We sequenced 640,000 human exomes and estimated associations of rare coding variants with body mass index (BMI). We identified 16 genes for which the burden of rare non-synonymous variants was associated with BMI at the exome-wide level of statistical significance (p<3.6×10-07; GPR75, CALCR, MC4R, GIPR, GPR151 and 11 other genes). This included an association with lower BMI (per-allele beta,-1.8 [95% CI, -2.5, -1.2] kg/m2 of BMI; p=2.6×10-08) and protection against obesity (per-allele odds ratio, 0.46; 95% CI, 0.31, 0.67; p=6.9×10-05) for ultra-rare protein truncating variants in GPR75, a G-protein coupled receptor. The association was due to multiple independent rare variants predicted to truncate GPR75 across the protein, was driven by an association with lower overall body fat mass, was independent of other common or rare alleles at the locus and was additive to that of a genome-wide polygenic score for BMI. Knock-out of Gpr75 in mice resulted in resistance to weight gain in a high-fat diet model, which was allele-dose dependent (25% lower weight gain for Gpr75-/+ and 44% lower weight gain for Gpr75-/- mice compared with wild-type Gpr75+/+ mice) and accompanied by improved metabolic parameters. These results suggest that inhibition of GPR75 may be a novel therapeutic strategy for obesity and illustrate the power of massive-scale human exome sequencing for the identification and validation of drug targets.