Distributed engrams enable parallelized orthogonal computations within and across brain regions

Authors: Douglas Feitosa Tomé, Chenchen Shen, Basile Confavreux, Madhumathan Mukherjee, Ying Zhang, Dheeraj S. Roy, Tim P. Vogels

Presentation type: Poster at SNUFA 2025 online workshop (5-6 Nov 2025)

Abstract

Our ability to form and retrieve memories relies on a network of connected brain regions. In particular, sparse neuronal ensembles across regions form distributed memory engrams that have structural and functional connectivity. However, the computations supported by distributed engrams remain unclear. Here, we demonstrate that distributed engrams enable memory generalization and discrimination in parallel within and across regions. Orthogonal and complementary, generalization and discrimination are computations that must be balanced for adaptive, memory-guided behavior. For instance, while animals need to generalize threat-predictive cues to novel stimuli with shared features, they also need to discriminate between threat-predictive and neutral cues. By modeling multi-region spiking neural networks with state-dependent and region-specific synaptic plasticity, our model predicted that neural representations in different regions either align or rotate over time with correspondingly small or large representational drift. In addition, our model predicted that, within individual regions, distinct engrams generalize and discriminate in parallel by bringing the representations of training and novel stimuli either closer together or further apart, respectively. Critically, our model found a relationship between generalizing and discriminating engrams that indicated that both can predict memory-guided behavior. Using in vivo longitudinal calcium imaging in mouse hippocampus, thalamus, and cortex, we conducted associative learning experiments that supported our model’s predictions. We also found computational evidence that generalizing and discriminating engrams remain (nearly) non-overlapping and that distinct forms of co-active synaptic plasticity can account for the neural representations and computations observed in our experiments. Our results revealed that the distributed organization of memory enables parallelized orthogonal computations throughout the brain — a potentially fundamental computational principle of distributed neural representations across perception, cognition, and behavior.