Analytical Performance of the Roche LightCycler® Mycobacterium Detection Kit for the Diagnosis of Clinically Important Mycobacterial Species

Background The LightCycler® Mycobacterium Detection Kit based on real-time PCR technology for the detection of Mycobacterium tuberculosis, Mycobacterium avium and Mycobacterium kansasii was recently developed. This study evaluated its analytical sensitivity, specificity and reproducibility. Methodology/Principal Findings Plasmid standards were prepared and used to determine the limit of detection. The assay was also performed against organisms other than mycobacteria, other mycobacterial strains and interfering substances to exclude cross-reactivity and interference. Reference standards were prepared and tested to assess the assay's reproducibility. All PCR assays were performed using the LightCycler® 2.0 Instrument. The detection limit for M. tuberculosis was 28 copies per microlitre. Neither cross-reactivity nor interference occurred with non-mycobacterial organisms and substances tested. Overall reproducibility for consecutive measurements, run-to-run, lot-to-lot, day-to-day and laboratory-to-laboratory achieved a coefficient of variance of less than two percent. Significance The LightCycler® Mycobacterium Detection kit has shown to be a robust and accurate assay with the potential to be used as a rapid TB diagnostic test.


Introduction
In 2009, 5.8 million cases of tuberculosis were reported to the national tuberculosis control programmes globally where only 57% of the globally notified pulmonary cases were smear positive. Furthermore, the World Health Organization estimates that close to 1.3 million people die as a consequence of the disease. Therefore serious efforts are needed to strengthen the control of tuberculosis [1].
The occurrence of nontuberculous mycobacteria (NTM) from pulmonary samples is occurring with greater frequency driven mainly by the high rates of people living with HIV [2]. The commonest NTM's associated with pulmonary infection are M. avium complex, M. kansasii, M. abscessus and M. fortuitum [2][3][4].
Since tuberculosis is a disease for which the vaccine has failed to provide complete protective immunity [5], its control has relied on early diagnosis and treatment. However, smear microscopy which is the primary tool has a low sensitivity and culture has a very lengthy turnaround time despite its sensitivity [6]. Several nucleic acid amplification tests are available and all aimed at improving the diagnosis of tuberculosis [6][7][8][9]. However, none of these assays have performed adequately to become the new standard for diagnosis. Recently the XpertH MTB/Rif (Cepheid, USA) test, based on real-time polymerase chain reaction (PCR), has received WHO endorsement [10].
Real-time PCR provides great hope as it is a highly sensitive, specific, and rapid molecular method. In addition, an assessment of the dissociation-characteristics of double-stranded DNA during heating known as melting curve analysis can provide species identification [11][12][13]. The LightCyclerH System (Roche Diagnostics, Germany) has been widely used as a platform for real-time PCR. This platform has the advantage of a high-throughput capacity which is important in a high-burden setting.
A new commercial molecular diagnostic assay the LightCyclerH Mycobacterium Detection Kit (Roche Products (Pty) Ltd, Randburg, South Africa), was developed for use on respiratory samples. It is based on the real-time PCR technology for the detection of M. tuberculosis, M. avium and M. kansasii within 90 minutes and amplifies the 16s ribosomal RNA (rRNA) gene which includes the hypervariable region A. The products are detected using the fluorogenic hybridization probes and species identification by melting curve analysis on the LightCyclerH 2.0 (Roche Diagnostics, Germany). A synthetic internal control has been integrated into the reaction. This study evaluated the assay's analytical sensitivity, specificity and reproducibility.

Methods
This study was performed at 4 centres viz. the HELIOS Klinikum E. v. Behring (Germany), Institute of Medical Microbiology Hygiene -University of Regensburg (Germany), Department of Medical Microbiology -University of Pretoria (South Africa) and the National Health Laboratory Service -Kimberley laboratory (South Africa).
These centres were involved in analytically evaluating the LightCyclerH Mycobacterium Detection Kit by preparation of plasmid standards to determine the limit of detection, a challenge panel consisting of other mycobacterial species, non-mycobacterial organisms and interfering substances to assess specificity and mycobacterial standards to measure reproducibility. The assay's reproducibility in terms of consecutive measurements (intra-run), run-to-run, day-to-day, lot-to-lot and laboratory-to-laboratory was performed at the two South African centres only.

