The ubiquitous polyomavirus JC virus may be reactivated during prolonged immune suppression from a latent infection in kidney or other organs to a productive infection of oligodendrocytes known as progressive multifocal leukoencephalopathy (PML). Events leading to brain pathology are not known, but may include or result in alterations of the regulatory region. We have compared the promoter~enhancer structure of JC virus isolates from eleven PML brains. The duplications and deletions of the regulatory region were different in each patient, and usually only one sequence was found in each. The sites of strand breakage in the promoter were not random: four or five preferred sites or areas exist. Alignment of the JCV prototype Mad~1 regulatory region with the unduplicated archetypal structure defines six blocks of sequence, A through F. The preferred sites of strand breaks delineate these regions, although Mad~1 is an unusual promoter which contains a break site not observed in other isolates, and an additional site is targeted in several promoters. Region A, containing the TATA box, and the first half of region C, containing several enhancer elements, and region E, are consistently retained. Region B, the 23~bp """"""""insertion"""""""" in the archetypal structure relative to Mad~1, was also retained in all 11 isolates. Region D, the 66~bp """"""""insertion"""""""", was retained in isolates from 3 patients. Regions A and D were never duplicated, whereas regions C and E usually were duplicated or triplicated. Variation in the exact point of breakage within the preferred sites, alternate use of the sites result in sequences which are unique in each case. At the same time, the limited choice of break sites and the characteristic fates of the regions themselves result in three broad patterns of repeat sequences. The patterns do not correspond to the viral genotypes 1 and 2 defined by coding region base changes, and do not appear to be a stable feature of the virus. Rather, rearrangements appear to be generated in the host from a basic archetypal sequence. Kidneys from three of these patients contain JCV genomes with mainly archetypal promoter structures, and the low level of rearranged promoter sequences which were seen had identical sequence to that from the brain. Sequences of the coding regions from these kidneys were identical to those from the brain of the same patient, demonstrating that the same viral genome in different tissues can have two different rearranged forms of the promoter. Type~specific PCR primers were used to show that the kidneys did not contain a low level of a different type virus. Only three out of six kidneys from PML patients were JCV~positive, suggesting that the kidney is not the sole site of latent infection.