|Home | About | Journals | Submit | Contact Us | Français|
After isoniazid and rifampin (rifampicin), the next pivotal drug class in Mycobacterium tuberculosis treatment is the fluoroquinolone class. Mutations in resistance-determining regions (RDR) of the rpoB, katG, and gyrA genes occur with frequencies of 97%, 50%, and 85% among M. tuberculosis isolates resistant to rifampin, isoniazid, and fluoroquinolones, respectively. Sequences are highly conserved, and certain mutations correlate well with phenotypic resistance. We developed a pyrosequencing assay to determine M. tuberculosis genotypic resistance to rifampin, isoniazid, and fluoroquinolones. We characterized 102 M. tuberculosis clinical isolates from the Philippines for susceptibility to rifampin, isoniazid, and ofloxacin by using the conventional submerged-disk proportion method and validated our pyrosequencing assay using these isolates. DNA was extracted and amplified by using PCR primers directed toward the RDR of the rpoB, katG, and gyrA genes, and pyrosequencing was performed on the extracts. The M. tuberculosis H37Rv strain (ATCC 25618) was used as the reference strain. The sensitivities and specificities of pyrosequencing were 96.7% and 97.3%, 63.8% and 100%, and 70.0% and 100% for the detection of resistance to rifampin, isoniazid, and ofloxacin, respectively. Pyrosequencing is thus a rapid and accurate method for detecting M. tuberculosis resistance to these three drugs.
Rifampin (rifampicin), isoniazid, and the fluoroquinolones are the most important initial drug markers for extensively drug-resistant Mycobacterium tuberculosis strains, defined as multidrug-resistant (MDR) isolates (resistant to both isoniazid and rifampin) with additional resistance to a fluoroquinolone and to one of the injectable drugs (2). The fluoroquinolones have become an essential part of treatment regimens for MDR tuberculosis (7, 25). Due to their potency and safety, the new-generation fluoroquinolones are now even being evaluated as first-line medications for tuberculosis (3, 13, 20). Wang et al. further suggested that routine fluoroquinolone resistance testing may have a clinical impact by showing a significant correlation between development of fluoroquinolone and first-line M. tuberculosis drug resistance in an area in which resistant strains are highly endemic (28).
The spontaneous acquisition of DNA sequence mutations is the primary genetic basis for the development of M. tuberculosis drug resistance (14). Since sequences are highly conserved, certain mutations correlate well with phenotypic resistance, and a limited number of mutations account for the majority of phenotypic resistance to the important antituberculosis medications, various methods of genotypic testing have successfully been used for the rapid detection of M. tuberculosis resistance (16, 22). The sites that most frequently contain mutations associated with phenotypic resistance, called resistance-determining regions (RDR), differ depending on the drug tested. Among rifampin-resistant isolates worldwide, 95 to 97% harbor mutations in the rifampin RDR, an 81-bp target encompassing codons 507 to 533 of the 3,519-bp rpoB gene (17). Isoniazid resistance has a more complex mechanism, involving several gene targets, the most important of which is codon 315 of the 2,223-bp katG gene, in which mutations are found in up to 50% of resistant isolates (18). Likewise, M. tuberculosis has a quinolone RDR which spans codons 88 to 94 of the 2,517-bp gyrA gene. Mutations in this region, particularly in codon 88, 90, 91, or 94, correlate with high-level resistance and are seen in 42 to 85% of resistant clinical isolates (6).
Pyrosequencing, a method of DNA sequencing by synthesis, has been applied to the rapid detection of M. tuberculosis resistance to rifampin, isoniazid, and ethambutol (9, 29). Its main advantage is a much shorter turnaround time than that of conventional drug susceptibility testing, the latter taking 2 to 4 weeks from the time an isolate is obtained in pure culture.
After isoniazid and rifampin, the next pivotal drug class in M. tuberculosis treatment is the fluoroquinolone class, as previously discussed (3, 7, 13, 20, 25, 28). Given that most resistance to the latter is determined by mutations that are generally limited to the quinolone RDR of the gyrA gene, it should be feasible and clinically more relevant to develop an assay for rapid resistance testing which includes fluoroquinolone resistance in addition to rifampin and isoniazid resistance.
We developed a pyrosequencing assay to determine M. tuberculosis genotypic resistance to rifampin, isoniazid, and fluoroquinolones, which we validated against the conventional submerged-disk proportion method. We also improved on the previously reported pyrosequencing assay by reducing the number of primers required to sequence for rifampin resistance (9).
One hundred two Mycobacterium tuberculosis clinical isolates were selected from the strain collection of the TB Research Laboratory, Department of Medicine, the University of the Philippines-Philippine General Hospital. No more than one strain was obtained from the same patient. All isolates, previously stored at −80°C, were regrown on Lowenstein-Jensen slants for retesting. Reference strain H37Rv (ATCC 25618) was used as a control for the wild-type sequence.
