Functional redundancy of the posterior hippocampi is selectively disrupted in non-demented older adults with β-amyloid deposition

Abstract

Several neural mechanisms underlying resilience to Alzheimer's disease (AD) have been proposed, including redundant neural connections between the posterior hippocampi and all other brain regions, and global functional connectivity of the left frontal cortex (LFC). Here, we investigated if functional redundancy of the hippocampus (HC) and LFC underscores neural resilience in the presence of early AD pathologies. From the ADNI database, cognitively normal older adults (CN) (N = 220; 36 % Aβ+) and patients with Mild Cognitive Impairment (MCI) (N = 143; 51 % Aβ+) were utilized. Functional redundancy was calculated from resting state fMRI data using a graph theoretical approach by summing the direct and indirect paths (path lengths = 1-4) between each region of interest and its 263 functional connections. Posterior HC, but not anterior HC or LFC, redundancy was significantly lower in Aβ+ than Aβ-groups, regardless of diagnosis. Posterior HC redundancy related to higher education and better episodic memory, but it did not moderate the Aβ-cognition relationships across the diagnostic groups. Together, these findings suggest that posterior HC redundancy captures network disruption that parallels selective vulnerability to Aβ deposition. Further, our findings indicate that functional redundancy may underscore a network metric different from global functional connectivity of the LFC.

Keywords: Aging; Alzheimer's disease; Beta-amyloid plaques; Functional magnetic resonance imaging; Graph theory; Resilience.

Fig. 1

Brain nodes and a priori defined regions of interest. (A) 263 nodes distributed across 14 networks displayed in glass brain. (B) A priori defined regions of interest displayed in radiological view. Regions of interest included the anterior and posterior bilateral hippocampi and the left frontal cortex. HC = hippocampus; AUD = auditory network; COP = cinguloopercular network; DAN = dorsal attention network; DMN = default mode network; FPN = frontoparietal network; MTL = medial temporal lobe network; PMN = parietomedial network; SAL = salience network; REW = reward network; SMD = somatomotor network dorsal; SML = somatomotor network lateral; UNA = unassigned; VAN = ventral attention network; VIS = visual network.

Fig. 2

Cognitive performance by diagnostic group and Aβ-status on composite scores of (A) memory, and (B) executive function. Plotted composite scores were normalized via z-transformation and bars represent Mean ± SE.

Fig. 3

Functional redundancy by diagnostic group and Aβ-status in five a priori regions of interest. Plotted values represent functional redundancy of the (A) left anterior HC, (B) right anterior HC, (C) left posterior HC, (D) right posterior HC, and (E) left frontal cortex. Functional redundancy is summarized by AUC across the density thresholds 0.025–0.25 [steps of 0.025]) and bars represent Mean ± SE. ∗p < .05.

Fig. 4

Different relationships between Aβ deposition and hippocampal redundancy by Aβ-status and diagnosis. (A) Right anterior HC redundancy-Aβ SUVR relationship by amyloid-status in CN. Continuous amyloid SUVR (x-axis) and functional redundancy in the right anterior HC (y-axis) are plotted as a function of Aβ-status. Aβ− CN is depicted in purple and Aβ+ CN is depicted in green. A best fitting regression line for each CN group is shown in the scatterplot (A). (B) Left posterior HC redundancy-Aβ SUVR relationship by diagnosis. Continuous amyloid SUVR (x-axis) and functional redundancy in the left posterior HC (y-axis) are plotted as a function of diagnosis. CN is depicted in blue and MCI is depicted in yellow. A best fitting regression line for each diagnostic group is shown in the scatterplot (B).

Fig. 5

Relationships between posterior hippocampal redundancy and years of education. Redundancy values are represented on the x-axis and the years of education are represented on the y-axis for the (A) left posterior hippocampus and (B) right posterior hippocampus. The best fitting regression line is depicted in each plot.

Fig. 6

Posterior hippocampal redundancy relations with memory performance. Plotted are composite memory scores on the y-axis and redundancy on the x-axis for the (A) left posterior hippocampus, and (B) right posterior hippocampus. Aβ− is depicted in purple and Aβ+ is depicted in green collapsing both CN and MCI groups. Regression coefficients of the relationships between regional redundancy and memory in each Aβ group are displayed in the bottom right portion of each plot.