Premium
Transcriptomic profiles underlying functional brain networks at different stages of Alzheimer’s disease
Author(s) -
Yu Meichen,
Nho Kwangsik,
Risacher Shan L.,
Chumin Evgeny J.,
West John D.,
Tharp Matt,
Zhu Jian,
Wen Qiuting,
Eastman Bobi,
Shen Li,
Apostolova Liana G.,
Wu YuChien,
Sporns Olaf,
Saykin Andrew 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.046163
Subject(s) - gene , transcriptome , biology , gene expression , computational biology , gene regulatory network , gene expression profiling , genome wide association study , microarray analysis techniques , microarray , genetics , neuroscience , single nucleotide polymorphism , genotype
Background Specific functional network connectivity patterns have been shown at different stages along the Alzheimer’s disease (AD) continuum, but the contribution of regional gene expression to molecular mechanisms remains unknown. Method We used resting‐state functional MRI to construct functional brain networks in 47 cognitively normal (CN), 46 subjective cognitive decline (SCD), 34 mild cognitive impairment (MCI) and 22 AD participants. Regional gene expression profiles for 20736 protein‐coding genes were derived from brain‐wide microarray‐based transcriptome data from the Allen Human Brain Atlas. We performed one hypothesis‐driven (using pre‐selected AD susceptibility genes) and two data‐driven analyses (using all 20736 genes) to uncover the spatial associations between brain‐wide gene expression levels and connectivity strength (gene‐to‐connectivity associations). In the hypothesis‐driven analysis, we computed the spatial gene‐to‐connectivity associations for 51 AD susceptibility genes selected from large‐scale GWAS for the four clinical groups, separately. In a data‐driven analysis, we determined the gene‐to‐connectivity associations for average gene expression levels of co‐expressed gene modules in the gene co‐expression network. We also performed gene‐set enrichment analysis (GSEA) to identify biochemical pathways associated with the connectivity patterns in AD (1000 permutations; P FDR < 0.05; gene sets from the Gene Ontology). Finally, we analyzed commonalities among the most relevant genes of the identified pathways. Result We identified 15 AD susceptibility genes showing significant gene‐to‐connectivity associations after multiple testing adjustment (P FDR < 0.05). In particular, ECHDC3 and HS3ST1 showed the strongest positive and negative gene‐to‐connectivity associations, respectively, and the strength of the correlations increased as a function of disease stage (CN < SCD < MCI < AD). In addition, two of five gene co‐expression modules showed consistently significant gene‐to‐connectivity associations across the four clinical groups. GSEA identified 31 positively enriched pathways including cellular response to metal ion and fatty acid binding, and 11 negatively enriched pathways including enzyme activity and cellular metabolism. Conclusion Our findings may open novel avenues for improving the interpretation of genetic and imaging biomarkers, for facilitating the detectability of AD trajectories, and for developing effective therapeutic strategies at earlier stages of the disease spectrum when connectivity changes may be among the earliest features.