A dendritic switch between Hebbian and anti-Hebbian plasticity at distal neocortical synapses


Per Jesper Sjöström* and Michael Häusser

Wolfson Institute for Biomedical Research, University College London, London WC1E 6BT, UK


Backpropagating action potentials (APs) are required for the induction of synaptic plasticity in many neuronal types. We modulated AP backpropagation and investigated its effects on synaptic plasticity in L5 pyramidal neurons in neocortical slices at 34C. Paired recordings were made with presynaptic pyramidal neurons in L2/3 or L5. Synaptic contacts located in reconstructions indicated that 20-80% EPSP rise time (RT) predicts input location accurately (r=0.88). L2/3-to-L5 synapses tended to be more distal than their L5-to-L5 counterparts, as indicated by RT: 3.10.5 versus 2.00.2 ms (p<0.05). Pairing APs and EPSPs (5 at 50 Hz, +10 ms interval) reliably evoked Hebbian LTP at proximal inputs (RT<3ms; after/before=1376%, n=36), but anti-Hebbian LTD was produced at distal inputs (RT>3ms; 836%, n=20; p<0.001). Distal LTP was rescued by EPSPs larger than those of unitary connections (>0.9mV; 1406%, n=14; p<0.001) and by dendritic depolarizing current injection (1608%, n=4; p<0.001), presumably because both conditions promote AP backpropagation. In agreement with this view, supralinear Ca2+ signals were generated throughout the distal dendritic tree by pairing APs with large EPSPs (supralinearity=10030%, n=37, p<0.001), or with dendritic depolarizing current injection (21060%, n=6, p<0.001), but not with small EPSPs (-28%, n=9; 0.370.2mV). The supralinear Ca2+ signal increased with distance, described by a power law ranging from 20% proximally, to 500% at 900 m from the soma. Interestingly, synaptic stimulation alone, but not APs alone, resulted in LTD (706%, n=10 vs 1026%, n=10; p<0.01). This form of LTD was abolished by the cannabinoid receptor CB1 blocker AM251 (0.9M; 993%, n=6; p<0.01). These results indicate that long-term plasticity rules in L5 pyramidal neurons depend on dendritic location, and distal plasticity can be switched between a Hebbian and an anti-Hebbian mode. This novel learning rule may help equalize the efficacy of distal inputs.


Supported by the Wellcome Trust, EU, JSPS and Gatsby Foundation.