Pitsiani, Margarita; (2022) Computations in L4 Barrel Cortex During Texture Discrimination. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text pitsiani_thesis_for_submission.pdf - Accepted Version Download (16MB) | Preview

Abstract

A major question in neuroscience is how sensory signals are processed and integrated with ongoing activity in cortical areas. Specifically, it is unclear how local cortical network computations modulate sensory signals. Mouse barrel cortex is an area of primary somatosensory cortex that is known to encode whisker-related information. The first cortical layer to receive the information from the whiskers is the layer 4 (L4) in barrel cortex. L4 is topographically organised in barrels with every barrel receiving information from one whisker. It has been reported that the largest fraction of excitatory synaptic input to a given L4 cell originates from other L4 cells of the same barrel that are recurrently connected to it. The role of such recurrent activity in densely connected layers like L4 barrel cortex is unclear. Recurrent cortical activity may contribute to signal amplification, modulate signal correlations, or assist the neuronal network to retain working memory. The presence of a single primary source of external input, the thalamus, makes L4 barrel cortex a suitable network to study the combinatorial effect of recurrent and feedforward activity. In this work, we investigate the interplay of feedforward and recurrent activity by combining large-scale imaging and electrophysiological recordings with biologically inspired computational modelling. Firstly, we utilised two-photon calcium imaging of population activity in L4, which allowed monitoring of the activity of a large proportion of cells while a head-fixed mouse performed a texture discrimination task with two choices. This revealed that the majority of the L4 cells are tuned to at least one of the whisker kinematics with other behavioural elements such choice and trial type having very little contribution in neuronal firing. In addition, given that this work involved one of the first uses of two-photon imaging of L4 neurons in barrel cortex, we also performed a spatial analysis of the tuning patterns which uncovered that the neurons are not spatially organised, but they are orchestrated in a salt and pepper fashion. Next, we investigated the contribution of L4 recurrent activity to the tuning properties that were characterised with two-photon imaging. Specifically, we combined Neuropixels recordings with perturbation experiments by silencing a proportion of L4 excitatory cells to assess the effect of recurrent connections on the functional characteristics on individual neurons. Although L4 has been primarily characterised as an amplification layer, our experimental results suggested a more nuanced function of the network, as some cells exhibited increased activity after photo-inhibition. One possible explanation is that the L4 network could be organised to compensate for neuronal loss and, in turn, to prevent information loss. Lastly, computational modelling of the L4 network was attempted to study the role of recurrent activity beyond experimental limitations. For this purpose, a biologically inspired spiking network of a single barrel was initially built and fit to two-photon imaging data collected during free whisking. Although previous studies suggested that interneurons suppress baseline activity, our model showed that whisking information is required to capture neuronal modulation. However, it became apparent that a non-linear spiking network might not be the best approach to study network motifs. Therefore, in future a simpler non-spiking network is recommended for analysis on recurrent functionalities. Overall, this work investigates the computations that take place in L4 by combining large scale recordings during behaviour, perturbations experiments and computational analysis, providing new insights into L4 functionality that goes beyond the amplification consensus.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item