Cortical activity arises from the interplay of recurrently connected excitatory (E) and inhibitory (I) neurons. While theoretical models of E-I interactions have existed for many years (e.g., Wilson and Cowan, 1972), quantitative experimental analysis of E-I dynamics has become possible only recently, thanks to developments in population recording and optogenetics. To study and manipulate E-I dynamics in a quantitative manner, we combined electrophysiological recording using multisite silicon probes with optogenetic stimulation of pyramidal and parvalbumin-expressing (Pvalb) neurons in primary visual cortex (V1) of quietly awake mice.
To separately stimulate excitatory and inhibitory populations we used multi-wavelength optogenetics. We crossed the Thy18 mouse line (which expresses ChR2 in a subpopulation of excitatory neurons) with a Pvalb-Cre driver line, and we expressed the red-shifted opsin C1V1 in Pvalb neurons by injecting a conditional virus. Brief blue (445 nm) and green (561 nm) laser pulses were delivered to stimulate the excitatory and Pvalb populations.
We summarized the network activity at each moment by two population rates, representing the total firing rates of all recorded wide-spiking (presumed to be mainly excitatory) and narrow-spiking (putative Pvalb) neurons. We derived these population rates directly from the recorded data via a novel analysis technique that did not require spike sorting: for each spike, we used locality-sensitive hashing to estimate a smooth waveform whose width could then be measured accurately.
A single blue-light pulse caused an initial brief activation in the excitatory population, which in turn excited the putative Pvalb population, followed by prolonged suppression of both populations. Responsiveness to a second pulse was suppressed to various degrees depending on the interpulse interval. Similarly, a single green-light pulse reliably increased the population activity of putative Pvalb neurons, followed by a less noticeable quiet period. We are currently developing a dynamical system model to predict single-trial responses as a function of spontaneous activity preceding the stimulation.
These initial results indicate that in V1 of quietly awake mice, repeated
delivery of single laser pulses or pairs of pulses at controlled intervals is
effective in probing the dynamics of the network. It may thus be possible to
predict the dynamics of excitatory and Pvalb neuronal populations
quantitatively on a trial-by-trial basis.