Restriction fragment length polymorphism analysis of Korean clinical isolates of Mycobacterium tuberculosis using a 245 bp fragment of IS6110 revealed a conserved 3.5 kb Pvull fragment. Attempts to clone this 3.5 kb fragment resulted in the serendipitous discovery of a novel repeat sequence present within a separate 3.5 kb Pvull genomic fragment. Nucleotide sequencing of a 823 bp region containing the putative repeat sequence revealed the presence of three small direct repeats, three palindromes and a 453 bp region that was analogous to 455 bp of a M. tuberculosis sequence previously reported. The presence of this 453 bp repeat sequence was demonstrated in standard mycobacterial strains belonging to the M. tuberculosis complex, including the H37Rv, H37Ra, Erdman, and Canetti strains and M. bovis and M. bovis bacille Calmette-Guérin (BCG). Other mycobacterial species (M. kansasii, M. smegmatis, M. simiae, M. fortuitum, M. scrofulaceum, M. intracellulare, M. avium, and M. haemophilum) did not contain this sequence, suggesting that the 453 bp repeat sequence was specific to the M. tuberculosis complex. Of the 13 Korean and 12 other clinical isolates of M. tuberculosis tested, all contained three to four copies of the repeat sequence. The Southern blot patterns of the various M. tuberculosis strains allowed classification into five different groups. The most frequent pattern was the 'BCG-type' (4.7, 3.5, and 2.4 kb bands); the second most frequent pattern was the '4-band-type' (13, 4.7, 3.5, and 2.4 kb), observed only in the Korean clinical isolates, and the third most common pattern was the M. tuberculosis H37Rv/H37Ra/M. bovis-type (13, 4.7, and 3.5 kb bands). Upstream sequences indicate proximity to the rhamnose biosynthesis (rfb) cluster of M. tuberculosis. Our results indicate that the repeat sequence may be useful for the design of probe and polymerase chain reaction primers for the identification and epidemiological testing of members of the M. tuberculosis complex.