TMS over the pre-SMA enhances semantic cognition via remote network effects on task-based activity and connectivity

Abstract

Background: The continuous decline of executive abilities with age is mirrored by increased neural activity of domain-general networks during task processing. So far, it remains unclear how much domain-general networks contribute to domain-specific processes such as language when cognitive demands increase. The current neuroimaging study explored the potential of intermittent theta-burst stimulation (iTBS) over a domain-general hub to enhance executive and semantic processing in healthy middle-aged to older adults.

Methods: We implemented a cross-over within-subject study design with three task-based neuroimaging sessions per participant. Using an individualized stimulation approach, each participant received once effective and once sham iTBS over the pre-supplementary motor area (pre-SMA), a region of domain-general control. Subsequently, task-specific stimulation effects were assessed in functional MRI using a semantic and a non-verbal executive task with varying cognitive demand.

Results: Effective stimulation increased activity only during semantic processing in visual and dorsal attention networks. Further, iTBS induced increased seed-based connectivity in task-specific networks for semantic and executive conditions with high cognitive load but overall reduced whole-brain coupling between domain-general networks. Notably, stimulation-induced changes in activity and connectivity related differently to behavior: While stronger activity of the parietal dorsal attention network was linked to poorer semantic performance, its enhanced coupling with the pre-SMA was associated with more efficient semantic processing.

Conclusions: iTBS modulates networks in a task-dependent manner and generates effects at regions remote to the stimulation site. These neural changes are linked to more efficient semantic processing, which underlines the general potential of network stimulation approaches in cognitive aging.

Keywords: Aging; Functional connectivity; Language; Semantic cognition; TMS; fMRI.

Fig. 1

Experimental Design. (A) Participants completed three sessions: a baseline fMRI session and two iTBS + fMRI sessions with effective and sham stimulation. (B) Per fMRI session, two task runs were completed. Blocks of the semantic judgment and the tone judgment task were interspersed with rest blocks. (C) We simulated the average electrical field at our target site, the pre-SMA (visualized on the normalized cortical surface). (D) Example trials for the semantic and the tone judgment task are shown. Participants heard a short phrase or a sequence of two tones. At the offset of the auditory stimulus, a picture of an object or an arrow appeared. Participants indicated via button press whether auditory and visual stimuli matched. RMT: resting motor threshold, WPM: word-picture matching, FPM: feature-picture matching.

Fig. 2

Behavioral results. (A) Results for accuracy and reaction time for each condition at each session. Boxplots show median and 1.5 × interquartile ranges. Half-violin plots display distribution and dotted lines show changes of mean values across sessions. (B) Individual data for effect of stimulation sessions on accuracy and reaction time for each condition. Results of mixed-effects regression (predicted marginal effects) for (C) accuracy and (D) reaction time. Full results output of both models can be found in Table S4. Cong: congruent items, Incong: incongruent items.

Fig. 3

Univariate activation results for experimental conditions during the baseline session. (A) Results for feature-picture matching (FPM) contrasted with tone judgment, (B) results for word-picture matching (WPM) contrasted with tone judgment, and (C) contrasting both semantic conditions, WPM and FPM, with each other. Results are FWE-corrected at peak level p < 0.05 (t > 6.21) with a minimal cluster size k = 10 voxels.

Fig. 4

Effect of stimulation on brain activation. After effective stimulation, stronger activation was found for (A) feature-picture matching (FPM) compared with rest (implicit baseline) and tone judgment. We extracted percent signal change (PSC) for significant clusters and correlated the difference in PSC between effective and sham sessions with the difference in behavior between effective and sham sessions. (B) For Δ of accuracy of FPM, a negative correlation with the difference in PSC in the left superior parietal lobe (SPL) was detected. fMRI results are thresholded at p < 0.05 at peak level (t > 1.71) and FWE-corrected at p < 0.05 at cluster level.

Fig. 5

Seed-based and whole-brain functional connectivity. (A) Two seeds showed increased whole-brain coupling after effective relative to sham stimulation. Only the left ventral SPL seed showed stronger coupling for the task contrasts FPM > WPM and tone judgment > FPM. fMRI results are thresholded at p < 0.001 at peak level (t > 3.55) and FWE-corrected at p < 0.05 at cluster level. (B) Binary PPI activation maps plotted onto a seven-networks functional connectivity parcellation (Yeo et al., 2011). C) Parameter estimates extracted for significant clusters of functional connectivity results for each seed and respective contrast. Parameter estimates reflect beta values of each task condition against rest. (D) Chord diagram displays significant differences in functional coupling between effective and sham stimulation. Results for the task contrast FPM > WPM revealed reduced functional connectivity between domain-general networks after effective relative to sham stimulation. The color intensity and width of a connection indicate its correlational strength.

Fig. 6

Relationship between stimulation-induced changes in functional connectivity and behavior. Reduced coupling of pre-SMA and left ventral SPL after effective iTBS was associated with slower reaction times (RT) during the feature-picture matching (FPM) condition after effective relative to sham iTBS.

Fig. 7

iTBS over a hub of the domain-general multiple-demand network enhanced coupling with other cognitive control networks distal to the stimulation site. Stimulation-induced increases in activity of cognitive control networks were linked to poorer performance, while increased connectivity between task-relevant networks was associated with more efficient processing during semantic judgments.