Association of Cortical β-Amyloid Protein in the Absence of Insoluble Deposits With Alzheimer Disease

Abstract

Importance: β-Amyloid deposits are a pathologic hallmark of Alzheimer disease (AD). However, the extent to which cortical β-amyloid protein in the absence of insoluble deposits is associated with classic features of AD appear to be unknown.

Objective: To examine the associations of cortical β-amyloid protein in the absence of insoluble deposits with cognitive decline, neurofibrillary tangles, other age-associated neuropathologic conditions, and APOE.

Design, setting, and participants: This analysis combines data from 2 community-based clinicopathologic cohort studies of aging. The Religious Orders Study started in 1994, and the Rush Memory and Aging Project started in 1997. Both studies are ongoing. Participants without known dementia were enrolled and agreed to annual clinical evaluations and brain donation after death. Primary analyses focused on individuals without β-amyloid deposits. Data analyses occurred in mid-September 2018.

Main outcomes and measures: β-Amyloid protein abundance was measured by targeted proteomics using selected reaction monitoring. β-Amyloid deposits were detected using immunohistochemistry. Other neuropathologic indices were quantified via uniform structured evaluation. Linear mixed models were used to examine the association of β-amyloid protein with cognitive decline. Regression models examined the protein associations with neuropathologic outcomes and the APOE genotype.

Results: By mid-September 2018, 3575 older persons were enrolled, and 1559 had died and undergone brain autopsy. Proteomic data were collected in 1208 individuals, and 5 with missing cognitive scores were excluded. Of the remaining 1203, primary analyses focused on 148 individuals (12.3%) without β-amyloid deposits. In this group, the mean (SD) age at death was 87.0 (7.0) years, and 84 individuals (56.8%) were women. In the absence of β-amyloid deposits, we did not observe an association of β-amyloid protein with decline in episodic memory, but the protein was associated with faster rates of decline in processing speed (mean [SE] change, -0.014 [0.005]; P = .008) and visuospatial abilities (mean [SE] change, -0.013 [0.005]; P = .006). We did not observe protein association with paired helical filament tau tangle density. The protein was associated with amyloid angiopathy (odds ratio, 1.38 [95% CI, 1.15-1.67]; P < .001) but no other brain pathology. The associations with cognitive decline were unchanged after controlling for amyloid angiopathy. Neither APOE ε4 nor a polygenic Alzheimer risk score was associated with β-amyloid protein.

Conclusions and relevance: Cortical β-amyloid protein was associated with faster cognitive decline in the absence of β-amyloid deposits, which supports the role of cortical soluble β-amyloid as a neurotoxic agent in aging. The lack of protein association with paired helical filament tau tangles, episodic memory decline, or strong genetic drivers of deposited β-amyloid suggests an underlying neuropathologic change that may differ from that of AD.

Figure 1.. Distribution of Cortical β-Amyloid Protein Level

A, The histogram reveals a bimodal pattern in the distribution of cortical β-amyloid protein level (n = 1203). B, A histogram of protein level with β-amyloid deposits (n = 1055). C, A histogram of protein level without β-amyloid deposits (n = 148). D, Overlays of B and C further illustrate that the higher peak was exclusively made up of individuals with β-amyloid deposits; almost all individuals without β-amyloid deposits were clustered in the lower peak.

Figure 2.. β-Amyloid Protein and Cognitive Decline in the Absence of β-Amyloid Deposits

A-F, Model expected values of cognition over time for representative participants (female participants aged 85 years with 15 years of education) with 25th (orange), 50th (dark gray), and 75th (blue) percentiles of β-amyloid level in global cognition (A), episodic memory (B), semantic memory (C), working memory (D), processing speed (E), and visuospatial ability (F). The expected values over time are calculated using point estimates of both intercept and slope as derived from the linear mixed models.