We also conducted Summary-data-based Mendelian Randomization (SMR) and Heterogeneity In Dependent Instruments (HEIDI) tests23 to prioritize likely causal genes and variants by integrating summary statistics from our AAOS GWAS and the Cardiogenics study (Supplementary Table 10). SMR/HEIDI analysis was performed for the SPI1/CELF1 locus using rs1057233, rs10838698, rs10838699, rs7928163, rs10838725 and rs1377416 as candidate causal variants. In both monocytes and macrophages, SPI1 was consistently identified as the most likely gene whose expression levels are associated with AD survival because of causality/pleiotropy at the same underlying causal variant (rs1057233, rs10838698, rs10838699, or rs7928163 in the same LD block) (SMR P < 4.90E-04, Bonferroni-corrected threshold for 6 SNPs tested against 17 probes and HEIDI P ≥ 0.05, Supplementary Fig. 6). Neither conditional analysis nor SMR/HEIDI analysis could definitively identify a single functional variant among this set of 4 SNPs in high LD. Functional analyses will be necessary to determine which SNPs in this LD block directly affects SPI1 expression. Overall, rs1057233 and tagging SNPs are associated with AD risk and survival, and CSF Aβ42. The strong cis-eQTL effects and colocalization results point to SPI1 as the most likely candidate gene underlying the disease association at the SPI1/CELF1 locus.
