-PE, which Held, the existence of two releasable vesicle pools replenishes slowly, is readily releasable and is the with different kinetic characteristics: (1) a pool main contributor in evoked release during low- rapidly released that refills slowly and (2) a pool frequency stimulation. That second pool depends on slowly released but rapidly refilled. Remarkably, they insoluble Munc13-1 molecules, which reside at the have shown that the refilling of the rapidly releasing active zone and are associated with the release pool is dramatically accelerated by Ca2 þ /CaM, sug- machinery. The high Pvr pool constitutes the RRP of gesting that indeed this process might be mediated by synaptic vesicles. The low Pvr pool is used to increase the Munc13/CaM complex. the number of primed vesicles through DAG-mediated Munc13-1 is generally distributed throughout the activation during periods of high synaptic activity.139 brain and knockout mice die almost immediately after A likely regulatory mechanism for Munc13 proteins is birth,124,130 as it happens for the knockin mutation, that Ca2 þ influx increase during high synaptic activity which abolishes DAG/beta-PE to the C1 domain. Such leads to DAG production through Ca2 þ -dependent- a phenotype does not allow to study the behavioral PLC activation. The lethal phenotype of knockin mice changes, as for example learning and memory altera- with a mutated Munc13 C1 domain indicates that the tions, which might happen in the absence of Munc13- activity-dependent priming mediated by Munc13-1 is 1. It would be of great interest to evaluate in vivo the probably a crucial phenomenon for the proper func- effect of the mutation that abolishes the formation of tion of neuronal networks, for example neuronal the Munc13/CaM complex, although it might also circuits involved in basic body functions. turn out to be lethal. Interestingly, Munc13-2 mice are It has been recently reported that an important viable but no behavioral studies have been reported functional regulation of Munc13 proteins is mediated yet on those mice. However, the isoform Munc13-3 is by calmodulin binding. Junge et al135 have demon- almost exclusively expressed in the cerebellum, and strated that Munc13-1 and the ubiquitously expressed knockout mice for Munc13-3 survive to adulthood.144 splice variant of Munc13-2, ubMunc13-2, binds Munc13-3 protein is localized to the synaptic neuro- calmodulin in a Ca2 þ -dependent manner. In order to phil of the cerebellar molecular layer, and since it is study the functional consequences of that interaction not found in Purkinje cell dendrites, it is most likely a in central synapses, Junge et al have overexpressed presynaptic protein at parallel fiber–Purkinje cell normal Munc13 proteins or mutant versions insensi- synapses. Munc13-3 deletion mutants exhibit in- tive to calmodulin in neurons from double knockout creased PPF at parallel fiber–Purkinje cell synapses, mice lacking Munc13-1 and Munc13-2. The resulting which suggest a decrease in Pr probably due to a rescued phenotypes revealed that mutants with priming defect. Augustin et al proposed that Munc13- disrupted calmodulin binding sites abolish the 3 functions as a priming factor, although with less short-term synaptic plasticity properties of Munc13 efficiency than Munc13-1. Interestingly, Munc13-3 proteins: (1) Munc13-1 W464R mutant displayed a knockout mice display normal spontaneous motor stronger depression and reduction of augmentation in activity but have an impaired ability to learn complex comparison with Munc13-1 wild type and (2) motor tasks. As it has been discussed above, in Munc13-2 W387R mutant displayed no facilitation parallel fiber-Purkinje cell synapses, LTP is induced and reduction of augmentation in comparison with and expressed presynaptically. Munc13-3 and its Munc13-1 wild type. The rescue with the wild-type priming action might underlie the potentiation of

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 193 neurotransmitter release upon LTP induction, a mice did not have any PPF phenotype, the function of hypothesis that could be tested in the Munc13-3 RIM1-alpha in those synapses is probably indepen- knockout mice. dent of Rab3a. Longer lasting forms of short-term synaptic plasticity were also altered in excitatory RIM1-alpha synapses lacking RIM1-alpha, which had an increase RIMs (Rab3 interacting molecules) constitute a family in post-tetanic potentiation and facilitation and less of multidomain adaptor proteins that were initially depression than wild-type controls at moderate discovered as putative effectors for the synaptic frequency stimulation (14 Hz). Synapses lacking vesicle protein Rab3a.145 In vertebrates, RIMs are Rab3A had a similar phenotype at 14 Hz stimulation encoded by four different genes and generate several but no changes for PTP, while Munc-13 heterozygous isoforms by alternative splicing.146 RIM1-alpha and synapses behaved as wild-type controls. Those RIM2-alpha consist of an N-terminal Zn2 þ -finger increases in short-term synaptic plasticity in RIM1- domain, a central PDZ and C2A domains, and a C- alpha mutants suggested a decrease in Pr, a fact that terminal C2B domain (Figure 2a). RIM1-alpha, the Schoch et al evidenced by a significantly slower rate best studied isoform, is located at the active zone of of inhibition with the open-channel blocker MK-801. the presynaptic terminals and forms a protein scaffold Those observations point toward RIM1-alpha as a interacting with key presynaptic molecules. In addi- regulator of Pr, especially during synaptic plasticity. tion to Rab3a, RIM1-alpha binds active zone proteins A recent study has approached a more detailed Munc13-1147,148 and ERCs,149,150 the synaptic vesicle investigation of the role of RIM1-alpha in presynaptic protein synaptotagmin1, a calcium sensor for fast mechanisms of neurotransmitter release in cultured synchronous release,151,152 and other synaptic pro- hippocampal neurons.158 Calakos et al have found teins as alpha-liprins.152 Indirectly, RIM1-alpha binds that neurons lacking RIM1-alpha have a reduction in to L-type calcium channels153 through the so-called synaptic transmission, characterized by a reduction RIM-BPs (RIM binding proteins).150 Insightful func- in the size of EPSCs charge. This is consistent with tional studies of RIM have been recently performed in the reported decreased of Pr152 and the authors C. elegans154 and in knockout mice.152,155,156,126 attributed it to an observed reduction of B50% in Schoch et al152 have shown that mice lacking RIM1- the size of the RRP of synaptic vesicles measured alpha survived to adulthood with slighter high upon stimulation with hypertonic solutions. Interest- mortality compared to wild-type littermates and ingly, Calakos et al found a normal Pvr in synapses displayed a poor maternal and nurturing behavior. lacking RIM1-alpha, which means that the primed They did not show any structural abnormalities in the vesicles in the remaining RRP undergo normal brain and did not have significant changes in the exocytosis. Since Munc13-1 is a key priming fac- levels of synaptic proteins, except for the active zone tor,124,125,130 they concluded that the missing 50% of protein Munc13-1 which was decreased by B60% the RRP could be explained because in RIM1-alpha probably because, since Munc-13 binds to RIM1- knockout mice, there is a substantial reduction in alpha, it degrades when RIM1-alpha is absent.147,157 Munc13-1 levels.152 That implies that there are Schoch et al systematically studied synaptic trans- probably two different subpools of synaptic vesicles mission at Schaffer collateral/commissural fibers within the RRP. The missing subpool of the RRP in excitatory synapses on CA1 pyramidal cells on RIM1-alpha KO mice probably corresponds in normal hippocampal slices from RIM1-alpha knockout mice conditions to a pool