Neuronal selectivity for stimulus information determines prefrontal LFP gamma power regardless of task execution

Abstract

Local field potential (LFP) power in gamma frequency has been thought to be modulated by cognitive variables during task execution. We sought to understand how the properties of neurons and LFPs from same sites representing information before and after training in cognitive tasks. We therefore trained monkeys to perform working memory tasks and analyzed the neurons and LFPs in different prefrontal subdivisions. In agreement with previous studies, we found that sites containing neurons selective for stimuli held in memory exhibit increased gamma power during and after the presentation of stimuli compared to the baseline period. Unexpectedly, we also found that the elevation of gamma power above the baseline was diminished after training, compared to power observed prior to training in the same monkeys; despite an overall in firing rate observed after training. Learning to perform the task did alter some properties of LFP gamma power, it decreased at the time that subjects needed to make a judgment about whether two stimuli were same or not, and differential gamma power was observed for matching and nonmatching stimuli. Our results indicate that gamma power elevation in the prefrontal cortex represents automatic stimulus processing, not necessarily tied to a cognitive task being executed.

Brain-wide human oscillatory LFP activity during visual working memory

“…Beta oscillations are readily detectible in other extracellular field recordings (such as EEG or MEG) and are also a reliable marker of underlying cognitive processes impacting neural circuit interactions, just as gamma oscillations are (Uhlhaas and Singer, 2011; Roux and Uhlhaas, 2014), including working memory and top-down control (Siegel et al, 2012; Helfrich and Knight, 2016). Consistent with the aforementioned animal studies (Singh et al, 2022; Singh et al, 2023), in our study, decrement of beta oscillations was detected during the task execution, and differed systematically between areas, at least around the time of stimulus presentations and early in the delay period of the task.…”

“…During presentation of stimuli, correlated bottom-up inputs can serve to synchronize population neuronal spiking, and phases of synchronized excitation by pyramidal neurons followed by inhibition by interneurons can thus produce rhythmicity specifically in the gamma frequency range (Fries, 2009). Less precisely-timed or correlated inputs may fail to generate gamma oscillations, and indeed recent animal studies have suggested more prominent changes in the beta rather than gamma frequency range after learning to perform working memory tasks (Singh et al, 2022; Singh et al, 2023). Similarly, a recent human study of activation patterns in auditory working memory with intracranial recordings demonstrated that frontal and temporal regions with high decoding accuracy were not accompanied by significant increases in gamma power (Uluç et al, 2023).…”

“…We then constructed a 1-way or 2-way ANOVA representing the task epoch and region. In every case, the analysis was performed for the beta and high-gamma frequency band, defined as 16-40 Hz and 100-150 Hz, respectively, based on prior studies of working memory (Lundqvist et al, 2016; Haller et al, 2018; Singh et al, 2023). Post-hoc pairwise comparisons on any significant main effects were performed with Tukey’s method.…”

“…Neurophysiological experiments in non-human primates have identified neuronal activation of the prefrontal cortex as central for the maintenance of working memory (Riley and Constantinidis, 2016). Persistent activity generated by the spiking of prefrontal neurons is also associated with distinct patterns of spectral power, identifiable in local field potential recordings (Singh et al, 2022; Singh et al, 2023).…”