DNA extraction
Mycobacterial reference standards and simulated respiratory samples were decontaminated using N-acetyl-L-cysteine-Sodium hydroxide (NaLc-NaOH) [14] and resuspended in 1.5 ml of phosphate buffer containing 0.5% Tween 80 which allowed better homogenization and even distribution of the bacilli within the suspension prior to extraction.
The COBASH Amplicor Respiratory Sample Preparation Kit (Roche Diagnostics, Germany) was used according to the manufacturer's instructions. Briefly, a 100 ml aliquot of the NALC-NaOH decontaminated sample was mixed with the kit's wash solution and centrifuged (14,0006 g for 10 min). After centrifugation, the supernatant was removed and the lysis reagent was added to the pellet. After brief vortexing, the suspension was incubated at 60uC for 45 min to complete lysis of the mycobacteria. Thereafter the neutralizing reagent was added making a total volume of 200 ml neutralized DNA sample. Four microlitres (4 ml) of the neutralized sample was used as template DNA for subsequent PCR testing.

PCR assay
All PCR assays were performed using the LightCyclerH 2.0 (Roche, Germany) real-time instrument. For amplification and detection, the LightCyclerH Mycobacterium Detection Kit (Roche Diagnostics) was used according to the manufacturer's instructions. Samples were amplified as follows: initial denaturation step at 95uC for 10 min to activate the FastStart Taq DNA polymerase and 45 cycles of denaturation at 95uC for 10 sec, annealing at 50uC for 10 sec, and an extension at 72uC for 20 sec.
Following the amplification phase, a melting curve analysis at 40uC to 65uC was performed with a temperature transition rate of 0.1uC/sec to determine the melting temperature (T m ) values for the sequences targeted by the hybridization probes. T m values were either automatically or manually assigned from a plot generated by the instrument. A T m of 53.5-56.5uC indicated M. tuberculosis, 48-51uC M. avium and 57-60uC M. kansasii.
The generation of target amplicons for each sample was monitored between the annealing and the elongation steps at 640 nm. Samples positive for the specific amplicon (an approximate 200 bp fragment of the mycobacterial 16S rDNA) were identified by the instrument at the cycle number where the fluorescence attributable to the target sequences exceeded that measured for background. Samples scored as positive by the instrument were confirmed by visual inspection of the amplification curve (cycle number versus fluorescence value) generated by the instrument.

Preparation of simulated respiratory samples
Samples of bronchoalveolar lavage (BAL) originating from patients without any clinical signs or symptoms of tuberculosis or other mycobacterial infections were used as template material for preparation of the simulated respiratory samples for the reproducibility study. The BAL samples were spiked with the prepared mycobacterial reference standards. Culture of the spiked samples yielded 13 000 CFU/ml for M. tuberculosis, 39 000 CFU/ml for M. avium and 30 000 CFU/ml for M. kansasii. These stock solutions were used to prepare the dilution series.

Analytical Sensitivity (Limit of Detection)
Plasmid standards using a single gene copy were used to determine the Limit of Detection (LOD) of the assay. Plasmid standards for M. tuberculosis were tested in a 6 fold dilution series: 130, 65, 33, 16, 7 and 2 copies per microlitre. All measurements were performed in 20-fold replicates (5 replicates in 4 independent runs on the same LightCyclerH Instrument and the same day). For verification of the measurements the reference standards were run in parallel on the FDA approved CobasH Amplicor Mycobacterium Tuberculosis Test. The effect of interfering substances (table 1) on PCR inhibition was also assessed.

Analytical specificity
In order to check for cross reactivity, a challenge panel was prepared which consisted of other mycobacterial species and nonmycobacterial organisms (table 1); this was tested in duplicate using the LightCyclerH Mycobacterium Detection Kit. The concentrations of the template nucleic acids were determined by photometric measurement and quantitative in-house PCR assays [15]. The species identities of all bacterial and fungal strains were verified by complete 16S rDNA sequencing.

Reproducibility Study
The reproducibility study was designed to measure the variability of the T m and CP measurements of the assay. For this reproducibility study the following standards were used M. tuberculosis at concentrations of 1625 cp/ml, 325 cp/ml, 65 cp/ml and 13 cp/ml, M. avium at 4875 cp/ml, 487.5 cp/ml and 48.75 cp/ ml, and M. kansasii concentrations of 3750 cp/ml, 375 cp/ml and 37.5 cp/ml. In each run, these standards were tested in triplicate for M. tuberculosis and duplicate for both M. avium and M. kansasii. The study was performed at two laboratories; each using a single technician, with a total of 60 runs completed over a 10 day period. The reproducibility of consecutive measurements (intra-run), runto-run, day-to day, lot-to-lot and laboratory-to-laboratory was measured. A coefficient of variance (CV) of less than ten percent was expected.

