Ancestry‐Specific eQTL Associations in AA/NHW Alzheimer Disease Cohorts

Abstract Background Although the genetic etiology of Alzheimer’s Disease (AD) has been shown to vary among ancestrally diverse populations, the influence of genetic variation on the transcriptome within these populations has not yet been extensively investigated. Multi‐ancestry AD datasets featuring both genetic information and gene expression measurements are needed to support expression quantitative trait loci (eQTl) studies in historically underrepresented populations. In this work, we have compared eQTL signals in case/control AD sample populations of both African American (AA) and non‐Hispanic white (NHW) ancestries. Method We analyzed cohorts of AA and NHW ancestries (N aa = 232, N nhw = 241) with corresponding array‐based genotypes and RNA‐seq mRNA expression levels from whole blood, along with clinical AD diagnoses (N aa_case = 115, N aa_control = 117, N nhw_case = 121, N nhw_control = 120). Transcriptome‐wide eQTL associations between genetic variants and gene expression levels were assessed via regression across 17,638 gene units, adjusting for age, sex, experimental factors, and genetic principal components. This was repeated across all sample set combinations of ancestry‐AD pairings (ancestry = [aa, nhw, both] * AD = [case, control, both]). Result The number of significant (p‐value < 10e‐7) eQTL signals varied by ancestry, with the AA cohort showing 76,742 raw eQTL associations, the NHW cohort 19,734, and both ancestries combined 81,763. When further sub‐setting sample sets to AD cases only, 770 and 788 significant eQTL signals were observed for AA and NHW cohorts respectively, which impacted 135 (AA) and 130 (NHW) genes. Interestingly, these ancestry‐aware, case‐only eQTLs were remarkably distinct between the two ancestries, with only 71 (∼9.1%) of the eQTLs and only 19 (∼14.3%) of the impacted genes being shared by both ancestral cohorts. Of particular interest is the appearance of WWOX ( a gene of wide neurological importance, and known AD risk loci), as differentially significant in the AA cohort ( P = 8.07e‐7) as compared to the NHW cohort ( P = 9.06e‐3). Conclusion We have observed differences in eQTL signals relative to AD diagnosis status between AA and NHW sample populations, which may indicate the presence of novel ancestry‐specific effects impacting the genetic etiology of AD through changes in gene expression.