Functional specialisation of multisensory temporal integration in the mouse superior colliculus

Abstract

Our perception of the world depends on the brains ability to integrate information from multiple senses, with temporal disparities providing a critical cue for binding or segregating cross-modal signals1,2. The superior colliculus (SC) is a key site for integrating sensory modalities, but how cellular and network mechanisms in distinct anatomical regions within the SC contribute to multisensory integration remains poorly understood. Here, we recorded responses from over 5,000 neurons across the SCs anatomical axes of awake mice during presentations of spatially coincident audiovisual stimuli with varying temporal asynchronies. Our findings revealed that multisensory neurons reliably encoded audiovisual delays and exhibited nonlinear summation of auditory and visual inputs, with nonlinearities being more pronounced when visual stimuli preceded auditory stimuli, consistent with the natural statistics of light and sound propagation. Nonlinear summation was crucial for population-level decoding accuracy and precision of AV delay representation. Moreover, enhanced population decoding of audiovisual delays in the posterior-medial SC, facilitated temporal discriminability in the peripheral visual field. Cross-correlation analysis indicated higher connectivity in the medial SC and functional specific recurrent connectivity, with visual, auditory, and multisensory neurons preferentially connecting to other neurons of the same functional subclass, and multisensory neurons receiving approximately 50 percent of the total local input from other multisensory neurons. Our results highlight the interplay between single-neuron computations, network connectivity, and population coding in the SC, where nonlinear integration, distributed representations and regional functional specialisations enables robust sensory binding and supports the accurate encoding of temporal multisensory information. Our study provides new insights into how the brain leverages both single-neuron and network-level mechanisms to represent sensory features by adapting to the statistics of the natural world.