Mycobacterium tuberculosis and Rifampin Resistance, United Kingdom

A national diagnostic service identified M. tuberculosis and rifampin resistance in primary clinical specimens faster than conventional techniques.

T he increasing incidence of multidrug-resistant tuberculosis (MDRTB), defined as resistance to at least rifampin and isoniazid, is a notable global health problem (1). The rapid identification of patients with MDRTB enables early institution of appropriate treatment, which is associated with improved survival (2,3), and infection control procedures to minimize risk of transmission (4). The Centers for Disease Control and Prevention recommends that the culture/identification and susceptibility testing of Mycobacterium tuberculosis complex (MTBC) be completed within 21 and 30 days of specimen receipt, respectively (5). Molecular assays based on the genetics of drug resistance may considerably reduce these turnaround times.
In the United Kingdom, 82.5% of rifampin-resistant isolates are also resistant to isoniazid (6), making rifampin resistance a useful surrogate marker for MDRTB. Most rifampin-resistant MTBC strains have mutations in an 81bp region of the rpoB gene that encodes the RNA polymerase β subunit (7). This region is therefore an ideal target for molecular tests for rifampin resistance.
The United Kingdom Health Protection Agency Mycobacterium Reference Unit (MRU) offers a national molecular diagnostic service (Fastrack) for detection of MTBC and rifampin resistance (8) by using the INNO-LiPA Rif.TB assay (Innogenetics, Zwijndrecht, Belgium) and supplemented by DNA sequencing as needed. This assay is based on reverse hybridization between rpoB amplicons and membrane-bound capture probes (1 specific for MTBC, 5 overlapping wild-type probes spanning the rpoB target region, and 4 of the most common mutations). Genotypic resistance is indicated by absence of hybridization with wild-type probes or hybridization with resistance mutation probes (9).
A review of the line probe assay (LiPA) (10) found that most previous evaluations focused on mycobacterial isolates and culture-positive (mainly respiratory) specimens (9,(11)(12)(13)(14), but relatively little data exist on nonrespiratory and smear-negative specimens, which are often collected in routine clinical practice (8,15,16). The Fastrack service was initially targeted at smear-positive respiratory samples and mycobacterial isolates, but in response to widespread demand from other laboratories, was extended to all specimens, regardless of acid-fast bacilli (AFB) status. In January 2002, an in-house polymerase chain reaction (PCR) assay targeting the IS6110 insertion sequence (17) replaced LiPA for testing cerebrospinal fluid (CSF) samples. Therefore, CSF samples were not included in this study. This study evaluated LiPA in the context of a nontrial clinical service in one of the largest reported samples of

Mycobacterium tuberculosis
and Rifampin Resistance, United Kingdom 1,997 primary clinical specimens (including 658 nonrespiratory) and 290 clinical isolates tested from 1999 to 2002.

Clinical Specimens
From January 1999 through December 2002, the MRU received 2,287 consecutive non-CSF specimens from 2,110 patients (comprising 1,997 primary clinical specimens and 290 clinical isolates) from 152 centers in the United Kingdom and Ireland for Fastrack analysis. Specimens are submitted for analysis at the discretion of individual referring laboratories, usually when the diagnosis of MTBC is uncertain or when rifampin resistance is suspected. When multiple specimens were received from a single patient, each specimen was processed separately. Of the primary specimens, 1,339 respiratory specimens were sputum, bronchial washings, and bronchial and tracheal aspirates; 658 were nonrespiratory specimens. Samples were received only on weekdays, and routine processing and culture were initiated within 24 hours of receipt. Turnaround times for completion of analysis, culture, and identification of MTBC and drug-susceptibility testing were calculated from date of specimen receipt (5).