of synaptic vesicles activated by a and compared it in parallel with available Rab3A priming reaction where the interaction between homozygous and Munc13-1 heterozygous knockout Munc13-1 and RIM1-alpha is required. The remaining mice. Rab3A and Munc13-1 interact with RIM1- RRP in RIM1-alpha KO mice is a subpool likely alpha. Since Munc13-1 heterozygous mice had a independent of RIM proteins because immunocyto- 50% decrease in Munc13-1 levels, they turned out chemical analysis revealed that RIM2-alpha was to be a very appropriate mouse control to assess if absent from hippocampal cultures. These findings phenotypical alterations in RIM1-alpha knockout reveal a role for RIM1-alpha as an enhancer of mice are directly due to the lack of RIM1-alpha or to neurotransmitter release through the potentiation of the secondary decrease in Munc13-1 levels. Although synaptic vesicle priming.126 In addition Calakos et al all the mutants displayed robust evoked postsynaptic showed that RIM1-alpha participates in two other responses, they have different phenotypes in short- aspects of transmitter release: (1) the regulation of term synaptic plasticity. RIM1-alpha and Rab3a short-term synaptic plasticity and (2) the asynchro- knockout mice presented a similarly large increase nous Ca2 þ -triggered release. in PPF at short interstimulus intervals (50 ms) in Similar to what happened in brain slices,152 excitatory synapses, while Munc13-1 heterozygous cultured neurons lacking RIM1-alpha displayed a have no alterations in PPF. In contrast, RIM1-alpha short-term synaptic plasticity characterized by less knockout inhibitory synapses had a rather decreased paired-pulse depression and a sustained response PPF ratio, indicating that similarly to what has been during 14-Hz train stimulations. A decrease in paired- reported for Munc13-1, RIM1-alpha might have pulse depression is expected for a synapse, as RIM1- different functions in excitatory and inhibitory sy- alpha KO synapses, with a low Pr.1,2 However, in RIM napses. Since inhibitory synapses in Rab3A knockout KO synapses, the maintenance of the synaptic

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 194

Figure 2 RIM1-alpha is an essential protein for presynaptic plasticity. (a) RIM1-alpha scaffolds key proteins at the interface between the synaptic vesicle and the active zone. The domain structure of RIMs incorporates an N-terminal zinc-finger, central PDZ and two C domains (C2A and C2B). Asterisk indicates a key PKA phosphorylation site (Serine 413) (modified from Schoch et al;152 with permission; copyrightr 2002 Nature Publishing Group); (b) Mossy fiber LTP is abolished in RIM1a KO mice. Mossy fiber LTP was induced by a single tetanus in wild-type mice (open circles) in contrast to RIM1-alpha KO mice, where mossy LTP is abolished (modified from Castillo et al;155 with permission; copyrightr 2002 Nature Publishing Group); (c) PKA phosphorylation in Serine 413 of RIM1-alpha is essential for cerebellar LTP. RIM1-alpha KO neurons were transfected with wild-type RIM1-alpha or mutant RIM1-alpha containing substitutions in two PKA phosphorylation sites. RIM1-alpha wildtype (not shown) or RIM1-alpha with mutated serine 1548 (open and filled squares) rescued LTP. In contrast, RIM1-alpha with mutated serine 413 failed to rescue LTP in RIM1-alpha deficient neurons (scale bars, 5 ms, 8 pA) (modified from Lonart et al;156 with permission; copyrightr 2003 Elsevier); (d) Poor performance of RIM1-alpha deficient mice in Morris water maze. Wild type mice exhibit a progressively shorter latency to find a hidden platform in contrast to RIM1-alpha KO mice that expend longer time looking for the platform, indicating an impairment in spatial memory. (e) Fear learning defect in RIM1-alpha deficient mice. Wildtype mice learn to recognize a novel context evidenced by a longer time in a freezing state. Shorter times in freezing state reflect a deficient fear learning in RIM1-alpha KO mice (modified from Powell et al;178 with permission; copyrightr 2004 Elsevier).

response during the prolonged train of stimulations have compared RRP size before and upon 14-Hz which, in contrast, leads to a marked depression in stimulation and found that the RRP, after stimulation, wild-type control neurons is remarkable. What could recovers its original size in both RIM1-alpha KO and be the explanation for such a phenotype? Calakos et al wild-type synapses. Therefore, the sustained synaptic

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 195 response cannot be explained for a growth of RRP size function in all synapses studied with specific roles during maintained stimulation. They then compared depending on the specific synapse characteristics, the action potential evoked responses before and after probably reflecting the different RIM1-alpha binding the train and computed Pvr. Interestingly, they found partners present in each synapse. Consistently, RIM1- that in synapses lacking RIM1-alpha, Pvr increased by alpha is essential for LTP in synapses that express effect of the train stimulation, while in wild-type presynaptic LTP as mossy fibers and parallel fibers. neurons, Pvr did not change. To understand the As it has been previously discussed, an important details of how Pvr increase is accomplished a deeper open question in understanding the molecular details study would be necessary. Nevertheless a possible of presynaptic plasticity is the identification of PKA interpretation is that RIM1-alpha might function as a substrates that become phosphorylated upon Ca2 þ - regulator of short-term synaptic plasticity, able to dependent PKA activation. Synapsins and rabphilin even decrease Pvr during maintained synaptic activ- are both phosphorylated by PKA but previous studies ity. RIM1-alpha is a putative effector of the synaptic on mutant mice in which both synapsinI and II, or vesicle protein Rab3A. Mice lacking Rab3A also have rabphilin, were deleted showed normal an increased PPF with no changes in Pr. Based on LTP.156,165,167,168 On the other hand, Rab3A, although those observations, Rab3A has been proposed to essential for mossy fiber LTP, is not a PKA substrate. function as a negative regulator of synaptic vesicle Within that scenario, RIM1-alpha was an excellent exocytosis.159 It might be possible that RIM1-alpha candidate as a PKA dependent inductor (and expres- acting as a key regulator of synaptic plasticity might ser) of presynaptic LTP. Lonart et al have recently set a ceiling in neurotransmitter release in situations shown that RIM1-alpha is readily phosphorylated by where a massive release could lead to presynaptic PKA upon elevation of cAMP levels in synaptosomes exhaustion or postsynaptic overstimulation. from mossy fibers. RIM1-alpha is phosphorylated at After the arrival of an action potential at an two sites, serine 413 and serine 1548. Interestingly, excitatory synapse, neurotransmitter release displays analysis with phosphospecific antibodies revealed an two components:160 a fast and synchronous compo- increase in in vivo phosphorylation in serine 413 in nent triggered by the Ca2 þ -sensor synaptotagmin synaptosomes treated with the phosphatase inhibitor 1161,162 and a slow, asynchronous component that is okadaic acid; in contrast, no phosphorylation changes also Ca2 þ -dependent but not well understood from appeared in serine 1548. Lonart et al showed that the molecular point of view. Calakos et al found that impaired LTP cerebellar granule cell-Purkinje sy- neurons from RIM1-alpha KO mice have an B50% napses in neuron cultures from RIM1-alpha knockout reduction in the asynchronous component of release, mice got rescued upon transfection with RIM1-alpha which implies RIM1-alpha in a postpriming step full length and got blocked if the rescued cells were related to Ca2 þ -triggered vesicle fusion. incubated with a PKA inhibitor. These