Mechanisms of distributed working memory in a large-scale network of macaque neocortex

Abstract

Neural activity underlying working memory is not a local phenomenon but distributed across multiple brain regions. To elucidate the circuit mechanism of such distributed activity, we developed an anatomically constrained computational model of large-scale macaque cortex. We found that mnemonic internal states may emerge from inter-areal reverberation, even in a regime where none of the isolated areas is capable of generating self-sustained activity. The mnemonic activity pattern along the cortical hierarchy indicates a transition in space, separating areas engaged in working memory and those which do not. A host of spatially distinct attractor states is found, potentially subserving various internal processes. The model yields testable predictions, including the idea of counterstream inhibitory bias, the role of prefrontal areas in controlling distributed attractors, and the resilience of distributed activity to lesions or inactivation. This work provides a theoretical framework for identifying large-scale brain mechanisms and computational principles of distributed cognitive processes.

A connectome-based model of conscious access in monkey cortex

“…To reproduce these latter experimental activity patterns, we found that feedback connections must largely target inhibitory neurons. This is consistent with recent reports that top-down attention has a net-inhibitory effect on sensory cortex (Huang et al 2019; Yoo et al 2021) and with previous models predicting the need for feedback inhibition to sustain distributed persistent activity (Mejias and Wang 2021). A parsimonious explanation of the contrasting data on early sensory activity is that many tasks can be performed by maintaining a representation of only higher-order aspects of the stimuli, and therefore sustained activity is not required in early sensory cortex.…”

“…Feedforward and feedback connections may also target different cell types. If both feedforward and feedback connections mainly target excitatory cells, models often fail to propagate activity, or display global network synchrony and instability (Joglekar et al 2018; Mejias and Wang 2021). This issue can be alleviated by assuming that feedback connections have a slight net bias towards targeting inhibitory cells in the so-called counterstream inhibition hypothesis (Mejias and Wang 2021).…”

“…If both feedforward and feedback connections mainly target excitatory cells, models often fail to propagate activity, or display global network synchrony and instability (Joglekar et al 2018; Mejias and Wang 2021). This issue can be alleviated by assuming that feedback connections have a slight net bias towards targeting inhibitory cells in the so-called counterstream inhibition hypothesis (Mejias and Wang 2021). Therefore, in our model, except when otherwise stated, feedforward connections principally target excitatory cells, and feedback connections have a bias towards inhibitory cells (Fig 1).…”

“…In order to keep the spontaneous activity level constant, the local NMDA current targeting I population is defined as followed (Mejias and Wang 2021) with and …”

“…In this way we estimated the spine counts on pyramidal cells in 24 of the 40 injected regions in the Lyon atlas. Based on the strong positive correlation between spine count and cortical hierarchy (r = 0.61, p = 0.001), and following previous work (Chaudhuri et al 2015; Froudist-Walsh et al 2021a; Mejias and Wang 2021), we inferred the spine count for the remaining regions based on the hierarchy using linear regression.…”