2.2.1. Minimal Requirements and Characteristics

In the past decade, long-term changes in the strength of inhibitory synapses have been reported in a large number of developing and adult structures, including Mauthner cells of the goldfish [34], the hippocampus [9,22,24,40,41,49,54,56], the cortex [21,30], the cerebellum [23], the deep cerebellar nucleus [43], the lateral superior olive [36], lateral amygdal [2] and the brain stem [4,17] (Figure 2.1 and Figure 2.2). In all these structures, attention had be focused to demonstrate that GABAergic and glycinergic synapses themselves undergo long-term changes in synaptic efficacy. Long-term plasticity was observed on pharmacologically isolated inhibitory postsynaptic potentials (IPSPs) or currents (IPSCs) or on unitary IPSCs evoked by direct stimulation of the interneurons.

FIGURE 2.1

Long-term changes in the strength of inhibitory synapses in the developing brain. (a) In the neonatal rat hippocampus, high-frequency stimulation leads to LTD and LTP of GABAergic synapses. Both forms of synaptic plasticity require a membrane depolarization (more...)

FIGURE 2.2

Long-term changes in the strength of inhibitory synapses in the adult brain. (a) In the adult rat hippocampus, repetitive activation of Schaffer collaterals triggers heterosynaptic LTD that is induced via the activation of postsynaptic group I metabotropic (more...)

To date, most studies on plasticity at GABAergic synapses have focused on GABA_A-receptor mediated postsynaptic potentials (GABA_A-PSPs). However, activation of either pre- or post-synaptic GABA_B receptors during the conditioning protocol appears to have a key role in the induction of long-term plasticity at both GABA_Aergic and glycinergic synapses (Figure 2.1a and Figure 2.2c) [4,22,31,35,49,53], although in one case activation of GABA_B receptors has been reported to prevent the induction of GABAergic synaptic plasticity [25,26]

In most studies, plasticity at inhibitory synapses can be induced by high- or low-frequency stimulations [15]. Although such conditioning protocol provides a good tool to study the mechanisms by which plasticity are triggered and expressed, such patterns of activity are rather unlikely to occur in vivo. Thus, the relevance of the conditioning protocol to physiological or pathological conditions must be taken into account to gain insight into the role that this plasticity might serve in the developing and adult nervous system. Compelling evidence shows that conditioning protocols relevant to physiological or pathological conditions can also induce plasticity at GABAergic and glycinergic synapses. The best example was provided by Korn and colleagues, who showed that auditory stimulations in the goldfish trigger LTP of glycinergic synapses in vivo [47] that shares similar properties with LTP induced by tetanic stimulation [34,46]. More recently, Cherubini and colleagues have reported that spontaneously occurring network oscillations, which constitute a hallmark of developing networks, can trigger LTP_GABA-A in the developing rat hippocampus [24]. In the hippocampus and cortex, plasticity of inhibitory synapses can also be triggered by post-synaptic firing of the target cells [5,38] or by coincidence of back-propagating action potentials with pre-synaptic stimulation of inhibitory terminals [21,56]. Such conditioning protocols are related to the sort of activity that pyramidal neurons might experience in vivo.

When stimulation of presynaptic inhibitory terminals is required, LTP and LTD is only expressed by the conditioned fibers. In the adult rat hippocampus, LTD_GABA-A is a highly localized phenomenon that spreads less than 20 μm away from the conditioned fibers along the apical dendrite [9]. In this particular example, the local regulation of GABAergic synaptic strength is due to the restricted spread of the retrograde signal required for LTD_GABA-A induction to synapses impinging on a small portion of the CA1 pyramidal cell dendrite. The interneurons exhibited a great heterogeneity. One of the most striking differences is found on the restricted and specific laminar distribution of their axonal terminals on dendritic and perisomatic regions of their target cells [14]. This heterogeneity provides the possibility that the different interneuronal populations may selectively influence the efficacy of afferent inputs and the emergence and maintenance of network oscillations. Thus, depending on the type of interneurons, local changes in the efficacy of inhibitory synapses will have different consequences on the input-output relationship of the target neurons.

