Neurophysiological signatures of default mode network dysfunction and cognitive decline in Alzheimer's disease

Abstract

Neural hyperexcitability and network dysfunction are neurophysiological hallmarks of Alzheimer's disease (AD) in animal studies, but their presence and clinical relevance in humans remain poorly understood. We introduce a perturbation-based approach combining transcranial magnetic stimulation and electroencephalography (TMS-EEG), alongside resting-state EEG (rsEEG), to investigate neurophysiological basis of default mode network (DMN) dysfunction in early AD. While rsEEG revealed global neural slowing and disrupted synchrony, these measures reflected widespread changes in brain neurophysiology without network-specific insights. In contrast, TMS-EEG identified network-specific local hyperexcitability in the parietal DMN and disrupted connectivity with frontal DMN regions, which uniquely predicted distinct cognitive impairments and mediated the link between structural brain integrity and cognition. Our findings provide critical insights into how network-specific neurophysiological disruptions contribute to AD-related cognitive dysfunction. Perturbation-based assessments hold promise as potential markers of early detection, disease progression, and target engagement for disease-modifying therapies aiming to restore abnormal neurophysiology in AD.

Fig. 1.. Experimental design, TMS targets, and overview of TMS-EEG measure extraction in source space.

(A) Schematic representation of the experimental design. (B) Individual TMS targets for AD (top) and HC (bottom) groups projected onto a common brain template for each stimulation site (M1, blue dots; IPL, red dots). (C) Individual E-field simulation maps (left) were thresholded at the highest % 0.01 to generate personalized local masks (right) on the cortical surface for each TMS site. These local masks were used to extract regional EEG activity and measure cortical excitability at the site of stimulation. (D) Canonical rsfMRI maps derived from group-averaged fMRI connectivity analysis (left) were projected onto individual cortical surfaces, and a DMN mask (middle) was used as a region of interest. The TMS-targeted DMN node (IPL) is highlighted within the squared region, with the blue dot representing the TMS site (middle). This region is expanded on the right to show overlap between the IPL-DMN (red) and the E-field–based local map (orange). The bottom panel shows distinct regions at frontal (blue), temporal (green), and precuneus (brown) sites within the DMN mask. These masks were used to measure network connectivity. (E) Representative example of TMS-EEG measures. Top panel shows an individual cortical surface model with the DMN mask (shaded red regions) and source-reconstructed EEG activity for IPL-TMS from a representative HC participant. Thresholded EEG activity (>70%) at 37 ms following TMS is localized over the stimulated region (Left-IPL), representing local cortical excitability (left), while EEG activity at 79 ms (right) is localized over the frontal DMN, representing network connectivity. The middle panel shows scalp TEPs used for source reconstruction. The bottom panel shows time series of normalized averaged EEG activity (in z scores) extracted from local masks for cortical excitability (orange) and nonstimulated DMN masks for network connectivity measures.

Fig. 2.. Increased early-local DMN excitability in AD.

(A) TEPs and their spatial-temporal dynamics from one representative HC (left) and one AD participant (right). Top panels show TEPs of all EEG channels. Middle and bottom panels show corresponding scalp topographies and source reconstructed activations for the selected peaks. (B) Average current density time series in the AD group showing site (IPL versus M1) and condition specificity (TMS versus Sham) of TEPs. (C) Averaged evoked responses following IPL-TMS (top left) and M1-TMS (bottom left) at the site of stimulation both for AD and HC groups. Solid colored lines show group-averaged current density time series (in z scores) extracted from individualized E-field–based local masks for IPL and M1. Shaded regions show interindividual response variation with SE of measurements. Green colored blocks at the bottom of each panel show significant time points between AD and HC groups that survived cluster permutations, while black colored blocks indicate significant time points that did not survive permutation tests. Violin plots on the right show average activity within 15 to 65 ms following IPL-TMS (top right) and M1 (bottom right) for both groups. White dots in violin plots represent median value and gray-colored dots represent individual responses. Colored horizontal lines and gray vertical bars represent grand average values and interquartile ranges, respectively. * in upper violin plot denotes statistical significance at P < 0.05 with corresponding effect size calculated using Cohens’ d. (D) Cortical maps for grand averaged current densities (in z scores) between 15 and 65 ms on MNI template following IPL-TMS (left) and M1-TMS (right) for AD (top) and HC (middle) groups. Bottom panels show statistical results of thresholded cluster–based permutation t tests (cluster P < 0.05) with hot colors indicating AD > HC and cold colors indicating AD < HC.

Fig. 3.. Increased temporal and reduced frontal DMN connectivity in AD.

