Copy number variant scores are associated with cerebrovascular pathology in aging

Late-onset Alzheimer's disease (LOAD) is the most common cause of dementia in older adults, with specific genomic copy number variants (CNVs) implicated in its pathology. However, the aggregate burden of genome-wide CNVs in dementia and age-related neuropathologies is uncharacterized. This study investigated the association between genome-wide CNV scores (CNV-S) and dementia, as well as LOAD-related neuropathologies, in 1011 elderly participants (mean age 88.06) from two ongoing US-based longitudinal clinical-pathological cohort studies who were initially dementia-free and consented to brain donation upon death. Participants exhibited varying cognitive statuses at death (429 dementia, 258 mild cognitive impairment, 324 cognitively normal). We evaluated effects of (1) eight individual-level CNV-S based on gene loss intolerance and dosage sensitivity; (2) a single nucleotide polymorphism (SNP)-based LOAD polygenic score (PGS-LOAD) calculated using Bayesian continuous shrinkage; and (3) covariates (age, sex, and education). Outcomes included cognitive scores across 19 tests, clinical diagnoses of Alzheimer's disease or mild cognitive impairment, and four LOAD-related neuropathologies assessed postmortem. Analyses identified 4867 CNVs (3918 deletions, 949 duplications) mapped to 3211 genes. Higher deletion CNV-S were significantly associated with increased cerebrovascular pathologies (pLI: β = 0.14, 95% CI [0.08, 0.21]; LOEUF: β = 0.14, 95% CI [0.08, 0.20]; pHI: β = 0.15, 95% CI [0.08, 0.21]; binarized pHI: β = 0.14, 95% CI [0.08, 0.21]). Models predicting cerebral atherosclerosis that included deletion CNV-S significantly outperformed models based on only PGS-LOAD (R² increase: 0.02). These findings suggest that genome-wide CNV burden, particularly deletions in dosage-sensitive genes, contributes to cerebrovascular pathology in aging. CNV-S may augment existing LOAD genetic risk models by capturing vascular pathways distinct from traditional SNP-based risk.