Hebbian spike-timing dependent plasticity at the cerebellar input stage

Spike-timing dependent plasticity (STDP) is a form of long-term synaptic plasticity exploiting the time relationship between postsynaptic action potentials (AP) and EPSPs [1-4]. Surprisingly enough, very little was known about STDP in the cerebellum [5-8], although it is thought to play a critical role for learning appropriate timing of actions. We speculated that low-frequency oscillations observed in the granular layer may provide a reference for repetitive EPSP/AP phase coupling. Here we show that EPSP-spike pairing at 6Hz ( 60 times) can optimally induce STDP at the mossy fiber - granule cell synapse (Fig. 1). When the AP followed the EPSP, EPSP/AP pairing with 0<t<25 ms induced long- term potentiation of EPSC amplitude (st-LTP: +47.4 ± 11.7%, n=11, p<0.05). When the AP preceded the EPSP, EPSP/AP pairing with 0<t<-25 ms induced long-term depression of EPSC amplitude (st-LTD: -37.7 ± 8.5%, n=13, p<0.005). In order to verify the STDP requirement for a phased-locked EPSP/AP pairing, in a series of recordings the time between the EPSP onset and the AP peak was varied randomly (Fig. 2). After random EPSP/AP pairing, EPSC amplitudes were not significantly changed (-1.4 ± 1.9%, n=5; p=0.8), showing that STDP induction was critically dependent on the maintenance of a precise EPSP/AP phase relationship. Since EPSPs led APs in st-LTP while APs led EPSPs in st-LTD, STDP was Hebbian in nature. STDP occurred at 10 Hz but vanished below 1 Hz. Figure 3 shows the time course of EPSC changes with 10 Hz and 1 Hz pairing. With 10 Hz pairing, STDP was still present showing st-LTP at positive EPSP/AP pairing (t=+25 ms, 23.9 ± 5.4%, n=4; p<0.05) and st-LTD at negative EPSP/AP pairing (t=-25 ms, -18.0 ± 3.1%, n=5; p<0.05). Conversely, with 1 Hz pairing, STDP disappeared leaving only LTD both at positive EPSP/AP pairing (t=+25 ms, -33.5 ± 12.1%, n=5; p<0.05) and at negative EPSP/AP pairing (t=-25 ms, -30.9 ± 8.0%, n=5; p<0.005). In a different series of recordings, in order to investigate whether STDP depended on postsynaptic Ca2+ concentration ([Ca2+]i) changes, the pipette intracellular solution was supplemented with the calcium buffer, 10 mM BAPTA. Figure 4 shows the time course of EPSC changes. The high concentration of BAPTA prevented both st-LTP (-5.3 ± 5.7%, n=4; p=0.4) and st-LTD (+3.7 ± 10.1%, n=4; p=0.6). In order to examine the induction mechanism underlying STDP (Fig. 5), we evaluated the involvement of NMDARs and MGluRs, which are primarily responsible for the postsynaptic [Ca2+]i changes required for both LTP and LTD at several glutamatergic synapses [9-11]. Both st-LTP and st-LTD required NMDA receptors, but st-LTP also required reinforcing signals mediated by mGluRs. When the NMDAR blockers D-APV (50 μM) and 7-Cl-Kyn acid (50 μM), were added to the extracellular solution, STDP protocols failed to induce either st-LTP (+4.7 ± 2.9%, n=5; p=0.2) or st-LTD (-10.6 ± 5.5%, n=6; p=0.2). During extracellular application of the mGluR antagonist AIDA (15 μM), protocols used for st-LTD induction still caused a significant EPSC reduction (-60.4 ± 11.3%, n=4; p<0.05. However, during a similar AIDA application, protocols used for st-LTP induction proved inefficient and a significant st-LTD was observed in turn (-16.9 ± 5.4%, n=5; p<0.05). Importantly, st-LTP and st-LTD were significantly larger than LTP and LTD obtained by modulating the frequency and duration of mossy fiber bursts [9, 10], probably because STDP expression involved postsynaptic in addition to presynaptic mechanisms. The mechanism of STDP expression was first assessed by analyzing the paired-pulse ratio (PPR, interstimulus interval 20 ms) and the coefficient of variation of EPSCs (CV) [12, 13](Fig. 6). .During st-LTP, PPR showed a significantly reduction by 18.2 ± 6.5 % (p<0.05, n=5), while during st-LTD PPR showed a significantly increase by 26.3 ± 9.4% (p<0.01, n=4). During st-LTP, CV showed a significantly reduction by –27.9 ± 5.4% (p<0.005, n=9), while during st-LTD CV showed a significant increase by +37.1 ± 11.9% (p<0.05, n=9). The PPR and CV changes followed the time course of EPSC amplitude (see Fig. 1), suggesting that the a bidirectional modification in quantum content took part to mossy fiber-granule cell STDP. Additional evidence for a pre- or postsynaptic mechanism of expression for STDP can be obtained by analyzing miniature EPSCs (mEPSCs) before and after induction of STDP [14-16]. After EPSP/AP pairing with t=+25 ms, the EPSCs showed +22.0 ± 6.3% increase (p<0.05, n=5) attesting effective st-LTP induction. Interestingly, in the same recordings the mEPSCs showed significant increase in both amplitude (+16.9 ± 6.3 %, n=5; p<0.05; Fig.) and frequency (+18.1 ± 8.7%, n=5; p<0.05). After AP-EPSP pairing with t=-25 ms, the EPSCs showed -43.8 ± 4.0 % change (p<0.05, n=4) attesting effective st-LTD induction. In this case, the mEPSCs showed a decrease in both amplitude (-20.1 ± 8.8%, n=4; p<0.05) and frequency (-29.8 ± 11.1%, n=4; p<0.05; Fig. 7). Altogether, these results suggested that a modification in quantum size took part to mossy fiber - granule cell STDP. This work reveals the existence of STDP at the rat cerebellar mossy fiber – granule cell synapse. While st-LTP was induced when EPSP led AP, st-LTD was induced when AP led EPSP according to the Hebbian principle of coincidence detection between pre- and postsynaptic activity. To our knowledge, this is the first evidence that Hebbian STDP occurs in the cerebellum, since previous observations only reported either non-Hebbian STDP at the parallel fiber-Purkinje-like cell synapse [5] or anti-Hebbian STDP at the corresponding synapses of cerebellum-like structures [6-8]. Mossy fiber - granule cell STDP was detected using repeated cycles on the theta-frequency band and depended on precise positive or negative phase locking, such that it was abolished by random pairing. STDP was optimal in the theta-frequency band (6-10 Hz). This is actually a range of frequencies at which coherent oscillations have been detected in the granular layer [17, 18]. Thus, STDP at mossy fiber – granule cell synapse could provide a mean to coordinate learning at the cerebellar input stage with activity generated in extracerebellar structures like the neocortex and striatum, that show coherent oscillations with the cerebellum [19-21].