Short term plasticity at hippocampal mossy fiber synapses

Short-term synaptic plasticity refers to the regulation of synapses by their past activity on time scales of milliseconds to minutes. Hippocampal mossy fiber synapses onto CA3 pyramidal cells (Mf-CA3 synapses) are endowed with remarkable forms of short-term synaptic plasticity expressed as facilitation of synaptic release by a factor of up to ten-fold. Three main forms of short-term plasticity are distinguished: 1) Frequency facilitation, which includes low frequency facilitation and train facilitation, operating in the range of tens of milliseconds to several seconds; 2) Post-tetanic potentiation triggered by trains of high frequency stimulation, which lasts several minutes; 3) Finally, depolarization-induced potentiation of excitation, based on retrograde signaling, with an onset and offset of several minutes. Here we describe the proposed mechanisms for short-term plasticity, mainly based on the kinetics of presynaptic Ca 2+ transients and the Ca 2+ sensor synaptotagmin 7, on cAMP-dependent mechanisms and readily releasable vesicle pool, and on the regulation of presynaptic K + channels. We then review evidence for a physiological function of short-term plasticity of Mf-CA3 synapses in information transfer between the dentate gyrus and CA3 in conditions of natural spiking, and discuss how short-term plasticity counteracts robust feedforward inhibition in a frequency-dependent manner. Although DG-CA3 connections have long been proposed to play a role in memory, direct evidence for an implication of short-term plasticity at Mf-CA3 synapses is mostly lacking. The mechanistic knowledge gained on short-term plasticity at Mf-CA3 synapses should help in designing future experiments to directly test how this evolutionary conserved feature controls hippocampal circuit function in behavioural conditions.


Introduction
Synapses are regulated by their past activity on time scales ranging from tens of milliseconds to days and longer, albeit most attention has focused on long-term synaptic plasticity (> 30 minutes), which is most commonly expressed as changes in postsynaptic properties and is thought to underlie memory encoding, storage and persistence (Takeuchi et al., 2014).Nevertheless, numerous types of short-term, activity-dependent synaptic plasticity are expressed at a presynaptic level, which involves the regulation of neurotransmitter release.Short-term synaptic plasticity has been ascribed important roles in sensory information processing and in gain control (Zucker and Regehr, 2002).Different types of synapses employ a variety of mechanisms to encode incoming spiking activity by virtue of the complement of short-term synaptic plasticity mechanisms they possess (Zucker and Regehr, 2002).
Hippocampal mossy fiber synapses, which connect the granule cells of the dentate gyrus (DG) to CA3 pyramidal cells, have long been renowned for their high dynamic range of short-term synaptic plasticity encompassing time scales from tens of milliseconds to seconds and to minutes (Nicoll and Schmitz, 2005).Over the past three decades, hippocampal mossy fiber synapses onto CA3 pyramidal cells (here Mf-CA3 synapses) have been viewed as an ideal model synapse to study mechanisms of short-term plasticity.Amongst the unique properties that have allowed to fine tune their analysis with a growing combination of methodologies, (i) Mf-CA3 synapses are strictly stratified in the stratum lucidum of CA3, (ii) the segregated input from DG granule cells enables selective genetic engineering of presynaptic proteins, and (iii) the Mf-CA3 terminal boutons are directly accessible to patch-clamp recordings due to their large volume (1-5 µm in diameter) (Rollenhagen et al., 2007).In addition, the conservation of the unique morpho-functional properties of Mf-CA3 synapses in all vertebrates, from rodents (Nicoll and Schmitz, 2005) to humans (Pelkey et al., 2023), plead for a major function of the large dynamic range of presynaptic plasticity at Mf-CA3 synapses for hippocampal circuits, in memory and in spatial navigation.Indeed, CA3 region is essential for encoding and recall of episodic memory and is thought to be involved in pattern completion (Kesner and Rolls, 2015).In contrast to CA3, the DG is proposed to play a role in pattern separation, in which the large number of DG granule cells provide sparse representations that enable similar inputs to be transformed into distinct outputs (Borzello et al., 2023;Kesner and Rolls, 2015).

