Stimulation of distinct parietal locations differentiates frontal versus hippocampal network involvement in memory formation

Abstract

Adjacent regions of parietal cortex are thought to affiliate with distinct large-scale networks and thereby make different contributions to memory formation. We directly tested this putative functional segregation within parietal cortex by perturbing activity of anterior versus posterior parietal areas. We applied noninvasive theta-burst transcranial magnetic stimulation to these locations immediately before a semantic encoding task, and subsequently tested recollection memory. Consistent with previous findings, fMRI activity in left inferior frontal gyrus during semantic encoding correlated with subsequent high memory accuracy and strong subjective recollection. Stimulation of the posterior parietal cortex decoupled its network - the hippocampal-cortical network - from left inferior frontal gyrus. Furthermore, posterior parietal stimulation reduced highly accurate subjective recollection. Critically, both of these changes occurred relative to stimulation of the anterior parietal cortex. Stimulating anterior versus posterior parietal cortex therefore differentiated hippocampal network involvement in episodic memory. This provides direct evidence that distinct territories within close proximity of each other in parietal cortex make functionally distinct contributions to memory formation. Further, noninvasive stimulation has the spatial resolution required to differentially modulate the interaction of these adjacent parietal locations with distributed large-scale brain networks.

Keywords: Episodic memory; Parietal cortex; Semantic memory; Subjective recollection; Theta-burst stimulation.

Graphical abstract

Fig. 1

Experiment design. (A) Before the experimental stimulation sessions, subjects completed resting-state fMRI and structural MRI scans used to guide TMS delivery. fMRI connectivity analyses were used to define subject-specific left posterior and anterior parietal targets. Posterior parietal sites (purple) were defined based on subject-specific connectivity to left hippocampal seeds (indicated by arrow). Anterior parietal sites (orange) were defined based on subject-specific connectivity to left IFG seeds (indicated by arrow). We further used vertex stimulation as a control condition. (B) Each subject participated in three experimental sessions. Each session began with one of the three stimulation conditions, followed by a resting-state MRI scan and a word encoding phase. This portion of the task occurred during the typical expected duration of stimulation-related effects on neural function. Recognition memory for words and word locations studied during encoding were tested after a delay, such that retrieval likely occurred after stimulation effects had decayed. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 2

Hippocampal and frontal networks of interest. Spheres centered at the peaks of clusters robustly connected with the hippocampal seed are plotted left, and spheres centered at the peaks of clusters robustly connected with the IFG seed are plotted right, creating group-level hippocampal and frontal networks, respectively.

Fig. 3

Memory performance linked to item and source memory confidence. (A) Subjects performed a semantic categorization task (i.e., judging whether words were animals) while fMRI was recorded. Following a delay, a 2-stage memory test was administered whereby item recognition then location memory was tested. For the item recognition prompt, subjects attempted to discriminate studied words from novel foils while simultaneously rating confidence using a 4-point scale (confident/guess). After registering the recognition response, location memory was tested for each studied word (studied-left/studied-right) with simultaneous reporting of confidence (confident/guess). (B) Subjects were significantly more accurate in discriminating studied from novel words when they self-reported confident item memory. (C) Subjects were significantly more accurate when they reported they were confident making both item and source memory judgments. ***p < 0.001.

Fig. 4

Posterior parietal stimulation reduces the proportion of strong memory trials. Subjects had fewer strong explicit memory trials following posterior relative to anterior parietal stimulation. Colored dots indicate individual subjects; box-and-whisker plots include whiskers marking first and third quartiles. *p < 0.05.

Fig. 5

Posterior parietal stimulation reduces connectivity between the hippocampal network and task-relevant brain regions. Resting-state fMRI connectivity between the hippocampal network and task-relevant brain regions – (A) IFG and (B) posterior temporal cortex – was reduced following posterior parietal relative to anterior parietal or control stimulation. Lines are drawn between hippocampal network regions and the peaks of task-relevant IFG and temporal brain activity. Dots on bar plots indicate individual subjects; means are plotted alongside standard errors. **p < 0.01, *p < 0.05.

Fig. 6

Posterior parietal stimulation reduces connectivity between the hippocampal network and task-relevant brain regions. Hippocampal network fMRI connectivity with IFG and temporal-task relevant regions was consistently reduced following posterior parietal stimulation versus anterior parietal (left) and control (right) stimulation conditions. Dots indicate individual subjects, and correlations are plotted with regression lines including confidence intervals. *p < 0.05.

Fig. 7

Stimulation condition does not affect connectivity between the frontal network and task-relevant brain regions. Resting-state fMRI connectivity between the frontal network and task-relevant brain regions – (A) IFG and (B) posterior temporal cortex – did not differ between stimulation conditions. Lines are drawn between frontal network regions and the peaks of task-relevant IFG and temporal brain activity. Dots on bar plots indicate individual subjects; means are plotted alongside standard errors. **p < 0.01, *p < 0.05.