results Calakos et al showed that RIM1-alpha was not suggested that PKA phosphorylation of the trans- required for synaptogenesis, normal quantal size, fected RIM1-alpha might be essential for LTP, but did spontaneous release, the synchronous component of not prove it. They then showed that transfection with evoked release or refilling of the PPR. In summary, RIM1-alpha carrying a mutation at serine 413 did not their data identified RIM1-alpha as a crucial regulator rescue LTP, in contrast to a mutation at serine 1548, of synaptic vesicle maturation at the active zone which behaved as the wild-type rescue (Figure 2c). implied in priming and Ca2 þ triggering of synaptic Furthermore, similar results were obtained upon LTP vesicle fusion. induction with cAMP analogs activators of PKA. In addition, the transfection in wild-type neurons of Rim1-alpha in LTP RIM1-alpha mutated at serine 413 selectively attenu- Since hippocampal mossy fiber LTP is abolished in ated LTP. Rab3A knockout mice,163 to explain Rab3A role in Those results demonstrated that the phosphoryla- presynaptic plasticity the molecular models pointed tion of RIM1-alpha at a single site, serine 413, is toward the two main Rab3A effectors: rabphilin and required for LTP of neurotransmitter release at RIM1-alpha. Surprisingly, rabphilin-deficient cerebellar parallel fiber synapses. These results mice164,165 turned out to have normal synaptic demonstrate that LTP induction might be mediated transmission and exhibited none of the phenotypic by a direct modification of a single active zone protein traits of Rab3A knockout mice. The systematic study and they constitute an important step forward in of LTP in RIM1-alpha deficient mice yielded key understanding of the molecular mechanisms of pre- findings (Schoch et al152, Castillo et al155 reviewed in synaptic LTP. Lonart166: (1) neurotransmitter release and short-term synaptic plasticity in mossy fiber terminals were RIM1-alpha is critical for learning and memory normal, in contrast to the changes found in the CA1 Since RIM1-alpha has been shown to be a key region where Pr was decreased and PPF increased; (2) protein for regulation of short- and long-term hippocampal mossy fiber LTP and cerebellar parallel synaptic plasticity, the question arising is if indeed fiber LTP were abolished in contrast to normal LTP at those changes translate into a learning and memory the CA1 region (Figure 2b and c). Remarkably, these phenotype. Powell et al have recently carried results revealed that RIM1-alpha plays a central out a systematic behavioral analysis of RIM1-alpha

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 196 deficient mice and analyzed associative learning and impairs the Ca2 þ -dependent phospholipid binding memory, locomotor activity, motor coordination and to the C2A domain and translates into an B50% anxiety-like responses. RIM1-alpha knockout mice decreased of Pr and increased short-term synaptic exhibit a dramatic deficit in associative learning and plasticity in cultured hippocampal neurons.162 In locomotor responses to novelty but they are normal in contrast to the learning deficits in RIM1-alpha KO other behavioral tests. RIM1-alpha knockout mice mice, the synaptotagmin R233Q knockin did not have performed very poorly in the Morris water maze, any alterations in learning and memory. Interestingly, revealing a deficit in spatial learning (Figure 2d). In a those results would indicate that a generalized test to evaluate conditioning of fear to a novel context, decrease in Pr does not translate into a learning normal mice learned after a single pairing to recog- deficit. Powell et al considered that an equally nize an experimental chamber with a foot shock. Fear balanced decreased of Pr in excitatory and inhibitory learning was scored better the longer the time that the synapses, as presumably occurs in Syt R233Q mutant mice spent in the freezing state. As shown in Figure mice, might allow plasticity mechanisms to get 2e, RIM1-alpha knockout mice spend significantly adapted to general low Pr values without impair- lower times in the freezing state in comparison to ments of cognitive function. In contrast, since RIM1- wild-type littermates, indicating a clear deficit in fear alpha might have subtle different effects on excitatory conditioning. and inhibitory neurons,152 RIM1-alpha absence could RIM1-alpha knockout mice display a decreased lead to an unbalanced equilibrium between excitatory context-dependent fear conditioning, a behaviour that and inhibitory components of synaptic function, requires integrity of both hippocampus and amygdala. which would translate into a deficient computation Furthermore, the mutants have a deficit in cue- of cognitive signals. RIM1-alpha mice also present dependent fear conditioning, which requires the increased post-tetanic potentiation (PTP) and in- integration in the amygdala of a sensory stimulus creased PPF in hippocampal area CA1, which might associated to the fear stimulus. These observations contribute to the spatial learning phenotype. In any point to a potential defect in synaptic plasticity in case, the authors propose that the RIM1-alpha KO amygdala in RIM1-alpha knockout mice. As dis- phenotype reflects that normal cognitive function cussed above, the heterosynaptic associative type of requires a coordinated integration of Pr, short- and LTP in amygdala involves a presynaptically induced long-term plasticity. and expressed form of LTP mediated by presynaptic Powell et al’s study is the first genetic demonstra- NMDA receptors.4 RIM1-alpha might also participate tion showing that the alteration of a presynaptic in presynaptic LTP in amygdala similarly as it does in protein specifically alters cognitive function. As the hippocampal mossy fibers and cerebellar parallel authors emphasized, it is indeed surprising that a fibers. It would be very interesting to quantify protein that seems to play a general role in the synaptic plasticy in lateral amygdala of RIM1-alpha molecular organization of the active zone and ex- knockout mice. presses in many different brain regions affects only However, an important point to analyze is which the associative learning with no alterations in other features of the synaptic plasticity alterations might be behaviors. responsible for the associative learning phenotype. The important deficit in mossy fiber LTP or the low Pr Conclusions could be hypothesized to explain the impairment in spatial learning. Powell et al analyzed RIM1-alpha In the present review, we have described several knockout mice in parallel with Rab3A deficient mice exemplary synapses where presynaptic mechanisms that lack mossy fiber LTP and display an increased of neurotransmitter release enhancement underlie short-term synaptic plasticity with no changes of Pr at changes in synaptic plasticity, in particular presy- the CA1 region of hippocampus. They found that naptic LTP. The progress in the knowledge of the Rab3A knockout mice present no detectable defect in synaptic vesicle cycle, in particular how molecular learning and memory. These data indicate that networks build up different physiological outputs, deficient mossy fiber LTP is probably not the reason promises to be a fruitful path to advance in the for the spatial learning defect of RIM1-alpha mutants, understanding of molecular mechanisms of presynap- in agreement with other studies in mutant mice tic plasticity. Elucidation of the complete set of real lacking Rab3A169 or with defects in PKA activity key players for LTP is a challenging task.175,176 The in Huang et al;170 see also Villacres et al,171 Wu et al172 vivo approaches dealing with a combination of and Otto et al173 genetics and quantitative physiology and behavioral In order to investigate whether the decrease in Pr analysis in mice are setting a powerful experimental could be the cause of the behavioral phenotype, scenario to advance in the right direction. Resident Powell et al included in the analysis synaptotagmin 1 molecules at the active zone such as Munc13 proteins R233Q knockin mice to compare them with RIM1- and RIM1-alpha, which plays a fundamental role in alpha mutants. Synaptotagmin 1 is a Ca2 þ sensor for the maturation of synaptic vesicles to reach compe- fast and synchronous exocytosis.162,122,174 Synaptotag- tence for Ca2 þ -dependent fusion, have been demon- min 1 R233Q knockin mice bear a punctual mutation strated to be crucial targets of biochemical cascades in the C2A domain of synaptotagmin 1, which activated during synaptic activity and responsible for

