Rhythmic variations in prefrontal inter-neuronal correlations, their underlying mechanisms and their behavioral correlates

Abstract

20Functional neuronal correlations between pairs of neurons are thought to play an important role 21 in neuronal information processing and optimal neuronal computations during attention, 22 perception, decision-making and learning. These noise correlations are often assumed to be 23 stable in time. However, recent studies suggest that cognitive processes are rhythmic, this 24 rhythmicity accounting for variations in overt behavioral performance. Whether this 25 rhythmicity coincides with variations in shared noise variability is unknown. Here, we perform 26 simultaneous recordings from the macaque frontal eye fields, while animals are engaged in a 27 spatial memory task. We report that noise correlations in prefrontal cortex fluctuate 28 rhythmically in the high alpha (10-16Hz) and beta (20-30Hz) frequency ranges. Importantly, 29 these rhythmic modulations in shared neuronal variability account for dynamic changes in overt 30 behavioral performance. They also coincide with increased spike-LFP phase coupling in these 31 specific frequency ranges, the spatial profile of which vary between superficial and deep 32 cortical layers. Finally, we demonstrate, using an artificial neuronal model, that rhythmic 33 variations in noise correlation oscillations parsimoniously arise from long range (LFP) and local 34 spike-LFP phase coupling mechanisms. Thus a significant portion of noise correlation 35 fluctuations can be attributed to long-range global network rhythmicity. 36 37 65 rhythmic modulations in noise correlations account both for overt behavioral performance and 66 for layer specific modulations in spike-field phase coupling. Based on an artificial model, we 67 demonstrate that rhythmic variations in noise correlation oscillations parsimoniously arise from 68 long range (LFP) and local spike-LFP phase coupling mechanisms. 69 70 4 Results 71Neuronal recordings were performed in the prefrontal cortex, specifically in the frontal 72 eye field (FEF, figure 1A), a structure known to play a key role in covert spatial attention 28,33-73 35 . In each session, multi-unit activity (MUA) and local field potential (LFP) were recorded 74 bilaterally, while monkeys performed a memory guided saccade task ( figure 1B). Specifically, 75 monkeys were required to hold the position of a spatial cue in memory for 700 to 1900ms and 76 to perform a saccade towards the memorized spatial location on the extinction of the fixation 77 point that served as a go signal. In the following, noise correlations between the different 78 prefrontal signals of the same hemisphere were computed during the time interval running from 79 300ms to 1500ms following cue offset, on neuronal activities averaged over 200ms sliding 80 windows (step of 10ms). As shown by previous studies, noise correlations decrease as a 81 function of cortical distance ( Figure S1A, 1-way ANOVA, p<0.001, Wilcoxon rank sum test, 82 p<0.001 for 750 µm, p<0.001 for 1000 µm, 23,36,37 and are significantly lower among neuronal 83 pairs with different spatial selectivity than neu...

Interneuronal correlations dynamically adjust to task demands at multiple time-scales

“…Consistent with our previous study 24 , in all of the three tasks, behavioral performance, defined as the proportion of correct trials as compared to error trials, varied as a function of alpha and beta noise correlation oscillations. Indeed, on a session by session basis, we could identify an optimal alpha (10-16Hz) phase for which the behavioral performance was maximized, in antiphase with a bad alpha phase, for which the behavioral performance was lowest (figure 5C).…”

“…In figure 4 we show that noise correlations dynamically adjust to the probabilistic structure of the trial, and are lowest at the time of highest behavioral demand in the trial. In figure 5D-F, as well as in our independent study 24 , noise correlations and SFC show a strong pattern of co-variation. We propose that noise correlations arise from specific changes in SFC coupling.…”

“…In an independent study 24 , we demonstrate that oscillations in noise correlations arise from specific phase coupling mechanisms between long-range incoming LFP signals and local spiking mechanisms, independently from phase-amplitude coupling mechanisms. Figure 5D represents SFC between spiking activity and LFP signals (see Materials and Methods) computed during a 1200ms time interval starting 300ms after either fixation onset (Fixation and Target detection task) or cue offset (Memory guided saccade task).…”

“…In an independent study 24 , we show that noise correlations in the prefrontal cortex fluctuate rhythmically in the high alpha (10-16Hz) and beta (20-30Hz) frequency ranges. This is reproduced here in three distinct tasks (figure 5A).…”

“…Several sources of noise correlations have been proposed, ranging from shared connectivity 16 , to global fluctuations in the excitability of cortical circuits 17, 18 , feedback signals 19 or internal areal dynamics 20–22 , or bottom-up peripheral sensory processing 23 . In an independent study 24 , we for example show that noise correlations in the prefrontal cortex fluctuate rhythmically in the high alpha (10-16Hz) and beta (20-30Hz) frequency ranges. We further show that these rhythmic fluctuations co-occur with changes in spike-LFP phase coupling and can be segregated into a long-range component, possibly reflecting global fluctuations in the excitability of the functional network of interest, as well as to local internal areal dynamics.…”