Both homosynaptic and heterosynaptic plasticity at inhibitory synapses have been reported (Figure 2.1 and Figure 2.2). In the neonatal rat hippocampus, the induction of LTP_GABA-A requires a membrane depolarization, provided by the activation of GABA_A receptors during the conditioning protocol [18,41]. This depolarization is strong enough to allow activation of voltage-dependent calcium channels (VDCCs), which will in turn trigger a cascade of events leading to homosynaptic LTP_GABA-A (Figure 2.1a). In the adult and neonatal rat hippocampus, tetanic stimulation triggers heterosynaptic LTD_GABA-A. In both cases, LTD_GABA-A is triggered post-synaptically via the activation of glutamatergic receptors during the conditioning protocol. That is, glutamate released during repetitive stimulation activates group I metabotropic glutamate receptors [9] on adult CA1 pyramidal cells (Figure 2.2a) or N-methyl-D-aspartate (NMDA) receptors on neonatal CA3 pyramidal cells [6,42] (Figure 2.1a) or adult CA1 pyramidal cells (Figure 2.2b) [40,55]. Activation of these receptors will, in turn, trigger LTD_GABA-A. Such heterosynaptic plasticity will locally affect dendritic integration of synaptic input, for instance, by decreasing or increasing the inhibitory shunt of neighboring glutamatergic afferents. Moreover, several studies have reported that a conditioning protocol can trigger LTP at excitatory synapses and, concomitantly, LTD at inhibitory synapses [40,55]. This opposite change in excitatory and inhibitory synaptic strength can change the ability of the excitatory synaptic potential to discharge an action potential [40,55], thereby changing the excitability of neuronal networks.

As with glutamatergic synaptic plasticity, a rise in intracellular Ca²⁺ concentration is important in shaping the strength of inhibitory synapses, although the source and location of this calcium rise and the consequences on inhibitory synaptic strength will vary depending on the structure considered and the conditioning protocol used (Figure 2.1 and Figure 2.2). In the neonatal rat hippocampus, the induction of LTP_GABA-A requires a calcium influx through postsynaptic VDCCs (Figure 2.1a) [5,18], while in the cortex (Figure 2.1c) [31] and cerebellum (Figure 2.2d) [20], activation of postsynaptic calcium stores is required. An influx of calcium through VDCCs triggers LTD in the cortex [38] and deep cerebellar nuclei (Figure 2.1b) [44]. An influx of calcium through postsynaptic NMDA channels also triggers LTD_GABA-A in the neonatal and adult rat hippocampus (Figure 2.1a and Figure 2.2b) [6,40,55] as well as in the neocortex (Figure 2.1b) [32]. Interestingly, the initial depolarization required for the removal of the Mg²⁺ block of NMDA channels depends on the developmental stage. This depolarization is provided by activation of postsynaptic GABA_A receptors in the neonates (Figure 2.1a) and by α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptors (Figure 2.2b) in the adult. Release of calcium by intracellular calcium stores could also trigger LTP_GABA-A in the neocortex (Figure 2.1c) [31] and hippocampus (Figure 2.2c) [22]. In these peculiar examples, GABA released during the conditioning protocol activates postsynaptic metabotropic GABA_B receptors, which then causes or favors intracellular release of calcium.

In most cases, the calcium rise that triggers plasticity of inhibitory synapses occurs in the post-synaptic target cells, as post-synaptic loading with a calcium chelator prevents its induction. However, the calcium rise could also occur in the presynaptic terminal [7,16] or in neighboring astrocytes (Figure 2.2c) [22]. Thus, in the adult hippocampus, repetitive firing of a single interneuron triggers a calcium- dependent LTP_GABA-A on pyramidal cells. The calcium rise is triggered on neighboring astrocytes following the activation of GABA_B receptors by GABA released during the conditioning protocol [22]. This calcium rise then causes the release of a retrograde messenger acting on GABAergic terminals [22].

