INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Historically tuberculosis (TB) has been and today it remains the single greatest cause of mortality due to an infectious agent, and with the increasing prevalence of TB's resistance to the drugs of choice (4, 30) the problem posed by TB to public health should not be underestimated. This was reflected in 1993 when the World Health Organization declared TB to be a global emergency (22). Estimates that one-third of the world's population is infected with Mycobacterium tuberculosis, the causative organism, leading to 3 million deaths annually, are frequently quoted (8, 10, 22).

In the United Kingdom, notification of TB became mandatory in 1913. Changes in social conditions contributed to a decline in the incidence of TB, and the advent of Mycobacterium bovis BCG vaccination and effective chemotherapeutic agents in the 1950s was accompanied by a 10-fold decrease in notifications between 1948 and 1987 (18). However, this decline ceased in 1987 and from then until 1993 notifications increased by 15% in England and Wales and by 34% in London (18). These increases have been attributed in part to immigration from countries with high incidences of infection (19), although the real threat to the success of national tuberculosis control programs globally, and the single most important risk factor for the development of TB, is coinfection with the human immunodeficiency virus (9). The problem is compounded by the rising incidence of drug resistance and particularly the emergence of multidrug-resistant TB (MDR-TB) (13). The World Health Organization initiated the Global Program in Drug Resistance in 1994 in order to estimate the true global rate of drug resistance in TB (4, 30). Primary MDR-TB was found in every country surveyed except Kenya. In the United Kingdom a surveillance system, MYCOBNET, was created in 1993 to collate drug resistance data from all the reference laboratories in the country. In England and Wales initial MDR-TB rates from 1993 to 1995 increased from 0.6 to 1.5%, with the combined clinical prevalence rate for MDR-TB increasing from 0.6 to 1.7% (2).

The primary aim of prompt diagnosis and treatment of pulmonary TB is to cure the individual, rendering him or her noninfectious and so interrupting the chain of transmission. Quadruple therapy with isoniazid, rifampin, pyrazinamide, and ethambutol is designed to achieve this and to prevent the emergence of MDR-TB (9, 13). In the United Kingdom it is estimated that 90% of rifampin-resistant isolates are also resistant to isoniazid (unpublished data). Rifampin resistance therefore serves as a useful surrogate marker for the detection of multidrug resistance. Furthermore, rifampin resistance means that short-course therapy is no longer an option and that second- and third-line drug susceptibilities are required in order to make an informed choice for alternative therapy (9, 13).

The genetic basis for rifampin resistance in the majority of rifampin-resistant isolates of the M. tuberculosis complex is mutation in the rpoB gene, which codes for the  subunit of the RNA polymerase (14, 26-28). Approximately 95% of rifampin-resistant isolates have a mutation in the 69-bp region corresponding to codons 511 to 533 of the rpoB gene. Current genotypic methods of testing for rifampin resistance rely on detection of mutations in this region. The mechanism of resistance in the remaining 5% of resistant isolates remains undetermined with the exception of further mutations at codons 381 (25), 481 (21), 505 (17), 508 (17), and 509 (21) of the rpoB gene. Molecular assays that have been used to screen the rpoB gene for rifampin resistance mutations include DNA sequencing (14), heteroduplex analysis (31), PCR single-stranded conformational polymorphism (PCR-SSCP) (27, 28), line probe assay (LiPA) (6, 7), and mismatch analysis (21).

Current methods of drug susceptibility testing with primary specimens can take 6 to 8 weeks, although once the specimens have been cultured this time is reduced to 7 to 10 days (12). In this study two genotypic methods, a LiPA (7) (INNO-LiPA Rif.TB; Innogenetics, Zwijndrecht, Belgium) and an RNA-RNA mismatch assay (21) (MisMatch Detect II; Ambion, Austin, Tex.), using cultures of M. tuberculosis and atypical mycobacteria were compared to standard culture methods of identification and susceptibility testing. A third assay, the PhaB assay (32), was applied to rifampin resistance detection in cultures of M. tuberculosis.

