An Exploration of Neural Heterogeneity and its Consequences on Network Dynamics and Neuromodulation

Abstract

Heterogeneity is a pervasive and ubiquitous aspect of physical and biological systems. This is especially true in the brain, where heterogeneity exists across all scales, from genetics and ion channel distribution to cell types and patterns of neuron connectivity. This heterogeneity has been linked to stable, persistent behaviour, increased information transmission, and effective neural encoding. Importantly, although neural heterogeneity is often regarded as a static property and modelled as a fixed meta-parameter, this is not the case in the brain. Rather, heterogeneity is, in many instances, a dynamic property of the brain that alters in response to persistent activity and brain state which has been associated with increased stability and robustness. However, the importance of specific manifestations of heterogeneity, in static or dynamic forms, is less clear. That is, recent research has shown that, despite many decades of research exploring animal-to-animal and cell-to-cell heterogeneity, it may have no bearing on the net electrical output of the cells and that neurons possess considerable electrophysiological overlap despite their heterogeneous composition. In this thesis, we explore heterogeneity in both its malleable and static forms. We first propose a model for a firing rate-based homeostatic adaptation of neural excitability (e.g. gain). Endowing a simple network model of excitatory neurons with this candidate mechanism and heterogeneous excitability. Measuring the net heterogeneity as a function of the average pairwise difference between neurons, we explore how this heterogeneity may be up- or down-regulated in response to external stimulation and network-related factors such as degree of connectivity, and presynaptic firing rate and weights. Maintaining heterogeneity in neural excitability has been experimentally and computationally demonstrated to be crucial to having persistently stable network dynamics. Notably, these observations have been made in very generalized frameworks, where the heterogeneity in neural excitability is assumed to arise from similar distributions for all cell types. However, highly detailed datasets of human cortical neurons have demonstrated this to be inaccurate. Rather, subtypes of cells have characteristic groupings of responsiveness and input sensitivity (e.g. excitability). Leveraging one such dataset made freely available by the Allen Institute for Brain Science, we investigate the effects of distinct excitabilities between the excitatory and inhibitory neurons on network dynamics. Beyond excitability, we consider the effects of morphological (e.g. neuron architecture) heterogeneity on neurons’ responses to external stimulation. Using biophysically accurate, freely available models of mouse primary visual cortex neurons, we do a detailed characterization of their whole cell (e.g. length, volume, etc.) and compartmental (e.g. branch thickness and length, etc.) traits. Simulating the neuron models under a uniform electric field, we record the somatic membrane potential response and link it to the strength of the applied field to define the susceptibility of each neuron to simulation. Thereafter, correlation and partial correlation analyses are performed to explore the importance of morphological traits to neuron susceptibility. The null results of these efforts then suggest that the high degree of heterogeneity in neuron morphology may limit the selectivity with which they can be targeted, and support the notion that neurons demonstrate high electrophysiological overlap.