Hippocampal mossy fiber synapses
CA3 mainly receives cortical information through two parallel inputs from the entorhinal cortex (EC): a direct pathway through the perforant path (PP) and an indirect disynaptic input through the DG.Mossy fibers (Mf) -the axons of DG cells -provide powerful inputs via 'giant' Mf boutons to CA3 PCs (Rebola et al., 2017;Rollenhagen et al., 2007).The Mfs establish large, highly specialized synapses with the proximal dendrites of CA3 pyramidal cells (CA3 PCs) predominantly on thorny excrescences.Mfs also contact various GABAergic interneurons in the hilus and stratum lucidum (CA3 INs) (Acsády et al., 1998;Szabadics and Soltesz, 2009) forming the structural basis of feedforward inhibition (Torborg et al., 2010).Each DG granule neuron gives rise to 10-15 Mf terminal boutons onto CA3 PCs, and in turn each CA3 PC receives synaptic inputs from circa 50 DG neurons (Acsády et al., 1998) (Gonzales et al., 2001).On average a single Mf bouton contacts the postsynaptic thorny excrescence on CA3 PC at 25 (up to 40) active zones (Rollenhagen et al., 2007) (Figure 1).Mf-CA3 synapses have very recently been the topic of an authoritative review (Vandael and Jonas, 2024).Here we choose to provide an in-depth review on short-term synaptic facilitation, a highly distinctive property of Mf-CA3 synapses.
Mf-CA3 PC synapses display a wide array of pre-and postsynaptic plasticity features (Nicoll and Schmitz, 2005;Rebola et al., 2017).In particular, Mf-CA3 synapses are highly regulated on a second to minute timescale through presynaptic short-term plasticity (Figure 2).It is important to note that DG cells in awake mice fire action potentials sparsely, and preferentially in bursts, at frequencies that are compatible with these various forms of shortterm plasticity (Diamantaki:2016ju;Vandael et al., 2020).Short-term presynaptic plasticity at Mf-CA3 synapses is expressed as a robust facilitation of synaptic transmission, which can potentiate up to a factor 10 and even more.Facilitation of glutamate release upon repeated stimulation has been directly monitored using the optical glutamate sensor iGluSnFR {Rama:2019bi}.Robust facilitation at Mf-CA3 synapses is consistent with the low initial release probability (Pr), which characterizes each individual synaptic release site at Mf terminal bouton.The low Pr can be readily evidenced by the presence of failures to synaptic transmission at low (around 0.1 Hz) Mf stimulation rates (Lanore et al., 2012).Quantal analysis of Mf-CA3 synapses has indeed estimated Pr at individual release sites to be around 1-2% (Jonas et al., 1993;Lanore et al., 2012).The high dynamic range of presynaptic plasticity is thus made possible by the low initial Pr at each individual release site, and by the high number of release sites at each Mf-CA3 bouton.
Presynaptic facilitation can be categorized according to the temporal parameters of the triggers and the duration of the potentiation, which act in a temporal continuum (Figure 2).Repetitive Mf stimulation at low frequencies (0.1 to 3 Hz), referred to as low frequency facilitation, triggers a progressive enhancement of synaptic transmission on a time scale of seconds, which decays with a time course of tens of seconds (Salin et al., 1996).Short trains of Mf stimulation at 10 to 100 Hz (including paired-pulse facilitation, a form of train facilitation) induce short-lived potentiation, which is referred to as high frequency facilitation.These short-lived (decay in seconds to tens of seconds) forms of presynaptic plasticity will be referred to as Frequency Facilitation (FF).
Longer trains of high frequency (>20 Hz) stimulation can give rise to post-tetanic potentiation (PTP) of synaptic transmission (Figure 2), which peaks rapidly within seconds after the inductive stimulation to more than 5-fold (Griffith, 1990).PTP decays down to baseline with time constants of a few minutes (Griffith, 1990).Repeating these long high-frequency trains of stimulation (3 times) leads to a presynaptic form of long-term potentiation (LTP) (Monday et al., 2018;Nicoll and Schmitz, 2005;Rebola et al., 2017), which will not be reviewed here.
At Mf-CA3 synapses, there is evidence for an additional form of short-term presynaptic potentiation of synaptic transmission called DPE (for depolarization-induced potentiation of excitation)(Figure 2), triggered by the depolarization of postsynaptic CA3 PCs and dependent on a retrograde signaling mechanism.DPE lasts and decays within a few minutes (Carta et al., 2014).Contrary to PTP, DPE takes a few minutes to reach a plateau of 2-fold potentiation, and decays within 5 to 10 minutes.
We will examine these various forms of short-term plasticity and the studies which have unraveled their underlying mechanisms, taking advantage of targeted methods (see Table 1) in conjunction with genetic mutations in mice (see Table 2).There is a continuum and a potential overlap in underlying mechanisms.Considering the natural activity patterns of presynaptic DG neurons (Diamantaki et al 2016;Vandael et al., 2020), and CA3 pyramidal cells in the case of DPE, it is evident that the overall physiological effect of presynaptic plasticity on information transfer between the DG and CA3 combines these different forms of short-term plasticity.Finally, it is important to indicate that synaptic excitation is strongly counterbalanced by feedforward inhibition (Rebola et al., 2017;Torborg et al., 2010, Prince et al, 2016), which shapes the properties of spike transfer between the DG and CA3.

Mechanisms of short-term plasticity at hippocampal mossy fibers
Residual Ca 2+ in the presynaptic bouton Upon invasion of a terminal presynaptic bouton by an action potential, there is a brief (<1ms) and large entry of Ca 2+ leading to a local cytosolic concentration of tens to hundreds of micromolar (Augustine et al., 1991).This large Ca 2+ signal can activate fast, low-affinity calcium sensors such as synaptotagmin 1 (Syt1) (Bacaj et al., 2013) resulting in vesicle fusion.However, Ca 2+ rapidly diffuses and equilibrates with various Ca 2+ buffers within presynaptic boutons, giving rise to a residual Ca 2+ signal.Synaptic facilitation has long been hypothesized to be the consequence of a larger amount of presynaptic Ca 2+ upon invasion of the terminal by a second and by subsequent action potentials (Zucker and Regehr, 2002).Following this hypothesis, it has been proposed that the lower concentration of residual Ca 2+ binds to a facilitation sensor distinct from Syt1, with higher binding affinity and slower kinetics, that interacts with the vesicle release machinery (Jackman and Regehr, 2017).
The amplitude and duration of presynaptic Ca 2+ transients are thus key parameters controlling the release of synaptic vesicles and short-term synaptic facilitation.Early work using fura-2 in guinea pig hippocampal slices has provided evidence for a presynaptic Ca 2+ enhancement with a decay time constant around 1 second (Regehr et al., 1994).Further work with rapid presynaptic Ca 2+ imaging has confirmed that presynaptic Ca 2+ elevations in Mf-CA3 terminal boutons comprise a fast and a slow component (Chamberland et al., 2020;Scott and Rusakov, 2006).A single action potential induces a bi-phasic elevation of free Ca 2+ : a brief transient to an estimated range of 8 -9 µM, followed by an elevation to 320 nM, which slowly decays to a resting level of 110 nM (Scott and Rusakov, 2006).The measured Ca 2+ removal rate is on the order of 10 times slower than in small Mf presynaptic terminals in the hilus (Jackson and Redman, 2003).The slow decay of presynaptic Ca 2+ transients is in part explained by the progressive saturation of endogenous fast Ca 2+ buffers, such as calbindin-D28k (CB), which is highly expressed in Mf terminal boutons, and calmodulin (Blatow et al., 2003;Chamberland et al., 2018).CB is likely effective at lowering initial release probability at Mf synapses because of its fast on-rate and high affinity for Ca 2+ (290 nM), further enabled by the loose coupling between Ca 2+ channels and release sensors at Mf synaptic release sites (Vyleta and Jonas, 2014).By analysing the effect of the slow and fast Ca 2+ buffers EGTA and BAPTA at different concentrations on EPSC amplitude in paired Mf-bouton to CA3 PC recordings, it was possible to estimate coupling distance between Ca 2+ channels and Ca 2+ sensors of circa 75 nm (Vyleta and Jonas, 2014).The increase of Ca 2+ in the presynaptic Mf bouton following an incoming action potential appears to arise primarily from highly localized Ca 2+ entry though N-type voltage-gated Ca 2+ channels (VGCCs), which controls glutamate release at a limited number of release sites; in contrast, Ca 2+ entry via P/Q-type VGCCs leads to spatially homogeneous Ca 2+ elevation throughout the Mf bouton (Chamberland et al., 2017).This seemingly distinct spatial organization of VGCCs has been proposed to play a role in short-term facilitation by acting either to favour multivesicular release or to recruit additional release sites (Chamberland et al., 2017).Sequestration by Ca 2+ stores contributes to maintaining a low level of presynaptic Ca 2+ and rapid extrusion from stores represents up to 20% of the presynaptic Ca 2+ transient evoked by a brief train of action potentials (Scott and Rusakov, 2006).Finally, mitochondria, which are abundant in Mf-CA3 terminal boutons (Rollenhagen et al., 2007), participate in the uptake of presynaptic Ca 2+ through the mitochondrial Ca 2+ uniporter in basal conditions, and following high-frequency trains (Devine et al., 2022;D. Lee et al., 2007;S. H. Lee et al., 2020), thereby controlling the extent of PTP (D. Lee et al., 2007).The slow decay kinetics of presynaptic Ca 2+ concentration allows for summation of Ca 2+ signals upon repetitive incoming action potential, and some experimental evidence suggests that this summation is causally related to short-term presynaptic plasticity during short trains (10-20Hz) (Regehr et al., 1994;Scott and Rusakov, 2006).More work is needed to understand how the kinetics of presynaptic Ca 2+ reflect synaptic facilitation to explain the different forms of short-term plasticity.