Interestingly, the same conditioning protocol can lead to both LTP and LTD, depending on the experimental conditions (Figure 2.1a). The source or the magnitude of the postsynaptic Ca²⁺ rise has been proposed to determine the polarity of the plasticity. In the neonatal rat hippocampus, high-frequency stimulation can lead to either LTP_GABA-A or LTD_GABA-A, depending on whether or not NMDA receptors were activated during the conditioning protocol [41]. Given that both forms of plasticity require a postsynaptic rise in calcium [41], these observations suggest that the polarity of synaptic changes might be determined by the source of calcium influx, i.e., an influx through VDCCs triggers LTP_GABA-A [5], whereas an influx through NMDA channels triggers LTD_GABA-A (Figure 2.1a) [6]. Alternatively, the magnitude of the calcium rise could determine the polarity of synaptic changes, as documented in the deep cerebellar nucleus. In this structure, the number of spikes and the related amount of Ca²⁺ that enters the post-synaptic neurons during the post-inhibitory rebound depolarization seems to determine whether LTP_GABA-A or LTD_GABA-A is induced with large numbers of spike leading to LTP_GABA-A [1].

2.2.2. Expression

Long-term changes in the strength of synaptic efficacy can be accounted for by at least four nonexclusive mechanisms: modifications in the number or properties of receptors at functional synapses; changes in the reversal potential of post-synaptic responses; modifications in the probability of transmitter release; and modifications in the number of functional synapses though either pre- or post-synaptic changes. Probably because fewer studies have been performed to date, the locus of expression of plasticity at inhibitory synapses appears less controversial than at their excitatory counterparts.

If the amount of GABA released during synaptic stimulation reaches the concentration that saturates postsynaptic GABA_A receptors, the total number or properties of receptors at functional synapses will be an important limiting factor. In the adult dentate gyrus, a direct relationship between synaptic GABA_A receptor number and quantal size at potentiated GABAergic synapses has been reported in an experimental model of temporal lobe epilepsy [45]. In this model, insertion of new GABA_A receptors underlies the increase in amplitude of unitary IPSCs. Other studies have reported that the pathway through which this Ca²⁺ rise is translated into long-term changes of synaptic efficacy involves changes in the properties of post-synaptic GABA_A receptors through activation of protein kinases or phosphatases [29]. Thus, in the adult hippocampus, the expression of LTD_GABA-A involves a down-regulation of GABA_A receptors by the calcium-sensitive phosphatase, calcineurin (Figure 2.2b) [40,55]. In the cerebellum, the expression of LTP_GABA-A required the activation of post-synaptic calcium-calmodulin-dependent kinase II (CaMKII, Figure 2.2d) [23,26]. In the lateral superior olive, both post-synaptic CaMKII and protein kinases A and C participate in the LTD of inhibitory synapses [37].

Changes in the strength of inhibitory inputs can also result from modification in the reversal potential of GABAergic synaptic responses, as documented in the hippocampus [56]. In this structure, coincident pre- and post-synaptic spiking leads to a persistent decrease in GABAergic synaptic strength associated with a depolarizing shift of the reversal potential of GABA_A receptor-mediated synaptic potentials (E_GABA). Similarly, pairing exogenously applied GABA with postsynaptic depolarization leads to a long-lasting transformation of hyperpolarizing GABAergic responses into depolarizing responses [12]. GABA_A receptors are permeable to chloride and the intracellular concentration is regulated by different chloride cotransporters [50]. In their study, Woodin et al. have shown that coincident pre- and post-synaptic spiking decreases the cation/chloride cotransporter KCC2 activity, resulting in the shift of E_GABA to more positive values [56]. Long-lasting change in E_GABA have been also reported in epileptic tissue [10,27], supporting the contribution of such phenomenon in the emergence and maintenance of pathological network activity

However, recent studies have reported that saturation of GABA_A receptors does not occur at all sites of GABA release [19]. In this case, change in the probability of GABA release or in the number of functional release sites might underlie the expression of synaptic plasticity. Compelling evidence has accumulated in support of this hypothesis. Thus, long-term changes in the strength of inhibitory synapses are often associated with changes in the failure probability of unitary inhibitory postsynaptic currents (IPSCs) [22,24,46]; changes in the coefficient of variation of evoked IPSCs [6,16,46]; changes in the paired pulse ratio of evoked IPSCs [9,24]; and changes in the frequency, but not amplitude, of miniature IPSCs [5,22] or asynchronous quantal IPSCs evoked in the presence of strontium [6].

