Leveraging Spatial Transcriptomics and Machine Learning to Study Adult Neurogenesis Dynamics in the Mouse Brain

Neural stem cell (NSC) dynamics in the adult brain are tightly regulated within specialized niches, yet the molecular mechanisms governing their plasticity remain largely unexplored. This thesis investigates two key aspects of adult neurogenesis using cutting-edge spatial transcriptomics technology and advanced data analysis methods. The first project explores the transient nature of pregnancy-generated neurons in the maternal olfactory bulb (OB). These neurons, essential for maternal olfactory adaptation, disappear post-lactation through a mechanism which is yet to be understood. Using Visium spatial transcriptomics and spatial cell-type deconvolution techniques, we test the hypothesis that cellular senescence underlies their elimination, examining affected cell populations and potential interactions with glial cells. Previous immunostaining experiments revealed an increase in microglia activation in mothers, and the spatial transcriptomics analysis confirms significant microglia activation in ensheathing cells and the glomerular layer, external plexiform layer, and tufted cells layer. This suggests that microglia may be responsible for clearing senescent cells. Supporting this hypothesis, senescence scores indicate higher levels of senescence in the outer layers of the olfactory bulb, mirroring microglial activation. The second project compares the spatial transcriptomic landscapes of the two neurogenic niches of the adult brain. We leverage the novel Visium HD technology to study neural stem cell behavior at subcellular resolution. Using deep-learning-based cell segmentation and custom binning, we map the interaction of stem cells with their environment in situ to uncover shared and distinct molecular mechanisms regulating NSC maintenance and activation, with a level of resolution and scale not previously possible with MERFISH-based technologies. Our results show that cell segmentation improves the granularity and specificity of cluster annotation compared to the standard 8μm binned data. This approach allows us to spatially resolve transcriptomic differences, providing insights into potentially shared NSC pools or niche elements across the two main neurogenic regions of the brain.