Synaptotagmin 7
Several candidate proteins have been proposed to act as a specialized Ca 2+ sensor that translates the smaller, long-lasting Ca 2+ signals between action potentials into increased synaptic vesicle release.A major breakthrough in the field came with the demonstration that synaptotagmin 7 (Syt7) mediates synaptic facilitation by directly increasing the probability of release (Jackman et al., 2016).Syt7 is expressed in presynaptic compartments, binds Ca 2+ with high affinity and slow kinetics, and it was previously thought to participate in asynchronous release and recovery from synaptic depression (Jackman and Regehr, 2017).Syt7 is abundantly expressed in the stratum lucidum where Mfs make synaptic contacts onto CA3 pyramidal cells (Barthet et al., 2018).FF (1 to 40 Hz) at Mf-CA3 synapses is practically eliminated in Syt7 knock-out mice without any detectable change in residual Ca 2+ in Mf boutons (Jackman et al., 2016).Accordingly, we have recently observed that the selective deletion of Syt7 in dentate granule cells leads to a marked decrease of facilitation at Mf-CA3 synapses in frequency ranges from 1 Hz to 20 Hz; in contrast, no change in the extent or duration of PTP was observed (Marneffe et al, unpublished).

Regulation of presynaptic Kv channels
Voltage-gated potassium (Kv) channels are involved in action potential (AP) repolarization in excitable cells.The broadening of presynaptic APs by inactivation of Kv channels leads to a nonlinear increase in Ca 2+ influx within the nerve terminal (Bischofberger et al., 2002;Geiger and Jonas, 2000).Presynaptic Mf bouton recordings have shown that APs broaden during repetitive spiking by cumulative inactivation of Kv channels, thus boosting presynaptic Ca 2+ influx (Geiger and Jonas, 2000).However, this is unlikely to be a significant component of facilitation except in conditions of high stimulation frequency (>30 Hz) (Scott and Rusakov, 2006).Presynaptic Kv7 channels expressed in Mf boutons appear to restrict neurotransmitter release at all frequencies of stimulation and decrease paired-pulse facilitation (Martinello et al., 2019).
Regulation of presynaptic Kv channels is a core mechanism of DPE, a form of short-term plasticity which is triggered by the spiking activity of postsynaptic CA3 pyramidal cells (Carta et al., 2014) (Figure).DPE depends on the activity-dependent release of arachidonic acid from postsynaptic CA3 PC, which acts as a retrograde messenger, inducing a robust facilitation of Mf-CA3 synaptic transmission over several minutes (Carta et al., 2014).Arachidonic acid directly inactivates presynaptic Kv channels at Mf-CA3 synapses leading to the broadening of the AP and to enhanced synaptic transmission by increasing the AP-driven Ca 2+ influx in presynaptic terminals (Carta et al., 2014).

