Since Robert Koch described the cause of tuberculosis in 1882, the natural history of the disease after primary infection has been subject to debate. Only approximately 10% of infected individuals develop active disease, which may appear years to decades after infection. Late onset has been attributed to the endogenous reactivation of dormant bacteria. However, this has not been documented by molecular means for latencies of more than a few years. In Denmark, we have recently recultured 205 freeze-dried Mycobacterium tuberculosis strains obtained from 1961 through 1967. These "historical" strains are analyzed by DNA restriction fragment-length polymorphism testing, and their DNA patterns are compared with those of 4008 recently obtained clinical specimens. This has, surprisingly, yielded molecular evidence of M. tuberculosis reactivation after 33 years of latent infection. A father and son who developed tuberculosis in 1961 and in 1994, respectively, were the only patients infected with strains that share an identical DNA pattern.

SUMMARYA co-operative numerical taxonomic analysis of rapidly growing mycobacteria of Runyon's group IV is reported. There was no limitation on the number, nature, or method of performance of the test characters contributed by each of the 12 participants. Initially 415 test characters were coded for analysis; deletion of irrelevant and repetitious data resulted in a final 195 characters used to generate the matching matrix. All nine major clusters defined in the study were of named species; however, three of these were noted to contain two or more species and the reduction of some of these to synonymy with prior epithets is discussed.

Two hundred three freeze-dried strains of Mycobacterium tuberculosis collected during the 1960s were compared with 4102 strains collected during the 1990s, and 14 DNA patterns identified among the "historical strains" were 100% identical to patterns identified among the "recent strains." They were isolated from 41 and 40 patients who had tuberculosis during the 1960s and 1990s, respectively. The patients' mean age differed by >30 years, a finding strongly suggesting that the patients from the 1990s experienced reactivation of M. tuberculosis infection acquired during the 1960s. The half-life of IS6110 DNA patterns during latency was estimated to be 36 years (95% confidence interval, 25-54 years). Thus, this comparison of historical and recent strains yields molecular epidemiologic evidence of M. tuberculosis reactivation spanning decades and suggests that the rate of change of DNA patterns during latency is much longer than that during active disease. This has important implications for the interpretation of clustering, especially for the extent of recent transmission.

The DNA‐DNA homology percentages obtained in this study indicate that M. avium and M. paratuberculosis belong to one species. Consequently, M. paratuberculosis ought to be considered a variant of M. avium, and the following designations are proposed: Mycobacterium avium, subsp. avium. Mycobacterium avium, subsp. paratuberculosis. Identification and classification of “wood pigeon mycobacteria” occurring in wild animals have been problematic due to their dysgonic and mycobactin‐dependent growth. DNA‐DNA homology percentages indicate that these bacteria are closely related to reference strains both of M. avium and of M. paratuberculosis.“Wood pigeon mycobacteria” should therefore be classified as atypical strains of M. avium, and the following designation is proposed: Mycobacterium avium subsp. columbae.