Target-cell-specific short-term plasticity in local circuits

Abstract

Short-term plasticity (STP) denotes changes in synaptic strength that last up to tens of seconds. It is generally thought that STP impacts information transfer across synaptic connections and may thereby provide neurons with, for example, the ability to detect input coherence, to maintain stability and to promote synchronization. STP is due to a combination of mechanisms, including vesicle depletion and calcium accumulation in synaptic terminals. Different forms of STP exist, depending on many factors, including synapse type. Recent evidence shows that synapse dependence holds true even for connections that originate from a single presynaptic cell, which implies that postsynaptic target cell type can determine synaptic short-term dynamics. This arrangement is surprising, since STP itself is chiefly due to presynaptic mechanisms. Target-specific synaptic dynamics in addition imply that STP is not a bug resulting from synapses fatiguing when driven too hard, but rather a feature that is selectively implemented in the brain for specific functional purposes. As an example, target-specific STP results in sequential somatic and dendritic inhibition in neocortical and hippocampal excitatory cells during high-frequency firing. Recent studies also show that the Elfn1 gene specifically controls STP at some synapse types. In addition, presynaptic NMDA receptors have been implicated in synapse-specific control of synaptic dynamics during high-frequency activity. We argue that synapse-specific STP deserves considerable further study, both experimentally and theoretically, since its function is not well known. We propose that synapse-specific STP has to be understood in the context of the local circuit, which requires combining different scientific disciplines ranging from molecular biology through electrophysiology to computer modeling.

Keywords: development; network models; short-term plasticity; synapse formation; synapse specificity; synaptic disease.

Figure 1

In local circuits, target-cell specific STP remaps spiking across the somato-dendritic axis. (A) Pyramidal cell (PC) inputs to basket cells (BC) are short-term depressing, whereas those to Martinotti cells (MC) are facilitating (Markram et al., 1998). As a result, high-frequency PC firing (Larkum et al., ; Murayama et al., 2009) activates MCs later than BCs, an effect that is amplified by presynaptic NMDA receptors (Figure 3 and Buchanan et al., 2012). BCs in turn innervate PCs perisomatically (Buchanan et al., 2012), whereas MCs contact the apical dendrite (Silberberg and Markram, 2007). (B) In cerebellum, synapses between parallel fibers (PF) and Purkinje cells (PuC) facilitate, as do connections to stellate cells (SC). In contrast, PF synapses onto BCs depress, so high-frequency PF activity triggers SCs later than BCs, leading to early onset somatic and late-onset dendritic PuC inhibition (Bao et al., 2010). (C) In hippocampus CA1, PCs connect to two distinct stratum oriens IN types with contrasting STP. Onset-transient BCs receive depressing input and target PCs and other INs perisomatically, whilst late-transient Martinotti-like cells (MLC) receive facilitating input and target dendrites (although see Hefft and Jonas, ; Glickfeld and Scanziani, 2006). During 50 Hz firing therefore, inhibition of PCs shifts from somatic to dendritic (Pouille and Scanziani, 2004). Early-onset BC inhibition of MLCs may additionally amplify this effect (Lovett-Barron et al., 2012). All synaptic traces were simulated based on data in (Pouille and Scanziani, ; Bao et al., ; Buchanan et al., 2012).

Figure 2

Postsynaptic Elfn1 controls presynaptic transmitter release. (A) Five stimuli at 20 Hz in the alveus produce short-term depression in stratum pyramidale or oriens PV INs (black). PV neurons overexpressing Elfn1-GFP no longer short-term depress (green). Left: example traces, right: ensemble normalized EPSC amplitude. (B) SOM INs normally facilitate (black), but facilitation is attenuated by GluR6-selective kainate receptor antagonist NS102 (left, red, n = 8). ShElfn1 expression reduces NS102 effect (right, n = 14). Top: example responses, bottom: ensemble normalized EPSC amplitude. (C) Proposed Elfn1 mechanism: CA1 PCs contact PV (blue) and SOM (red) INs (middle). PC synapses onto PV neurons lacking Elfn1 have short-term depression (blue). However, SOM INs expressing Elfn1 transsynaptically reduce probability of release, leading to facilitating PC-SOM IN connections (red). From (Sylwestrak and Ghosh, 2012). Reprinted with permission from AAAS. (A) ^*P < 0.05 by Mann-Whitney U test. (B) ^*P < 0.05; ^**P < 0.01 by ANOVA with Tukey's post-hoc test.

Figure 3

Presynaptic NMDARs reroute spiking in neocortical microcircuits. (A) Presynaptic NMDARs are specifically expressed at PC-PC and PC-MC synapses (closed symbols), but not at the other connections (open symbols). Circles denote inhibitory synapses. (B) A network model with tuned synaptic dynamics predicted that during 70-Hz presynaptic PC firing (vertical lines) blockade of preNMDARs would affect late MC but not early BC inhibition. Traces indicate the model's prediction of synaptic dynamics before (blue) and after (red) NMDAR blockade with AP5. (C) In this experiment, 70-Hz spiking in cell 1 resulted in both MC and BC-mediated inhibition in cell 3 when intermediate BC 2 was subthreshold depolarized. Intermediate MC (X) was not patched. (D) As predicted, AP5 wash-in affected the amplitude and latency of MC but not BC inhibition, indicating that preNMDARs at PC-MC but not PC-BC synapses boost neurotransmission during high-frequency firing. Reproduced from Buchanan et al. (2012) with permission from Elsevier.