Heterogeneity of functional cellular properties for neurons in mouse cerebral cortex

Biological systems are known to exhibit a high degree of heterogeneity in their constituent components and their organization, and neuronal systems are no exceptions. To understand the functional impact of this heterogeneity in the brain, network models need to consider how this is manifested at the level of cellular properties, and thus how to consider variability in reported experimental data. Many studies have pointed out the variability in neuronal density, structural organization or synaptic connectivity across different neuronal networks and populations. Similarly, neuronal physiological properties are known to greatly vary across neuronal populations. Yet, the characterization of electrophysiological diversity has mainly relied on descriptions of firing properties (e.g. bursting, spike frequency adaptation) with various quantitative definitions of the boundaries between neuronal classes (e.g. fast spiking, regular spiking). Furthermore, lab specific implementations of experimental design and data analysis are an obstacle for comparisons between studies. In this context, the quantitative consideration of neuronal variability across commonly accepted neuronal classes provides an objective approach to describe neuronal physiological heterogeneity. We analyzed several publicly available databases to characterize the variability of linear and input/output properties of cortical neurons, according to multiple factors covering the entire cortical neuronal population. We assessed the variability of the main cortical neuron types (Excitatory, PValb, Sst, Htr3a, Vip), revealing their heterogeneity as function of cortical area (primary visual, motor and somato-sensory areas), including between layers within a given area. Our comparative database study revealed that different experimental conditions (e.g., in-vitro vs. in-vivo, recording temperature) can influence the properties of any given cell type, while preserving overall differences between types. We find that considering the input to a given neuron in terms of the effective voltage response of a linear model can account for some of the heterogeneity of I/O properties, and suggest that these properties are directly linked to cell input resistance, thus cell size. This works constitute a strong foundation for the consideration of detailed neuronal electrophysiological heterogeneity in future large-scale modeling works.