cAMP and PKA-dependent mechanisms
Cyclic adenosine monophosphate (cAMP) is a crucial second messenger involved in both short-and long-term synaptic plasticity (potentiation and depression), and this has been well exemplified at Mf-CA3 synapses in which cAMP is a primary mediator of Mf-LTP (Shahoha et al., 2022).Mechanisms for short-term facilitation (FF or PTP) and LTP may share common mechanisms, but they have been shown in many instances to be distinct (Shahoha et al., 2022).Here we review evidence for the involvement of cAMP in presynaptic short-term plasticity and the underlying mechanisms which have been proposed.
Forskolin, an agonist of adenylate cyclase, has been widely used to show that an elevation of cAMP leads to facilitation of Mf-CA3 synaptic transmission (Weisskopf et al., 1994).There are several experimental evidence indicating that an elevation of cAMP is not involved in shortterm facilitation (FF) in response to trains (with durations in the 100 ms -1s time range).Albeit the time course of presynaptic cAMP increase following forskolin activation of adenylate cyclase is likely to be on the order of seconds (Klarenbeek et al., 2015), forskolin-induced Mf-CA3 synaptic potentiation takes minutes (e.g.Ben-Simon et al., 2015;Breustedt et al., 2010;Castillo et al., 1997;2002;Contractor et al., 2001;Fernandes et al., 2015;Kaeser-Woo et al., 2013;Viotti and Dresbach, 2019;Weisskopf et al., 1994).Mf-CA3 transmission can be potentiated by light activation of synaptoPAC, a presynaptically targeted version of the photoactivated adenylate cyclase bPAC in Mf terminals (Oldani et al., 2021); Mf-fEPSP amplitudes progressively increased within minutes after the start of the optical induction protocol (Oldani et al., 2021).More work is however needed to evaluate in a quantitative manner the kinetics of cAMP in the presynaptic Mf bouton in relation with changes in synaptic transmission.
In Mf-CA3 boutons, Ca 2+ entry during repetitive high frequency stimulation leads to elevation of presynaptic cAMP mediated by the Ca 2+ -dependent adenylate cyclases AC1 (Villacres et al., 1998) and AC8 (Wang et al., 2003).Surprisingly, the deletion of AC1, which eliminates Mf-LTP does not alter paired-pulse facilitation (Villacres et al., 1998), whereas the deletion of AC8, which only accounts for a minor fraction of cAMP accumulation in the hippocampus, impairs Mf-LTP as well as paired-pulse facilitation and PTP (Wang et al., 2003).The elevation of cAMP at Mf-CA3 synapses leads to molecular cascades starting with the activation of protein kinase A (PKA) (Fukaya et al., 2023;Shahoha et al., 2022).In most instances, the genetic deletion of PKA protein targets important for Mf-LTP (Rab3A, RIM1a, Rabphilin, Synaptotagmin-12) (Figure 4), does not lead to any change in paired-pulse facilitation and short-term facilitation (FF) in the 100 ms -second time range (Shahoha et al., 2022).Similarly, no alteration in FF in mice lacking Epac2, another effector of cAMP expressed at Mf-CA3 synapses was observed (Fernandes et al., 2015).It should be noted that the genetic deletion of tomosyn-1, another target of PKA, impairs both Mf-LTP and short-term facilitation; this is likely explained through an increase in basal Pr resulting in a reduction of the level of synaptic potentiation (Ben-Simon et al., 2015).Overall, the mechanisms of short-term facilitation in response to short bursts of repetitive stimulations (FF) appears to be largely independent of cAMP elevation induced by Ca 2+ influx into presynaptic Mf terminals and of consequent PKA activation.
In contrast to FF, PTP is fully prevented by antagonists of PKA (Vandael et al., 2020;Weisskopf et al., 1994).Paired recordings between a Mf-CA3 bouton and a postsynaptic CA3 PC have been used to show that PTP can be attributed to an increase in the size of the readily releasable vesicle pool (RRP), which is generated by enhanced pool refilling during the high frequency stimulation train (Vandael et al., 2020).In contrast, change in Pr only seems to play a minor contribution to enhanced vesicular release (Vandael et al., 2020).In addition, flashand-freeze electron microscopy (Maus et al., 2020) revealed that the number of docked synaptic vesicles increases rapidly following a tetanic stimulation (Vandael et al., 2020), or following a 15 min forskolin application (Orlando et al., 2021).It should be noted however that other experiments using forskolin to enhance synaptic transmission at Mf-CA3 synapses rather point to a rapid accumulation of Ca 2+ channels near synaptic release sites, which may underlie cAMP-induced potentiation of synaptic transmission (Fukaya et al., 2021;Midorikawa and Sakaba, 2017).The molecular mechanisms by which PKA activation may either increase the RRP or change the coupling between Ca 2+ channels and vesicle release sensors remain largely unknown.

Presynaptic autoreceptors
Presynaptic ionotropic glutamate receptors of the kainate type control synaptic transmission and presynaptic short-term plasticity at Mf-CA3 synapses (Contractor et al., 2001;Kamiya et al., 2002;Lauri et al., 2001;Mulle and Crépel, 2021;Pinheiro et al, 2007;Schmitz et al., 2001)(but see (Kwon and Castillo, 2008)).Presynaptic short-term plasticity is markedly decreased in both GluK2- (Contractor et al., 2001) and GluK3-deficient mice (Pinheiro et al., 2007).It has been hypothesized that kainate receptors act as autoreceptors located close to glutamate release sites to facilitate synaptic transmission by rapidly boosting Ca 2+ influx into the presynaptic Mf bouton following a single AP (Pinheiro et al., 2007;Schmitz et al., 2001;Scott et al., 2008), and possibly by activating Ca 2+ stores (Lauri et al., 2003;Scott et al., 2008).Finally, endogenous activation of presynaptic NMDA receptors, also acting as autoreceptors, during short trains of stimulation enhances short-term synaptic plasticity at Mf-CA3 synapses, with a significant contribution to Ca 2+ rise in the presynaptic Mf boutons (Lituma et al., 2021).It is however not known what would be the physiological conditions under which presynaptic NMDA receptors are activated by endogenously released glutamate.
Presynaptic metabotropic glutamate receptors of group II mGluRs (mGluR2/3) are also present in Mf-CA3 terminals and their activation can transiently suppress synaptic transmission by a mechanism which likely involves inhibition of presynaptic voltage-gated Ca 2+ channels (Pinheiro and Mulle, 2008).Blocking mGluR2/3 does not affect basal synaptic transmission but increases facilitation in a rather narrow frequency window (1-5 Hz), suggesting that this negative feedback may be counterbalanced by the substantial Ca 2+ influx resulting from higher frequency trains (Kwon and Castillo, 2008).

Modulation of short-term plasticity
Since the extent of short-term facilitation depends on the initial Pr, all conditions which alter initial Pr, may alter presynaptic short-term plasticity.For instance, increased short-term facilitation at Mf-CA3 synapses is observed in mice lacking the proteins Mover (Viotti and Dresbach, 2019), or in the absence of presynaptic tomosyn1 (Ben-Simon et al., 2015), for which there is evidence of decreased initial Pr.Similarly, Munc13-2 one of the Munc13 family member, is an essential presynaptic constituent required for basal synaptic transmission at Mf-CA3 synapses.Mf-CA3 short-term plasticity (i.e., paired-pulse and 1 Hz low frequency facilitation) is increased in constitutive Munc13-2 KO mice because of a reduction in Pr accompanied with an augmentation in failure rate (Breustedt et al., 2010).Presynaptically expressed LTP at Mf-CA3 synapses results in an increased Pr and recruitment of new release sites (Nicoll and Schmitz, 2005).Accordingly, Mf-CA3 LTP reduces the dynamic range of shortterm plasticity (Gundlfinger et al., 2007b).Finally, transsynaptic retrograde mechanisms, which may impact Pr also modulate presynaptic short-term plasticity (Carta et al., 2014;Vandael et al., 2021;Makani et al., 2021).
Subthreshold somatic depolarization of DG cells propagates several hundreds of micrometers along the Mf axon and modulates presynaptic neurotransmitter release at Mf-CA3 synapses by an analog coding mechanism (Alle et al., 2006), but this does not rely on changes in presynaptic Ca 2+ (Scott et al., 2008).It would be interesting to assess whether and how analog signaling may impact short-term plasticity, in particular in physiological conditions of DG cell bursting activity.