The second part of the study was a blinded analysis of consecutive sputum and bronchealveolar lavage specimens (BALs) submitted to the laboratory for molecular testing and isolation of mycobacteria as part of our special Fastrack diagnostic service over a 12-month period. Samples were accepted for analysis only when patients had significant risk factors for MDR-TB or if there was a clinical suspicion of resistance. Results were compared prospectively to standard primary culture results and subsequent biochemical identification and drug susceptibility test results.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Preliminary evaluation of methods. For the initial evaluation of the methods 49 isolates were selected from a bank of stored cultures. These included 16 isolates of M. tuberculosis which were resistant to rifampin by resistance ratio drug susceptibility testing, 15 isolates which were sensitive to rifampin, 15 nontuberculous mycobacteria (NTM), and 3 other members of the M. tuberculosis complex (M. bovis, M. bovis BCG, and M. africanum). All isolates were tested by the LiPA and the mismatch assay. Only M. tuberculosis isolates were tested by the PhaB assay. Susceptibility testing and biochemical identification were repeated by standard methods. To avoid observer bias, results were interpreted by a third party with no knowledge of the species identity or drug susceptibility of the cultured isolates.

Evaluation of assays in real-time diagnostic setting. The LiPA and mismatch assay were applied prospectively to primary specimens and cultures submitted to the laboratory as part of a national molecular testing service. A total of 59 submissions met the minimum scientific criteria (i.e., quality and quantity) for testing. Of these, 38 were smear-positive sputa or BALs and 21 were cultures on Löwenstein-Jensen (LJ) slopes or in BACTEC fluid (n = 1). Results of the molecular assays were returned to the requesting clinician within 72 h of receipt of the specimen. The molecular results were not available to personnel involved in the interpretation of the biochemical identification and drug susceptibility tests. The PhaB assay was not included in this part of the study to avoid compromising diagnosis by splitting the sample among too many tests.

Processing of primary specimens. Sputum and BAL smears were prepared without prior concentration and were stained with auramine-phenol (12) and Ziehl-Neelsen (12) strains. Decontamination was performed by the NaOH-NALC method (5). One milliliter of decontaminated specimen was transferred to a sterile screw-cap microcentrifuge tube for DNA extraction. The remaining fluid was used to inoculate an MB/BacT rapid culture vial (Organon Teknika Corp., Durham, N.C.) and pyruvate- and glycerol-containing LJ slopes.

Preparation of DNA extracts. DNA was extracted from mycobacteria by a simple and rapid method using chloroform to assist in disrupting cells and to precipitate proteins. All steps, including addition of DNA extracts to PCR mixtures, were performed in a class 1 biological safety cabinet under strict category 3 biohazard containment conditions.

Resistance ratio method for drug susceptibility testing, MIC testing by plate microdilutions, and proportional testing using a radiometric system. Drug susceptibility testing by resistance ratio, microdilution MIC, and proportional methods was performed as described by Heifets and Good (12), Telles and Yates (29), and Siddiqi (24).

PCR. The PCR mix was the same for each primer pair used. The 40-µl reaction mixture contained 1 U of BIOTAQ DNA polymerase (Bioline, London, United Kingdom). The supplied reaction buffer was used at the recommended working strength with 1.5 mM MgCl₂. Each reaction mixture contained 0.2 mM (each) dATP, dCTP, and dGTP and 0.1 mM (each) dTTP and dUTP. Each reaction mixture contained 20 pmol of each primer. The primers and cycling parameters for each reaction are described in the appropriate section below.

LiPA. The LiPA is based on the solid-phase reverse hybridization principle. Briefly, target DNA is amplified in a PCR using biotinylated primers which label the PCR product. After gel analysis, PCR products can be denatured and hybridized to membrane-bound capture probes, followed by a color detection step which involves the addition of a conjugate (streptavidin) to which is bound the enzyme alkaline phosphatase. Chromogenic substrates (5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium) are added and are converted into an insoluble purple-blue product by the alkaline phosphatase. Interpretation of the banding pattern allows positive identification of mycobacteria belonging to the TB complex and detection of mutations which would confer resistance to rifampin.

Mismatch analysis. The mismatch assay is based on the ability of double-stranded RNA to withstand digestion with RNase A. Target DNA is amplified by using primers which incorporate T7 RNA polymerase and SP6 RNA polymerase promoters in opposite directions, allowing RNA to be transcribed by using the PCR product as a template. A rifampin-sensitive wild-type strain (H37Ra) is also amplified by using the same primers but with the SP6 and T7 promoters incorporated in the strands complementary to the test strain. The test PCR product and reference PCR product are combined in a transcription reaction using either T7 or SP6 RNA polymerase. The complementary transcripts from the test and reference PCR products are allowed to hybridize, and the resulting hybrids are treated with RNase. Any mutations in the test transcript will not pair with the reference transcript, and so the hybrid will be cleaved at that point. Undigested transcripts and cleavage products can be detected by analysis using agarose gel electrophoresis.

