LETTER

Top Letter References

In their recent article (7), McHugh and colleagues assert that Southern hybridization analysis of restricted DNA from nontuberculous mycobacteria confirms their earlier observations (3, 6) and conclusively demonstrates the presence of multiple copies of IS3-like elements with homology to IS6110. Clearly, the preliminary observations presented in their paper do not justify this conclusion, and a more reasonable interpretation would be that their results are due to nonspecific hybridization.

Standard conditions for use of the Amersham ECL nucleic acid detection system require hybridization buffer containing 6 M urea and a temperature no greater than 42°C to avoid inactivation of the horseradish peroxidase linked to the probe. The salt concentration of the buffer must also be carefully adjusted for the specific probe being used. Similar conditions apply for the stringency wash. Details of the hybridization and washing conditions employed by McHugh et al. are not stated, although the temperature used was apparently 50°C in the absence of urea. These conditions are not more stringent than the standard method for restriction fragment length polymorphism (RFLP) analysis of Mycobacterium tuberculosis (8), as the authors claim, and are known to produce high levels of background hybridization. This is apparent since under the conditions they employed both a commonly used probe for RFLP typing of M. tuberculosis and their own 181-bp probe cross-reacted extensively with phage lambda marker DNA. We assume the authors do not believe that phage lambda contains multiple copies of an insertion element with homology to IS6110. With one exception (Fig. 4a, lane 4), the vague hybridization patterns shown in the paper are not representative of DNA fingerprints obtained by most workers using IS6110 and, despite the authors' contention, would not be acceptable for inclusion in any database. From the results presented, one would be forced to conclude that all the nontuberculous mycobacteria tested possess dozens of copies of putative IS3-like elements. Nonspecific hybridization under conditions of low stringency is a more plausible explanation. This would have been apparent to the authors had they also tested appropriate positive and negative controls. These would have included DNA isolated from strains of M. tuberculosis with established copy numbers of IS6110 and from unrelated organisms which are known to possess IS3-like elements.

In our laboratory and in many others throughout the world, the specificity of IS6110-based amplification assays has been amply demonstrated. Indeed, we obtained no false-positive results using DNA isolated from 27 strains of nontuberculous mycobacteria supplied by the authors of the paper under discussion (3, 4, 6). Of these strains, 22 were purported to contain DNA sequences with homology to IS6110, yet despite our challenge to the authors to provide sequence data to support their claims they have repeatedly failed to produce any convincing evidence that significant homology truly exists. A search of the GenBank database with the BLAST algorithm has revealed that the sequence with closest homology to the 181-bp PCR product obtained by McHugh and colleagues is an insertion element from Escherichia coli (5). However, homology is confined to a 157-bp region which shows only 63% identity in nucleic acid sequence with IS6110, while the longest contiguous stretch of homology is limited to a mere 14 bp.

In support of their assertions, McHugh et al. cite a multicenter study conducted in France which examined the specificity and sensitivity of IS6110-based PCR assays for the detection of M. tuberculosis in sputum (2). However, the authors of that study concluded that "false-positive results were reported by only five of the nine participating laboratories, and that no duplicate false-positive result was reported, suggesting that false-positivity might well be a problem of contamination rather than a lack of specificity." Undoubtedly amplification systems for the diagnosis of infectious diseases must be designed with care and undergo exhaustive testing, yet there is now overwhelming evidence that the IS6110-based assays we have developed are both highly specific and sensitive when conducted by appropriately qualified personnel in suitably equipped laboratories (3). Recently, Dalovisio et al. demonstrated that the specificity relative to culture of an IS6110-based PCR assay for detection of M. tuberculosis in respiratory specimens was 99% (1). This figure is higher than that obtained with either the Roche Amplicor system or the GenProbe Amplified Mycobacterium Tuberculosis Direct Test, both of which utilize alternative nucleic acid targets and have been approved for clinical use by the U.S. Food and Drug Administration.

IS6110 has been used by many experienced investigators both for DNA fingerprinting of M. tuberculosis and diagnosis of disease, resulting in a litany of publications. Obviously, new data that challenge existing dogma are welcomed, but studies proclaiming such advancements must be performed carefully with close regard to appropriate controls. The study by McHugh and colleagues does not approach this standard.