Atrophy of amygdala and abnormal memory-related alpha oscillations over posterior cingulate predict conversion to Alzheimer's disease

Abstract

Synaptic dysfunction, a key pathophysiological hallmark of Alzheimer's disease (AD), may account for abnormal memory-related EEG patterns in prodromal AD. Here, we investigate to what extent oscillatory EEG changes during memory encoding and/or retrieval enhance the accuracy of medial temporal lobe (MTL) atrophy in predicting conversion from amnestic mild cognitive impairment (aMCI) to AD. As expected, aMCI individuals that, within a 2-year follow-up period, developed dementia (N = 16) compared to healthy older (HO) (N = 26) and stable aMCI (N = 18) showed poorer associative memory, greater MTL atrophy, and lower capacity to recruit alpha oscillatory cortical networks. Interestingly, encoding-induced abnormal alpha desynchronized activity over the posterior cingulate cortex (PCC) at baseline showed significantly higher accuracy in predicting AD than the magnitude of amygdala atrophy. Nevertheless, the best accuracy was obtained when the two markers were fitted into the model (sensitivity = 78%, specificity = 82%). These results support the idea that synaptic integrity/function in the PCC is affected during prodromal AD and has the potential of improving early detection when combined with MRI biomarkers.

Figure 1. Memory performance in the recognition task and associations with hippocampal and amygdala volumes.

(A) Mean accuracy and RT in the face-location memory task. Vertical lines on the bars refer to mean standard error. Asterisks indicate significant differences between the two conditions of congruency and between the aMCI-c group and the remaining two groups. (B) Scatter plots depicting correlation between residuals of associative d’ and residuals of hippocampal or amygdala volume either on the left or right hemisphere after regressing these variables on age and total ICV. Each color represents one group. Given that regressions only reached significance when all participants were included in the analysis, only one fitted regression line was shown for each scatter plot.

Figure 2. Topographic distribution of group differences in alpha ERD and correlations with memory performance.

(A) Statistical nonparametric maps showing higher alpha ERD in aMCI-s than in aMCI-c during encoding and retrieval. (B) Scatter plots illustrating the relationship between residuals of alpha ERD in a voxel showing local maxima in the group contrast and residuals of memory performance in the recognition task (associative d’), for both aMCI-s and aMCI-c after regressing these variables on age. L = left; R = right; LMTG = left middle temporal gyrus; LPCC = left posterior cingulate; LAngG = left angular gyrus; *p < 0.05; **p < 0.005; ***p < 0.001.

Figure 3. Group differences in correlations between memory and beta ERD.

(A) Statistical nonparametric maps showing significant differences between aMCI-s and aMCI-c in correlations between beta ERD and memory performance (semantic d’ and associative d’). Blue areas indicate that the higher the beta ERD over these regions the better the memory performance in aMCI-s with respect to aMCI-c. Orange areas indicate that the higher the beta ERD over these regions the lower the benefit from semantic congruence in aMCI-c with respect to aMCI-s. (B) Scatter plots illustrating the relationship between residuals of beta ERD and residuals of memory performance in the recognition task in a voxel showing local maxima in the correlation, for both aMCI-s and aMCI-c after regressing these variables on age. L = left; R = right; LIPL = left inferior parietal lobe; LACC = left anterior cingulate; RPrec = right precuneus; *p < 0.05; **p < 0.005; ***p < 0.001.