Detection of Mycobacterium tuberculosis pncA Mutations by Genoscolar pza tb ii.pdf

SUSCEPTIBILITY...

SUSCEPTIBILITY crossm Detection of Mycobacterium tuberculosis pncA Mutations by the Nipro Genoscholar PZA-TB II Assay Compared to Conventional Sequencing Melisa J. Willby,a Maria Wijkander,b Joshua Havumaki,c Kartee Johnson,a* Jim Werngren,b Sven Hoffner,b Claudia M. Denkinger,c James E. Poseya a Centers for Disease Control and Prevention, Atlanta, Georgia, USA b The Public Health Agency of Sweden, Department of Microbiology, Unit of Highly Pathogenic Bacteria, Solna, Sweden c FIND, Geneva, Switzerland ABSTRACT Pyrazinamide (PZA) is a standard component of first-line treatment regimens for Mycobacterium tuberculosis and is included in treatment regimens for drug-resistant M. tuberculosis whenever possible. Therefore, it is imperative that susceptibility to PZA be assessed reliably prior to the initiation of therapy. Currently available growth-based PZA susceptibility tests are time-consuming, and results can be inconsistent. Molecular tests have been developed for most first-line antituberculosis drugs; however, a commercial molecular test is not yet available for rapid detection of PZA resistance. Recently, a line probe assay, the Nipro Genoscholar PZA-TB II assay, was developed for the detection of mutations within the pncA gene, including the promoter region, that are likely to lead to PZA resistance. The sensitivity and specificity of this assay were evaluated by two independent laboratories, using a combined total of 249 strains with muta- tions in pncA or its promoter and 21 strains with wild-type pncA. Overall, the as- say showed good sensitivity (93.2% [95% confidence interval, 89.3 to 95.8%]) and moderate specificity (91.2% [95% confidence interval, 77.0 to 97.0%]) for the identification of M. tuberculosis strains predicted to be resistant to PZA on the basis of the presence of mutations (excluding known PZA-susceptible mutations) in the pncA coding region or promoter. The assay shows promise for the molec- ular prediction of PZA resistance. KEYWORDS line probe assay, pncA, pyrazinamide, tuberculosis Received 7 September 2017 Returned for modification 25 September 2017 Accepted 23 October 2017 R apid diagnosis of tuberculosis (TB) and timely administration of effective treatment are critical for successful control of this disease. TB treatment regimens typically include a combination of at least four different antibiotics, with the drugs selected Accepted manuscript posted online 30 October 2017 Citation Willby MJ, Wijkander M, Havumaki J, depending on a number of factors, including susceptibility, tolerance, and availability Johnson K, Werngren J, Hoffner S, Denkinger CM, Posey JE. 2018. Detection of (1). Growth-based tests are considered the gold standard for evaluation of Mycobacte- Mycobacterium tuberculosis pncA mutations by rium tuberculosis drug susceptibility, although, due to the slow growth of M. tubercu- the Nipro Genoscholar PZA-TB II assay losis, these tests can take weeks to months to complete. Pyrazinamide (PZA), an compared to conventional sequencing. Antimicrob Agents Chemother 62:e01871-17. important antibiotic for treating both drug-susceptible and drug-resistant TB, is a https://doi.org/10.1128/AAC.01871-17. prodrug converted to its active form through the action of the mycobacterial enzyme Copyright © 2017 American Society for pyrazinamidase (PZase). PZase is active under acidic conditions; thus, growth-based Microbiology. All Rights Reserved. PZA susceptibility testing must be performed under similar conditions. This environ- Address correspondence to James E. Posey, ment is not optimal for M. tuberculosis growth, however, which may lead to false [email protected]. susceptibility results (2–4), while overinoculation of the organism in an attempt to * Present address: Kartee Johnson, Georgia Department of Public Health, Atlanta, counteract this inhibition of growth may lead to false resistance results (5). Importantly, Georgia, USA. PZA is a valuable component of new regimens under investigation (6–8). Thus, accurate January 2018 Volume 62 Issue 1 e01871-17 Antimicrobial Agents and Chemotherapy aac.asm.org 1 Willby et al. Antimicrobial Agents and Chemotherapy and timely evaluation of PZA susceptibility is crucial not only for ensuring the effec- tiveness of current regimens but also for designing clinical trials to assess novel regimens. Molecular diagnostic approaches are increasingly being used for prediction of drug resistance. These tests can provide rapid, accurate, and