Dissecting multi-population interactions across cortical areas and layers

The functional organization of cerebral cortex comprises several motifs, including feedforward, feedback, and horizontal interactions across areas and layers. Modern recording techniques now allow us to record from many neurons across these populations. However, especially as we consider multiple (more than two) populations, new conceptual and statistical frameworks are needed to reveal the population-level nature of these organizational motifs and how they contribute to brain function. Here, we advance the use of the dimensionality reduction approach group factor analysis (GFA), alongside large-scale electrophysiological recordings, to dissect multi-dimensional interactions across visual cortical areas and layers. GFA automatically determines not only the number of latent dimensions across a collection of neuronal populations, but also which subset of populations each latent involves. We first validated GFA's ability to recover multi-population interactions in synthetic data and in spiking neural activity since, to our knowledge, this work is the first such application of GFA. GFA not only performed well on realistic-scale simulated neural activity, but also reproduced key results from a prior study of areas V1 and V2: that the two areas interact via a communication subspace. We then used GFA to study interactions across select laminar compartments of macaque visual areas V1, V2, and V3d, recorded simultaneously with multiple Neuropixels probes. We found that (1) Laminar compartments of V1 share activity patterns with each other that are distinct from those shared with V2, consistent with the presence of distinct inter-areal and inter-laminar signaling pathways; and (2) V1-V2 and V1-V3d share unique activity patterns, consistent with the presence of distinct signaling pathways from V1 to these two downstream areas. The presence of unique inter-areal activity patterns depended on retinotopic alignment. This work establishes a framework for dissecting the multi-population interactions that underlie nearly all sensory, cognitive, and motor functions.