Routine Microscopy, Culture, Identification, and Susceptibility Testing
Samples were decontaminated by using the NaOH/Nacetyl-L-cysteine method in a 2-mL suspension, and AFB staining was performed with auramine-phenol and the Ziehl-Neelsen procedure (18,19). DNA was extracted from 1 mL of decontaminated specimen by using a previously described chloroform extraction technique (20), and the remaining 1 mL was added to 1 MB/BacT rapid culture vial (bioMérieux UK Ltd., Basingstoke, UK) and 1 Lowenstein-Jensen slope. Cultures were incubated for at least 8 weeks. Mycobacterial cultures were identified by microscopic and macroscopic appearances, biochemical tests, and DNA hybridization with Accuprobe (GenProbe, San Diego, CA, USA). Drug-susceptibility testing was carried out by the resistance ratio method (18).

LiPA
LiPA was performed according to manufacturer's instructions. The first round of a nested PCR was performed with 10 µL of DNA extract and outer primers (LiPA OP1, 5′-GAGAATTCGGTCGGCGAGCTGATCC-3′ and LiPA OP2, 5′-CGAAGCTTGACCCGCGCGTA-CACC-3′) for 30 cycles at 95°C for 60 s, 58ºC for 30 s, and 72ºC for 90 s. One microliter of first-round product was transferred to a 40-µL PCR mixture containing inner primers (LiPA IP1, 5′-GGTCGGCATGTCGCGGATGG-3′ and LiPA IP2, 5′-GCACGTCGCGGACCTCCAGC-3′), which were biotinylated at the 5′ end, for the second round of amplification for 30 cycles at 95°C for 20 s, 65ºC for 30 s, and 72°C for 30 s. Each PCR run included a duplicate and an inhibition control (100 genome copies of Mycobacterium bovis bacillus Calmette-Guérin [BCG]) for each sample, 5 extracted, water, negative controls, decontaminated, extracted, negative and positive controls (a known culture-positive clinical sample), and a positive control with a low amount of DNA (10 genome copies of BCG in 10 µL). A 260-bp band on agarose gel electrophoresis confirmed successful amplification. The hybridization assay to determine genotypic rifampin resistance was then performed and analyzed as previously described (13). The MTBC result was then reported as positive (accompanied by a rifampin-susceptibility result), negative, equivocal, or inhibited. Results were considered equivocal if a sample tested PCR positive on 1 of 2 duplicates on 2 separate occasions. Extracted DNA was stored for retesting equivocal and inhibited results and for future resolution of discrepant susceptibility results.

Sequencing of rpoB PCR Product
Cultures of MTBC with discordant rifampin-susceptibility results by phenotypic and LiPA testing underwent automated sequencing of the rpoB PCR products with either the Long Read Tower System (Visible Genetics, Suwanee, GA, USA) or the CEQ 8000 Genetic Analysis System (Beckman Coulter, High Wycombe, UK). DNA was extracted from cultures and amplified in a PCR containing the outer primers OP1 and OP2 and sequenced with the inner primers IP1 and IP2.

Statistical Analysis
Data were entered into Microsoft Access (Microsoft Corp., Redmond, WA, USA) and analyzed with Microsoft Excel. Detection of MTBC and rifampin resistance by LiPA was compared with results by the accepted standards of culture and phenotypic susceptibility testing. Concordance, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated. We excluded 85 (3.7%) samples from primary analysis because LiPA results could not be compared with culture results. These samples had equivocal PCR results (n = 27, 1.2%), were inhibitory to PCR (n = 22, 1.0%), could not be cultured (e.g., because of insufficient volume or histologic samples embedded in paraffin wax; n = 6, 0.3%), or were contaminated with bacteria or fungi (n = 30, 1.3%).

Microscopy and Culture
Of the primary specimens tested by LiPA, the AFB smear microscopy was positive in 1,137 (56.9%), negative in 821 (41.1%), and not performed in 39 (2.0%). Specimen types are shown in Tables 1 and 2. Culture identification and drug susceptibility results are shown in Table 3 Tables 1 and 2, respectively, show the results of LiPA in detecting MTBC from primary specimens and rifampin resistance from specimens that grew MTBC. Data on antituberculous treatment were incomplete, but when reported, 195 (9.8%) samples were from patients receiving treatment currently or within the last 3 months. A total of 309 (15.5%) had a history of antituberculous treatment (Tables 5 and 6).