The clinical isolates were tested using the submerged-disk proportion method (10, 11). Commercially available disks (BBL Sensi disk; Becton Dickinson, Cockeysville, MD) impregnated with standardized concentrations of rifampin, isoniazid, and ofloxacin were aseptically placed onto a quadrant of a sterile petri dish. Five milliliters of Middlebrook 7H10 medium supplemented with oleic acid-albumin-dextrose complex (BBL/Difco) was then added to each quadrant, resulting in the following final drug concentrations per quadrant: 0.2 and 1.0 μg/ml isoniazid, 1.0 μg/ml rifampin, and 2.0 μg/ml ofloxacin (15). A drug-free quadrant was also included. The microorganism inoculum was adjusted to a 1 McFarland turbidity standard and then diluted to 10−2 and 10−4 concentrations. The drug-free (control) and disk-containing quadrants were then each inoculated with 0.1 ml of suspension and subsequently incubated at 37°C under 5 to 10% CO2 for 21 days. The numbers of colonies in each quadrant were counted, and resistance was reported if the ratio of the number of colonies in the drug-containing quadrant to the number of colonies in the control was >0.01.
DNA extraction was performed at the Medical Research Laboratory at the University of the Philippines-Philippine General Hospital by using the heat lysis method by suspending a loopful of bacterial colony in TE buffer containing 1% Triton X-100, 0.5% Tween 20, 10 mM Tris-HCl (pH 8.0), and 1 mM EDTA by using a standard 5-μl inoculation loop and heating at 100°C for 15 min in a water bath. The crude lysates were then sent to the Cleveland Clinic for further testing.
PCR was performed using real-time PCR with the 2× Sensimix DNA kit (Quantace, Norwood, MA) on the Rotor-Gene 3000 cycler (Qiagen, Hilden, Germany). The PCR mixture consisted of target DNA (2 μl), 0.20 μM each of forward and reverse primers, 1× SYBR green, and 1× Sensimix in a 25-μl reaction volume. Amplification of the isoniazid RDR was done under the following conditions: incubation at 95°C for 10 min, after which we performed 45 cycles of amplification, consisting of 95°C for 10 s, 58°C for 30 s, and 72°C for 15 s. For amplification of the fluoroquinolone RDR, the reaction was initiated at 95°C for 10 min, followed by 45 cycles of amplification as follows: 95°C for 10 s, 55°C for 30 s, and 72°C for 15 s. The melt protocol consisted of a stepwise melt from 45 to 95°C in 1°C steps, holding for 30 s after the first step and then 5 s after each subsequent step. For both PCRs, the presence of amplified products was indicated by positive quantification curves and a single melt peak.
One set of forward and reverse primers was used to amplify the target region by PCR, with the intention of using two separate primers to sequence the relatively long region of interest within the amplified product. After amplification using the same PCR conditions outlined in the preceding section, the first sequencing reaction worked well, but we encountered difficulties sequencing the second half of the 81-bp RDR (codons 521 to 533) because of multiple background peaks and artifacts. We thus modified our initial PCR conditions by replacing dGTP with dITP in the PCR mixture, which then consisted of the following: target DNA (5 μl of lysate), 0.25 μM each of forward and reverse primers, 1× PCR buffer, 2 mmol/liter MgCl2 (Applied Biosystems, Branchburg, NJ), 0.125 μmol/liter each of dATP, dTTP, dCTP, and dITP-dGTP in a 3:1 ratio (USB Corp., Cleveland, OH, for dITP, and Roche, Mannheim, Germany, for other nucleotides), and 1.5 U Taq polymerase (Applied Biosystems) in a total volume of 50 μl. Amplification was done on GeneAmp PCR system 9600 (Applied Biosystems) under the following conditions: denaturation and enzyme activation at 94°C for 2 min and 50 cycles of 95°C for 20 s, 63°C for 30 s, and 72°C for 30 s, followed by extension at 72°C for 2 min. The PCR product by this modified protocol was used for sequencing of this region.