Consequences of Mf-CA3 short-term plasticity for circuit function Detonator properties in vivo
The DG provides sparse connections to CA3: each CA3 PC receives approximately 50 Mf inputs, and each DG granule cell contacts circa 15 CA3 PC.By contrast, there are many more, weaker, direct perforant path inputs from layer 2 of the entorhinal cortex onto each CA3 PC.Modeling studies have long proposed that the CA3 network works as an auto-associative network under the control of the Mf-CA3 connections, which act as "detonator" synapses (McNaughton and Morris, 1987).Even though the low initial Pr at Mf-CA3 synapses adds to the sparseness of Mf-CA3 connections, the "detonator" properties under specific physiological conditions, such as during learning, may make Mf-CA3 connections efficient in forcing a new pattern of firing onto the CA3 cells during memory encoding (Kesner and Rolls, 2015).Short-term facilitation (and mainly high frequency facilitation) endows excitatory Mf-CA3 synapses with these detonator properties.
Both tonic and train firing patterns result in facilitation of Mf-CA3 transmission in vivo, as described with bulk electrical stimulation protocols and extracellular recordings (Klausnitzer and Manahan-Vaughan, 2008).A major experiment in the field has quantified spike transmission between a single DG granule cell recorded and activated intracellularly, and postsynaptic CA3 PCs, in vivo in rats (Henze et al., 2002).Spike transmission between the DG and CA3 can occur, provided that the presynaptic cell fires with trains of spikes; spike transmission probability depends on the frequency of granule cell firing and number of spikes (Henze et al., 2002).This experiment confirms predictions that Mf-CA3 synapses are powerful and act as conditional detonators due to their short-term plasticity properties.Due to their low initial release and prominent facilitation, Mf-CA3 synapses thus act as high-pass filters to convey information from a single DG cell to a subset of CA3 PCs.
Mfs from DG granule cells also provide strong inputs onto CA3 interneurons, remarkably even more than onto PCs (Acsády et al., 1998), which form the structural basis for feedforward inhibition.Feedforward inhibition is a major component of spike transmission at Mf-CA3 synapses.DG granule cells contact at least four populations of interneurons, with distinct synaptic properties and features of short-term plasticity (Prince et al., 2016;Toth et al., 2000).A frequency-dependent switch from inhibition to excitation was previously observed in paired recordings of DG granule cells and CA3 pyramidal cells in organotypic slices by increasing presynaptic stimulation frequency (Mori et al., 2004).In vivo patch-clamp recordings from CA3 pyramidal cells showed that optogenetic stimulation of DG granule cells led to Mf-CA3 synaptic responses consisting predominantly of an IPSP at low stimulation frequency (0.05 Hz) (Zucca et al., 2017).Short-term plasticity of Mf-CA3 EPSPs with increasing frequency of presynaptic activity allows inhibition to be overcome to reach spike transmission in CA3 PCs, with a measured optimum around 10 Hz (Zucca et al., 2017).Albeit Mf synapses onto interneurons present various forms of short-term plasticity (Toth et al., 2000), in vivo presynaptic stimulation delivered in trains from 1 to 100 Hz did not enhance spike transmission onto CA3 interneurons (Henze et al., 2002;Zucca et al., 2017).

Natural spiking conditions
The physiological conditions under which short-term synaptic plasticity operates directly depend on the pattern of activity of DG granule cells in vivo, in behaving conditions.In vivo electrophysiological recordings or Ca 2+ imaging of putative DG granule cells in rodents have revealed that only a small minority of DG cells is active during behavior, whereas the rest are silent (Diamantaki et al., 2016;Pernia-Andrade and Jonas, 2014;Gundlfinger et al., 2010;Jung and McNaughton, 1993;Neunuebel and Knierim, 2012;Pilz et al., 2016).The use of juxtacellular recording techniques has allowed to morphologically identify the DG granule cells which are either silent (<15%) or active, in the rodent DG, and showed that the active pool is contributed by a subset of mature GCs with complex dendritic trees (Diamantaki et al., 2016).In the active neurons, a large fraction of spikes occurs within bursts with a large range of intraburst average frequency (from a few Hz to 100 Hz) (Diamantaki et al., 2016;Pernía-Andrade and Jonas, 2014;Vandael et al., 2020).
Natural patterns of DG granule cell activity used to stimulate Mfs while recording from CA3 PCs in vivo (Henze et al., 2002), and in slices (Gundlfinger et al., 2010;Mistry et al., 2011;Sachidhanandam et al., 2009) were shown to be effective in driving spike transmission.Suppressing short-term facilitation in GluK2 or GluK3 knock-out mice led to a profound reduction of spike transmission in response to a presynaptic stimulation pattern mimicking in vivo DG granule cell activity in a place field (Sachidhanandam et al., 2009).DG granule cells exhibit a wide range of spike patterns in behavioral conditions, some of which (or repetitions thereof) can induce long lasting potentiation of Mf-CA3 synaptic transmission (Chamberland et al., 2018;Gundlfinger et al., 2010;Mistry et al., 2011).Similarly, long bursts of spiking activity recorded in vivo (containing 10-24 APs) can trigger significant PTP in slices (Vandael et al., 2020).Finally, it is interesting to highlight the fact that a single sequence of spiking from a CA3 PC from a rat entering in a place field (<20 s), when delivered to a postsynaptic CA3 PC is sufficient to trigger robust DPE (Carta et al., 2014).At variance with other forms of presynaptic plasticity, which are specific to the Mf-CA3 synapse stimulate, DPE induced in a single spiking CA3 PC propagates to all presynaptic terminal Mf boutons connecting this cell (Carta et al., 2014).Hence, DPE and PTP, which share comparable durations, likely provide distinct computational rules to the DG-CA3 circuit.