sensitive results when the molecular mechanism of drug resistance is known. An estimated 83 to 93% of PZA resistance is associated with mutations in pncA, the gene encoding PZase, including its promoter (9–12). However, unlike drug resistance scenarios in which resistance is attributed to one or a few specific mutations within a small genomic region, mutations in the entire pncA open reading frame and its promoter region are associated with PZA resistance (10, 13). Therefore, an effective molecular diagnostic assay for PZA resistance would need to interrogate the entire length of pncA (561 bp) and its upstream regulatory region, instead of a few specific mutations, making real-time PCR-based molecular diagnostic approaches impractical. While DNA sequencing could adequately interrogate pncA for mutations, sequencing is expensive and requires a high degree of technical expertise and therefore may be difficult to implement in many settings. A line probe assay (LPA) could be an effective option for the rapid identification of pncA mutations, as demonstrated previously (14). The World Health Organization (WHO) endorsed the use of commercial LPAs for identification of the M. tuberculosis complex and detection of mutations associated with resistance to rifampin, isoniazid, and some second-line antituberculosis drugs in smear-positive specimens and strains (15). By 2014, 92 countries and territories had the capacity to perform LPAs, including 13 of the 27 countries with high burdens of multidrug-resistant (MDR) TB (16). In 2015, two new LPAs were evaluated for detection of MDR-TB, with favorable outcomes, underscoring the potential of this technology (17). Capitalizing on the current LPA infrastructure and previous experience with the Nipro NTM⫹MDRTB detection kit 2, the Nipro Corpora- tion (Osaka, Japan) developed the Genoscholar PZA-TB II assay, a LPA that focuses solely on the detection of PZA resistance and includes probes encompassing the entire pncA coding region and 18 nucleotides upstream. In this report, we present an evaluation of the Genoscholar PZA-TB II assay by two independent laboratories to assess the performance of this LPA in detecting pncA mutations using a panel of M. tuberculosis strains harboring a variety of mutations within pncA and its promoter as identified by Sanger sequencing. RESULTS Analysis by strain. The ability of this LPA to identify strains with pncA mutations and therefore likely PZA resistance was assessed using Sanger sequencing as the reference standard. For these analyses, we used the term wild type (WT) to indicate all strains that would be expected to be susceptible to PZA on the basis of Sanger sequencing results (n ⫽ 34), including strains with no mutations within pncA, only a synonymous single-nucleotide polymorphism (sSNP), or a nonsynonymous single- nucleotide polymorphism (nSNP) that has been shown not to confer resistance (18). Strains with all other nSNPs or insertions or deletions (indels) or heteroresistant strains with a combination of WT and single-nucleotide polymorphism (SNP)-containing pncA alleles (n ⫽ 231) were considered mutant and likely resistant to PZA. Results were divided into four categories, as follows. A true-negative (TN) result occurred when all probes were present for a strain determined to be WT by Sanger sequencing (using the definition of WT presented above), a false-positive (FP) result occurred when a probe was absent for a WT strain, a true-positive (TP) result occurred when any probe was absent for a mutant strain, and a false-negative (FN) result occurred when all probes were present for a mutant strain. Any strain with discordance between genotype and LPA results was subjected to repeat Sanger sequencing to confirm the genotype. If discordance persisted, then the LPA was repeated, and the second LPA result was used in the data analysis. The overall sensitivity and specificity for accurate identification of strains with a mutant pncA genotype were 93.2% and 91.2%, respectively; the values for the Centers January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 2 Evaluation of the Nipro Genoscholar PZA-TB II Assay Antimicrobial Agents and Chemotherapy TABLE 1 Analysis by strain No. of strains Sensitivity Specificity Site TP result TN result FP result FN result (95% CI) (%) (95% CI) (%) Overall 220 31 3 16 93.2 (89.3–95.8) 91.2 (77.0–97.0) CDC 117 14 3 10 92.1 (86.1–95.7) 82.4 (59.0–93.8) PHAS 103 17 0 6 94.5 (88.5–97.5) 100 (81.6–100) for Disease Control and Prevention [CDC] strains were 92.1 and 82.4%, respectively, and those for the