© 2005 American Thoracic Society doi: 10.1164/rccm.2506008
Polyclonal Tuberculosis and the Emergence of Drug ResistanceMcGill University Health Centre, Montreal, Quebec, Canada The development of a standardized tuberculosis (TB) strain-typing methodology just over a decade ago permitted an unprecedented opportunity to reevaluate the epidemiology of this ancient scourge (1). Do clinical relapses represent treatment failures, or reinfection? Do transmission chains extend beyond the appreciated outbreak? Do clinic-based definitions of primary versus acquired drug resistance faithfully reflect spread of resistant organisms between hosts as opposed to the development of resistance within a host? With the advent of IS6110-based restriction fragment length polymorphism (RFLP), these and other questions could be resolved on an agarose gel, providing an apparently fresh and unbiased vantage to track how TB spreads. The principle was, and remains, simple: the microbiologic culture from a patient with TB is grown for several weeks in the laboratory, whole genomic DNA is extracted, this DNA is digested and separated, and then it is probed for a transposable element that varies in number and site between strains. The pattern so derived is then used to represent the bacterial population of that patient and this DNA fingerprint used for within-patient and between-patient comparisons. However, all of this work was based on the widely accepted premise that TB is caused by a monoclonal bacterial infection. What if this is not always the case? In most RFLP databases, the presence of a mixed infection was only occasionally evident, suggesting that polyclonal infection was extremely rare (2). Alternatively, if the DNA of a minority bacterial population is insufficient to generate bands on the autoradiograph, the technique might be insensitive to the detection of mixed infections. More recently, the availability of genome sequence information for the tubercle bacillus has facilitated the development of highly sensitive, PCR-based assays that can be tailored to the detection of specific strains. PCR primers can be designed to only generate an amplicon when a mobile element is at a precise insertion point (3) or alternatively to permit amplification only when a specific genomic region has been deleted (4). Such strain-specific PCR assays can be used to interrogate for the presence of minute quantities of particular strains, and also used in parallel to assay for the presence of distinct strains in the same culture, or even the same sputum sample. Using a strain-specific PCR, a group of investigators based in Capetown recently reported in this journal that up to 20% of patients could simultaneously secrete two different bacterial strains in their sputum (3). This somewhat unexpected finding can in part be explained by considering the statistical likelihood of a patient with TB being exposed to a different strain from another patient with TB. In Capetown, the incidence of reinfection with the disease after successful treatment is on the order of 2% per annum (5), providing compelling evidence that TB cases in that community are indeed exposed to multiple Mycobacterium tuberculosis strains. These findings therefore beg certain questions about the interpretation of molecular epidemiologic studies. When serial samples from the same patient produce different RFLP patterns, does this truly indicate reinfection, or could this instead occur with a change of the dominant bacterial population? For patient management, is it possible that persons with both drug-sensitive and drug-resistant strains might present potentially confusing clinical microbiologic results? In the current issue of the Journal (pp. 636–642), van Rie and colleagues consider these questions in a report describing strain-specific typing of patients in whom RFLP patterns suggested reinfection (6). By applying a PCR-based method, the authors demonstrated several patients in whom standard anti-TB therapy was associated with a change from the dominant, susceptible bacterial population to the previously obscure, antibiotic-resistant population. Less frequently, they observed the converse occurrence, in which second-line treatment was associated with a bacterial reversion to the susceptible population. To a certain extent, these data merely represent a novel molecular variant on a familiar theme (7). The basis of multidrug TB chemotherapy is that within a patient, given a sufficient bacterial burden, there is an appreciable risk of the existence of spontaneous drug-resistant mutants of that patient's strain before treatment. Therefore, the distinction in this current report is that the drug-resistant minority population is a different bacterial clone. The practical implication is that unlike spontaneous mutations, which may occur at the frequency of 10–6 to 10–8, the minority population in the current report may occur at 10–2 frequency, providing a greater likelihood for rapid emergence on standard treatment. Fortunately, even in an extremely high transmission setting, the number of such occurrences remained relatively low, with the patients reported in this study representing less than 1% of the clinical cohort. Beyond the quantitative distinction, there may be biological reasons for mixed infections to behave differently. In the classical model, the organisms are otherwise isogenic. Therefore, the susceptible and resistant forms of a strain would compete based on the selective advantage in the face of antibiotics versus the fitness disadvantage of the mutant form. For instance, the absence of M. tuberculosis KatG catalase-peroxidase renders the organisms resistant to isoniazid (8) but susceptible to peroxides generated by the phagocyte NADPH oxidase (9). With mixed infections, the inherent biologic differences between bacterial clones adds complexity to this model. Until recently, genetic differences between M. tuberculosis strains have served more as molecular tags in epidemiology studies than as a basis for functional studies of phenotypic variability (10). This notion has been refuted by the demonstration that unlike most circulating strains of M. tuberculosis, a defined subset of the highly prevalent Beijing strain family produce a phenolic glycolipid that results in a hypervirulence phenotype in mouse and rabbit models (11, 12). From this it follows that the competition of multiple strains within a host will now involve considerations of the particular bacterial strains, based on their relative capacity to survive and thrive in the host environment. If M. tuberculosis strains are phenotypically different and, in the opportune environment, can coinfect the same host, a number of questions about bacterial genetics and evolution present themselves. While antibiotic pressure, as described by van Rie and colleagues, provides a clear opportunity to study the emergence of different bacterial populations, a number of other selective pressures can be envisioned, including different within-host environments, the impact of novel vaccines, and the natural history of superinfection in a host already latently infected (13). Just as the development of RFLP permitted us to re-ask old questions about TB epidemiology a decade ago, the ability to specifically detect the presence of different strains now permits us to revisit a number of questions about the pathogenesis of TB infection and disease. To this end, the work by Warren and colleagues in developing strain-specific genotyping provides an important starting point toward exploring these and other questions, in both experimental infections and natural epidemiologic cohorts. FOOTNOTES Conflict of Interest Statement: M.A.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. REFERENCES
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
