Figure 6. Illustration of a potential neuronal mechanism of repetition-induced gamma changes.

(a) Tuning curves of four example excitatory neurons (colored lines) and the drive they receive (colored arrows) for a stimulus of a given orientation (shown above the panel). (b) Local average excitatory inputs (black solid curve) and inhibitory inputs (gray dashed curve), adapted from Figure 6D of Salkoff et al., 2015, during a gamma cycle. Inhibitory inputs systematically lag excitatory inputs by a few milliseconds. Colored vertical lines indicate mean spike times of the four example neurons, color-coded according to (a). Their spike latencies during the gamma cycle are determined by their stimulus drive (Vinck et al., 2010). (c) A Hebbian STDP kernel, aligned to the average time of inhibitory spiking. The relative E-I spike timing between strongly driven/weakly driven excitatory neurons and the inhibitory neuron pool induces E-to-I synaptic weight increases/decreases, respectively. Note that the spike times shown in (b) are illustrations of the mean spike times of the respective neurons during the gamma cycle, whereas experimentally observed spike time distributions show substantial cycle-by-cycle variability. Thereby, for the two neurons with the strongest (yellow) and weakest (blue) drive, spike times occur almost exclusively during the positive or negative part of the STDP kernel, respectively. For the neuron with the second-strongest drive (red), spike times mostly overlap with the positive part, yet also partly with the negative part of the STDP kernel, and the reverse holds for the neuron with the second-weakest drive (purple). (d) The proposed mechanism should result in a modified E-I dynamic: Strengthened synaptic weights from strongly driven excitatory neurons to the inhibitory neuron pool accelerate the excitation-driven inhibition, thereby shortening the gamma cycle and increasing MUA-LFP gamma locking. (e) The proposed mechanism strengthens synaptic weights from strongly driven excitatory neurons to the local inhibitory interneuron pool. Furthermore, it strengthens inhibitory synaptic weights from the local inhibitory interneuron pool to the more weakly driven excitatory neurons.