PhaB assay. The PhaB assay was performed as described by Wilson et al. (32). The assay is based upon the ability of viable M. tuberculosis bacilli to support the replication of infecting mycobacteriophage and protect them from inactivation with a phagicidal chemical. After replication in the viable bacilli the bacteriophage are released and detected by mixing with fast-growing M. smegmatis on an agar plate. The bacteriophage in turn infect, replicate within, and lyse the M. smegmatis cells, and the lysis is detected as plaques in a lawn of bacterial growth. The number of plaques is directly related to the number of protected mycobacteriophage, which is in turn related to the number of viable bacilli present after treatment with the drug. Interpretation is based on the proportional method, by comparing the numbers of plaques on plates where the M. tuberculosis bacilli have been incubated with and without drugs.

DNA sequencing. DNA sequencing was performed with ABI PRISM dideoxy chain terminator technology with AmpliTaq DNA polymerase FS on an ABI 373 DNA sequencer. DNA was extracted from cultures and amplified in a PCR containing the outer primer from the LiPA. The PCR product was purified on spin columns (Wizard PCR Preps; Promega, Southampton, United Kingdom) and sequenced with the inner LiPA primers in the forward and reverse orientations. DNA sequencing results were only used to confirm the identity of mutations and were not used as part of the diagnostic service.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Preliminary evaluation of assays. Using the resistance ratio method as the "gold standard," 14 of 15 rifampin-sensitive isolates (93.3%) were scored as sensitive by the LiPA and 15 of 15 sensitive isolates (100%) were scored as sensitive by the PhaB and mismatch assays. A total of 16 of 16 resistant isolates (100%) were scored as resistant (presence of mutation) by the LiPA and the PhaB assay. A total of 15 of 16 resistant isolates (93.8%) were scored as resistant by the mismatch assay.

1

Testing of Fastrack specimens. All Fastrack isolates were correctly scored by the LiPA and mismatch assays applied prospectively6 of 6 resistant isolates, 12 of 12 sensitive isolates, and 3 of 3 NTM isolates. From the primary specimens submitted, 10 patients with rifampin-resistant isolates, 24 with sensitive isolates, and 4 with NTM isolates were subsequently detected based on culture and drug susceptibility testing by the resistance ratio method. The LiPA correctly scored 8 of 10 (80%) of the resistant specimens, 24 of 24 (100%) of the sensitive specimens, and 4 of 4 (100%) of the NTM specimens. The mismatch assay correctly scored 10 of 10 (100%) of the resistant specimens, 24 of 24 (100%) of the sensitive specimens, and 4 of 4 (100%) of the NTM specimens. The overall correlations of the LiPA and mismatch assays, when performed on primary specimens, with subsequent culture-based identification and drug susceptibility assays were 94.7 and 100%, respectively.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

This study was initiated as a result of the growing demand from clinicians for rapid molecular diagnostics for patients for whom clinical details and history suggested the presence of drug-resistant TB, i.e., previous TB, recent immigration from or travel to an area with high prevalence of drug resistant TB, failure to respond to therapy, or contact with a known MDR-TB patient. Overall, the performances of the LiPA and mismatch assay were similar, with both correlating to 97.8% of culture results in the initial evaluation. The in-house phenotypic method, the PhaB assay, when used on a subset of cultures for the rapid detection of rifampin resistance, compared favorably with the two genotypic methods, correlating with culture in 100% of specimens. The PhaB assay was originally developed for use in developing countries and is particularly suited to that environment as a result of its low cost and its relative simplicity of use, features which are not shared by the genotypic assays.

In the initial evaluation of the assays, two isolates (RM6 and RM10) possessing a CTG-to-CCG mutation at codon 511 were identified. With both these isolates the mismatch assay failed to detect the mutation, although the assay as originally described by Nash et al. (21) was capable of identifying this mutation. However, it is known that G-U mismatches are relatively resistant to cleavage with RNase A (20) and it seems likely that RNase digestion conditions are more critical for identification of mutations of this type. These two isolates were the only ones sequenced that possessed a T-to-C mutation.

Although the substitution of proline for leucine at codon 511 has been shown to confer resistance in a number of studies (3, 27), the two isolates in this study that possessed such a substitution gave variable results when tested by different phenotypic methods of drug susceptibility testing. Furthermore, Taniguchi et al. (25) demonstrated sensitivity to rifampin in one of three isolates with a leucine-to-proline substitution at codon 533. It can be postulated that sensitivity to rifampin in an isolate with a known resistance mutation is an artifact of subculture on artificial media. However, subsequent to the completion of this study we received a smear-positive sputum specimen for detection of M. tuberculosis and screening for rifampin resistance mutations. An S1 band deletion was obtained by the LiPA, but the cultured isolate proved to be sensitive to all first-line antituberculous drugs by resistance ratio and proportional drug susceptibility testing in the BACTEC system. It is logical that a mutation would not be selected for in a population if it did not confer some form of selective advantage, and it is therefore reasonable to assume that a leucine-to-proline amino acid substitution at codon 511 does indeed confer some level of resistance to rifampin. The clinical significance of this mutation is not clear and reinforces the importance of a measured approach to rapid methods of diagnosis in which, as in any area of medicine, synthesis of several pieces of evidence is used to manage patients. It is essential that rapid tests always be verified by standard culture-based methods of diagnosis.