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 197 plastic changes in neurotransmitter release through 15 Malgaroli A, Ting AE, Wendland B, Bergamaschi A, Villa A, Tsien priming regulation in short-term synaptic plasticity. RW et al. Presynaptic component of long-term potentiation Nevertheless, LTP establishment in addition requires visualized at individual hippocampal synapses. Science 1995; 268: 1624–1628. structural changes at the presynaptic terminal. In 16 Ryan TA, Ziv NE, Smith SJ. Potentiation of evoked vesicle particular, RIM1-alpha functions as an scaffold for turnover at individually resolved synaptic boutons. Neuron 1996; different types of proteins in the active zone. One of 17: 125–134. these proteins represent alpha-liprins,152 which reg- 17 Zakharenko SS, Zablow L, Siegelbaum SA. Visualization of ulate active zone size in the nematode.177 A proposed changes in presynaptic function during long-term synaptic plasticity. Nat Neurosci 2001; 4: 711–717. exciting idea is that RIM1-alpha phosphorylation 18 Emptage NJ, Reid CA, Fine A, Bliss TV. Optical quantal analysis might trigger LTP through structural changes based reveals a presynaptic component of LTP at hippocampal Schaffer- on molecular interactions with proteins as alpha- associational synapses. Neuron 2003; 38: 797–804. liprins.156 It is certainly predictable that active zones 19 Harris EW, Cotman CW. Long-term potentiation of guinea pig mossy fiber responses is not blocked by N-methyl D-aspartate will become a fascinating land to investigate mole- antagonists. Neurosci Lett 1986; 70: 132–137. cular mechanisms of presynaptic plasticity. 20 Zalutsky RA, Nicoll RA. Comparison of two forms of long-term potentiation in single hippocampal neurons. Science 1990; 248: 1619–1624. Acknowledgements 21 Ito I, Sugiyama H. Roles of glutamate receptors in long-term potentiation at hippocampal mossy fiber synapses. Neuroreport We are grateful to Drs G Alvarez de Toledo, TC Su¨ dhof 1991; 2: 333–336. and L Tabares for advice and support and Drs A Lu¨ thi, 22 Katsuki H, Kaneko S, Tajima A, Satoh M. Separate mechanisms of ML Montesinos and L Tabares for comments on the long-term potentiation in two input systems to CA3 pyramidal neurons of rat hippocampal slices as revealed by the whole-cell manuscript. Work in the RFC laboratory is supported patch-clamp technique. Neurosci Res 1991; 12: 393–402. in part by grants from the Spanish Ministry of 23 Castillo PE, Weisskopf MG, Nicoll RA. The role of Ca2 þ Education and Science (BFI2001-4959-E, BFI2002- channels in hippocampal mossy fiber synaptic transmission 01686) and EMBO-Young Investigator Programme, and long-term potentiation. Neuron 1994; 12: 261–269. 24 Maeda T, Kaneko S, Akaike A, Satoh M. Direct evidence for and fellowships from the Spanish Foundation for increase in excitatory amino acids release during mossy fiber LTP Health Research (FIS) (PGJC) and the Spanish in rat hippocampal slices as revealed by the patch sensor Ministry of Education and Science (PLC). methods. Neurosci Lett 1997; 224: 103–106. 25 Langdon RB, Johnson JW, Barrionuevo G. Posttetanic potentiation and presynaptically induced long-term potentiation at the mossy References fiber synapse in rat hippocampus. J Neurobiol 1995; 26: 370–385. 26 Mellor J, Nicoll RA. Hippocampal mossy fiber LTP is indepen- 1 Regehr WG, Stevens CF. Physiology of synaptic transmission and dent of postsynaptic calcium. Nat Neurosci 2001; 4: 125–126. short term synaptic plasticity. In: Cowan WM, Su¨ dhof TC, 27 Williams S, Johnston D. Long-term potentiation of hippocampal Stevens CF (eds). The Synapse. Johns Hopkins University Press: mossy fiber synapses is blocked by postsynaptic injection of Baltimore, 2000. calcium chelators. Neuron 1989; 3: 583–588. 2 Zucker RS, Regehr WG. Short-term synaptic plasticity. Annu Rev 28 Yamamoto C, Sawada S, Kamiya H. Enhancement of postsynaptic Physiol 2002; 64: 355–405. responsiveness during long-term potentiation of mossy fiber 3 Bliss TV, Lomo T. Long-lasting potentiation of synaptic transmis- synapses in guinea pig hippocampus. Neurosci Lett 1992; 138: sion in the dentate area of the anaesthetized rabbit following 111–114. stimulation of the perforant path. J Physiol 1973; 232: 331–356. 29 Yeckel MF, Kapur A, Johnston D. Multiple forms of LTP in 4 Humeau Y, Shaban H, Bissiere S, Luthi A. Presynaptic induction hippocampal CA3 neurons use a common postsynaptic mechan- of heterosynaptic associative plasticity in the mammalian brain. ism. Nat Neurosci 1999; 2: 625–633. Nature 2003; 426: 841–845. 30 Wang J, Yeckel MF, Johnston D, Zucker RS. Photolysis of 5 Bliss TV, Collingridge GL. A synaptic model of memory: long- postsynaptic caged Ca2 þ can potentiate and depress mossy term potentiation in the hippocampus. Nature 1993; 361: 31–39. fiber synaptic responses in rat hippocampal CA3 pyramidal 6 Nicoll RA, Malenka RC. Contrasting properties of two forms of neurons. J Neurophysiol 2004; 91: 1596–1607. long-term potentiation in the hippocampus. Nature 1995; 377: 31 Staubli U, Larson J, Lynch G. Mossy fiber potentiation and long- 115–118. term potentiation involve different expression mechanisms. 7 Kandel ER. The molecular biology of memory storage: a dialogue Synapse 1990; 5: 333–335. between genes and synapses. Science 2001; 294: 1030–1038. 32 Huang YY, Li XC, Kandel ER. cAMP contributes to mossy fiber 8 Malenka RC, Siegelbaum SA. Synaptic plasticity. In: Cowan WM, LTP by initiating both a covalently mediated early phase and Su¨dhof TC, Stevens CF (eds). The Synapse. Johns Hopkins macromolecular synthesis-dependent late phase. Cell 1994; 79: University Press: Baltimore, 2000, pp 455–518. 69–79. 9 Malenka RC. The long-term potential of LTP. Nat Rev Neurosci 33 Xiang Z, Greenwood AC, Kairiss EW, Brown TH. Quantal 2003; 4: 923–926. mechanism of long-term potentiation in hippocampal mossy- 10 Grimwood PD, Martin SJ, Morris RGM. Synaptic plasticity and fiber synapses. J Neurophysiol 1994; 71: 2552–2556. memory. In: Cowan WM, Su¨dhof TC, Stevens CF (eds). The 34 Rosenmund C, Clements JD, Westbrook GL. Nonuniform prob- Synapse. Johns Hopkins University Press: Baltimore, 2000, ability of glutamate release at a hippocampal synapse. Science pp 519–570. 1993; 262: 754–757. 11 Malinow R, Malenka RC. AMPA receptor trafficking and synaptic 35 Hessler NA, Shirke AM, Malinow R. The probability of plasticity. Annu Rev Neurosci 2002; 25: 103–126. transmitter release at a mammalian central synapse. Nature 12 Bredt DS, Nicoll RA. AMPA receptor trafficking at excitatory 1993; 366: 569–572. synapses. Neuron 2003; 40: 361–379. 36 Weisskopf MG, Nicoll RA. Presynaptic changes during mossy 13 Stevens CF, Wang Y. Changes in reliability of synaptic function as fibre LTP revealed by NMDA receptor-mediated synaptic a mechanism for plasticity. Nature 1994; 371: 704–707. responses. Nature 1995; 376: 256–259. 14 Bolshakov VY, Siegelbaum SA. Hippocampal long-term depres- 37 Weisskopf MG, Castillo PE, Zalutsky RA, Nicoll RA. Mediation of sion: arachidonic acid as a potential retrograde messenger. hippocampal mossy fiber long-term potentiation by cyclic AMP. Neuropharmacology 1995; 34: 1581–1587. Science 1994; 265: 1878–1882.

