The neuronal mechanisms of parametric working memory – the short-term storage of graded stimuli to guide behavior – are not fully elucidated. We have designed a working memory task where rats compare two sequential vibrations, S1 and S2, delivered to their whiskers (Fassihi et al, 2014). Vibrations are a series of velocities sampled from a zero-mean normal distribution. Rats must judge which stimulus had greater velocity standard deviation, σ (e.g. σ1 > σ2 turn left, σ1 < σ2 turn right). A critical operation in this task is to hold S1 information in working memory for subsequent comparison. In an earlier work we uncovered this cognitive capacity in rats (Fassihi et al, 2014), an ability previously ascribed only to primates. Where in the brain is such a memory kept and what is the nature of its representation? To address these questions, we performed simultaneous multi-electrode recordings from barrel cortex – the entryway of whisker sensory information into neocortex – and prelimbic area of medial prefrontal cortex (mPFC) which is involved in higher order cognitive functioning in rodents. During the presentation of S1 and S2, a majority of neurons in barrel cortex encoded the ongoing stimulus by monotonically modulating their firing rate as a function of σ; i.e. 42% increased and 11% decreased their firing rate for progressively larger σ values. During the 2 second delay interval between the two stimuli, neuronal populations in barrel cortex kept a graded representation of S1 in their firing rate; 30% at early delay and 15% at the end. In mPFC, neurons expressed divers coding characteristics yet more than one-fourth of them varied their discharge rate according to the ongoing stimulus. Interestingly, a similar proportion carried the stimulus signal up to early parts of delay period. A smaller but considerable proportion (10%) kept the memory until the end of delay interval. We implemented novel information theoretic measures to quantify the stimulus and decision signals in neuronal responses in different stages of the task. By these measures, a decision signal was present in barrel cortex neurons during the S2 period and during the post stimulus delay, when the animal needed to postpone its action. Medial PFC units also represented animal choice, but later in the trial in comparison to barrel cortex. Decision signals started to build up in this area after the termination of S2. We implemented a regularized linear discriminant algorithm (RDA) to decode stimulus and decision signals in the population activity of barrel cortex and mPFC neurons. The RDA outperformed individual clusters and the standard linear discriminant analysis (LDA). The stimulus and animal’s decision could be extracted from population activity simply by linearly weighting the responses of neuronal clusters. The population signal was present even in epochs of trial where no single cluster was informative. We predicted that coherent oscillations between brain areas might optimize the flow of information within the networks engaged by this task. Therefore, we quantified the phase synchronization of local field potentials in barrel cortex and mPFC. The two signals were coherent at theta range during S1 and S2 and, interestingly, prior to S1. We interpret the pre-stimulus coherence as reflecting top-down preparatory and expectation mechanisms. We showed, for the first time to our knowledge, the neuronal correlates of parametric working memory in rodents. The existence of both positive and negative codes in barrel cortex, besides the representation of stimulus memory and decision signals suggests that multiple functions might be folded into single modules. The mPFC also appears to be part of parametric working memory and decision making network in rats.
|Autori interni:||Esmaeili, Vahid|
|Titolo:||Neuronal correlates of tactile working memory in rat barrel cortex and prefrontal cortex|
|Relatore:||Diamond, Mathew Ernest|
|Data di pubblicazione:||5-dic-2014|
|Appare nelle tipologie:||8.1 PhD thesis|