Takayasu arteritis (TAK) is a systemic vasculitis with severe complications that affects the aorta and its large branches. HLA-B*52 is an established susceptibility locus to TAK. To date, there are still only a limited number of reports concerning non-HLA susceptibility loci to TAK. We conducted a genome-wide association study (GWAS) and a follow-up study in a total of 633 TAK cases and 5,928 controls. A total of 510,879 SNPs were genotyped, and 5,875,450 SNPs were imputed together with HLA-B*52. Functional annotation of significant loci, enhancer enrichment, and pathway analyses were conducted. We identified four unreported significant loci, namely rs2322599, rs103294, rs17133698, and rs1713450, in PTK2B, LILRA3/LILRB2, DUSP22, and KLHL33, respectively. Two additional significant loci unreported in non-European GWAS were identified, namely HSPA6/FCGR3A and chr21q.22. We found that a single variant associated with the expression of MICB, a ligand for natural killer (NK) cell receptor, could explain the entire association with the HLA-B region. Rs2322599 is strongly associated with the expression of PTK2B. Rs103294 risk allele in LILRA3/LILRB2 is known to be a tagging SNP for the deletion of LILRA3, a soluble receptor of HLA class I molecules. We found a significant epistasis effect between HLA-B*52 and rs103294 (P = 1.2 × 10−3). Enhancer enrichment analysis and pathway analysis suggested the involvement of NK cells (P = 8.8 × 10−5, enhancer enrichment). In conclusion, four unreported TAK susceptibility loci and an epistasis effect between LILRA3 and HLA-B*52 were identified. HLA and non-HLA regions suggested a critical role for NK cells in TAK.