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 198 38 Tong G, Malenka RC, Nicoll RA. Long-term potentiation in function in plasticity and cognition. J Neurosci 2004; 24: cultures of single hippocampal granule cells: a presynaptic form 8410–8415. of plasticity. Neuron 1996; 16: 1147–1157. 62 Hinds HL, Goussakov I, Nakazawa K, Tonegawa S, Bolshakov VY. 39 Lopez-Garcia JC, Arancio O, Kandel ER, Baranes D. A presynaptic Essential function of alpha-calcium/calmodulin-dependent pro- locus for long-term potentiation of elementary synaptic transmis- tein kinase II in neurotransmitter release at a glutamatergic sion at mossy fiber synapses in culture. Proc Natl Acad Sci USA central synapse. Proc Natl Acad Sci USA 2003; 100: 4275–4280. 1996; 93: 4712–4717. 63 Ninan I, Arancio O. Presynaptic CaMKII is necessary for synaptic 40 Kullmann DM. Silent synapses: what are they telling us about plasticity in cultured hippocampal neurons. Neuron 2004; 42: long-term potentiation? Philos Trans Roy Soc London B Biol Sci 129–141. 2003; 358: 727–733. 64 Wang H, Storm DR. Calmodulin-regulated adenylyl cyclases: 41 Voronin LL, Cherubini E. ‘Deaf, mute and whispering’ silent cross-talk and plasticity in the central nervous system. Mol synapses: their role in synaptic plasticity. J Physiol 2004; 557(Part Pharmacol 2003; 63: 463–468. 1): 3–12. 65 Huang YY, Kandel ER. Modulation of both the early and the late 42 Hallermann S et al. A large pool of releasable vesicles in a phase of mossy fiber LTP by the activation of beta-adrenergic cortical glutamatergic synapse. Proc Natl Acad Sci USA 2003; receptors. Neuron 1996; 16: 611–617. 100: 8975–8980. 66 Kamiya H, Umeda K, Ozawa S, Manabe T. Presynaptic Ca2 þ 43 Ramon y Cajal S. Histologie du Systeme Nerveux de l’Homme entry is unchanged during hippocampal mossy fiber long-term et des Vertebres. Maloine: Paris, 1911. potentiation. J Neurosci 2002; 22: 10524–10528. 44 Henze DA, Urban NN, Barrionuevo G. The multifarious hippo- 67 Bortolotto ZA, Clarke VR, Delany CM, Parry MC, Smolders I, campal mossy fiber pathway: a review. 2000; 98: Vignes M et al. Kainate receptors are involved in synaptic 407–427. plasticity. Nature 1999; 402: 297–301. 45 Jonas P, Major G, Sakmann B. Quantal components of unitary 68 Schmitz D, Mellor J, Breustedt J, Nicoll RA. Presynaptic EPSCs at the mossy fibre synapse on CA3 pyramidal cells of rat kainate receptors impart an associative property to hippocampal hippocampus. J Physiol 1993; 472: 615–663. mossy fiber long-term potentiation. Nat Neurosci 2003; 6: 46 Reid CA, Dixon DB, Takahashi M, Bliss TV, Fine A. Optical 1058–1063. quantal analysis indicates that long-term potentiation at single 69 Rodriguez-Moreno A, Sihra TS. Presynaptic hippocampal mossy fiber synapses is expressed through in- facilitation of glutamate release involves protein kinase A in the creased release probability. recruitment of new release sites. and rat hippocampus. J Physiol 2004; 557(Part 3): 733–745. activation of silent synapses. J Neurosci 2004; 24: 3618–3626. 70 Schmitz D, Mellor J, Frerking M, Nicoll RA. Presynaptic kainate 47 Reid CA, Fabian-Fine R, Fine A. Postsynaptic calcium transients receptors at hippocampal mossy fiber synapses. Proc Natl Acad evoked by activation of individual hippocampal mossy fiber Sci USA 2001; 98: 11003–11008. synapses. J Neurosci 2001; 21: 2206–2214. 71 Lerma J. Roles and rules of kainate receptors in synaptic 48 Alvarez de Toledo G, Fernandez-Chacon R, Fernandez JM. transmission. Nat Rev Neurosci 2003; 4: 481–495. Release of secretory products during transient vesicle fusion. 72 Bortolotto ZA, Lauri S, Isaac JT, Collingridge GL. Kainate Nature 1993; 363: 554–558. receptors and the induction of mossy fibre long-term potentia- 49 Ales E, Tabares L, Poyato JM, Valero V, Lindau M, Alvarez de tion. Philos Trans Roy Soc London B 2003; 358: 657–666. Toledo G. High calcium concentrations shift the mode of 73 Engelman HS, MacDermott AB. Presynaptic ionotropic receptors exocytosis to the kiss-and-run mechanism. Nat Cell Biol 1999; and control of transmitter release. Nat Rev Neurosci 2004; 5: 1: 40–44. 135–145. 50 Gandhi SP, Stevens CF. Three modes of synaptic vesicular 74 Hansel C, Linden DJ, D’Angelo E. Beyond parallel fiber LTD: the recycling revealed by single-vesicle imaging. Nature 2003; 423: diversity of synaptic and non-synaptic plasticity in the cerebel- 607–613. lum. Nat Neurosci 2001; 4: 467–475. 51 Aravanis AM, Pyle JL, Tsien RW. Single synaptic vesicles fusing 75 Bear MF, Linden DJ. The mechanisms and meaning of long-term transiently and successively without loss of identity. Nature synaptic depression. In: Cowan WM, Su¨dhof TC, Stevens CF 2003; 423: 643–647. (eds). The Synapse. Johns Hopkins University Press: Baltimore, 52 Fernandez-Chacon R, Alvarez de Toledo G. Cytosolic calcium 2000, pp 455–517. facilitates release of secretory products after exocytotic vesicle 76 Linden DJ. Neuroscience. From molecules to memory in the fusion. FEBS Lett 1995; 363: 221–225. cerebellum. Science 2003; 301: 1682–1685. 53 Hartmann J, Lindau M. A novel Ca(2 þ )-dependent step in 77 Koekkoek SK, Hulscher HC, Dortland BR, Hensbroek RA, exocytosis subsequent to vesicle fusion. FEBS Lett 1995; 363: Elgersma Y, Ruigrok TJ et al. Cerebellar LTD and learning- 217–220. dependent timing of conditioned eyelid responses. Science 2003; 54 Scepek S, Coorssen JR, Lindau M. Fusion pore expansion in 301: 1736–1739. horse eosinophils is modulated by Ca2 þ and protein kinase C 78 Medina JF, Nores WL, Ohyama T, Mauk MD. Mechanisms of via distinct mechanisms. EMBO J 1998; 17: 4340–4345. cerebellar learning suggested by eyelid conditioning. Curr Opin 55 Choi S, Klingauf J, Tsien RW. Postfusional regulation of cleft Neurobiol 2000; 10: 717–724. glutamate concentration during LTP at ’silent synapses’. Nat 79 Sakurai M. Calcium is an intracellular mediator of the climbing Neurosci 2000; 3: 330–336. fiber in induction of cerebellar long-term depression. Proc Natl 56 Renger JJ, Egles C, Liu G. A developmental switch in neuro- Acad Sci USA 1990; 87: 3383–3385. transmitter flux enhances synaptic efficacy by affecting AMPA 80 Sakurai M. Synaptic modification of parallel fibre-Purkinje cell receptor activation. Neuron 2001; 29: 469–484. transmission in in vitro guinea-pig cerebellar slices. J Physiol 57 Nielsen TA, DiGregorio DA, Silver RA. Modulation of glutamate (London) 1987; 394: 463–480. mobility reveals the mechanism underlying slow-rising ampar 81 Salin PA, Malenka RC, Nicoll RA. Cyclic AMP mediates a epscs and the diffusion coefficient in the synaptic cleft. Neuron presynaptic form of LTP at cerebellar parallel fiber synapses. 