The results obtained in this study compare favorably with those obtained by other methods used to detect mutations. Most screening methods use an amplification step, typically PCR, followed by an analysis step. The complexity of the post-PCR procedure can vary dramatically. PCR-SSCP has been performed successfully on cultures (26, 27) but with less success on primary specimens (28). In addition the results of PCR-SSCP can sometimes be hard to interpret (11). Distinctive results can be seen with the use of dideoxy fingerprinting (11), a modification of SSCP, which produces a unique pattern for each mutation. The LiPA provides very clear results, especially for the most common mutations, relying on probe binding and color detection. Similarly the mismatch assay provides clear results although, like SSCP, the actual mutation present is not identified. DNA sequencing is the "gold standard" (27) for mutation detection, because it provides a definitive identification of any mutation present. However, although DNA sequencing is simple for laboratories already performing it routinely, the costs of capital equipment and maintenance do not make it a cost-effective option for most clinical laboratories.

The assays used in this study are relatively simple post-PCR manipulations. The LiPA is the simplest to perform and interpret, requiring only a basic knowledge of molecular biology to perform successfully. In contrast the mismatch assay utilizes a transcription step and subsequent manipulation of RNA, although with proper precautions to avoid RNase contamination, this should not cause a problem.

The cost of an assay is an important factor in its applicability to a diagnostic setting. The LiPA kit costs $720 and provides 20 test strips, while MisMatch Detect II costs $555 and provides enough reagents for 120 tests. Other consumable costs are involved with both assays. Outer primers and the primer for incorporating the transcriptional promoter sites must be purchased for the mismatch assay. Although the LiPA kit includes sufficient inner primer for 30 PCRs, outer primers are required for a nested PCR when the starting material is a primary specimen. In addition insufficient inner primer was supplied for an adequate number of negative control reactions, necessitating the purchase of more inner primers, further increasing the cost per test of the LiPA. It is essential in any PCR that enough negative controls be performed to control for false positives (15), particularly in a nested PCR, to control for carryover contamination during the nesting procedure.

Although many studies specifically do not include samples from patients who are on therapy, the nature of this study meant that such samples would be received. An illustration of this is a particular patient from whom we initially received a culture for identification and susceptibility testing. The patient was not responding to standard first-line therapy, and so a request was made for screening for rifampin resistance mutations. A result was returned within 24 h, saying that the isolate was likely to be sensitive to rifampin. Sensitivity to rifampin was confirmed 2 weeks later by culture when isoniazid resistance was also identified. Therapy was modified accordingly, but when the patient had still failed to respond to therapy 5 weeks later, a sputum sample was sent for molecular testing and rapid culture. On this occasion a common rifampin resistance mutation (S531 to L) was identified 6 weeks before culture identified resistance to rifampin, isoniazid, and streptomycin. Subsequently, it became clear that the patient had not been compliant with therapy.

Of the samples submitted for molecular testing, NTM were present in 7 of 59 specimens (11.9%). Of the remaining samples which were subsequently identified as M. tuberculosis, rifampin resistance was identified in 16 of 52 specimens (30.8%) by culture-based methods. Overall, 40% of the samples submitted contained M. tuberculosis strains which were resistant to rifampin or were NTM strains. Although the NTM most commonly presenting as pulmonary disease will partially respond to therapy with the rifampin and ethambutol components of quadruple therapy for TB (1), such therapy will still be suboptimal. In addition a sputum smear-positive patient with an NTM infection can also be taken out of negative-pressure isolation.

Clearly there is a role for the identification of mycobacteria and detection of rifampin resistance by molecular methods. However, it is essential that the limits of the assays be recognized (16, 23) in order to avoid misdiagnosis through false-positive and -negative results. With limited resources and budgets for purchasing expensive molecular assays, it is important that laboratories focus on those samples from patients with significant risk factors for rifampin-resistant M. tuberculosis. We have found that encouraging clinicians to send high-quality specimens, accompanied by sufficient clinical information, has a key role to play in the effective use of these assays for diagnosis.