Colocalization analyses using coloc32 and SMR/HEIDI23 support the hypothesis that the same causal SNP(s) influence SPI1 expression and AD risk. However, neither conditional nor SMR/HEIDI analyses were able to pin-point an individual SNP; both approaches identified an LD block tagged by rs1057233, in which one or more SNPs may individually or in combination influence both SPI1 expression and AD risk. rs1057233 changes the target sequence and binding of miR-56933, and its tagging SNPs alter binding motifs of transcription factors including PU.1 itself (Supplementary Table 3 and Supplementary Fig. 7d). rs1377416, is located in a predicted enhancer in the vicinity of SPI1 and altered enhancer activity when assayed in vitro using a reporter construct transfected in BV2 cells19. However, rs1057233 remained significantly associated with AD after conditioning for either rs1377416 (P=1.2×10−3) or the previously reported IGAP GWAS top SNP rs10838725 (P=3.2×10−4) in the ADGC dataset. Further, the cis-eQTL association between rs1057233 and SPI1 expression remained significant after conditioning for either of these SNPs, whereas conditioning for rs1057233 abolished their cis-eQTL associations with SPI1 (Supplementary Table 9). Thus, rs1057233 and its tagging