Public Health Agency of Sweden [PHAS] strains were 94.5 and 100% (Table 1). Three FP results were obtained for strains tested at the CDC, due to the presence of nSNPs (Asp110Gly, Val163Ala, and Ala170Val) previously shown to be susceptible to PZA but not accounted for by probes in the LPA. A total of 16 FN results were reported (10 for CDC strains and 6 for PHAS strains), including 2 strains with nSNPs (Glu37Val and Val155Gly), 5 strains with mixtures of WT and mutant pncA alleles (as identified by Sanger sequencing), 6 strains with insertions within the pncA coding sequence, 2 strains harboring indels within the promoter region but upstream of the binding site for probe 1, and 1 strain with a deletion within the pncA coding sequence and a second deletion in the promoter region upstream of the binding site for probe 1. Analysis by probe. Results were also evaluated on a by-probe basis. Since 270 strains were tested, each probe was effectively interrogated 270 times. Between 1 and 18 strains (median, 5.5 strains) from our test pool contained a mutation within the region interrogated by each probe. Test results were divided into four categories, similar to the categories described above. The results for each probe are presented in Table 2. The collective sensitivity of all probes from all sites was 84.9%, while the sensitivity of individual probes ranged from 0 to 100% (average, 81.9%; median, 88.9%). The collective specificity of all probes from all sites was 99.9%, with the specificity of individual probes ranging from 98.9 to 100% (average, 99.9%; median, 100%). Sensi- tivity and specificity values were similar between the two test laboratories (data not shown). Sixteen silent mutations were correctly identified with no loss of band color- ization, including 12 Ser65Ser substitutions and 1 each of Leu35Leu, Ala38Ala, Ala39Ala, and Thr142Thr substitutions. One nSNP (Glu37Val) that does not result in PZA resis- tance was correctly identified as WT; however, the probe (probe 11) covering this sequence had 0% sensitivity in spite of the presence of four isolates with nSNPs within its binding region. Thus, the correct identification of this susceptible nSNP and sSNPs (Leu35Leu, Ala38Ala, and Ala39Ala) within the binding area for this probe could have been a coincidence associated with the low overall sensitivity of the probe. A total of 50 FN results and 18 FP results were reported in the by-probe analysis. One limitation of this study is the small number of mutations included in the study set for certain probes. If we consider only probes for which the study set included at least 5 mutations in the area covered by the probe (n ⫽ 26), then the overall sensitivity was 86.3% and values ranged from 37.5 to 100% (average, 85.0%; median, 86.6%), while the overall specificity was 99.9% (average, 99.9%; median, 100%) on a by-probe basis. DISCUSSION M. tuberculosis resistance to PZA is attributed in large part to mutations within pncA, the gene encoding the enzyme that converts PZA to its active form. In this study, the Nipro Genoscholar PZA-TB II assay was used to evaluate 270 M. tuberculosis strains for mutations within pncA and the 18 bases upstream of its start codon. Since the intent of this assay is to predict resistance to PZA based on detection of changes in the pncA sequence, as illustrated by the absence of colorization of one or more probes on a nitrocellulose strip, we expanded our definition of WT to include not only strains with unaltered pncA alleles but also those with sSNPs and PZA-susceptible nSNPs, based on Sanger sequencing (18), since all such alleles would result in PZA-susceptible strains. Conversely, we considered mutant the strains with all other nSNPs or indels in pncA or its promoter and heteroresistant strains with these mutations in combination with a WT January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 3 Willby et al. Antimicrobial Agents and Chemotherapy TABLE 2 Analysis by probe No. of strains Amino acids bound by TP TN FP FN Sensitivity Specificity Probe no. probe (nucleotides) result result result result (%) (%) 1 (⫺18)–1 8 261 0 1 88.9 100.0 2 1–5 7 263 0 0 100.0 100.0 3 4–9 14 255 0 1 93.3 100.0 4 7–11 9 261 0 0 100.0 100.0 5 11–16 9 259 0 2 81.8 100.0 6 14–18 4 263 2 1 80.0 99.2 7 17–21 6 264 0 0 100.0 100.0 8 22–26 3 266 1 0 100.0 99.6 9 27–31 3 267 0 0 100.0 100.0 10 29–34 1 266 3 0 100.0 98.9 11 34–41 0 265 1 4 0.0 99.6 12 41–46 2 265 2 1 66.7 99.3 13 47–52 16 251 1 2 88.9 99.6 14 49–53 14 255 1 0 100.0 99.6 15 52–56 2 267 0 1 66.7 100.0 16 56–61 16 253 1 0 100.0 99.6 17 