2004; 42: 757–771. Neuron 1996; 16: 797–803. 58 Malenka RC, Nicoll RA. Long-term potentiation–a decade of 82 Hirano T. Depression and potentiation of the synaptic transmis- progress? Science 1999; 285: 1870–1874. sion between a granule cell and a Purkinje cell in rat cerebellar 59 Nicoll RA. Expression mechanisms underlying long-term poten- culture. Neurosci Lett 1990; 119: 141–144. tiation: a postsynaptic view. Philos Trans Roy Soc London B Biol 83 Hirano T. Differential pre- and postsynaptic mechanisms for Sci 2003; 358: 721–726. synaptic potentiation and depression between a granule cell 60 Griffith LC. Regulation of calcium/calmodulin-dependent protein and a Purkinje cell in rat cerebellar culture. Synapse 1991; 7: kinase II activation by intramolecular and intermolecular inter- 321–323. actions. J Neurosci 2004; 24: 8394–8398. 84 Crepel F, Jaillard D. Pairing of pre- and postsynaptic activities in 61 Elgersma Y, Sweatt JD, Giese KP. Mouse genetic approaches to cerebellar Purkinje cells induces long-term changes in synaptic investigating calcium/calmodulin-dependent protein kinase II efficacy in vitro. J Physiol 1991; 432: 123–141.

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 199 85 Shibuki K, Okada D. Cerebellar long-term potentiation under ment of synaptic transmission in the cortical pathway to the suppressed postsynaptic Ca2 þ activity. Neuroreport 1992; 3: lateral amygdala. Neuron 2002; 34: 289–300. 231–234. 109 Bailey CH, Giustetto M, Huang YY, Hawkins RD, Kandel ER. Is 86 Linden DJ. Long-term potentiation of glial synaptic currents in heterosynaptic modulation essential for stabilizing Hebbian cerebellar culture. Neuron 1997; 18: 983–994. plasticity and memory? Nat Rev Neurosci 2000; 1: 11–20. 87 Linden DJ. Synaptically-evoked glutamate transport currents may 110 Pare D. Presynaptic induction and expression of NMDA- be used to detect the expression of long-term potentiation in dependent LTP. Trends Neurosci 2004; 27: 440–441. cerebellar culture. J Neurophysiol 1998; 79: 3151–3156. 111 Farb CR, Ledoux JE. Afferents from rat temporal cortex synapse 88 Linden DJ, Ahn S. Activation of presynaptic cAMP-dependent on lateral amygdala neurons that express NMDA and AMPA protein kinase is required for induction of cerebellar long-term receptors. Synapse 1999; 33: 218–229. potentiation. J Neurosci 1999; 19: 10221–10227. 112 Ohno-Shosaku T, Maejima T, Kano M. Endogenous cannabinoids 89 Chavis P, Mollard P, Bockaert J, Manzoni O. Visualization of mediate retrograde signals from depolarized postsynaptic neu- cyclic AMP-regulated presynaptic activity at cerebellar granule rons to presynaptic terminals. Neuron 2001; 29: 729–738. cells. Neuron 1998; 20: 773–781. 113 Kreitzer AC, Regehr WG. Cerebellar depolarization-induced 90 Sherman SM, Guillery RW. The Synaptic Organization of the suppression of inhibition is mediated by endogenous cannabi- Brain. In: Shepherd GM (ed). Thalamus. Oxford University Press: noids. J Neurosci 2001; 21: RC174. Oxford, 2004, pp 311–360. 114 Kreitzer AC, Regehr WG. Retrograde inhibition of presynaptic 91 Guillery RW, Sherman SM. Thalamic relay functions and their calcium influx by endogenous cannabinoids at excitatory role in corticocortical communication: generalizations from the synapses onto Purkinje cells. Neuron 2001; 29: 717–727. visual system. Neuron 2002; 33: 163–175. 115 Takahashi KA, Linden DJ. Cannabinoid receptor modulation of 92 Frigyesi TL. Intracellular recordings from neurons in dorsolateral synapses received by cerebellar Purkinje cells. J Neurophysiol thalamic reticular nucleus during capsular. basal ganglia and 2000; 83: 1167–1180. midline thalamic stimulation. Brain Res 1972; 48: 157–172. 116 Wilson RI, Kunos G, Nicoll RA. Presynaptic specificity of 93 Scharfman HE, Lu SM, Guido W, Adams PR, Sherman SM. endocannabinoid signaling in the hippocampus. Neuron 2001; N-methyl-D-aspartate receptors contribute to excitatory postsy- 31: 453–462. naptic potentials of cat lateral geniculate neurons recorded 117 Tzingounis AV, Nicoll RA. Presynaptic NMDA receptors get into in thalamic slices. Proc Natl Acad Sci USA 1990; 87: the act. Nat Neurosci 2004; 7: 419–420. 4548–4552. 118 Silver RA, Momiyama A, Cull-Candy SG. Locus of frequency- 94 Steriade M, Timofeev I. Short-term plasticity during intrathala- dependent depression identified with multiple-probability fluc- mic augmenting responses in decorticated cats. J Neurosci 1997; tuation analysis at rat climbing fibre-Purkinje cell synapses. 17: 3778–3795. J Physiol 1998; 510(Part 3): 881–902. 95 von Krosigk M, Monckton JE, Reiner PB, McCormick DA. 119 Llano I, Gonzalez J, Caputo C, Lai FA, Blayney LM, Tan YP et al. Dynamic properties of corticothalamic excitatory postsynaptic Presynaptic calcium stores underlie large-amplitude miniature potentials and thalamic reticular inhibitory postsynaptic poten- IPSCs and spontaneous calcium transients. Nat Neurosci 2000; 3: tials in thalamocortical neurons of the guinea-pig dorsal lateral 1256–1265. geniculate nucleus. Neuroscience 1999; 91: 7–20. 120 Galante M, Marty A. Presynaptic ryanodine-sensitive calcium 96 Castro-Alamancos MA, Calcagnotto ME. Presynaptic long-term stores contribute to evoked neurotransmitter release at the basket potentiation in corticothalamic synapses. J Neurosci 1999; 19: cell-Purkinje cell synapse. J Neurosci 2003; 23: 11229–11234. 