Lithium is an effective mood stabilizer for bipolar disorder patients and its therapeutic effect may involve inhibition of inositol monophosphatase activity. In humans, the enzyme is encoded by two genes, IMPA1 and IMPA2. IMPA2 maps to 18p11.2, a genomic interval for which evidence of linkage to bipolar disorder has been supported by several reports. We performed a genetic association study in Japanese cohorts (496 patients with bipolar disorder and 543 control subjects). Interestingly, we observed association of IMPA2 promoter single nucleotide polymorphisms (SNPs) (-461C and -207T) with bipolar disorder, the identical SNPs reported previously in a different population. In vitro promoter assay and genetic haplotype analysis showed that the combination of (-461C)-(-207T)-(-185A) drove enhanced transcription and the haplotypes containing (-461C)-(-207T)-(-185A) contributed to risk for bipolar disorder. Expression study on post-mortem brains revealed increased transcription from the IMPA2 allele that harbored (-461C)-(-207T)-(-185A) in the frontal cortex of bipolar disorder patients. The examination of allele-specific expressions in post-mortem brains did not support genomic imprinting of IMPA2, which was suggested nearby genomic locus. Contrasting to a prior report, therapeutic concentrations of lithium could not suppress the transcription of IMPA2 mRNA, and the mood-stabilizing effect of lithium is, if IMPA2 was one of the targets of lithium, deemed to be generated via inhibition of enzymatic reaction rather than transcriptional suppression. In conclusion, the present study suggests that a promoter haplotype of IMPA2 possibly contributes to risk for bipolar disorder by elevating IMPA2 levels in the brain, albeit the genetic effect varies among populations.