Implications for hippocampal function and behavior
There is yet no direct demonstration of any consequence of suppressing the different forms of presynaptic facilitation for hippocampal function and for related behavior, such as spatial navigation and memory.Because the DG supports spatial pattern separation during learning, a process enabling the hippocampus to store different memories of similar events (Kesner and Rolls, 2015), it would be particularly relevant to investigate how the selective abrogation of presynaptic short-term plasticity impacts memory in this context.Computational modelling has proposed that Mf-CA3 connections are involved in the activation of subpopulations of CA3 PCs to form nonoverlapping representations of new memories via the strengthening of the associational/commissural (A/C) synapses (Kesner and Rolls, 2015;McNaughton and Morris, 1987).Reversible inactivation of Mfs in mice impaired the storage of spatial representation, but not the consolidation and recall of spatial memories, indicating that Mfs are preferentially involved in new learning (Lassalle et al., 2000).The impact of short-term facilitation of direct excitatory and indirect feed-forward inhibitory responses depends on connectivity the pattern of DG-CA3 and CA3-CA3 connections.Indeed if a single DG granule cell preferentially contacts a set of CA3 PC cells connected to each other by virtue of recurrent connections, short-term facilitation will have a strong influence on CA3-CA3 autoassociative networks, which are thought to be important for episodic memory encoding (Kesner and Rolls, 2015).On the other hand, it seems that feedforward inhibition is randomly wired from individual granule cells onto CA3 PCs (Neubrandt et al., 2017).It is interesting to note that brief spike trains in Mf-CA3 synapses in combination with A/C synaptic activity, can induce LTP within the CA3 autoassociative network (Kobayashi and Poo, 2004).This heterosynaptic interaction is compromised in conditions in which Mf-CA3 short-term facilitation is suppressed by the genetical deletion of either GluK2 or GluK3 KAR subunits (Sachidhanandam et al., 2009).This highlights the importance of train facilitation at Mf-CA3 synapses in the contribution to associative LTP at CA3-CA3 synapses.Feedforward inhibition is an important parameter for driving spiking activity in an ensemble of CA3 PCs following trains of spiking activity in DG granule cells.Interestingly, Mf-CA3 IN connections, which form the structural basis for feedforward inhibition, are subject to fast and robust structural remodeling upon learning (Ruediger et al, 2011), even though experimental evidence for the consequences of structural plasticity in terms of activity of CA3 circuits is mostly missing.The question also arises whether STP at Mf-CA3 synapses undergo homeoplastic changes following learning to compensate for increased feedforward inhibition.Testing the specific role of the predominant short-term facilitation at Mf-CA3 synapses for memory encoding should be made possible by the discovery that Mf-CA3 PC synapses do not facilitate in the absence of Syt7 (Jackman et al., 2016).

Mf-CA3 short-term plasticity in aging and in pathological conditions
Aging-related functional decline in both physiological and pathological conditions is thought to impact on short term plasticity mechanisms (Villanueva-Castillo et al., 2017;Briggs et al., 2017;Singh et al., 2018).A marked decrease in paired-pulse facilitation is observed at Mf-CA3 synapses in aged mice, although it is not clear whether this is due to a change in the facilitation mechanism or to an increase in initial Pr (Villanueva-Castillo et al., 2017).Such a decrease in paired-pulse facilitation at Mf-CA3 synapses is also observed in extracellular field recordings of mouse slices incubated with Aβ oligomers from human brain (Jin et al., 2022).In contrast, no significant impairment of short-term plasticity at Mf-CA3 synapses (low frequency facilitation, paired-pulse facilitation, PTP) was observed in the APP/PS1 mouse model of Alzheimer's disease (AD), neither at early stages nor at later (12 months) stages (Maingret et al., 2017;Viana da Silva et al., 2019).However, the extent of DPE showed a marked decrease in 6 months-old APP/PS1 mice, possibly because defects in lipid transport and metabolism are involved in the pathogenesis of AD (Viana da Silva et al., 2019).Contrasting results were obtained in the Tg2576a mouse model of AD; at early stages, PTP at Mf-CA3 synapses was compromised, whereas paired-pulse facilitation was not affected at any inter event interval (Lee et al., 2012).Impaired PTP was associated with an alteration of Ca 2+ uptake by mitochondria, and both defects were restored with the administration of an antioxidant (Lee et al., 2012).
Disruption of mitochondrial Ca 2+ homeostasis also accompanies deficits in train facilitation and PTP at Mf-CA3 synapses of mice deleted for presenilins (PS1 and PS2).Mutations of presenilins, which form the catalytic subunit of y-secretase, give rise to familial forms of AD (S. H. Lee et al., 2017).Deletion of PS1 and PS2 selectively from DG granule cells results in a robust impairment in FF and in vesicle replenishment rate at Mf-CA3 synapses, without any apparent change in the initial Pr (Barthet et al., 2018).This impairment of short-term plasticity is causally related to a selective downregulation of presynaptic Syt7 at Mf-CA3 synapses (Barthet et al., 2018).Hence, it seems that the aging process and AD, one of the most common forms of dementia, have as a common denominator deficit of Mf-CA3 short-term plasticity.
Overall, it has been proposed that in both models of schizophrenia and autism spectrum disorders, alteration of short-term synaptic plasticity may play a dominant role in dysfunctional information processing (Crabtree and Gogos, 2014).In-depth analysis of shortterm presynaptic plasticity at Mf-CA3 synapses has not yet been performed in the context of autism spectrum disorder.Mice with a truncation of the endogenous Disc1 ortholog to model the phenotypes of schizophrenia display changes in short-term plasticity at Mf-CA3 synapses (Kvajo et al., 2011).Interestingly, the presynaptic deficit is restricted to low frequency facilitation (<0.33 Hz), and is accompanied with an acceleration of recovery from facilitation (Kvajo et al., 2011).It has been suggested that the Disc1 mutation may induce a change in Ca 2+ handling by buffers such as calcineurin (Crabtree and Gogos, 2014).

