Cis-eQTL analyses of AAOS-associated SNPs revealed limited associations when using data from brain tissue homogenates, yet identified multiple candidate genes when using data from myeloid cells, the top candidate causal cell types for AD based on the stratified LD score regression analysis of AD heritability presented here. This calls attention to careful selection of relevant cell types in eQTL studies of disease associations. In particular, by conducting cis-eQTL analyses using monocyte and macrophage datasets, we discovered associations of AAOS-associated SNPs with the expression of SELL, SPI1, MYBPC3, NUP160, MS4A4A, MS4A6A and SUN2 (Table 3). Furthermore, we replicated the cis-eQTL associations of rs1057233 with SPI1, MYBPC3, rs7930318 with MS4A4A, MS4A6A and rs2272918 with SELL in an independent monocyte dataset. We further showed that SPI1 myeloid cis-eQTLs and AAOS-associated SNPs are not likely to be colocalized by chance and thus may be in the causal path to AD (Fig. 1). Notably, the minor allele of rs1057233 (G) is suggestively associated with lower AD risk (P=5.4×10−6, 5.9×10−7 in IGAP stage I, stage I and II combined, respectively)1, later AAO (P=8.4×10−6) and significantly associated
