Premium
Translating genetic risk variants in disease‐associated enhancers into novel mouse models of Alzheimer’s disease
Author(s) -
Preuss Christoph,
Chen Xi,
Chen Kathleen,
Theesfeld Chandra,
Cofer Evan,
Uyar Asli,
Cary Gregory A.,
Pandey Ravi S.,
Garceau Dylan,
Kotredes Kevin P.,
Logsdon Ben,
Mangravite Lara M.,
Howell Gareth,
Sasner Michael,
Troyanskaya Olga,
Carter Gregory W.
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.040529
Subject(s) - genome wide association study , enhancer , biology , computational biology , disease , expression quantitative trait loci , gene , genetics , phenotype , genome , cell type , single nucleotide polymorphism , gene expression , cell , medicine , genotype , pathology
Background The enrichment of late‐onset Alzheimer’s disease (LOAD) GWAS variants in noncoding regions of the genome reveals new potential for modeling disease risk. Yet, identifying noncoding causal variants and the cell types in which they are functional remains challenging. Translating noncoding variants into novel mouse models can elucidate phenotypic effects of those variants through specific perturbations of gene enhancers associated with LOAD risk. Method We developed a multi‐tier approach to translate noncoding GWAS risk variants into novel mouse models of LOAD. First, we used publicly available GWAS summary statistics and eQTL data from the AMP‐AD consortium (ROS/MAP, Mayo Clinic) across two brain regions (DLPFC, TCX) to prioritize LOAD risk variants. Then, a deep learning approach (DeepSEA) was implemented to identify regulatory variants in conserved enhancers that are likely to disrupt expression of nearby candidate genes in both human and mouse. Further, ATAC‐seq data from human and mouse was integrated to predict cell‐type specific effects of the prioritized regulatory elements. Finally, the top‐predicted disease‐associated variants were used to develop novel mouse models as part of the MODEL‐AD initiative. Result Our integrative approach for prioritizing disease‐associated variants revealed a set of conserved risk variants that mediate the expression of nearby LOAD candidate genes in a cell type‐specific manner. These included several well‐known LOAD risk loci ( BIN1 , PT2KB, SCIMP, CD2AP ). In addition, we identified a genome‐wide significant risk variant (rs4575098, p = 2.05 × 10(‐11), Jansen et al. 2019), which decreased expression of the novel candidate FCER1G in the brain of LOAD patients (ROS/MAP eQTL p = 1.81 × 10(‐05), beta = ‐0.27). This highly conserved variant resides in a microglia specific enhancer and overlaps with a binding site of the transcriptional regulator PU.1 ( SPI1 ). We created a 2kb enhancer deletion mouse model for the FCER1G associated variant and knock‐in models for variants linked to increased expression in three well‐known GWAS loci ( BIN1, SCIMP, PT2KB ). We provide initial functional characterization for our novel models based on RNA‐seq data. Conclusion Our findings expand the list of LOAD candidate genes likely to be impacted by noncoding variants and provide a novel set of mouse models to study their effects in vivo.