SUBCELLULAR MECHANISMS OF EXPERIENCE-DEPENDENT MODULATION IN PYRAMIDAL NEURONS DURING MULTISENSORY INTEGRATION

Pyramidal neurons, the principal excitatory cells in the cortex, receive synaptic inputs from multiple sources distributed across their dendritic arbors. This spatial segregation is accompanied by active dendritic mechanisms, including local, experience-dependent plasticity that selectively modulates specific input pathways. Such subcellular plasticity offers a mechanism through which learning reshapes sensory representations without global changes in neuronal responsiveness. Input-specific processing is particularly relevant for multisensory integration, where inputs from distinct sensory modalities must be dynamically weighted according to behavioral context to support learned behaviors. Despite evidence of learning-dependent dendritic plasticity, the role of subcellular changes in multisensory integration within associative cortical areas remains poorly understood. One promising candidate for such integration is the rostrolateral area (RL), a higher-order parietal region that receives convergent visual and somatosensory input. To investigate this, we employed longitudinal two-photon calcium imaging with sparse neuronal labeling to monitor dendritic and somatic activity in L2/3 pyramidal neurons in both primary visual cortex (V1) and RL of awake, behaving mice. Using a visuotactile detection task, we found no amplification of dendritic or somatic responses to visual, tactile, or combined visuotactile stimuli during learning. Instead, learning was associated with a redistribution of stimulus responsiveness, with a larger fraction of regions of interest (ROIs) responding preferentially to tactile and visuotactile stimulation, indicating recruitment of task-relevant activity. Additionally, some ROIs demonstrated multisensory enhancement and increased lateralized selectivity, reflecting experience-dependent modulation of convergent inputs.Collectively, these results identify a subcellular mechanism by which learning reshapes sensory representations through selective reweighting of convergent inputs.