A modification in the number of functional release sites has been directly demonstrated in the goldfish [8]. Thus, in dual recordings of presynaptic glycinergic interneurons and postsynaptic Mauthner cells, 25% of the interneurons produced no detectable postsynaptic response. Morphological examination revealed that the number of synaptic contacts made by these “silent” interneurons is similar to that of “functional” interneurons. After a tetanic stimulation of the VIIIth nerve, that produces LTP_gly , the “silent” interneurons become functional. Similarly, in the neonatal rat hippocampus, postsynaptic application of a conditioning protocol leading to LTP_GABA-A leads to the appearance of functional GABAergic synapses in previously “silent” CA3 pyramidal neurons [18]. It remains to be determined whether the modifications in the number of functional release sites is due to all-or none modifications in the number of postsynaptic receptors or to presynaptic switching on in transmitter release.

If plasticity is expressed at the postsynaptic level while induction requires a postsynaptic rise in calcium, the information should be transmitted back from the postsynaptic cell to the presynaptic inhibitory terminal. Such synaptic feedback provided by retrograde messengers has been demonstrated in the rat hippocampus. Chevaleyre and Castillo [9] recorded from adult hippocampal slices, and found that high frequency stimulation induces an NMDA-independent LTD_GABA-A. This LT-_GABA-A is triggered postsynaptically via activation of group I metabotropic glutamatergic receptors (mGluRs), but is expressed presynaptically (Figure 2A). Thus activation of postsynaptic group I mGluRs leads to the release of endocannabinoids, which then causes a persistent reduction of GABA release. The implication of a trans-synaptic messenger in the induction of LTP_GABA-A in the rat hippocampus has also been demonstrated in an elegant study by Kang and collaborators [22] (Figure 2.2c). In this study, repetitive firing of hippocampal interneurons leads to the activation of GABA_B receptors on astrocytes, and a subsequent increase in intracellular calcium concentration. This calcium rise causes the release of a messenger from astrocytes, probably glutamate, which in turn triggers an increase in the probability of GABA release. Similar feedback regulation of inhibitory synaptic transmission has been reported in the rat cerebellum, although this control occurs in a short (seconds to minutes) time scale. Thus, short depolarization of Purkinje cells leads to a transient (tens of seconds) decrease followed by a short-term (10 to 15 minutes) increase in the frequency of miniature inhibitory post-synaptic currents (mIPSCs). Both phenomena, termed respectively depolarization-induced suppression of inhibition (DSI) [39] and depolarization-induced potentiation of inhibition (DPI) [13], require a post-synaptic rise in calcium concentration and are mediated presynaptically. In DSI, the post-synaptic calcium rise initiates endocannabinoid synthesis that activates presynaptic CB1 receptors, resulting in inhibition of GABA release for tens of seconds [48]. In DPI, the post-synaptic rise in calcium induces the release of glutamate from Purkinje cells [13]. Glutamate activates pre-synaptic NMDA receptors, leading to calcium influx into the pre-synaptic terminal that, with the activation of ryanodine-sensitive calcium stores, induces a short-lasting increase of GABA release.

In all studies, modified synaptic strength seems to be maintained by a mechanism independent of neuronal activity. In contrast, in the developing rat visual cortex, the maintenance of LTP_GABA-A requires firing of pre-synaptic inhibitory terminals and pre-synaptic calcium influx through VDCCs (Figure 2.1c) [33]. Thus, if stimulation of the test pathway was stopped after LTP_GABA-A induction, potentiated responses returned to a baseline level. Because this plasticity could underlie experience-dependent refinement of visual inputs early in life, this observation suggests that this refinement might not persist unless strengthened synapses are activated by visual stimulation.