9090–9097. 121 Sudhof TC. The synaptic vesicle cycle: a cascade of protein- 97 Deschenes M, Hu B. Membrane resistance increase induced in protein interactions. Nature 1995; 375: 645–653. thalamic neurons by stimulation of brainstem cholinergic 122 Sudhof TC. The synaptic vesicle cycle. Annu Rev Neurosci 2004; afferents. Brain Res 1990; 513: 339–342. 27: 509–547. 98 Lindstrom S, Wrobel A. Frequency dependent corticofugal 123 Rosenmund C, Rettig J, Brose N. Molecular mechanisms of active excitation of principal cells in the cat’s dorsal lateral geniculate zone function. Curr Opin Neurobiol 2003; 13: 509–519. nucleus. Exp Brain Res 1990; 79: 313–318. 124 Varoqueaux F, Sigler A, Rhee JS, Brose N, Enk C, Reim K et al. 99 Tsumoto T, Creutzfeldt OD, Legendy CR. Functional organization Total arrest of spontaneous and evoked synaptic transmission but of the corticofugal system from visual cortex to lateral geniculate normal synaptogenesis in the absence of Munc13-mediated nucleus in the cat (with an appendix on geniculo-cortical mono- vesicle priming. Proc Natl Acad Sci USA 2002; 99: 9037–9042. synaptic connections). Exp Brain Res 1978; 32: 345–364. 125 Rosenmund C, Sigler A, Augustin I,, Reim K, Brose N, Rhee JS. 100 Casado M, Dieudonne S, Ascher P. Presynaptic N-methyl-D- Differential control of vesicle priming and short-term plasticity aspartate receptors at the parallel fiber-Purkinje cell synapse. by Munc13 isoforms. Neuron 2002; 33: 411–424. Proc Natl Acad Sci USA 2000; 97: 11593–11597. 126 Calakos N, Schoch S, Sudhof TC, Malenka RC. Multiple roles for 101 Casado M, Isope P, Ascher P. Involvement of presynaptic N- the active zone protein RIM1alpha in late stages of neurotrans- methyl-D-aspartate receptors in cerebellar long-term depression. mitter release. Neuron 2004; 42: 889–896. Neuron 2002; 33: 123–130. 127 Brose N, Hofmann K, Hata Y, Sudhof TC. Mammalian homo- 102 Duguid IC, Smart TG. Retrograde activation of presynaptic logues of Caenorhabditis elegans unc-13 gene define novel family NMDA receptors enhances GABA release at cerebellar interneur- of C2-domain proteins. J Biol Chem 1995; 270: 25273–25280. on-Purkinje cell synapses. Nat Neurosci 2004; 7: 525–533. 128 Richmond JE, Davis WS, Jorgensen EM. UNC-13 is required 103 LeDoux JE. Emotion circuits in the brain. Annu Rev Neurosci for synaptic vesicle fusion in C. elegans. Nat Neurosci 1999; 2: 2000; 23: 155–184. 959–964. 104 Stevens CF. A million dollar question: does LTP ¼ memory? 129 Aravamudan B, Fergestad T, Davis WS, Rodesch CK, Broadie K. Neuron 1998; 20: 1–2. Drosophila UNC-13 is essential for synaptic transmission. Nat 105 Huang YY, Kandel ER. Postsynaptic induction and PKA- Neurosci 1999; 2: 965–971. dependent expression of LTP in the lateral amygdala. Neuron 130 Augustin I, Rosenmund C, Sudhof TC, Brose N. Munc13-1 is 1998; 21: 169–178. essential for fusion competence of glutamatergic synaptic 106 Bauer EP, Schafe GE, LeDoux JE. NMDA receptors and L-type vesicles. Nature 1999; 400: 457–461. voltage-gated calcium channels contribute to long-term potentia- 131 Brose N, Rosenmund C, Rettig J. Regulation of transmitter release by tion and different components of fear memory formation in the Unc-13 and its homologues. Curr Opin Neurobiol 2000; 10: 303–311. lateral amygdala. J Neurosci 2002; 22: 5239–5249. 132 Betz A, Okamoto M, Benseler F, Brose N. Direct interaction of the 107 McKernan MG, Shinnick-Gallagher P. Fear conditioning induces rat unc-13 homologue Munc13-1 with the N terminus of a lasting potentiation of synaptic currents in vitro. Nature 1997; syntaxin. J Biol Chem 1997; 272: 2520–2526. 390: 607–611. 133 Richmond JE, Weimer RM, Jorgensen EM. An open form of 108 Tsvetkov E, Carlezon WA, Benes FM, Kandel ER, Bolshakov VY. syntaxin bypasses the requirement for UNC-13 in vesicle Fear conditioning occludes LTP-induced presynaptic enhance- priming. Nature 2001; 412: 338–341.

Molecular Psychiatry Presynaptic plasticity in the brain P Garcı´a-Junco-Clemente et al 200 134 Betz A, Ashery U, Rickmann M, Augustin I, Neher E, Sudhof TC 156 Lonart G, Schoch S, Kaeser PS, Larkin CJ, Sudhof TC, Linden DJ. et al. Munc13-1 is a presynaptic phorbol ester receptor Phosphorylation of RIM1alpha by PKA triggers presynaptic long- that enhances neurotransmitter release. Neuron 1998; 21: term potentiation at cerebellar parallel fiber synapses. Cell 2003; 123–136. 115: 49–60. 135 Junge HJ, Rhee JS, Jahn O, Varoqueaux F, Spiess J, Waxham MN 157 Wang X, Hu B, Zimmermann B, Kilimann MW. Rim1 and et al. Calmodulin and Munc13 form a Ca2 þ sensor/effector rabphilin-3 bind Rab3-GTP by composite determinants partially complex that controls short-term synaptic plasticity. Cell 2004; related through N-terminal alpha -helix motifs. J Biol Chem 2001; 118: 389–401. 276: 32480–32488. 136 Ashery U, Varoqueaux F, Voets T, Betz A, Thakur P, Koch H et al. 158 Bekkers JM, Stevens CF. Excitatory and inhibitory autaptic Munc13-1 acts as a priming factor for large dense-core vesicles in currents in isolated hippocampal neurons maintained in cell bovine chromaffin cells. EMBO J 2000; 19: 3586–3596. culture. Proc Natl Acad Sci USA 1991; 88: 7834–7838. 137 Rhee JS, Betz A, Pyott S, Reim K, Varoqueaux F, Augustin I et al. 159 Geppert M, Goda Y, Stevens CF, Sudhof TC. The small GTP- Beta phorbol ester- and diacylglycerol-induced augmentation of binding protein Rab3A regulates a late step in synaptic vesicle transmitter release is mediated by Munc13s and not by PKCs. Cell fusion. Nature 1997; 387: 810–814. 2002; 108: 121–133. 