Discrepant Results
There were 136 false-negative MTBC results, i.e., samples negative by LiPA that subsequently yielded MTBC on culture. There were 118 apparently false-positive MTBC results by LiPA, which were PCR positive but did not grow MTBC, although 88 were AFB positive. A total of 83 falsepositive samples were considered to have correct molecular results because they were from patients with a microbiologic diagnosis of MTBC made at MRU from another sample (n = 61) or from patients who were receiving antituberculous treatment currently or had received it within the last 3 months (n = 22). These 83 samples were excluded from statistical analysis to give adjusted values for specificity and PPV (Table 1). Ten specimens were from 6 patients with discrepant results for rifampin susceptibility (Table 7).
Eight specimens had wild-type rpoB, and 2 had mutations not associated with rifampin resistance.
When PCR was compared with culture for detecting MTBC, some false-positive results may, in fact, have been true-positive results. Of 118 samples classified as false positive, 83 were believed to be true positive on the basis of our planned protocol. These consisted of 61 samples from patients with a microbiologic diagnosis of MTBC at our laboratory in the last 18 or subsequent 3 months and an additional 22 samples from patients who were receiving antituberculous treatment or who had received it within the last 3 months. Patients who were receiving treatment currently or within the last 3 months were significantly less likely to have MTBC; of 195 samples from such patients, 70 (35.9%) had MTBC compared to 871 (49.3%) of 1,766 samples from patients with no reported treatment within the last 3 months (χ 2 = 12.7, p<0.001). Furthermore, a significantly higher proportion of rifampin-resistant MTBC was isolated from patients receiving treatment (12/70, 17.1%) compared with patients not reported to be receiving treatment (34/866, 3.9%, χ 2 = 24.2, p<0.001). In these 83 false-positive samples believed to represent true positive results, PCR detected nucleic acid from nonviable organisms (due to treatment) or viable organisms in insufficient numbers for successful culture. If these 83 samples are excluded from overall analysis, specificity improves for all primary specimens, AFB-positive specimens, and AFB-negative specimens from 88.2%, 70.8%, and 95.7%, respectively, to adjusted values of 96.2%, 91.8% and 97.7% (Table 1). PPV improves from 86.9%, 89.5% and 54.7%, respectively, to 95.7%, 97.5% and 70.0%. Other false-positive samples could probably be excluded; we only chose to exclude those with microbiologic diagnoses of MTBC at our laboratory because we had no data on microbiologic, histologic, or clinical diagnoses made by the other hospitals that submitted these samples. Furthermore, since relevant data were often not provided, many more patients likely were receiving antituberculous therapy that we were unaware of because treatment failure is a common reason for specimens being submitted for testing.
PCR-equivocal results were excluded from the primary analysis. However, a PCR-equivocal result may represent a lack of sensitivity. If PCR-equivocal results are considered PCR negative, the adjusted values for detecting MTBC in primary specimens were only marginally altered to 90.6%, 84.0%, 96.2%, 95.7%, and 85.7%, respectively, for concordance, sensitivity, specificity, PPV, and NPV.
A recent review of LiPA results reported that although little data on clinical specimens were available, sensitivity appeared lower than that of isolates (10). Our study confirmed this finding, with sensitivities of 85.2% for all clinical specimens and 98.7% for isolates. As with other PCR-based tests (21)(22)(23), sensitivities of LiPA for AFBnegative (29.4%) and nonrespiratory samples (61.1%) were low. Sensitivity was also reduced to 71.6% in patients receiving treatment at the time or within 3 months of the time the sample was obtained. Marttila et al. tested 75 clinical specimens with LiPA, including 66 from nonrespiratory sites, and reported a sensitivity of 58.8% compared with final clinical and pathologic diagnoses, whereas cultures showed a sensitivity of 35.3% (15). Several