(A) Temporal dynamics of TMS-evoked responses at temporal (green-colored brain regions on top left) and frontal DMN (dark blue–colored brain region on bottom left) following IPL stimulation. Solid colored lines show group-averaged current density time series (in z scores), while shaded regions showed variation with SE of measurements. Green-colored blocks at the bottom of each panel show significant cluster of time points between AD and HC groups. Violin plots on the right show total amount of evoked current densities between 60 and 85 ms (top) and 75 and 101 ms (bottom) following IPL-TMS for temporal and frontal DMN, respectively. Violin plots for 125 to 175 time window is provided in fig. S3. (B) Cortical maps of grand averaged current densities (in z scores) between 60 and 85 ms (left) and 75 and 101 ms (right) on a template brain model following IPL-TMS for AD (top) and HC (middle) groups. Bottom panels show statistical results of thresholded cluster–based permutation t tests (cluster P < 0.05) with hot colors indicating AD > HC and cold colors in indicating AD < HC. *P < 0.05, **P < 0.01.

Fig. 4.. Relationships between TMS-EEG measures of cortical excitability, network connectivity, and cognitive functions in AD participants.

(A) Scatter plots with regression lines showing bivariate correlations of cortical excitability (left) and network connectivity (right) with cognitive function. Color codes refer to different temporal windows with red showing early (15 to 65 ms) and blue showing late (75 to 101 ms) activations. Correlation coefficients (r) are provided for each regression line with asterisks indicating statistically significant correlations (P < 0.05). (B) Cortical maps showing vertex-wise correlations of early responses with ADAS-cog (top left) and RAVLT (bottom left) scores with hot colors indicating positive and cold colors indicating negative correlations. Statistically thresholded cortical maps (P < 0.05) on the right shows brain regions with significant correlations. (C) Cortical maps showing vertex-wise correlations of late responses with Digit-Span (top) and Fluency (bottom).

Fig. 5.. RsEEG Measures.

(A) Current density time series (top) were extracted from each vertex (n = 15,000) to compute power spectral density (PSD) from 1 to 50 Hz (middle). PSD exponent (red line in middle panel) is computed and removed from PSD (black line in middle panel) and relative power at canonical EEG frequency bands were calculated at each vertex on the individual cortical surface (bottom). SPR is calculated as the ratio of delta + theta power (cyan shaded region in middle panel) to the alpha + beta power (magenta shaded region in middle panel) for each vertex. (B) TMS coordinates (green circle in top panels) for Left-IPL were used to determine seed vertex on the individual cortical surface. The seed vertex is expanded over the cortical surface to define a seed region (blue-shaded region in the expanded top panel). PECs were computed between each vertex within the seed region (black dots within the blue-shaded region) and rest of the brain to generate individual connectivity matrices at delta-theta (2 to 7 Hz), alpha (8 to12 Hz), and beta (13 to 29 Hz) bands (middle with color-coded EEG time series). Connectivity matrices were averaged and mapped on individual cortical surface to estimate rsEEG connectivity of the IPL with the rest of the brain (bottom showing cortical connectivity maps).

Fig. 6.. Global slowing in resting-state neural oscillations in AD.

(A) Cortical maps showing vertex-wise distribution of spectral power in canonical frequency bands and (top left) and SPR (bottom left). Bottom panels show statistical results of thresholded cluster–based permutation t tests (cluster P < 0.05) with hot colors indicating AD < HC and cold colors in indicating AD > HC. (B) Violin plots showing SPR averaged across the entire cortical space (global), within individualized masks of IPL, M1, and frontal node of the DMN both in healthy (blue) and AD (red) participants. * in upper violin plot denotes statistical significance at P < 0.05 with corresponding effect size calculated using Cohens’ d.

Fig. 7.. Resting-state neural synchronization differences between AD and HC participants are not network specific.

(A) Violin plots showing alpha band neural synchrony averaged across the entire cortical space (global), within individualized masks of IPL, M1, and frontal node of the DMN both in healthy (blue) and AD (red) participants. * in top violin plot denotes statistical significance at P < 0.05 with corresponding effect size calculated using Cohens’ d. (B) Cortical maps showing vertex-wise distribution of neural synchrony in canonical frequency bands and (left). Right panels show statistical results of thresholded cluster–based permutation t tests (cluster P < 0.05) with hot colors indicating AD < HC and cold colors in indicating AD > HC.

Fig. 8.. Relationships between rsEEG measures of SPR, neural synchrony, and cognitive functions in AD participants.

(A) Scatter plots with regression lines showing bivariate correlations between rsEEG measures of SPR (left) and neural synchrony (right) with cognitive function in AD participants. Color codes refer to different cortical regions with red showing rsEEG measures from IPL mask and blue showing from the frontal node of the DMN and black showing the average measure computed across the entire cortex (global). Correlation coefficients (r) are provided for each regression line with asterisks indicating statistically significant correlations (P < 0.05). (B) Cortical maps showing vertex-wise correlations of SPR with ADAS-cog (top) and RAVLT (bottom) scores with hot colors indicating positive and cold colors indicating negative correlations. Statistically thresholded cortical maps (P < 0.05) shows brain regions with significant correlation