Relationships Between Tau and Glucose Metabolism Reflect Alzheimer's Disease Pathology in Cognitively Normal Older Adults

Abstract

Tau is associated with hypometabolism in patients with Alzheimer's disease. In normal aging, the association between tau and glucose metabolism is not fully characterized. We used [18F] AV-1451, [18F] Fluorodeoxyglucose, and [11C] Pittsburgh Compound-B (PiB) PET to measure associations between tau and glucose metabolism in cognitively normal older adults (N = 49). Participants were divided into amyloid-negative (PiB-, n = 28) and amyloid-positive (PiB+, n = 21) groups to determine effects of amyloid-β. We assessed both local and across-brain regional tau-glucose metabolism associations separately in PiB-/PiB+ groups using correlation matrices and sparse canonical correlations. Relationships between tau and glucose metabolism differed by amyloid status, and were primarily spatially distinct. In PiB- subjects, tau was associated with broad regions of increased glucose metabolism. In PiB+ subjects, medial temporal lobe tau was associated with widespread hypometabolism, while tau outside of the medial temporal lobe was associated with decreased and increased glucose metabolism. We further found that regions with earlier tau spread were associated with stronger negative correlations with glucose metabolism. Our findings indicate that in normal aging, low levels of tau are associated with a phase of increased metabolism, while high levels of tau in the presence of amyloid-β are associated with hypometabolism at downstream sites.

Keywords: Alzheimer’s disease; aging; amyloid-β; glucose metabolism; tau.

Figure 1.

Across-brain AV-1451–FDG correlations. Pearson partial correlations between AV-1451 and FDG ROIs in the PiB– (A) and PiB+ (B) groups. Adjusted P-values were computed from a permutation analysis, with a white dot in each cell indicating adjusted P < 0.05, and a white plus (+) indicating adjusted P < 0.10. Ent, Entorhinal Cortex; Hipp, Hippocampus; Parahipp, Parahippocampal Gyrus; Amyg, Amygdala; Mid, Middle; Temp, Temporal; Inf, Inferior; Rost, Rostral; Ant, Anterior; Cing, Cingulate; Caud, Caudal; Post, Posterior; Isth, Isthmus; Sup, Superior; BankSTS, Banks of the Superior Temporal Sulcus; Par, Parietal; Marg, Marginal; Front, Frontal; Orb, Orbital; Lat, Lateral; Occ, Occipital.

Figure 2.

Sparse canonical correlation analysis. Significant dimensions produced from the sparse canonical correlation analysis in the PiB– (A–C) and PiB+ (D–F) groups. Dimensions represent maximized correlations between sparse weightings of AV-1451 and FDG ROIs. Each dimension consists of AV-1451 weights, representing an increase (positive weights) in AV-1451, with the corresponding FDG weights, representing either an increase or decrease (negative weights) in FDG. Weights reduced to zero due to sparsity constraints are not included in the color scale. Bilateral ROIs are depicted on a left hemisphere template brain.

Figure 3.

Extent of Tau pathology and AV-1451–FDG correlations. The relationship between the order of tau spread and the range of AV-1451–FDG correlation coefficients is depicted as a bar plot in (A). AV-1451 ROIs are ranked by the order of tau spread from first to last in the PiB– (top) and PiB+ (bottom) groups separately. Bar graphs represent the mean Fisher’s z’ transformed correlation coefficient (± standard error) for each AV-1451–FDG pair associated with the corresponding AV-1451 ROI. The relationship between mean AV-1451 SUVR and the range of AV-1451–FDG correlation coefficients is depicted as a scatter plot in (B) for the PiB– (top) and PiB+ (bottom) groups separately. The mean SUVR of each AV-1451 ROI is plotted on the x-axis. Each individual Fisher’s z’ transformed correlation coefficient value is plotted corresponding with the mean SUVR of the AV-1451 ROI.