Population trajectory analysis reveals divergent state-space geometries across three cortical excitatory cell types

Understanding how cortical cell types differ in state-space geometry is central to linking circuit composition with population coding. Population neural trajectories provide a compact representation of high-dimensional activity, capturing both spatial configuration and temporal evolution. In the visual cortex, excitatory subtypes differ in laminar location, projection targets, and synaptic integration, yet most trajectory analyses pool mixed populations, obscuring subtype-specific contributions. Here, we compared trajectory geometries across three genetically defined excitatory subtypes— Cux2-CreERT2, Emx1-IRES-Cre, and Slc17a7-IRES2-Cre—in the mouse primary visual cortex, using large-scale two-photon calcium imaging data from the Allen Brain Observatory and five complementary structural metrics. Across all metrics, Cux2 neurons exhibited consistently smaller trajectory scales than the other two types, with medium effect sizes stable across time windows, stimulus orientations, and leave-one-out validation. Emx1 and Slc17a7 populations showed broadly overlapping profiles, with the largest differences emerging in the mid-response window, corresponding to the sensory integration phase of population dynamics. These findings reveal distinct geometric signatures imposed by excitatory subtypes: Cux2 circuits favor localized, stable representations, whereas Emx1 and Slc17a7 circuits support broader, distributed integration. This provides a framework for linking microcircuit composition to population-level dynamics under identical sensory conditions.