61–66 9 258 1 2 81.8 99.6 18 67–71 7 261 0 2 77.8 100.0 19 70–74 1 269 0 0 100.0 100.0 20 73–78 3 267 0 0 100.0 100.0 21 77–81 2 267 0 1 66.7 100.0 22 80–86 4 266 0 0 100.0 100.0 23 85–90 1 267 0 2 33.3 100.0 24 91–95 7 262 0 1 87.5 100.0 25 95–100 18 252 0 0 100.0 100.0 26 100–105 5 264 0 1 83.3 100.0 27 106–111 3 267 0 0 100.0 100.0 28 111–116 1 268 0 1 50.0 100.0 29 114–119 4 265 1 0 100.0 99.6 30 120–124 5 264 0 1 83.3 100.0 31 124–129 5 265 0 0 100.0 100.0 32 129–134 11 256 0 3 78.6 100.0 33 133–138 9 258 3 0 100.0 98.9 34 137–142 16 252 0 2 88.9 100.0 35 142–146 6 261 0 3 66.7 100.0 36 145–150 8 261 0 1 88.9 100.0 37 149–154 6 262 1 1 85.7 99.6 38 154–159 6 262 0 2 75.0 100.0 39 156–161 2 267 0 1 66.7 100.0 40 161–165 5 265 0 0 100.0 100.0 41 163–167 4 265 0 1 80.0 100.0 42 166–170 4 266 0 0 100.0 100.0 43 168–172 6 260 0 4 60.0 100.0 44 170–175 3 262 0 5 37.5 100.0 45 172–177 5 263 0 2 71.4 100.0 46 177–182 0 269 0 1 0.0 100.0 47 180–185 1 269 0 0 100.0 100.0 48 185 (568) 1 269 0 0 100.0 100.0 Overall 84.9 99.9 pncA allele. Overall, using these definitions, the sensitivity and specificity for identifi- cation of pncA mutant strains were 93.2% and 91.2%, respectively. Sensitivity values were comparable for the two individual test laboratories (CDC, 92.1%; PHAS, 94.5%). Of note, 5 of 16 false-negative strains were identified by Sanger sequencing as heterore- sistant, containing a mixture of WT and mutant pncA alleles. Strains with a mixture of mutant and WT pncA alleles generally have been shown to be resistant to PZA; therefore, heteroresistant strains were considered mutant (19). Other false-negative strains (n ⫽ 7) contained insertions and, although an insertion may disrupt the translation of the gene or alter the gene product and lead to resistance, it may not be identified by this assay since the WT sequence is still present. The CDC sample set January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 4 Evaluation of the Nipro Genoscholar PZA-TB II Assay Antimicrobial Agents and Chemotherapy included three susceptible nSNPs that were not specifically accounted for in the LPA and therefore resulted in the absence of the WT band, leading to a false-positive interpretation and a decrease in specificity for that test site (CDC, 82.4%; PHAS, 100%). We considered it important to include these susceptible SNPs in our WT definition, since this assay is intended to serve as a means to predict resistance and susceptibility. The prevalence of these or other susceptible SNPs within a test population would affect the specificity of the assay for that population. Our results are comparable to those from a recent evaluation of this assay, with a smaller and less diverse set of isolates from Bangladesh and Pakistan, by Driesen et al. (19). In that study, LPA results were compared to a composite reference based on a combination of Illumina and Sanger sequencing results and MGIT test results. The authors concluded that there was 94.3% agreement between LPA results and the composite reference. Furthermore, Driesen et al. reported false-negative results for heteroresistant isolates, as seen in the current study (19). Since the pncA sequences of the test collection strains had been determined previously, we were able to assess the sensitivity and specificity of each probe indi- vidually. Because the sites tested strains from their own collections, there was some variability in the strains assayed for each probe among the sites. We found that overall the probes were moderately sensitive (84.9%) but highly specific (99.9%). Again, these values were very similar for the two test sites. Fourteen probes had 100% sensitivity and specificity. Thirty-three probes had sensitivity and specificity of ⱖ80%. Probe 11 (covering amino acids 34 to 41) and probe 46 (covering amino acids 177 to 182) had sensitivity of 0%, given that there were no TP results for these probes despite the presence of strains in the test collections with mutations in these regions (probe 11, 4 strains; probe 46, 1 strain). Of note, the strain missed by probe 46 contained a mutation in amino acid 182, which is the very last amino acid covered by probe 46 and is also covered by probe 47. Redesign of probes for these regions could improve the sensi- tivity of this assay. There was a combined total of 68 false-positive and false-negative results. Twenty-two false-negative results occurred when the mutation was in either the first or last codon covered by a probe, suggesting that some probes may not detect mutations in these positions efficiently. Fortunately, in