Alterations in MRI-visible perivascular spaces precede dementia diagnosis by 18 years in autosomal dominant Alzheimer's disease

Abstract

Introduction: Perivascular space (PVS) alterations are traditionally linked to cardiovascular risk factors and aging, but may also play a direct role in Alzheimer's disease (AD). To reduce confounding from age-related comorbidities, we examined PVSs in autosomal dominant AD (ADAD).

Methods: In this cross-sectional study of 96 non-demented individuals (62 mutation carriers), we quantified PVS count fraction and mean diameter in white matter and basal ganglia using automated magnetic resonance imaging analysis. Linear mixed models assessed group differences along the disease course, adjusting for cardiovascular risk factors.

Results: Compared to non-carriers, mutation carriers showed lower PVS count fraction in white matter and basal ganglia, and larger PVS diameter in basal ganglia and the temporal lobe. Changes were evident up to 18 years before expected dementia onset and followed trajectories similar to amyloid beta 42 and tau biomarkers.

Discussion: ADAD is associated with early PVS alterations, suggesting perivascular changes may be integral to primary AD pathology.

Highlights: Autosomal dominant Alzheimer's disease (ADAD) mutation carriers have reduced magnetic resonance imaging-visible perivascular space (PVS) count fraction in the white matter and basal ganglia. ADAD mutation carriers show enlarged PVS in the basal ganglia and temporal white matter. PVS alterations start 18 years before the estimated time of dementia diagnosis. The spatial localization of PVS changes overlaps with regions of amyloid beta (Aβ) accumulation. The temporal evolution of PVS alterations aligns with Aβ and tau changes in the cerebrospinal fluid.

Keywords: Alzheimer's disease; autosomal dominant Alzheimer's disease; cerebral small vessel disease; dominantly inherited Alzheimer's disease; magnetic resonance imaging; perivascular spaces.

FIGURE 1

Differences in perivascular space count fraction and mean diameter in the white matter and basal ganglia. Estimated marginal means (solid dots) and 95% confidence intervals (error bars) for PVS count fraction (A, B) and PVS mean diameter (C, D) in non‐carriers (green) and mutation carriers (violet) in the BG (A, C) and WM (B, D). Mean estimates were obtained using linear mixed models including mutation status, estimated years from dementia diagnosis, and their interaction; sex, cardiovascular risk score; and presence of an APOE ε4 allele as fixed effects, while family membership was included as a random effect. Empty colored dots correspond to individual empirical observations. The asterisk(s) refer to the p value for the variable "mutation status" obtained from the adjusted linear mixed model. Insets in (B) and (D) show color‐coded brain maps illustrating model coefficients for "mutation status" in various brain regions against PVS count fraction or PVS mean diameter, respectively. Cooler colors (e.g., darker blue) represent regions with stronger negative associations, while warmer colors (e.g., red) indicate regions with stronger positive associations between mutation status and PVS biomarkers. Only regions with uncorrected p value < 0.05 are shown, while others are grayed out.*: 0.01 ≤ p < 0.05; **: 0.001 ≤ p < 0.01; ***: 0.0001 ≤ p < 0.001. APOE, apolipoprotein E; BG, basal ganglia; Cl, confidence interval; PVS, perivascular space; WM, white matter.

FIGURE 2

Time of first detectable changes in perivascular spaces count fraction and mean diameter. Mean and 95% confidence intervals of the estimated differences in PVS count fraction (A) and mean diameter (B) between mutation carriers and non carriers. Differences were calculated with linear mixed models at each estimated year from EDD (range: ‐20 to ‐10). Linear mixed models included mutation status, EDD, and their interaction; sex; cardiovascular risk score; and presence of APOE ε4 allele(s) as fixed effects, while family membership was included as a random effect. Estimated mean differences were averaged over levels of cardiovascular risk, sex, and APOE. The red bar indicates the earliest time point of abnormal PVS metrics for a particular brain region. APOE, apolipoprotein E; BG, basal ganglia; CI, confidence interval; EDD, expected dementia diagnosis; PVS, perivascular space; WM, white matter.

FIGURE 3

Changes in clinical, imaging, and biochemical biomarkers with respect to changes in perivascular space metrics as a function of estimated years from expected symptom onset. The main panel displays the SD between mutation carriers and non carriers over the expected course of the disease for various clinical, imaging, and biochemicat biomarkers, as reported in Bateman et al. These differences are compared to the SD in PVS count fraction in the WM (WM‐PVS count fraction, in tight red) and PVS mean diameter in the BG (BG‐PVS mean diameter, in red). The expected course of the disease is assessed using the estimated years from EYO. To reconcile the discrepancy between the definition of EYO onset used in Bateman et al. and our definition of EDD, we added 2.6 years to EDD before estimating SD for PVS metrics. This number represented the average time difference between DIAN‐determined EYO and our‐determined EDD (see main text). The inset highlights the comparison between WM‐PVS count fraction and CSF biomarkers (Aβ and tau). To facilitate visual comparisons of the relative magnitude and temporal course of changes, absolute values of SDs are shown. Aβ, amyloid beta; BG, basal ganglia; CDR‐SOB, Clinical Dementia Rating Sum of Boxes; CSF, cerebrospinal fluid; DIAN, Dominantly Inherited Alzheimer Network; EDD, expected dementia diagnosis; EYO, expected symptom onset; PVS, perivascular space; SD, standardized differences; WM, white matter.