Statistical analysis
The limit of detection (LOD) was determined by using the Probit method, where, the number of positive responses were converted to a ratio for each of the test concentrations and fitted to a regression model of binary response variables. The LOD was read off the generated graph at the 95% probability for response. For the determination of the measurement precision, the run-to-run, day-today, lot-to-lot and laboratory-to-laboratory variation, a variance component analysis for the crossing points (CP) and melting temperatures (Tm) of M. tuberculosis, M. avium and M. kansasii at different concentrations was carried out. The variance component analysis was performed. All statistical analyses were carried out using the statistical software package JMP 7.0 (SAS/STATH Software).

Analytical Sensitivity (LOD)
Plasmid standards with the target gene were used to determine the limit of detection of the assay. Based on Probit analysis the LOD at the 0.95 (95%) probability for response was calculated at 28 copies per ml [95%CI: 19,54] or 112 copies per reaction [95%CI: 76, 216] ( Figure 1). Furthermore, interfering substances had no effect on PCR performance.

Analytical Specificity
For all non mycobacterial organism tested the specific primer and probe set of the LightCyclerH Mycobacterium Detection Kit showed no cross-reactivity with non-mycobacterial organisms. However, two other mycobacterial species were detected M. flavescens with a T m around 50uC (expected for M. avium) and M. scrofulaceum appeared with a melting point around 58uC (expected for M. kansasii) and no cross-reactivity in the melting temperature range for M. tuberculosis.

Reproducibility
The coefficient of variance for the CP and T m values of M. tuberculosis, M. avium and M. kansasii at different concentrations were calculated. The T m values for the run-to-run, day-to-day and lotto-lot had a coefficient of variance (CV) of less than 2%. The majority of samples demonstrated CV's below 0.4% (Table 2) with a standard deviation of between 0.1-0.2uC. The highest variance for the T m of M. tuberculosis and M. avium occurred in the laboratory-to-laboratory component due to three outliers at the Pretoria site from a total of 760 measurements.
There was a variance in the mean T m of approximately 0.3uC-0.4uC across all the mycobacterial concentrations between the Pretoria and Kimberley sites ( Table 3). The standard deviation for the T m ranged between 0.1-0.2uC and these were independent of sample concentrations. However, in a single measurement from

Discussion
Improved diagnostic tests for tuberculosis are urgently needed and this has been demonstrated in several studies [8,16]. Diagnostics with better sensitivity and the ability to differentiate between M. tuberculosis and non-tuberculous mycobacteria are required. The application of nucleic acid amplification techniques for tuberculosis diagnosis is promising as it can be done directly on patient specimens with corresponding high sensitivity/specificity and a rapid turnaround time. Real-time PCR provides an advanced methodology due to its unique ability to monitor the complete DNA amplification process and identify specific products within the same reaction using fluorescence techniques. The initial high cost of this technology has decreased considerably with the continual improvement in their capability and accuracy [17].
This study showed the LightCyclerH Mycobacterium Detection kit to be a robust and accurate assay. The limit of detection as low as 28 copies per microlitre (112 copies/reaction) is promising when compared with other TB molecular assays although not as sensitive as the XpertH MTB/Rif assay (Cepheid, USA) [18][19][20]. The assay has the potential of being used as a rapid TB diagnostic in place of smear microscopy where resources are available. The lack of sensitivity in comparison to the XpertH MTB/Rif assay may be due to the initial specimen volume of a 100 ml in comparison to the 1.5 ml used by the Cepheid system [20]. Studies to evaluate changes in the input volume to improve test performance are required.
This study was reproducible as it showed good performance at two independent sites an academic laboratory and a routine   The assay did not cross react with any of the nonmycobacterial organisms tested nor was the assay inhibited by the tested interfering substances. All but two of the tested mycobacterial species cross-reacted within the T m range of M. avium and M. kansasii, however, no cross-reactivity occurred within the T m range for M. tuberculosis. The two mycobacterial species detected by the challenge panel were M. scrofulaceum which is usually isolated from lymph nodes and M. flavescens which has been rarely isolated in humans [21]. The LightCyclerH Mycobacterium Detection Kit proved to be specific for M. tuberculosis, M. avium and M. kansasii which are clinically important for respiratory infections especially in TB/HIV co-infection settings. The test can further be improved to detect other mycobacterial species based on their unique melting temperatures. This would have a direct impact on the early and correct management of the affected patients.
The LightCyclerH Mycobacterium Detection kit showed minimal variance when applied using different operators and environments highlighting its robustness. The overall, analytical performance make this kit a candidate tool in the raoid diagnosis of tuberculosis. Clinical evaluations in routine settings are required to validate its performance.