Prefrontal stimulation prior to motor sequence learning alters multivoxel patterns in the striatum and the hippocampus

Abstract

Motor sequence learning (MSL) is supported by dynamical interactions between hippocampal and striatal networks that are thought to be orchestrated by the prefrontal cortex. In the present study, we tested whether individually-tailored theta-burst stimulation of the dorsolateral prefrontal cortex (DLPFC) prior to MSL, can modulate multivoxel response patterns in the stimulated cortical area, the hippocampus and the striatum. Response patterns were assessed with multivoxel correlation structure analyses of functional magnetic resonance imaging data acquired during task practice and during resting-state scans before and after learning/stimulation. Results revealed that, across stimulation conditions, MSL induced greater modulation of task-related DLPFC multivoxel patterns than random practice. A similar learning-related modulatory effect was observed on sensorimotor putamen patterns under inhibitory stimulation. Furthermore, MSL as well as inhibitory stimulation affected (posterior) hippocampal multivoxel patterns at post-intervention rest. Exploratory analyses showed that MSL-related brain patterns in the posterior hippocampus persisted into post-learning rest preferentially after inhibitory stimulation. These results collectively show that prefrontal stimulation can alter multivoxel brain patterns in deep brain regions that are critical for the MSL process. They also suggest that stimulation influenced early offline consolidation processes as evidenced by a stimulation-induced modulation of the reinstatement of task pattern into post-learning wakeful rest.

Prefrontal stimulation disrupts motor memory consolidation at the micro timescale

“…In contrast to our previous work showing a differential effect of iTBS vs. cTBS on task-related brain patterns and functional connectivity (Gann et al, 2021b(Gann et al, , 2021a, the present analyses did not show such stimulation-specific effects. It is worth noting though that -similar to the present results -our previous studies did not highlight any effect of stimulation condition on brain activity during initial training (Gann et al, 2021a).…”

“…Importantly, it remains uncertain whether altering these neural signatures with experimental interventions can effectively influence consolidation. In line with a series of hippocampal-targeted non-invasive brain stimulation studies in the declarative memory domain (e.g., (Wang et al, 2014; Kim et al, 2018; Tambini et al, 2018; Freedberg et al, 2019; Hermiller et al, 2019, 2020; Warren et al, 2019)), we recently showed that learning-related responses in the hippocampus were influenced by stimulation of the dorsolateral prefrontal cortex (DLPFC) applied prior to initial MSL (Gann et al, 2021b, 2021a). Specifically, inhibitory - as compared to facilitatory - thetaburst stimulation (TBS) of the DLPFC altered functional connectivity patterns in fronto-hippocampal networks over the course of learning.…”

“…An approach that has shown promise to study such replay using fMRI data is multivariate pattern analyses. Specifically, multivoxel correlation structure (MVCS) analyses allow the examination of the persistence of task patterns into subsequent rest; a process that is thought to reflect the reactivation of learning-related patterns during offline epochs (Tambini and Davachi, 2013; Gann et al, 2021b; King et al, 2021). Thus, in a second set of exploratory fMRI analyses, we used MVCS approaches to test whether stimulation influenced the persistence of task patterns into inter-practice rest intervals.…”

“…To do so, we computed the level of similarity (similarity index) of multi-voxel correlation structures (MVCS) between task practice and inter-practice rest periods in our two ROIs and in a control region. The analysis pipeline, implemented in Matlab, is described below and followed similar procedures as in our previous work (Gann et al, 2021b; King et al, 2021).…”

“…The resulting co-registration parameters were then applied to the realigned functional images. To optimize voxel pattern analyses, functional and anatomical data remained in subject-specific (i.e., native) space, and no spatial smoothing was applied to functional images (Tambini and Davachi, 2013; Gann et al, 2021b; King et al, 2021). Additional preprocessing of the time series was performed prior to running the MVCS analyses.…”