Conclusions
Mf-CA3 synapses display unique structural and functional properties, among which presynaptic short-term plasticity, which show remarkable conservation from rodents to humans (Pelkey et al., 2023).There has been a large number of studies examining the mechanistic details of the different forms of short-term facilitation, which can be tentatively classified in three major types.Frequency facilitation (including low frequency facilitation and train facilitation) mainly depends on the slow kinetics of extrusion of Ca 2+ from the presynaptic bouton, and on the Ca 2+ sensor Syt7.PTP depends on cAMP signaling and is primarily caused by an increase in the readily releasable vesicle pool.DPE depends on lipid-mediated retrograde signaling, which is based on the modulation of presynaptic K + channels.There is likely a large interplay between the different forms of short-term plasticity, especially in physiological conditions of DG cell activity, to control spike transmission.Here we highlight some unresolved questions related to the remarkable features of short-term plasticity at Mf-CA3 synapses (Box 1), and the specific methods which have been developed to tackle these questions (Table 2).Whereas there has been key recent advancement in the understanding of the mechanisms, the physiological "raison-d'être" of this major property for hippocampal circuit function in behavioural conditions has not yet been clearly addressed.C. Low power electron microscopy image of a mossy fiber bouton (MfB), showing the surface of an individual MfB (in light yellow), its postsynaptic spiny excrescences (ex; in light blue) and surrounding astrocytic processes (in green), as revealed by glutamine synthetase preembedding immunohistochemistry. Active zones are highlighted in red and mitochondria (mi) in white.Scale bar: 1 μm.Taken from Rollenhagen et al., 2007.D. Schematic representation of a Mf-CA3 PC synapse, which form giant boutons that have an average of 20 release sites and make contact with a complex postsynaptic element in CA3 PCs, the thorny excrescence.In addition, filopodia emerging from mossy fiber boutons provide direct excitatory inputs onto CA3 interneurons (not shown).Glial cell processes surround the MfB but do not reach active zones.In vivo demonstration that Mf-CA3 synapses act as conditional detonators.Spike transmission between a DG granule cell recorded with an intracellular electrode and its pyramidal cell targets in CA3 recorded with an extracellular electrode in vivo.Left, camera lucida reconstruction of a biocytin-labeled granule cell and the extracellular electrode track where spike transmission was observed.Scale bar, 50 μm.m, molecular layer; g, granule cell layer; h, hilus; IC, intracellular electrode track; EC, extracellular electrode track.Center, superimposed (n 60) intracellularly evoked action potentials in a DG granule cell (bottom traces) and simultaneously recorded extracellular units in CA3.Right, mean percent of maximum spike transmission probability (± s.e.m.) onto CA3 PCs (n = 6) as a function of instantaneous frequency (1/ previous interspike interval) for each spike in the DG granule cell train shows that higher probability of spike transmission is associated with higher frequencies.Taken and adapted from Henze et al, 2002,    • Mechanisms of Syt7 action, a key player of presynaptic facilitation: investigate how exactly Syt7 controls docking with the use of genetic manipulation, channelrhodopsin stimulation and high pressure freezing.• Presynaptic residual Ca 2+ dynamics: because residual Ca 2+ is key for the activation of Syt7 in the context of STP, it is important to better understand the mechanisms, allowing the prolonged increase of presynaptic Ca 2+ , including the role of Ca 2+ buffers, and organelles such as mitochondria.

Bibliography Legends
• Targets of PKA underlying PTP: identify proteins of the vesicle release machinery, or presynaptic ion channels regulated by cAMP-dependent activation of PKA or Epac2.
• Presynaptic cAMP dynamics: explore how presynaptic cAMP (and cGMP) levels are regulated in the Mf boutons in physiological conditions or DG cell activity; in order to better understand the conditions under which PKA-dependent phosphorylation of presynaptic protein targets control Pr.
• Presynaptic autoreceptors: investigate the subcellular localization of presynaptic iGluRs (KARs, NMDARs) and the mechanisms by which they control STP at Mf-CA3 synapses.
• Neuromodulation: explore how neuromodulation (and which neuromodulators) is involved in the regulation of STP under physiological and behavioural conditions.

Physiology and behavioral consequences:
• Investigate the role of Mf-CA3 short-term plasticity in the transfer of information from the DG to CA3 in vivo in awake mice: use of high density electrophysiological recordings to study correlated circuit activity of DG and CA3 in conditions of suppression of shortterm plasticity, in relation with hippocampal oscillatory activity.• Investigate the role of the different forms of short-term plasticity at Mf-CA3 synapses on behaviour, including memory encoding, pattern separation and pattern completion dependent mechanisms, and on spatial information processing.

Figure 1 :
Figure 1: Morphological properties of Mf-CA3 synapses.A. Section of the mouse dorsal hippocampus showing the mossy fiber track labeled with a ZnT3 antibody (in red), running from the DG cell layer to the stratum lucidum in CA3.Neuronal cell bodies are labeled in green with a NeuN antibody.Courtesy of Adam Gorlewicz.B. Confocal microscopy images of CA3 PCs and mossy fibers boutons labeled by a membrane bound YFP, following in vivo infection of a myristoylated-palmitoylated YFP-expressing lentiviral vector.In insert, 3D reconstruction of thorny excrescences (in red).Adapted from Lanore et al, 2012, courtesy of Virginie Labrousse.C.Low power electron microscopy image of a mossy fiber bouton (MfB), showing the surface of an individual MfB (in light yellow), its postsynaptic spiny excrescences (ex; in light blue) and surrounding astrocytic processes (in green), as revealed by glutamine synthetase preembedding immunohistochemistry. Active zones are highlighted in red and mitochondria (mi) in white.Scale bar: 1 μm.Taken fromRollenhagen et al., 2007.D. Schematic representation of a Mf-CA3 PC synapse, which form giant boutons that have an average of 20 release sites and make contact with a complex postsynaptic element in CA3 PCs, the thorny excrescence.In addition, filopodia emerging from mossy fiber boutons provide direct excitatory inputs onto CA3 interneurons (not shown).Glial cell processes surround the MfB but do not reach active zones.

