Frequency-specific noninvasive modulation of memory retrieval and its relationship with hippocampal network connectivity

Abstract

Episodic memory is thought to rely on interactions of the hippocampus with other regions of the distributed hippocampal-cortical network (HCN) via interregional activity synchrony in the theta frequency band. We sought to causally test this hypothesis using network-targeted transcranial magnetic stimulation. Healthy human participants completed four experimental sessions, each involving a different stimulation pattern delivered to the same individualized parietal cortex location of the HCN for all sessions. There were three active stimulation conditions, including continuous theta-burst stimulation, intermittent theta-burst stimulation, and beta-frequency (20-Hz) repetitive stimulation, and one sham condition. Resting-state fMRI and episodic memory testing were used to assess the impact of stimulation on hippocampal fMRI connectivity related to retrieval success. We hypothesized that theta-burst stimulation conditions would most strongly influence hippocampal-HCN fMRI connectivity and retrieval, given the hypothesized relevance of theta-band activity for HCN memory function. Continuous theta-burst stimulation improved item retrieval success relative to sham and relative to beta-frequency stimulation, whereas intermittent theta-burst stimulation led to numerical but nonsignificant item retrieval improvement. Mean hippocampal fMRI connectivity did not vary for any stimulation conditions, whereas individual differences in retrieval improvements due to continuous theta-burst stimulation were associated with corresponding increases in fMRI connectivity between the hippocampus and other HCN locations. No such memory-related connectivity effects were identified for the other stimulation conditions, indicating that only continuous theta-burst stimulation affected memory-related hippocampal-HCN connectivity. Furthermore, these effects were specific to the targeted HCN, with no significant memory-related fMRI connectivity effects for two distinct control brain networks. These findings support a causal role for fMRI connectivity of the hippocampus with the HCN in episodic memory retrieval and indicate that contributions of this network to retrieval are particularly sensitive to continuous theta-burst noninvasive stimulation.

Keywords: TMS; hippocampus; neuromodulation; recognition; theta-burst stimulation.

© 2018 Wiley Periodicals, Inc.

Figures

FIGURE 1

Experiment overview. (a) One parietal stimulation location was selected for each subject based on high baseline resting-state fMRI connectivity with the hippocampus. Hippocampal volumes of interest are shown in an overlap map in which colorization indicates the number of subjects having each voxel included in the target volume. (b) Individualized parietal-cortex stimulation locations are shown as a green spheres for each subject. (c) Approximate timeline an experimental session with the approximate duration of each phase and transition (exact values and ranges are provided in the text). A different TMS condition was used for each of four sessions, administered in counterbalanced order across subjects. All TMS conditions were delivered to the same individualized location of parietal cortex for each subject [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 2

Networks of interest. (a) The HCN (dark purple) was defined as regions with robust fMRI connectivity to a hippocampal target volume (light purple). (b) The DAN (dark green) was defined as regions with robust fMRI connectivity to a seed in left superior frontal gyrus (light green). (c) The V1N (dark blue) was defined as regions with robust fMRI connectivity to a seed in dorsal central V1 (light blue) [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 3

Stimulation effects on retrieval. (a) Word recognition memory performance (d’) for each stimulation condition (left) and the change in d’ values for each active stimulation condition relative to sham (right). Error bars indicate SEM. Statistical results are given in the text and P values are replicated here for convenience. (b) Changes in d’ values due to each active stimulation condition relative to sham shown for each subject. Bar plots are ordered by subject, such that each subject’s change relative to sham for the three stimulation conditions can be identified by looking at the same row across the plots

FIGURE 4

θ-Burst effects on the relationship between hippocampal-HCN connectivity and retrieval relative to sham. Relative to sham, θ-burst increased the association between hippocampal-HCN connectivity and d’ for (a) posterior cingulate and retrosplenial cortex, (b) left medial frontal gyrus, and (c) right medial frontal gyrus areas of the HCN. Slices are labeled with the corresponding Talairach (RAI) slice. (d) This effect is illustrated by plotting the θ-burst versus sham difference in connectivity with the hippocampus against the difference in d’ values for all subjects, averaged over all supra-threshold voxels. (e) The positive relationship of difference values shown in figure (d) was due to positive correlation of hippocampal connectivity with d’ following θ-burst and essentially zero correlation following sham, which are plotted separately averaged over all supra-threshold voxels. Note that the relationships shown in figure (d) and (e) are statistically redundant with the voxel-wise analysis results and are shown for illustrative purposes only, with only descriptive statistics provided [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 5

θ-Burst effects on the relationship between hippocampal-HCN connectivity and retrieval relative to β-freq. (a) Relative to β-freq, θ-burst increased the association between hippocampal-HCN connectivity and d’ for the posterior-cingulate/precuneus area of the HCN. (b) This effect is illustrated by plotting the θ-burst versus β-freq difference in connectivity with the hippocampus against the difference in d’ values for all subjects, averaged over all supra-threshold voxels. (c) The positive relationship of difference values shown in figure (b) was due to positive correlation of hippocampal connectivity with d’ following θ-burst and essentially zero correlation following β-freq, which are plotted separately averaged over all supra-threshold voxels. Note that the relationships shown in figure (b) and (c) are statistically redundant with the voxel-wise analysis results and are shown for illustrative purposes only, with only descriptive statistics provided [Color figure can be viewed at wileyonlinelibrary.com]