Integration of sensory information in layer 2/3 of barrel cortex - a modelling approach

Neurons are the basic cellular units of the brain and are connected via synapses to form neural networks. One of the central questions in neuroscience is how particular tasks or "computations", are implemented by neural networks to generate behavior and how patterns of activity are stored during learning. In our work, we focus on mouse somatosensory cortex (known as barrel cortex), which is known to be associated to the mouse vibrissae system and helps rodents to sense the environment by discriminating between different textures and identifying objects. Object identification requires the integration of sensory information from multiple whiskers. However, where and how single whisker information is integrated to represent the palpated object is unclear. It is known that information from a single whisker enters cortex via layer 4 of the corresponding barrel column. Within the barrel column, whisker-specific activity is transmitted to more superficial layers like the layer 2/3. In contrast to layer 4, neurons in layer 2/3 of barrel cortex exhibit very low stimulus-evoked firing rates and low whisker selectivity. In addition, layer 2/3 is of particular interest for investigating temporal and spatial integration due to the aberrant lateral connectivity between the different barrel column. As part of my PhD project I want to combine two-photon calcium imaging of layer 2/3 and layer 4 of barrel cortex and network modelling to understand the integration of multi-whisker information in barrel cortex. To begin with, we are building an anatomically-realistic, in silico neuronal network model of layer 2/3 and layer 4 barrel cortex to simulate information processing with an emphasis on multi-barrel integration. The parameters used for the model are based on the available literature and we plan to compare results from these theoretical simulations with experimental data from calcium imaging experiments in layer 2/3 of mice performing a texture discrimination task, allowing us to test and refine the model further. Specifically, two-photon calcium imaging will allow us to capture the activity of a large population of layer 2/3 neurons and extract task-specific activity patterns. After fitting the model to the data we will have a tool to explore the network architecture and its critical components, as well as tractable coding schemes that layer 2/3 could be operating in. Furthermore, the model will provide predictions with respect to how perturbations of layer 2/3 barrel cortex could affect information coding and corresponding behaviours that we can test experimentally. These results will lead to new insights into the configuration of cortical circuits and the nature of cortical codes.