Figure 2 :
Figure 2: Different forms of short-term plasticity at Mf-CA3 synapses.A. Frequency facilitation (FF).On the left, representative traces of Mf-CA3 EPSCs recorded following induction by either low frequencies (0.1 and 1Hz) or high frequency train stimulation (20 Hz).On the right, a schematic representation of the extent and time course of FF.B. Post-tetanic potentiation (PTP) is induced by a train of high frequency stimulation (typically 25 to 100 Hz trains), and decays with a time constant of a few minutes.On the left, representative traces; on the right, schematic representation of the extent and time course.C. Depolarization induced potentiation of excitation (DPE) is trigerred by postsynaptic depolarization (or bursts of postsynaptic APs) in CA3 PCs, which results in a retrograde potentiation of excitation lasting up several minutes.On the left, representative traces; on the right, schematic representation of the extent and time course.

Figure 3 -
Figure 3 -Mechanisms of frequency facilitation.A. Field EPSP recordings of Mf-CA3 synaptic responses following train stimulation in control (black trace) and Syt7 KO (red trace) mice, showing a profound reduction of facilitation in the absence of Syt7 (right graph).Vertical scale bars: 100 µV.Taken from Jackmann and Regehr, 2016.B. Schematic representation of the time course of presynaptic voltage and APs (in black), of presynaptic Ca 2+ levels (blue), of Syt1 and Syt7 bound to the vesicule release machinery (in green and red, respectively).C. Schematic representation of a Mf active zone.Local Ca 2+ signal in the active zone increases following incoming action potentials (1) and decays with a fast and a slow component.The slowly decaying phase of Ca 2+ (2) is thought to allow summation of Ca 2+ concentration upon repetitive stimulation (3) which results in the activation of Syt7, a Ca 2+ sensor with high affinity and slow kinetics which promotes vesicle fusion.The orange symbols represent Ca 2+ buffers.

Figure 4 -
Figure 4 -Mechanisms of PTP.Schematic representation of a Mf-CA3 active zone illustrating the potentential mechanisms of PTP.PTP is induced by high frequency trains of APs, with the large influx of Ca 2+ leading to the Ca 2+--dependent activation of adenylate cyclase, which in turn leads to increased presynaptic cAMP and activation of either PKA and/or Epac2.The phosphorylation of PKA protein targets, which are not yet fully characterized in this context, results in the docking of vesicules and posterior facilitation of synaptic release.

Figure 5 -
Figure 5 -Mechanisms of DPE.Schematic representation of a Mf-CA3 active zone illustrating the potential mechanisms of DPE.Depolarization induced potentiation of excitation (DPE) is based on retrograde mechanisms involving membrane lipids (AA, arachidonic acid).By modulating presynaptic Kv channels, AA induces the broadening of action potentials and a subsequent increase of Ca 2+ influx, which facilitates synaptic release.

Figure 6 -
Figure6-Consequences of presynaptic facilitation on spike transfer at DG to CA3 connections: detonator properties.A. In vivo demonstration that Mf-CA3 synapses act as conditional detonators.Spike transmission between a DG granule cell recorded with an intracellular electrode and its pyramidal cell targets in CA3 recorded with an extracellular electrode in vivo.Left, camera lucida reconstruction of a biocytin-labeled granule cell and the extracellular electrode track where spike transmission was observed.Scale bar, 50 μm.m, molecular layer; g, granule cell layer; h, hilus; IC, intracellular electrode track; EC, extracellular electrode track.Center, superimposed (n 60) intracellularly evoked action potentials in a DG granule cell (bottom traces) and simultaneously recorded extracellular units in CA3.Right, mean percent of maximum spike transmission probability (± s.e.m.) onto CA3 PCs (n = 6) as a function of instantaneous frequency (1/ previous interspike interval) for each spike in the DG granule cell train shows that higher probability of spike transmission is associated with higher frequencies.Taken and adapted fromHenze et al, 2002, see details therein.B. Short-term facilitation underlies the detonator properties of Mf-CA3 synapses.Bottom left, representative traces of Mf-EPSPs recorded in CA3 PCs in slices of control and GluK2 -/-mice, in which Mf-CA3 short-term facilitation is strongly impaired.Top left, ten consecutive sweeps are shown as a raster plot indicating spikes, in response to the naturalistic stimulation pattern taken from in vivo DG granule cell recordings (illustrated at the top of the panel), in control and GluK2 -/-mice.Note the marked reduction in spiking activity in GluK2 -/-mice.Right, graph representing the number of spikes recorded in CA3 PCs as a function of the instantaneous frequency for control (black) and GluK2 -/-(red) mice.Taken and adapted fromSachidhanandham et al, 2009.
Figure6-Consequences of presynaptic facilitation on spike transfer at DG to CA3 connections: detonator properties.A. In vivo demonstration that Mf-CA3 synapses act as conditional detonators.Spike transmission between a DG granule cell recorded with an intracellular electrode and its pyramidal cell targets in CA3 recorded with an extracellular electrode in vivo.Left, camera lucida reconstruction of a biocytin-labeled granule cell and the extracellular electrode track where spike transmission was observed.Scale bar, 50 μm.m, molecular layer; g, granule cell layer; h, hilus; IC, intracellular electrode track; EC, extracellular electrode track.Center, superimposed (n 60) intracellularly evoked action potentials in a DG granule cell (bottom traces) and simultaneously recorded extracellular units in CA3.Right, mean percent of maximum spike transmission probability (± s.e.m.) onto CA3 PCs (n = 6) as a function of instantaneous frequency (1/ previous interspike interval) for each spike in the DG granule cell train shows that higher probability of spike transmission is associated with higher frequencies.Taken and adapted fromHenze et al, 2002, see details therein.B. Short-term facilitation underlies the detonator properties of Mf-CA3 synapses.Bottom left, representative traces of Mf-EPSPs recorded in CA3 PCs in slices of control and GluK2 -/-mice, in which Mf-CA3 short-term facilitation is strongly impaired.Top left, ten consecutive sweeps are shown as a raster plot indicating spikes, in response to the naturalistic stimulation pattern taken from in vivo DG granule cell recordings (illustrated at the top of the panel), in control and GluK2 -/-mice.Note the marked reduction in spiking activity in GluK2 -/-mice.Right, graph representing the number of spikes recorded in CA3 PCs as a function of the instantaneous frequency for control (black) and GluK2 -/-(red) mice.Taken and adapted fromSachidhanandham et al, 2009.

Table 2 -
Consequences of genetic manipulation on short-term plasticity at Mf-CA3 synapsesMechanisms of STP at Mf-CA3 synapses.