Neuromodulatory effects on early visual signal processing

Abstract

Understanding how the brain processes information and generates simple to complex behavior constitutes one of the core objectives in systems neuroscience. However, when studying different neural circuits, their dynamics and interactions researchers often assume fixed connectivity, overlooking a crucial factor - the effect of neuromodulators. Neuromodulators can modulate circuit activity depending on several aspects, such as different brain states or sensory contexts. Therefore, considering the modulatory effects of neuromodulators on the functionality of neural circuits is an indispensable step towards a more complete picture of the brain’s ability to process information. Generally, this issue affects all neural systems; hence this thesis tries to address this with an experimental and computational approach to resolve neuromodulatory effects on cell type-level in a well-define system, the mouse retina. In the first study, we established and applied a machine-learning-based classification algorithm to identify individual functional retinal ganglion cell types, which enabled detailed cell type-resolved analyses. We applied the classifier to newly acquired data of light-evoked retinal ganglion cell responses and successfully identified their functional types. Here, the cell type-resolved analysis revealed that a particular principle of efficient coding applies to all types in a similar way. In a second study, we focused on the issue of inter-experimental variability that can occur during the process of pooling datasets. As a result, further downstream analyses may be complicated by the subtle variations between the individual datasets. To tackle this, we proposed a theoretical framework based on an adversarial autoencoder with the objective to remove inter-experimental variability from the pooled dataset, while preserving the underlying biological signal of interest. In the last study of this thesis, we investigated the functional effects of the neuromodulator nitric oxide on the retinal output signal. To this end, we used our previously developed retinal ganglion cell type classifier to unravel type-specific effects and established a paired recording protocol to account for type-specific time-dependent effects. We found that certain retinal ganglion cell types showed adaptational type-specific changes and that nitric oxide had a distinct modulation of a particular group of retinal ganglion cells. In summary, I first present several experimental and computational methods that allow to study functional neuromodulatory effects on the retinal output signal in a cell type-resolved manner and, second, use these tools to demonstrate their feasibility to study the neuromodulator nitric oxide.