THALAMIC INHIBITION ALIGNS SENSORY CODING WITH LEARNING

Adaptive behavior depends on the brain learning to filter unimportant sensory inputs while optimizing the detection of behaviorally relevant cues, a function widely attributed to higher-order cortical networks. However, recent work shows that learning-related changes in computation already emerge in the sensory thalamus, raising the question of which circuit mechanisms enable response plasticity and relevance-based filtering at the level of sensory relays. Using deep brain two-photon imaging of the auditory thalamus (medial geniculate body, MGB) combined with circuit-specific neuronal manipulations in mice performing an audiovisual detection and discrimination task, we found that dynamic inhibitory input from the thalamic reticular nucleus (TRN) controls learning-related response plasticity in MGB. As animals learned stimulus-outcome associations, MGB neurons developed biased responses to reward-predicting cues, whereas silencing TRN input abolished this reward preference coding in MGB and impaired learned performance. Together, these findings identify TRN as a dynamic regulator of adaptive sensory coding at the level of sensory thalamus upon associative learning, establishing inhibition in MGB as a core computational mechanism that selectively suppresses non-relevant signals to align sensory representations with learned value and behavioral goals.