Circuits, Structure and Behavior in Mouse Early Vision

Primary visual cortex (V1) in rodents, despite being a central and well-characterized research model for systems and computational neuroscience, still lacks consensus on a correspondence of its functional organization and underlying circuitry, as well as the circuit pathways that merge sensory and non-visual modulatory signals. This dissertation collectively develops experimental and theoretical arguments to refine both aspects. The first and second part of the thesis revisit the long hold view of 'salt-and-pepper' spatial organization of orientation tuning in mouse V1. Using large-scale two-photon imaging we observe a fine spatial scale functional organization in both L2/3 and L4 of mouse V1: neurons within ~20μm exhibit remarkably strongly correlated orientation tuning, yet drops to random, uncorrelated levels beyond this range. We term this pattern 'micro-clusters'. To explain how 'micro-clusters' are inherited and stabilized in L2/3 despite underlying broad lateral connectivity, we develop a spatially extended excitation–inhibition (E/I) recurrent circuit model, and derive two qualitative and testable predictions: extra strong components of connection strength over a comparable fine spatial scale; balanced E/I condition over macro-spatial scales yet imbalanced condition over micro-spatial scales. Furthermore, these predictions are validated by various cell-type specific circuit mapping experiments. We next explore the developmental origin of 'micro-cluster' through thalamo-cortical inputs in L4. We employ a Hebbian/Oja rule learning model for LGN to L4 synaptic connections and via an analytical asymptotic analysis we show that L4 recurrent interactions primarily sets the fine spatial scale of clustered functional organization in L4. The third part of the thesis reassess the sources of non-visual input in mouse V1 and LGN via modulation by free, uninstructed body movement behavior in mouse V1. We collect calcium imaging data of both thalamo-cortical boutons (as feedforward inputs to V1) and V1 cortical neurons with simultaneous measurements of spontaneous facial movements. Using rigorous statistical procedures that control for spurious 'nonsense' correlations from slow trending neurophysiological and behavioral data, we reveal a robust movement behavioral modulation in visual stimulated feedforward pathway rather than exclusively from top-down pathway, which is driven by both eye movement and non-ocular movement features. Our analysis suggests that V1 cortical neurons exhibit a robust integration of non-visual signals from a top-down pathway that is active across visually and non-visually stimulated periods with a stimulus-gated bottom-up dLGN pathway. Population analyses further show that movement- and stimulus-related information occupy intertwined subspaces of V1 population activity, while the stimulus information can be properly and perfectly decoded without interference of behavioral variables. Altogether, these results reconcile the interplay between fine-scale functional organization and underlying circuitry within different theoretical frameworks, and establish the potential circuit pathways that merge sensory and free behavior information, re-framing mouse V1 as a context-aware, fine-grained network.