Premium
Multimodal genome‐wide meta‐analysis of brain amyloidosis reveals heterogeneity across CSF, PET, and pathological amyloid measures
Author(s) -
Archer Derek B.,
Mahoney Emily R.,
Dumitrescu Logan,
Jefferson Angela L.,
Jagust William J.,
Resnick Susan M.,
Bilgel Murat,
Johnson Sterling C.,
Engelman Corinne D.,
Cruchaga Carlos,
Zetterberg Henrik,
Blennow Kaj,
Deming Yuetiva,
Sperling Reisa A.,
Johnson Keith A.,
Buckley Rachel F.,
Larson Eric B.,
Mayeux Richard,
Bennett David A.,
Schneider Julie A.,
Kukull Walter A.,
Keene C. Dirk,
Montine Thomas J.,
Beecham Gary W.,
Schellenberg Gerard D.,
Hohman Timothy J.
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.046009
Subject(s) - amyloidosis , medicine , meta analysis , amyloid (mycology) , pathology , genome wide association study , pittsburgh compound b , positron emission tomography , oncology , alzheimer's disease , disease , biology , nuclear medicine , genetics , genotype , single nucleotide polymorphism , gene
Background Amyloidosis in Alzheimer’s disease (AD) is a common feature that can be measured via cerebrospinal fluid (CSF), positron emission tomography (PET), and autopsy measures of neuritic plaques. While individual genome‐wide association studies (GWAS) have identified more than 40 loci as potential biological drivers of AD, fewer drivers of amyloidosis have been identified due to the small sample sizes in endophenotype analyses. Combining unimodal measures of amyloid into a larger multi‐modal analysis could provide new insight into mechanisms driving amyloidosis. The goal of this study was to conduct a large meta‐analysis of CSF‐, PET‐, and pathology‐derived metrics of amyloid positivity. Method 11,941 non‐Hispanic white participants (CSF = 2505, PET = 3976, Autopsy = 5460) from 14 cohorts were analyzed. To facilitate cross‐modality harmonization, a binarized amyloid status (negative/positive) was determined on an individual cohort basis using two approaches: a Gaussian mixture model clustering algorithm for CSF/PET measurements and CERAD thresholds (CERAD>sparse or CERAD ≤ sparse) for autopsy measures of amyloid pathology. In total, our analysis included 5,634 amyloid‐negative and 6,307 amyloid‐positive individuals. GWAS using a logistic regression model covarying for age and sex were performed in each of the 14 cohorts, and a meta‐analysis was performed within each modality. These meta‐analyses were followed with a multi‐modal meta‐analysis. The standard genome‐wide threshold of statistical significance (p < 5 × 10 −8 ) was applied. Additionally, we performed candidate analysis for the 39 previously published clinical AD or amyloid risk loci. Result Aside from variants within the APOE region, our multi‐modal meta‐analysis did not identify any genome‐wide loci (Figure 1). A lack of convergent GWAS signals was also observed when stratifying by age, sex, and APOE carrier status. However, our candidate gene analysis demonstrated some consistency across modalities. Two prior loci were significant in 2/3 modalities, including ABCA7 (rs4147929; p = 1.98 × 10 −4 ) and CLU (rs4236671; p = 0.001). Conclusion In the largest GWAS of brain amyloidosis, we found a lack of convergence across modalities, suggesting substantial heterogeneity across measures and cohorts. Candidate analyses highlight some consistent signals in ABCA7 and CLU . Results suggest continued emphasis on increasing sample sizes for endophenotype analyses and on standardized methodologies to reduce sample heterogeneity to improve statistical power.