Cortical spike trains are highly irregular both during ongoing, spontaneous activity and when driven at high firing rates. There is uncertainty about the source of this irregularity, ranging from intrinsic noise sources in neurons to collective effects in large-scale cortical networks. Cortical interneurons display highly irregular spike times (coefficient of variation of the interspike intervals >1) in response to dc-current injection in vitro. This is in marked contrast to cortical pyramidal cells, which spike highly irregularly in vivo, but regularly in vitro. We show with in vitro recordings and computational models that this is due to the fast activation kinetics of interneuronal K + currents. This explanation holds over a wide parameter range and with Gaussian white, power-law, and OrnsteinUhlenbeck noise. The intrinsically irregular spiking of interneurons could contribute to the irregularity of the cortical network.inhibitory interneuron | bistability | neural noise | fluctuations | cortex W hen spike trains of cortical pyramidal neurons are recorded in vivo from the cortices of awake or sleeping cats (1-3) or awake monkeys (4), the interspike intervals (ISIs) are highly variable, with coefficients of variation (CV ISI = ISI SD/ISI mean) >1. However, when the main cortical excitatory (pyramidal) neurons are depolarized in vitro by injection of constant current to above firing threshold, their spike trains are substantially more regular (CV ISI <0.5). The irregularity of pyramidal neuron firing in vivo arises from the intense, ongoing, temporally correlated synaptic activity that bombards cortical neurons (1, 2, 5-7). The lower in vitro irregularity can be raised to in vivo levels by using fluctuating current injection (8), hyperosmolar solution (9), and a neuromodulatory mixture (10, 11).Computational models of irregular neuronal firing typically include network synaptic activity represented by stochastic aggregate processes. Fluctuating inputs have been represented by Brownian motion (12) or an Ornstein-Uhlenbeck process (8) in both single-and multicompartmental neuronal models (5). These are externally induced fluctuations and do not mechanistically explain the variety of irregular discharge patterns observed in cortical interneurons. In this study we combined in vitro recordings and computational models of cortical interneurons to show how neuronal conductances interact with the stochastic input to produce the variety of observed neuronal phenotypes.
ResultsRecordings from Cortical Inhibitory Interneurons. Cortical GABAergic interneurons have a range of functions including the gating (13) and entrainment (14) of neuronal firing, dendritic integration (15), synaptic plasticity (16), and the generation of network oscillations (10, 11). There are several morphologically and physiologically distinct classes of interneurons (17).In contrast to pyramidal neurons, a constant current injected into mouse visual cortex layer 2/3 interneurons in vitro frequently produced a spike train with a CV ISI that substant...