A Simplified Serotonin Neuron Model

Abstract

The serotonin (5HT) neurons of the dorsal raphe nucleus (DRN) play an impor- tant and nuanced role in regulating animal behaviour. They exhibit heterogenous and dynamic responses to rewards and punishments in vivo, and perturbations of their ac- tivity modulate diverse behavioural states. This functional complexity is reflected in the network architecture of the DRN, with its multiple cell types, local feed-forward and recurrent connections, and partially segregated input-output streams that span most of the forebrain. At the centre of this elaborate circuit, 5HT neurons them- selves are now believed to be highly electrophysiologically heterogenous. As a first step towards leveraging these observations to better understand the role of the DRN in regulating behaviour, we set out to produce a phenomenological 5HT neuron model capable of bridging the gap between single-neuron dynamics and network processing. We found that a class of leaky integrate-and-fire (LIF)-derived models that accurately replicate the firing behaviour of a variety of cortical neuron types could not capture the behaviour of 5HT neurons. This is because, unlike cortical pyramidal neurons, 5HT neurons exhibit pronounced nonlinearities in their subthreshold dynamics near action potential threshold due to a pair of voltage-dependent potassium currents operating on distinct timescales. Augmenting the LIF-derived model with both potassium currents resulted in a significantly improved description of 5HT neuron firing dynamics. Addi- tionally, we report that the distributions of the biophysical parameters that describe these potassium currents and other fundamental properties of 5HT neurons suggest the existence of a single, highly variable underlying population of these cells, rather than multiple distinct serotonergic types. Our simplified 5HT neuron model opens the door to understanding how the essential biophysical features of these cells and their cell-to-cell variability shape population-level encoding of behavioural variables in the DRN.