In GABAergic synapses in layer II/III of the immature visual cortex the probability of release is high, resulting in pronounced short-term depression (Jiang et al., 2005
; Tang et al., 2007
). Here we present evidence that the transition to a mature state with low release probability and reduced STD is an experience-dependent process that requires the recruitment of endocannabinoid-dependent mechanisms similar to those that subserve iLTD in vitro. In support of this proposition we demonstrate that 1) endocannabinoid-dependent presynaptic iLTD is expressed in pyramidal neurons of layer II/III only during a critical period, which can be arrested but not reversed by visual deprivation, 2) synapses made by fast-spiking interneurons exhibit developmentally constrained iLTD and 3) although iLTD reduces the probability of GABA release, it also reduces short-term depression and the variability of synaptic transmission during irregular patterns of stimulation, similar to the changes we observe in response to visual experience. Together this suggests a novel role of endocannabinoid-mediated synaptic plasticity in the maturation of GABAergic transmission in the mammalian visual cortex, which may serve to shift the relative efficacy of GABAergic transmission to higher frequencies.
Endocannabinoid-dependent forms of presynaptic LTD have been described at both excitatory and inhibitory synapses in multiple brain regions, and a diversity of mechanisms appears to link synaptic activity to endocannabinoid synthesis and LTD induction (Chevaleyre et al., 2006
; Hashimotodani et al., 2007
; Lovinger, 2008
). The rules of induction of iLTD in layer II/III pyramidal neurons appear to be similar to those described in hippocampal CA1 neurons (Chevaleyre and Castillo, 2003
), although we have not identified which endocannabinoid species is involved in layer II/III iLTD. In both cases iLTD is only induced in active GABAergic inputs, and glutamatergic release is required to activate mGluR5 receptors and PLC, suggesting a heterosynaptic induction mechanism. However, there are seemingly mechanistic differences: CA1 iLTD requires spontaneous firing in the interneurons to activate calcineurin (Heifets et al., 2008
) in the axon terminals, whereas FS(PV) cells rarely fire in cortical slices (Maffei et al., 2004
). It is unclear whether in visual cortex calcineurin is not necessary, or if it is sufficiently activated by baseline stimulation alone.
The mechanisms for the developmental loss of iLTD are more difficult to compare due the scarcity of data. Prior to our study, developmental changes in endocannabinoid-mediated LTD have only been reported in inhibitory synapses in CA1 (Corlew et al., 2007
; Kang-Park et al., 2007
) and in excitatory synapses in layer II/III of visual cortex (Huang et al., 2008
). The developmental reduction of iLTD in CA1 has been attributed to an increase in the tonic level of endogenous agonists around puberty (Kang-Park et al., 2007
). In contrast, we demonstrate that the developmental loss of iLTD in layer II/III pyramidal neurons is associated with the loss of response to endocannabinoid agonists. In FS-PV→pyr synapses, this loss of responsiveness coincides with the end of synaptogenesis. The relationship between synaptogenesis and iLTD is unclear, especially given that at this stage visual experience drives many unrelated changes in synaptic function. Nevertheless, a simple scenario to consider is that iLTD is induced in vivo
by visually evoked activity and somehow drives the loss of responsiveness to endocannabinoids. The target of CB1R agonists would therefore be only naïve synapses; consequently, the response to endocannabinoids disappears with the end of GABAergic synaptogenesis. Whether the CB1Rs remain in FS(PV) axon terminals after the end of the critical period, and whether iLTD becomes irreversible once induced remain open questions.
Our study is the first to report endocannabinoid modulation of synaptic function made by FS-PV interneurons in the mammalian cortex, although such modulation has been documented in the amygdala (McDonald and Mascagni, 2001
) and basal ganglia (Freiman et al., 2006
; Narushima et al., 2006
; Uchigashima et al., 2007
). In cortex it was reported that FS-PV neurons were immunocytochemically negative for CB1 receptors (Katona et al., 1999
). However, it is possible that immunocytochemistry is not sensitive enough to detect CB1Rs in PV-positive interneurons. There are several examples in which immunocytochemical methods did not detect the presence of CB1Rs, but cannabinoid modulation was confirmed with other methods (Hill et al., 2007
; Sjostrom et al., 2003
). Our finding that WIN does not affect the FS(PV)-mediated uIPSC at 1 μM but it depresses them at 10 μM suggest a low content of CB1 receptors in PV containing axon terminals. A relative lower CB1 content at the PV-axon terminal could also explain why 1 μM WIN depresses the CCK-mediated, but not the FS(PV)-mediated uIPSC (Galarreta et al., 2008
). Alternatively, higher concentrations of WIN might recruit additional processes (other than CB1Rs) that are also required for this form of iLTD. Regardless of the mechanism, the requirement for a high dose of agonist predicts that the threshold for iLTD induction would be rarely achieved in vivo
in the absence of visual experience.
Recent work has shown that presynaptic forms of iLTD and iLTP coexist at the same synapses (Nugent et al., 2007
; Pan et al., 2008
). Postsynaptic form of LTP and LTD are generally regarded as activity-dependent mechanisms that allow bi-directional modification of synaptic strength. Interestingly, the reduction in inhibitory function after iLTD in CA1 pyramidal neurons also lowers the threshold for activity-dependent postsynaptic LTP at excitatory synapses (Chevaleyre and Castillo, 2004
). Here we demonstrate that iLTD plays additional roles in the visual cortex. Our finding that iLTD reduces the responses evoked at low frequencies (0.03 to 0.05 Hz) but not at high frequencies (30 Hz) () is consistent with a redistribution model of presynaptic plasticity in which changes in synaptic strength depend on the stimulation frequency (Abbott et al., 1997
; Tsodyks and Markram, 1997
). Thus, iLTD might enhance rather than reduce GABAergic function during high firing rates. One proposed role of fast-spiking interneurons is to integrate network activity and to broadcast an inhibitory signal that confers homeostatic stability to the network. Such a negative feedback function would be better served by synapses with low probability of release that attenuate less during sustained firing than by synapses with high probability of release that rapidly deplete synaptic resources.
It is interesting to note that in parallel with the developmental decrease in the probability of release, fast-spiking GABAergic cells also increase the expression of potassium channels responsible for the repolarization of the action potential (Goldberg et al., 2005
). This shortens the duration of the action potential, thereby reducing the probability of the release per single action potential, but allows higher rates of firing and release. This suggests a developmental orchestration of mechanisms aimed at enhancing GABAergic performance during high frequency activation, at the expense of low frequency performance. During maturation, the reduction in the probability of release at low frequencies may be compensated by the 2 to 3-fold increase in the number of release sites. Whether such a developmental increase in the capacity fast transmission contributes to the closure of critical period remains to be determined. Finally, the principles we describe for the maturation of GABAergic inputs to layer II/III pyramidal cells might be more general. In all cases examined, glutamatergic synapses in sensory cortices undergo a developmental reduction in paired-pulse depression (Feldmeyer and Radnikow, 2009
; Frick et al., 2007
; Reyes and Sakmann, 1999
; Yanagisawa et al., 2004
) and also deprivation-induced increases in short-term depression (Finnerty et al., 1999
), similar to what we describe for GABAergic synapses. It is possible that endocannabinoid-mediated regulation of synaptic transmission may be a ubiquitous mechanism in the developmental maturation of synaptic transmission at both GABAergic and glutamatergic synapses.