2.2.3. Functional relevance

Most of our interests on long term changes in synaptic efficacy stem from the possible functions that LTP and LTD might serve in the developing and adult nervous system.

In the adult nervous system, the role of plasticity at inhibitory synapses was previously thought to regulate the occurrence of plasticity at excitatory synapses. However, plasticity of inhibitory synapses, in parallel with plasticity at excitatory synapses, can also affect the probability of the target neurons to fire action potentials, and will contribute to hyperexcitability of neuronal networks observed in pathological conditions. The balance between excitation and inhibition in neuronal networks can critically influence the level and type of spontaneous synaptic activity within the network, and changes in synaptic strength could lead to profound network modifications. For instance, in the adult rat hippocampus, LTD_GABA-A induced by tetanic stimulation revealed latent excitatory synaptic connections between hippocampal pyramidal cells. Such phenomenon likely contributes to the emergence or maintenance of epileptiform activity [54]. Plasticity at inhibitory synapses alone, without changes in glutamatergic synaptic strength, could also have behavioral consequences. This was directly addressed in the goldfish with the plasticity at glycinergic synapses onto Mauthner cells. Activation of the Mauthner cell system by sound is known to contribute to an escape reaction that orients the fish away from the predator [59]. In an elegant study performed in vivo, sound stimulation sufficient to produce LTP of glycinergic synapses, but inefficient to modify the excitatory inputs onto Mauthner cells, was reported to decrease the probability to escape in the conditioned goldfish [57,58], whereas the basic properties of the escape reflex were not modified.

Plasticity at inhibitory synapses could also play a crucial role in the developing brain. Thus, both LTP and LTD of inhibitory synapses have been described in different developing brain regions, including the lateral superior olive [36], the cortex [30] and the hippocampus [41]. In all these structures, plasticity is induced during a restricted period of development that closely matches the period of functional synaptic maturation. In the auditory system, the tonotopic organization of glycinergic projections is achieved through synapses elimination [51], a process involving activity-dependent mechanisms [52]. The structural refinement of axonal arbors emerges gradually and is preceded by a functional elimination and strengthening of GABA/glycine connections [28]. The period during which LTD is induced in this structure coincides with the period of functional elimination of inhibitory synapses [36].

To strengthen the link between activity-dependent maturation of inhibitory synapses and long-term plasticity showing that the activity involved in the development of inhibitory synapses is able to produce modifications in inhibitory synaptic strength is necessary. The presence of spontaneous pattered synaptic activity is a hallmark of the developing network [3]. In the neonatal hippocampus, this network activity, termed giant depolarizing potentials (GDPs), consists of bursts of action potentials associated with post-synaptic influx of calcium through VDCCs. Thus, the minimal requirements to induce LTP_GABA-A in the neonatal rat hippocampus are present during GDPs. This activity could therefore represent the physiological pattern of activity leading to the functional maturation of inhibitory synapses through LTP/LTD-like mechanisms [18]. In agreement with this hypothesis, pharmacological blockade of the GDPs prevents the functional maturation of GABAergic synapses [11], and application of a conditioning protocol that mimics, at least in part, the post-synaptic consequences of GDPs triggers LTP_GABA-A during a narrow postnatal time window [18]. Moreover, the same protocol also leads to the appearance of functional GABAergic synapses on previously silent cells, thus mimicking the functional maturation of GABAergic synapses occurring in vivo [5,18]. Finally, pairing evoked GABAergic synaptic responses with GDPs leads to LTP_GABA-A in the neonatal rat hippocampus [24].

A key issue that might uncover the link between activity-dependent functional maturation and synaptic plasticity is to show that the same cellular mechanisms are involved in both phenomena. In this context, neurotrophins and related Trk receptor-coupled protein tyrosine kinases (PTKs) have been implicated in synapse development and plasticity, and could likely represent the signal linking long-term plasticity and activity-dependent maturation of inhibitory synapses. In agreement with this hypothesis, recent results suggest that TrkB receptors participate in the induction of LTD_Gly in the developing auditory brain stem [35] and LTP_GABA-A in the developing rat hippocampus (Gaiarsa and Gubellini, unpublished results).