factors may explain the lower sensitivity of PCR-based methods in these samples. The mycobacterial load is lower, as demonstrated by the significantly shorter time taken to culture MTBC for AFB-positive samples than for AFB-negative samples (18.5 days vs. 29.5 days, z = 8.0, p<0.001), and respiratory samples than nonrespiratory samples (18.7 days vs. 25.0 days, z = 5.6, p<0.001). However, more respiratory samples were AFB positive (94.3% vs. 55.0%). Irregular clumping may take place within paucibacillary specimens, and small, suboptimal sample volumes often lead to sampling errors. Nonrespiratory specimens, especially pleural fluid, bone marrow, pus, and tissue biopsy specimens, may contain inhibitors of amplification (22,23). Inhibition rates in this study were 1.0% overall, with above-average rates in blood and feces (both 2/3, The nonrespiratory specimen types with the highest sensitivity rates were vertebral aspiratesbiopsy specimens (n = 30, sensitivity 83.3%), gastric aspirates (n = 18, sensitivity 80.0%), and lymph node aspirates/biopsy specimens (n = 144, sensitivity 72.5%). For pleural fluid, one of the most commonly submitted samples (n = 107), LiPA had one of the lowest sensitivity rates (21.7%) for detecting MTBC. The difficulties in detecting MTBC in pleural fluid are well recognized, with previous reported sensitivities of 20% (24) and 50% (23) with the Gen-Probe amplified M. tuberculosis direct test.
Detection of rifampin resistance by LiPA may be used as an early predictor of MDRTB before phenotypic susceptibilities are available, but this clearly depends on the prevalence of rifampin monoresistance in the study population. The diagnosis of rifampin monoresistance is also critical because this automatically invalidates the use of short-course chemotherapy (25). Of the 59 correctly identified rifampin-resistant MTBC isolates, 11 were rifampin monoresistant. The overall prevalence was 1.0% in this study, which was higher than the 0.3% reported in a national UK survey (6). This result reflects a common underlying reason for specimen referral for Fastrack analysis, i.e., failure of response to treatment.
For primary samples in which LiPA detected MTBC, diagnosis of tuberculosis was made an average of 15.2 days earlier than with automated liquid culture (14.8 days for AFB-positive specimens and 22.1 days for AFB-negative specimens). More days were saved with nonrespiratory samples (18.3 days) than with respiratory samples (14.7 days), although these samples had the lowest probability of detection. LiPA accurately determined rifampin susceptibility earlier than solid culture-based techniques by a mean of 30.7 days for all primary specimens. This compares favorably with a study that found that LiPA saved a median of 24 days compared with susceptibility testing with the BACTEC liquid culture system (Becton Dickinson, Sparks, MD, USA) and 54 days with solid media (11).
In summary, LiPA may be used with clinical samples for diagnosis of MTBC and rifampin resistance, saving, when positive results are obtained, an average of 15.2 days and 30.7 days, respectively, compared with conventional techniques. However, some limitations of LiPA are evident. As with other PCR-based assays, sensitivity is reduced in AFB-negative and nonrespiratory samples, such as paucibacillary forms of the disease, in which rapid diagnosis would be most helpful. Although the assay is a potential diagnostic route for patients receiving therapy, sensitivity is also reduced in these circumstances. The lower sensitivity rates for certain samples and the possibility of a PCR-equivocal or PCR-inhibited result also mean that conventional culture and sensitivity testing should still be used at the same time. Alternatives to LiPA may be useful, e.g., we used an IS6110-based PCR for diagnosis of tuberculous meningitis. Similarly, rifampin-resistance mutations can be detected by DNA sequencing (we now sequence all PCR products identified as MTBC with any form of rifampin probe mutations) or with noncommercial macroarrays (26,27). Thus, molecular results, as with any laboratory test, should be reviewed in the context of all clinical, microbiologic, and histologic results.