Robust visual feedforward and feedback signal processing in the mouse thalamus

Bauer, Yannik T. (2022): Robust visual feedforward and feedback signal processing in the mouse thalamus. Dissertation, LMU München: Graduate School of Systemic Neurosciences (GSN)

Abstract

Robust vision starts in the retina and is finally accomplished in the cortex – but what role does the dorsolateral geniculate nucleus of the thalamus (dLGN) play at the intermediate stage of the early visual processing pathway? In this thesis, I investigated how the dLGN in the awake mouse computes visual representations and how dLGN activity is shaped by retinal feedforward signals, cortico-thalamic feedback and behavioural state. A guiding hypothesis was that the dLGN is not a passive relay of retinal inputs, but an active signal transformer that may improve the reliability, efficiency, and robustness of the neural population code. In the first study included in this work, we investigated which functional retinal ganglion cell (RGC) types project to the dLGN and how multiple RGC types converge onto single dLGN relay cells. The second study explored the impact of global suppression of V1 cortico-thalamic feedback on dLGN responses to naturalistic stimuli, and compared the effects of feedback versus locomotion and natural versus artificial stimuli. Lastly, in the third study, we modelled dLGN activity to more complex movie stimuli and used a more selective optogenetic feedback suppression method and assessed if and how the model benefits from ad- ditional information about feedback, as well as locomotion and pupil size. To summarize our results, we first found that the majority of functional RGCs project to the dLGN, which displays a large response diversity, and that an average of five types converge onto a given relay cell, two of which exert the strongest functional impact. Secondly, global feedback suppression reduced dLGN firing rates and increased bursting, with stronger effects observed for naturalistic stimuli than artificial ones, and similar but independent effects of feedback versus locomotion. Lastly, the third study confirmed that dLGN mean firing rates are decreased by direct feedback suppression, and increased during periods of running and large pupil sizes. These observations are reflected in the model, whose predictions benefit mostly from additional feedback but not behavioural state information, but which nevertheless manages to extract dLGN spatio-temporal receptive fields (STRFs) for complex movies as well as artificial stimuli. In conclusion, in vivo mouse dLGN activity is shaped mostly by the influences of sparse functional retino-thalamic convergence, and is modulated to a lesser degree by cortico-thalamic feedback and behavioural state. This suggests that the dLGN is not a passive relay but instead actively transforms visual signals by combining its visual and extra-visual inputs, in agreement with the consensus view on the subject.

• Dokumententyp: Dissertationen (Dissertation, LMU München)
• Keywords: computational neuroscience, visual system, feedback, LGN
• Themengebiete: 500 Naturwissenschaften und Mathematik ; 500 Naturwissenschaften und Mathematik > 570 Biowissenschaften, Biologie
• Fakultäten: Graduate School of Systemic Neurosciences (GSN)
• Sprache der Hochschulschrift: Englisch
• Datum der mündlichen Prüfung: 24. Juni 2022
• 1. Bericht­erstatter:in: Busse, Laura