Background: Low lean body mass is the most important predictor of sarcopenia with strong genetic background. The aim of this study was to uncover genetic factors underlying lean mass development. Materials and methods: We performed a genome-wide association study (GWAS) of fat-adjusted leg lean mass in the Framingham Heart Study (FHS, N = 6587), and replicated in the Women's Health Initiative-African American sub-sample (WHI-AA, N = 847) and the Kansas City Osteoporosis Study (KCOS, N = 2219). We also cross-validated significant variants in the publicly available body mass index (BMI) summary results (N ~ 700,000). We then performed a series of functional investigations on the identified variants. Results: Four correlated SNPs at 6p21.1 were identified at the genome-wide significance (GWS, ? = 5.0 × 10-8) level in the discovery FHS sample (rs551145, rs524533, rs571770, and rs545970, p = 3.40-9.77 × 10-9), and were successfully replicated in both the WHI-AA and the KCOS samples (one-sided p = 1.61 × 10-3-0.04). They were further cross-validated by the large-scale BMI summary results (p = 7.0-9.8 × 10-3). Cis-eQTL analyses associated these SNPs with the NFKBIE gene expression. Electrophoresis mobility shift assay (EMSA) in mouse C2C12 myoblast cells implied that rs524533 and rs571770 were bound to an unknown transcription factor in an allelic specific manner, while rs551145 and rs545970 did not. Dual-luciferase reporter assay revealed that both rs524533 and rs571770 downregulated luciferase expression by repressing promoter activity. Moreover, the regulation pattern was allelic specific, strengthening the evidence towards their differential regulatory effects. Conclusions: Through a large-scale GWAS followed by a series of functional investigations, we identified 2 correlated functional variants at 6p21.1 associated with leg lean mass. Our findings not only enhanced our understanding of molecular basis of lean mass development but also provided useful candidate genes for further functional studies.