T o test the hypothesis that alterations in regulatory regions of the insulin gene occur in a subset of patients with non-insulin-dependent diabetes mellitus (NIDDM), the promoter region was studied by polymerase chain reaction (PCR) amplification directly from genomic DNA, followed by high-resolution polyacrylamide gel electrophoresis under nondenaturing conditions. By using this method a previously identified HincII polymorphism (GTTGAC to GTTGAG at position-56) in American Blacks was readily detected, indicating that single base changes could be observed. In the course of screening the insulin promoter from 40 American Black subjects with NIDDM, an apparent larger allele was found in two individuals. Both patients were shown to have in addition to a normal allele, a larger allele containing an 8-bp repeat, TGGTCTAA from positions -322 to -315 of the insulin promoter. To facilitate rapid screening for the 8-bp repeat, a high-resolution agarose gel electrophoretic analysis was adopted. DNA from American Black NIDDM subjects (n = 100) and nondiabetic subjects (n = 100) was PCR amplified and analyzed. The 8-bp repeat was present in five NIDDM subjects, and one nondiabetic subject. DNA from Mauritius Creoles, also of African ancestry, was analyzed, and the 8-bp repeat was present in 3 of 41 NIDDM subjects, and 0 of 41 nondiabetic subjects. Analysis of glucose metabolism in three presumed normal sibs of an NIDDM patient with an 8-bp repeat revealed that one sib had overt diabetes, and two sibs were glucose intolerant, but there was no consistent segregation of the insulin promoter variant with the diabetes phenotype. The variant promoter was not present in 35 Caucasian NIDDM patients or in 40 Pima Indians. To test the biological consequences of the 8-bp repeat sequence in the insulin promoter, a normal and variant promoter were subcloned into a luciferase plasmid, and reporter gene activity assessed by transient transfection into mouse insulinoma (beta TC1) and hamster insulinoma (HIT) cells. The promoter activity of the variant allele was found to be reduced to 37.9 +/- 10.3% of the activity of the normal promoter in HIT cells (P less than 0.01, n = 4), and 49.1 +/- 6.4% in beta TC1 cells (P less than 0.01, n = 6). These data thus suggest that a naturally occurring variant of the insulin promoter may contribute to the diabetes phenotype in 5-6% of Black NIDDM patients.