Prefrontal coding of naturalistic working memory: Mechanisms during normal maintenance and modelled disease
Abstract
Neural activity in the primate lateral prefrontal cortex (LPFC) has been causally linked to working memory (WM) — the brief maintenance and mental manipulation of information. Primates use WM to perform tasks in complex contexts; however, neural mechanisms of WM and the pathophysiology related to WM deficits have traditionally been studied using simple tasks that deviate from naturalistic conditions. This raises the question, how is WM processed in naturalistic conditions? To explore this, I trained two macaque monkeys on a spatial WM task set in a naturalistic virtual environment. During the task, a target was presented in 1 of 9 locations in a virtual arena. The target then disappeared and following a 2-second delay period, subjects navigated to the cued target location using a joystick. I recorded single neuron activity using two 96-channel Utah Arrays implanted in LPFC (areas 9/46 & 8a). During this task, single neurons are spatially selective for remembered target locations and neural populations contain large amounts of information about the target location over the duration of the task. Neural coding for WM is robust and distinct from signals related to perception and eye movement. Using ketamine to model the pathophysiology of schizophrenia, I demonstrate drastic deficits in WM performance. The decrease in performance is related to differential effects on putative excitatory and inhibitory neurons. Inhibitory neurons decrease their firing rate for their preferred location after ketamine injection, thus disinhibiting excitatory cells, resulting in distorted WM representations. Finally, I demonstrate a new neural code for maintaining WM in naturalistic conditions. Precise temporal patterns of population activity contain large amounts of information about target location and target trajectories in our virtual task. Ketamine distorts these neural sequences and their relationship to behavior. Together, these findings demonstrate that the LPFC relies on a robust neural code that uses firing rate and temporal information to maintain WM representations in the presence of incoming sensory signals and eye movement. These findings detail how the primate LPFC encodes WM representations in naturalistic conditions and models how WM deficits may arise in day-to-day life.