O4–01–03: Genome‐wide association study and admixture mapping of age‐related cognitive decline in African‐Americans

To leverage genome-wide data from several cohorts so as to identify genetic determinants of age-related cognitive decline among older African-American (AA) people. We examined genome-wide genotyping data for 4,426 unrelated older AA subjects who had at least two repeated measures of cognition from six (CHAP, IIDP, ROS, MAP, MARS, and WHICAP) prospective community-based studies. We compared them to 2,703 unrelated non-Hispanic European Americans (EAs) from the ROS, MAP and CHAP studies. Within each cohort, individual cognitive tests were combined to form an aggregate measure of global cognition. We used linear mixed effects models to characterize individual paths of change in cognition, controlling for age, sex and education as fixed effects. After quality control, imputation was performed using the 1000 Genomes Project Phase I combined reference panels of EA and AA ancestry. Common variants were meta-analyzed across cohorts using weighted fixed-effect models. HAPMIX software was used to estimate ancestry at each locus, and MIXSCORE software was used to test whether individuals with faster or slower rates of global cognitive decline had a proportion of ancestry different from the genome average. APOE ε4 haplotype was strongly associated with rate of cognitive decline in AAs (P= 1.92 x 10 -14) but the magnitude of the effect was weaker in AAs (β= -0.01) compared to EAs (β= -0.05). In AAs we replicated previously known loci in EAs including ABCA7 (P= 4.13 x 10 -4), PICALM (7.35 x 10 -4), and EPHA1 (P= 8.06 x 10 -4). We discovered one genome-wide significant association at the TRPS1 locus (chr 8q23), and replicated that association in EAs (P(DISC) = 2.52 x 10 -7; P(REP) = 9.79 x 10 -3; P(JOINT) = 1.37 x 10 -8). We also found another variant at the TEK locus (chr 9p21) with suggestive evidence of association (P(DISC) = 4.10 x 10 -7). Using admixture mapping, we did not detect any significant admixture peak. These results suggest that genetic factors alone may not explain differences in age-related cognitive decline between AAs and EAs. Further large-scale admixture mapping studies will be necessary to validate our findings.