Structural white matter properties and cognitive resilience to tau pathology

Abstract

Introduction: We assessed whether macro- and/or micro-structural white matter properties are associated with cognitive resilience to Alzheimer's disease pathology years prior to clinical onset.

Methods: We examined whether global efficiency, an indicator of communication efficiency in brain networks, and diffusion measurements within the limbic network and default mode network moderate the association between amyloid-β/tau pathology and cognitive decline. We also investigated whether demographic and health/risk factors are associated with white matter properties.

Results: Higher global efficiency of the limbic network, as well as free-water corrected diffusion measures within the tracts of both networks, attenuated the impact of tau pathology on memory decline. Education, age, sex, white matter hyperintensities, and vascular risk factors were associated with white matter properties of both networks.

Discussion: White matter can influence cognitive resilience against tau pathology, and promoting education and vascular health may enhance optimal white matter properties.

Highlights: Aβ and tau were associated with longitudinal memory change over ∼7.5 years. White matter properties attenuated the impact of tau pathology on memory change. Health/risk factors were associated with white matter properties.

Keywords: amyloid‐β; cognitive resilience; default mode network; limbic network; tau; white matter.

FIGURE 1

Methodology overview. The whole‐brain network was constructed using whole‐brain tractogram and brain parcellation (step 1); the limbic network and the DMN were extracted from the whole‐brain networks (step 2), and graph theory analysis was applied to calculate the global efficiency of the two networks, respectively (step 3–4); tracts that connected regions of the limbic network and the DMN were extracted and weighted average diffusion metrics within the tracts of both networks were calculated (step 3–4); multiple linear regression models were performed to examine the potential moderation effect of global efficiency/diffusion metrics of the limbic network and the DMN on pathological‐cognitive associations (step 5). Multivariate PLS analyses were conducted to examine the relationship between the macro‐/micro‐structural white matter measurements of the structural networks, and health/lifestyle factors (step 6). AD_T, free‐water corrected axial diffusivity; DMN, default mode network; FA_T, free‐water corrected fractional anisotropy; FW, free‐water index; MD_T, free‐water corrected mean diffusivity; RD_T, free‐water corrected radial diffusivity; PLS, partial least square; SVD, singular value decomposition

FIGURE 2

Longitudinal memory change and associations with AD pathology. Annual memory changes were calculated based on the longitudinal RBANS data using linear mixed‐effects models, and the average of annual changes for immediate memory and delayed memory are shown in (A, B). The associations between Aβ‐/tau‐pathology are displayed in (C, D). Higher levels of both Aβ and tau pathology were associated with worse immediate and delayed memory decline. Linear models were adjusted for age and sex. Uncorrected two‐sided p‐values are presented; † indicates adjusted p‐value ≤ 0.05 after FDR correction. β _st: standardized estimate β; Aβ, amyloid‐β; AD, Alzheimer's disease; FDR, false discovery rate; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status.

FIGURE 3

Association between Aβ/tau pathology and macro‐/micro‐structural white matter measurements. The scatter plots show the association between Aβ/tau pathology and global efficiency of the limbic network (A, B) and the DMN (C, D). Linear models were adjusted for age and sex. β _st estimates and 95% standardized CIs, obtained from linear regression models incorporating various white matter measures as dependent variables are displayed on the right of each panel. Light green represents significant β _st estimates for the association between Aβ/tau pathology and white matter measures, and pink represents non‐significant β _st estimates. Aβ, amyloid‐β; AD_T, free‐water corrected axial diffusivity; β _st, standardized estimate β; CI, confidence interval; DMN, default mode network; FA_T, free‐water corrected fractional anisotropy; FW, free‐water index; MD_T, free‐water corrected mean diffusivity; RD_T, free‐water corrected radial diffusivity.

FIGURE 4

Moderation by global efficiency on the association between tau pathology and longitudinal memory change. In the limbic network, higher global efficiency was associated with an attenuated effect of tau burden on both changes in immediate memory (A) and changes in delayed memory (B). The global efficiency of the DMN showed a non‐significant moderation effect on the association between tau pathology on both change in immediate memory (C) and change in delayed memory (D). Note that moderation effects were assessed using continuous values of global efficiency, and the data were subsequently divided into terciles for visualization purposes. Specifically, the lowest tercile contains the lower third of the data distribution (L; colored in yellow), the middle tercile spans the middle third (M; colored in light gray), and the upper tercile covers the upper third (H; colored in dark blue). Age, sex, and gray matter volume (divided by total intracranial volume) of regions for the limbic network or the DMN were adjusted for in linear models. Uncorrected two‐sided p‐values are presented; † indicates adjusted p‐value ≤ 0.05 after FDR correction. β _st: standardized estimate β; DMN, default mode network.

FIGURE 5

Moderation by diffusion metrics on the association between tau pathology and longitudinal memory change. The moderating effects of diffusion metrics within the limbic network (A, B) and DMN (C, D) on the impact of tau pathology on immediate and delayed memory change. The scatter plots in each panel show that FA_T in the tracts acts as a moderator on the association between tau burden and longitudinal memory changes. Note that moderation effects were assessed using continuous values of diffusion metrics (average diffusion metrics in the tracts of the limbic network or the DMN), and the data were subsequently divided into terciles for visualization purposes. Specifically, the lowest tercile contains the lower third of the data distribution (L; colored in yellow), the middle tercile spans the middle third (M; colored in light gray), and the upper tercile covers the upper third (H; colored in dark blue). Age, sex, and gray matter volume (divided by total intracranial volume) of regions for the limbic network or DMN were adjusted for in linear models. Uncorrected two‐sided p‐values are presented; † indicates adjusted p‐value ≤ 0.05 after FDR correction. β _st estimates and 95% standardized CIs, obtained from linear regression models incorporating various diffusion measures as interaction terms, are displayed on the right of each panel. Light green represents significant β _st estimates for the moderation effect of diffusion metrics and pink represents non‐significant β _st estimates. AD_T, free‐water corrected axial diffusivity; CI, confidence interval; DMN, default mode network; FA_T, free‐water corrected fractional anisotropy; β _st, standardized estimate β; FDR, false discovery rate; FW, free‐water index; MD_T, free‐water corrected mean diffusivity; RD_T, free‐water corrected radial diffusivity.

FIGURE 6

Results of the PLS analyses between white matter characteristics and health/risk factors. In each panel, the magnitude of the bootstrap ratios for white matter measurements within the limbic network (A) and DMN (B) are depicted using a gradient green color scale on the left. The significance threshold is indicated by dotted lines; bootstrap ratios of white matter measurements above/below the dotted lines are considered as significant variables contributing to the association pattern. On the right side of each panel, the strengths of the loadings for the health/risk factors are shown in a gradient orange hue, with error bars representing bootstrap‐estimated 95% confidence intervals; * indicates significant factors that contributed to the association pattern. AD_T, free‐water corrected axial diffusivity; DMN, default mode network; FA_T, free‐water corrected fractional anisotropy; FW, free‐water index; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; MD_T, free‐water corrected mean diffusivity; PLS, partial least square; RD_T, free‐water corrected radial diffusivity; WMH, white matter hyperintensity.