Pyrosequencing was performed on the 102 clinical isolates and on the ATCC reference strain according to the instructions of the manufacturer (Amersham Biosciences, Fairfield, CT). Briefly, PCR products were immobilized on streptavidin-coated Sepharose beads and served as single-stranded DNA templates. These beads were subsequently transferred to a 96-well plate containing annealing solution and sequencing primer. The areas of interest were sequenced using two sequencing primers for the rpoB gene and one each for the katG and gyrA genes (Table (Table1).1). Pyrosequencing was carried out on an automated PSQ 96MA system (Qiagen, Hilden, Germany) by using the PSQ Gold 96 SQA reagent kit, containing enzyme, substrate, and nucleotides. The reaction cascade primarily consisted of the incorporation of nucleotides into the growing DNA chain, culminating in the production of light. The pattern of emitted light in relation to the nucleotide dispensation order and number of nucleotides incorporated was subsequently illustrated on a pyrogram. The data were then analyzed by the pyrosequencing software. Manual interpretation of the pyrograms was performed for known homopolymer regions (three or more identical nucleotides in succession, e.g., GGG) when the number of nucleotides in the software read differed from the known wild-type sequence. The occurrence of occasional discrepancies with sequencing homopolymer regions has been a recognized issue with pyrosequencing (19).
Susceptibility results based on genotypic testing were compared with those of the submerged-disk proportion method.
Of the 102 isolates tested by the proportion method, 28 were susceptible to all three drugs, 54 were MDR, 1 was resistant to isoniazid and ofloxacin, 5 were resistant only to rifampin, and 14 were resistant only to isoniazid. Sixty-seven of the 69 isoniazid-resistant strains had high-level resistance (MIC ≥ 1). Among the 54 MDR isolates, 9 were also resistant to ofloxacin.
The pyrograms for the first region of the rifampin RDR, as well as the RDR for isoniazid and fluoroquinolones, were read by the software well, typically up to the 36th or 37th bp for the rpoB and gyrA targets. The katG sequences were interpretable by the software until the 20th bp (Fig. (Fig.2).2). For the rpoB and gyrA RDR sequences, even when the software gave a lower grade to the peaks beyond the 30th bp, software interpretation still gave accurate sequences, needing very few manual reads. This was important for the rpoB sequences, for which mutations of interest occurred as far as 37 bp from the sequencing primer.
The addition of dITP to the PCR mixture greatly improved the pyrograms for the second region of the rifampin RDR, with only minor issues encountered. The software read the sequences accurately up to codon 533 for most of the isolates. Thirty-two of the 102 isolates required a manual read to rectify a software misread in the GGGG homopolymer region around codon 523, at which the software misinterpreted the sequence as three Gs instead of four. This had been a recognized issue with pyrosequencing, and the methods included a provision for manual reads when this was encountered. In 28 of the 102 isolates (as well as the ATCC strain), an extra G in codon 532 was read by the software, yielding a sequence of GCGG instead of the expected GCG. This could not be satisfactorily resolved by manual reads. This erroneous reading at this particular site was confirmed to be an artifact by Sanger sequencing of the reference ATCC strain and 2 of these 28 isolates by using the same PCR primers, confirming that the correct sequence was indeed GCG.
Among the 59 isolates that were resistant to rifampin by the proportion method, 57 contained mutations in their rpoB RDR sequences compared to the ATCC strain (wild-type) sequence (Tables (Tables22 and and3).3). One isolate had two point mutations (at codons 512 and 516), while the rest had single point mutations. The most common regions harboring mutations were located in codons 531 (57%) and 526 (27%). Two of 43 isolates that were susceptible to rifampin had a point mutation in codon 533 (CTG to CCG; Leu to Pro). The same amino acid substitution was also observed in the sequence of one phenotypically resistant isolate which contained no other mutations.
Forty-four of 69 isolates phenotypically resistant to isoniazid had mutations in codon 315 of the katG RDR (Tables (Tables22 and and3).3). All the susceptible isolates had a wild-type sequence identical to that of the ATCC strain.
Seven of 10 ofloxacin-resistant isolates carried mutations in codons 90 and 94 of the quinolone RDR, while the remainder had wild-type sequences (Tables (Tables22 and and3).3). None of the 92 ofloxacin-susceptible isolates harbored any mutations. In addition, all 102 clinical strains exhibited the naturally occurring polymorphism (AGC to ACC) in codon 95 which has not been associated with phenotypic resistance.
The primers used in this study allowed successful sequencing of the RDR of the rpoB, katG, and gyrA genes. The rpoB region was successfully sequenced with two sequencing primers as opposed to four in a previously reported pyrosequencing assay (9). However, this required alternate PCR conditions, with partial substitution of dITP for dGTP to allow the GC-rich second half of the rifampin RDR to be successfully sequenced.
Our results correlate well with what is known about the genetic basis of M. tuberculosis drug resistance against rifampin, isoniazid, and the fluoroquinolones. The frequencies of mutations detected in resistant isolates (96.7% for rpoB in rifampin-resistant isolates, 64% for katG in isoniazid-resistant isolates, and 70% for gyrA in ofloxacin-resistant isolates) by our pyrosequencing assay are consistent with other reports in the literature. The observed predominance of mutations in codons 526 and 531 of the rpoB gene also paralleled what has been reported worldwide (14, 17).