160 Goda Y, Stevens CF. Two components of transmitter release at a 138 Rosenmund C, Stevens CF. Definition of the readily releasable central synapse. Proc Natl Acad Sci USA 1994; 91: 12942–12946. pool of vesicles at hippocampal synapses. Neuron 1996; 16: 161 Geppert M, Goda Y, Hammer RE, Li C, Rosahl TW, Stevens CF 1197–1207. et al. Synaptotagmin I: a major Ca2 þ sensor for transmitter 139 Brose N, Rosenmund C. Move over protein kinase C, you’ve got release at a central synapse. Cell 1994; 79: 717–727. company: alternative cellular effectors of diacylglycerol and 162 Fernandez-Chacon R, Konigstorfer A, Gerber SH, Garcia J, Matos phorbol esters. J Cell Sci 2002; 115(Part 23): 4399–4411. MF, Stevens CF et al. Synaptotagmin I functions as a calcium 140 Dittman JS, Regehr WG. Calcium dependence and recovery regulator of release probability. Nature 2001; 410: 41–49. kinetics of presynaptic depression at the climbing fiber to 163 Castillo PE, Janz R, Sudhof TC, Tzounopoulos T, Malenka RC, Purkinje cell synapse. J Neurosci 1998; 18: 6147–6162. Nicoll RA. Rab3A is essential for mossy fibre long-term 141 Stevens CF, Wesseling JF. Activity-dependent modulation of the potentiation in the hippocampus. Nature 1997; 388: 590–593. rate at which synaptic vesicles become available to undergo 164 Schluter OM, Schnell E, Verhage M, Tzonopoulos T, Nicoll RA, exocytosis. Neuron 1998; 21: 415–424. Janz R et al. Rabphilin knock-out mice reveal that rabphilin is not 142 Wang LY, Kaczmarek LK. High-frequency firing helps replenish required for rab3 function in regulating neurotransmitter release. the readily releasable pool of synaptic vesicles. Nature 1998; 394: J Neurosci 1999; 19: 5834–5846. 384–388. 165 Hosaka M, Hammer RE, Sudhof TC. A phospho-switch controls 143 Sakaba T, Neher E. Calmodulin mediates rapid recruitment of the dynamic association of synapsins with synaptic vesicles. fast-releasing synaptic vesicles at a calyx-type synapse. Neuron Neuron 1999; 24: 377–387. 2001; 32: 1119–1131. 166 Lonart G. RIM1: an edge for presynaptic plasticity. Trends 144 Augustin I, Korte S, Rickmann M, Kretzschmar HA, Sudhof TC, Neurosci 2002; 25: 329–332. Herms JW et al. The cerebellum-specific Munc13 isoform 167 Spillane DM, Rosahl TW, Sudhof TC, Malenka RC. Long-term Munc13-3 regulates cerebellar synaptic transmission and motor potentiation in mice lacking synapsins. Neuropharmacology learning in mice. J Neurosci 2001; 21: 10–17. 1995; 34: 1573–1579. 145 Wang Y, Okamoto M, Schmitz F, Hofmann K, Sudhof TC. Rim is a 168 Lonart G, Janz R, Johnson KM, Sudhof TC. Mechanism of action putative Rab3 effector in regulating synaptic-vesicle fusion. of rab3A in mossy fiber LTP. Neuron 1998; 21: 1141–1150. Nature 1997; 388: 593–598. 169 Hensbroek RA, Kamal A, Baars AM, Verhage M, Spruijt BM. 146 Wang Y, Sudhof TC. Genomic definition of RIM proteins: Spatial, contextual and working memory are not affected by the evolutionary amplification of a family of synaptic regulatory absence of mossy fiber long-term potentiation and depression. proteins(small star, filled). Genomics 2003; 81: 126–137. Behav Brain Res 2003; 138: 215–223. 147 Betz A, Thakur P, Junge HJ, Ashery U, Rhee JS, Scheuss V et al. 170 Huang YY, Kandel ER, Varshavsky L, Brandon EP, Qi M, Idzerda Functional interaction of the active zone proteins Munc13-1 and RL et al. A genetic test of the effects of mutations in PKA on RIM1 in synaptic vesicle priming. Neuron 2001; 30: 183–196. mossy fiber LTP and its relation to spatial and contextual 148 Wang Y, Dulubova I, Rizo J, Sudhof TC. Functional analysis of learning. Cell 1995; 83: 1211–1222. conserved structural elements in yeast syntaxin Vam3p. J Biol 171 Villacres EC, Wong ST, Chavkin C, Storm DR. Type I adenylyl Chem 2001; 276: 28598–28605. cyclase mutant mice have impaired mossy fiber long-term 149 Ohtsuka T, Takao-Rikitsu E, Inoue E, Inoue M, Takeuchi M, potentiation. J Neurosci 1998; 18: 3186–3194. Matsubara K et al. Cast: a novel protein of the cytomatrix at the 172 Wu ZL, Thomas SA, Villacres EC, Xia Z, Simmons ML, Chavkin C active zone of synapses that forms a ternary complex with RIM1 et al. Altered behavior and long-term potentiation in type I and munc13-1. J Cell Biol 2002; 158: 577–590. adenylyl cyclase mutant mice. Proc Natl Acad Sci USA 1995; 92: 150 Wang Y, Liu X, Biederer T, Sudhof TC. A family of RIM-binding 220–224. proteins regulated by alternative splicing: Implications for the 173 Otto C, Kovalchuk Y, Wolfer DP, Gass P, Martin M, Zuschratter W genesis of synaptic active zones. Proc Natl Acad Sci USA 2002; et al. Impairment of mossy fiber long-term potentiation and 99: 14464–14469. associative learning in pituitary adenylate cyclase activating 151 Coppola T, Magnin-Luthi S, Perret-Menoud V, Gattesco S, polypeptide type I receptor-deficient mice. J Neurosci 2001; 21: Schiavo G, Regazzi R. Direct interaction of the Rab3 effector 5520–5527. RIM with Ca2 þ channels, SNAP-25, and synaptotagmin. J Biol 174 Sorensen JB, Fernandez-Chacon R, Sudhof TC, Neher E. Chem 2001; 276: 32756–32762. Examining synaptotagmin 1 function in dense core vesicle 152 Schoch S, Castillo PE, Jo T, Mukherjee K, Geppert M, Wang Y exocytosis under direct control of Ca2 þ . J Gen Physiol 2003; et al. RIM1alpha forms a protein scaffold for regulating neuro- 122: 265–276. transmitter release at the active zone. Nature 2002; 415: 321–326. 175 Sanes JR, Lichtman JW. Can molecules explain long-term 153 Hibino H, Pironkova R, Onwumere O, Vologodskaia M, Hudspeth potentiation?. Nat Neurosci 1999; 2: 597–604. AJ, Lesage F. RIM binding proteins (RBPs) couple Rab3-interact- 176 Lisman J, Lichtman JW, Sanes JR. LTP: perils and progress. Nat ing molecules (RIMs) to voltage-gated Ca(2 þ ) channels. Neuron Rev Neurosci 2003; 4: 926–929. 2002; 34: 411–423. 177 Zhen M, Jin Y. The liprin protein SYD-2 regulates the 154 Koushika SP, Richmond JE, Hadwiger G, Weimer RM, Jorgensen differentiation of presynaptic termini in C. elegans. Nature EM, Nonet ML. A post-docking role for active zone protein Rim. 1999; 401: 371–375. Nat Neurosci 2001; 4: 997–1005. 178 Powell CM, Schoch S, Monteggia L, Barrot M, Matos MF, 155 Castillo PE, Schoch S, Schmitz F, Sudhof TC, Malenka RC. Feldmann N et al. The presynaptic active zone protein RIM1alpha is required for presynaptic long-term potentiation. RIM1alpha is critical for normal learning and memory. Neuron Nature 2002; 415: 327–330. 2004; 42: 143–153.

Molecular Psychiatry