Dendritic mechanisms underlying multiplexing and feature selectivity in cortical layer 5 pyramidal neurons

The integration of bottom-up sensory features and top-down feedback is essential for learning and executing sensorimotor cognitive tasks. The primary somatosensory cortex (S1) plays a central role in this process by integrating bottom-up feedforward inputs in its deep layers and feedback inputs in superficial layers. Layer 5 pyramidal neurons (L5 PNs) are key to this integration, with their extensive dendritic arborizations spanning both deep and superficial layers, and axons projecting to various cortical and subcortical regions. This unique morphology and connectivity enable L5 PNs to broadcast integration outcomes to downstream cortical and subcortical regions. However, the biophysical mechanisms underlying this integration at the subcellular level in individual L5 PNs remain poorly understood. To explore this cellular computation, I utilized a combination of holographic two-photon uncaging, electrophysiology, in vivo and ex vivo two-photon Ca²⁺ imaging, and biophysical simulations. My findings revealed that feedforward sensory features are multiplexed to "barcode" the somatic action potentials through the interplay between dendritic Na⁺ and NMDAR nonlinearities and somatic active ion channels. This rate coding of somatic spike trains dynamically integrates with feedback arriving at the apical dendritic arbor in a subtype-specific manner, enabling distinct integration outcomes to be routed to different subcortical pathways. Preliminary in vivo results from animals performing a sensory discrimination task suggest that apical dendritic spikes - hallmarks of translaminar integration of bottom-up and top-down inputs - evolve as animals learn the reward associations. Collectively, these findings highlight the synaptic and dendritic mechanisms underlying sensory feature multiplexing in L5 PNs, and shed light on how this integration evolves dynamically during cognitive learning behaviors.