A Genome‐Wide Search for Pleiotropy in Cognitive Domain Scores

Abstract

AbstractBackgroundGenome‐wide association studies (GWASs) have identified loci for Alzheimer’s disease (AD)‐related cognitive traits, and genetic correlations among these traits suggest that a single gene may affect them jointly, i.e., pleiotropy.MethodWe conducted GWASs for scores previously generated by factor analyses for cognitive domains—executive function, language, and memory—using longitudinal data from 7,503 Framingham Heart Study (FHS) and 6,758 National Alzheimer’s Coordinating Center (NACC) subjects who had an average of five and three exams, respectively. We averaged domain scores for each participant by weighting down scores at the visit without normal cognition and used the sum of weights to account for heterogeneous residual variances. The association of each domain score with 8.9 million genetic variants (MAF>0.01), imputed using the TOPMed reference panel (GRCh38), was tested separately in each dataset using linear models including age, sex, education level, and five principal components of ancestry as covariates. Results from FHS and NACC datasets were combined by meta‐analyses using the inverse‐variance method. Pleiotropy for each pair of cognitive domains was evaluated using PLACO (pleiotropic analysis under the composite null hypothesis).ResultScores for paired cognitive domains were moderately genetically correlated—executive function and language (ρP =0.46, ρG =0.69), executive function and memory (ρP =0.39, ρG =0.43), language and memory (ρP =0.44, ρG =0.46). Pleiotropy analyses revealed genome‐wide significant associations for all paired domain scores with many variants in the APOE region (p<3.91×10‐9 for each). For pleiotropy of executive function and language, we identified genome‐wide significant associations with DOK5 (rs28663252, p=4.51×10‐8) and suggestive associations with SOX14 (rs55829884, p=8.03×10‐8), IL20RB (rs11925531, p=9.26×10‐8), and RIMS2 (rs140291972, p=8.76×10‐7). We also found suggestive associations for pleiotropy of language and memory with LHFPL4 (rs2936482, p=5.08×10‐8) and for pleiotropy of executive function and memory with MC4R (rs1943220, p=7.54×10‐8) and CDH20 (rs78756286, p=1.50×10‐7).ConclusionWe identified significant evidence of pleiotropy for multiple cognitive domains, and these findings may improve the interpretation of the genetic basis of cognition and AD‐related traits. These results highlight the benefits of factor scores derived from groups of subjects assessed with various cognitive tests and pleiotropy analysis as a tool for identifying genes that contribute to performance in multiple cognitive domains.