The agreement between the presence of a mutation in a “hot-spot” region and phenotypic resistance by the agar proportion method was excellent. There were two rifampin-susceptible isolates, however, for which there was a discrepancy with genotypic results (a leucine-to-proline substitution corresponding to codon 533 of the rpoB gene). The same mutation was also found in one rifampin-resistant strain. This particular point mutation has previously been reported to not correlate well with phenotypic resistance (24). It has been described primarily from Asian strains and is considered by some to be a polymorphism that may serve as a marker for geographic origin (18).
Some obstacles in sequencing the rpoB target had to be overcome for this assay to be successful. We were unable to sequence the second half of the rpoB gene RDR by using the standard PCR conditions. We attributed the difficulty to increased self-priming and formation of secondary structures due to the high GC content. Jureen et al. reported a similar difficulty in their assay and added a poly(T) tail to the 5′ end of the forward PCR primer to resolve this issue (9). We modified our protocol by substituting dITP for dGTP in a 3:1 ratio in the PCR mixture. This modification yielded much cleaner and readable pyrograms. We did have an artifact whereby 28% of our extracts were read as having an extra G nucleotide in codon 532. Fortunately, this codon is not known to harbor resistance mutations. Lastly, we were unable to sequence the region from codon 518 to 521 of the rifampin RDR. This served as the hybridization site for our second sequencing primer, and pyrograms from the first primer could not be reliably read this far out (beyond 33 to 36 bp). However, this is a region in which resistance mutations are uncommon. The sensitivity of our assay remained comparable to that of other genotypic assays for M. tuberculosis drug resistance (1, 5, 26, 27).
All of our isolates had a polymorphism at codon 95 of the gyrA gene. This polymorphism has been reported in 15% of all M. tuberculosis isolates worldwide (23). It is not associated with resistance to the fluoroquinolones. It appears to be most commonly seen in strains from Asia, where the presence of this polymorphism is observed in most of the clinical isolates sequenced (8, 21).
Pyrosequencing has several advantages: greatly improved turnaround time compared to that of the conventional method, ability to perform large-volume testing in its 96-well tray format, and relatively low cost compared to those of molecular tests that utilize multiple probes. This method has been shown to exhibit excellent agreement with the line probe assay and Sanger sequencing in rapidly detecting M. tuberculosis resistance to rifampin compared to the phenotypic radiometric Bactec 460 method as the reference standard (9). In a prospective study by Marttila et al., pyrosequencing provided results of rifampin and isoniazid resistance testing in clinical practice an average of 19 days earlier than the conventional method, with sensitivities of 97.4% and 66.7%, respectively, and no false positives (12). Adding rapid testing for fluoroquinolone resistance would be expected to have an even greater clinical impact, particularly when there is resistance to rifampin and/or isoniazid.
Pyrosequencing, however, still cannot supplant conventional phenotypic methods of resistance testing. Where there are multiple genetic bases for resistance or where all the mechanisms of resistance have not been identified, genotypic methods have limitations. For instance, our assay looks for resistance only in the katG and gyrA RDR for isoniazid and fluoroquinolone resistance, respectively. For these drugs, there remain other important genetic mechanisms for resistance outside the katG and gyrA RDR. Where the genetic basis for resistance has been identified, it should be possible to develop genotypic assays to detect resistance. However, practical considerations limit the numbers of assays that can be run for each sample, and costs and complexities of testing additional sites have to be weighed against the incremental benefits of such testing in clinical practice.
The main advantage of our assay is its rapidity compared to that of conventional testing. This characteristic, coupled with the excellent specificity for detection of resistance to the three most important antituberculosis medications in the majority of cases, makes this assay potentially useful by allowing appropriate treatment several days or weeks sooner than would be possible with conventional testing. It should be most beneficial in areas or populations with some likelihood of resistant infection. Many countries in the former Soviet bloc already have MDR tuberculosis rates of 6.5% or higher, and 55 countries had reported cases of extensively drug-resistant tuberculosis by the end of 2008 (4). A mutation in our assay predicted resistance very well, with the exception of the leucine-to-proline substitution corresponding to codon 533, which may represent a polymorphism not truly indicative of resistance. Thus, our assay would be able to rapidly identify the majority of cases in which rifampin, isoniazid, and fluoroquinolones would not be effective, information that would have a significant clinical impact.
In conclusion, pyrosequencing is a rapid and accurate method for detecting Mycobacterium tuberculosis resistance to rifampin, isoniazid, and ofloxacin. Rapid availability of resistance data will allow appropriate treatment to be established substantially sooner than would be possible with conventional testing.
Published ahead of print on 21 October 2009.