A single spiking neuronal model to account for the diverse spontaneous firing patterns of lateral habenula neurons

It is hypothesised that the lateral habenula (LHb) plays an important role in reinforcement learning and mood regulation. Experiments have shown that the LHb is excited by aversive events, is causally involved in avoidance learning, and its dysfunctions are implicated in clinical depression. The LHb receives inputs from the limbic system and the basal ganglia, and projects to the major dopaminergic and serotonergic nuclei. Previously, we have explored the morphological and electrophysiological properties of LHb neurons in rat brain slices. The cells were categorised into neurogliaform, spherical, fusiform, polymorphic and vertical cells. Observed spontaneous firing patterns fell into 4 categories: silent, tonic regular, tonic irregular and rhythmic bursting (SIL, TR, TIR and BST). An interesting finding is that, with the exception of neurogliaform cells, the membrane properties of the neurons are very similar, and morphology does not correlate with the firing pattern. Importantly, the injection of small currents can change the pattern from SIL to TR to BST. Our current hypothesis is that these neurons are functionally similar, with slight variability resulting in heterogeneous spiking behaviour. In this work we explore this variability using computational modelling. We use the adaptive exponential integrate-and-fire neuronal model to replicate the firing patterns observed in LHb neurons. This model has been shown to be able to replicate a wide range of realistic spiking patterns, while remaining simple and very efficient. To identify parameters corresponding to the observed diversity of spiking patterns, a systematic computational approach is used. For each potential combination of parameter values, the model is tested and features such as frequency or number of spikes per burst are extracted and saved. Finally, a sliding window approach is used to identify a small region of the feature map containing the desired proportions of each of the observed pattern. The model demonstrates all behaviours so far observed in the LHb with very similar neuronal properties. In addition, small input variations can result in the LHb neurons crossing between regions with different behaviours, resulting in large output variations. This hints at neurocomputational properties similar to a switching mechanism. In further work, this single cell model will be expanded into a biologically inspired model of the LHb and its relations to other brain structures.In summary, our work suggests that LHb neurons may operate in the vicinity of multiple behavioural regimes. Our computational search algorithm can also be extended to solve other computational neuroscience problems.