Dynamic imbalances in cell-type-specific striatal ensembles reflect learned coupling between trajectory representations and locomotor dynamics

Basal ganglia models commonly propose that relative imbalances between direct and indirect pathway output shapes movement, but how such imbalances are expressed during behavior remains unclear. We simultaneously imaged identified direct-pathway and indirect-pathway spiny projection neurons (dSPNs and iSPNs) in dorsal striatum as mice locomoted through virtual visual environments for reward. Individual dSPNs and iSPNs encoded discrete locations within specific visual environments and, in a distance-based task, encoded distance traveled or elapsed time, revealing structured representations of goal-directed trajectories. At the population level, both pathways were broadly co-active and similarly correlated with locomotor speed, but their relative activity shifted systematically across learned trajectories: dSPNs dominated during early accelerating segments, and iSPNs dominated during later slowing segments. These imbalances were selectively expressed within ensembles tuned to spatial location or distance/time, depending on task structure, but were absent during comparable spontaneous locomotion outside the task context and during initial exposure to a novel environment. A computational model demonstrated that opponent plasticity driven by kinematics-linked teaching signals can reproduce the observed task-dependent imbalances through cell-type-specific plasticity of discrete trajectory-related inputs and can progressively organize locomotor kinematics over learning. Our findings indicate that direct/indirect pathway imbalances are not a general reflection of motor output, but are dynamic, state-dependent features of striatal activity that link structured trajectory representations to associated changes in behavioral vigor along repeated, goal-directed locomotor paths through learning.