most cases, probes overlap one another, with nucleotides toward the beginning or end of one probe also being present on an adjacent probe. This was the case for all 22 of the false-negative results described above and, consequently, none of those strains was incorrectly categorized with a FN result when data were analyzed at the strain level. Other sources of false-negative results included the presence of an insertion or deletion that did not disrupt probe binding (n ⫽ 13) or a pncA mutation in combination with the WT sequence (n ⫽ 5). Some false-positive results occurred when either the probe adjacent to the expected probe was recorded as missing instead of the expected probe (n ⫽ 3) or when the probe adjacent to the expected probe was recorded as missing along with the expected probe (n ⫽ 6). Importantly, in only 14 instances in which probe results were discordant with the sequencing results was the corresponding strain incorrectly iden- tified as susceptible; 7 of those 14 strains were strains with PZA heteroresistance, 6 were strains with indels, and 1 strain had a Val155Gly change. This study had several limitations. Only 21 strains with WT pncA sequences were included, although 13 additional strains with either silent mutations or susceptible nSNPs were considered WT for this study; therefore, the by-strain specificity values had wide confidence intervals (CIs). Since clinical strains contain such a variety of pncA mutations, the numbers of strains tested for each individual mutation were limited. While the overall test set contained at least one strain with a mutation in the binding site for each probe, the CDC test collection lacked strains with mutations corresponding to probe 48 and the PHAS collection lacked strains with mutations in 10 probe binding sites (probes 8, 10, 19, 27, 28, 39, 40, 41, 46, and 47). Several additional factors contributed to the imperfect performance of the LPA in both by-strain and by-probe analyses. For example, samples containing mutations located more than 18 nucleotides upstream of the pncA start codon were identified as January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 5 Willby et al. Antimicrobial Agents and Chemotherapy WT by the LPA, since probe 1 extended only 18 nucleotides upstream of pncA. Also, the test sample set included six strains that were identified by Sanger sequencing as containing a combination of WT and mutant sequences. These samples were included in the test set since heteroresistance is a real-world phenomenon. The presence of the WT sequence in a sample in combination with a mutant sequence may mask identifi- cation of mutations using this assay, and five such samples were identified as WT by the LPA, although they would likely be resistant to PZA; therefore, these were considered FN results in these analyses. Another factor contributing to discrepancies was the presence of some insertions within pncA that did not prevent the binding of probes, since the WT sequence was still present, but would be predicted to affect susceptibility. Furthermore, a recently published study described 10 pncA mutations, including 2 silent mutations and 8 nSNPs, that do not confer resistance (18). Our study set included 3 of those susceptible nSNPs and all of them were incorrectly identified as mutant, significantly affecting our by-strain specificity. Thus, the genetic diversity of locally circulating strains within a particular region, including the prevalence of strains with indels, susceptible nSNPs, or heteroresistance, could significantly affect the sensitivity and specificity of this assay, and local validation of the assay prior to its implementation is suggested. Added complications stem from the test strips themselves. Colorization of the individual probes is highly variable and can lead to confusion in test interpretation; therefore, interrater variability needs to be further evaluated. Five samples were thrown out of the study due to an inability to interpret their test strips. Moreover, the strips are short and the probes are situated very close to one another, making identification of the exact probe that is absent quite challenging. An automated reader might facilitate interpretation of the results, and reportedly Nipro is currently working on this solution. If an incorrect probe was recorded as absent due to miscounting, then the sensitivity and specificity of the by-probe analysis would be affected, although this would not influence the by-strain analysis. Additionally, some codons were included on two probes. At times, one of the expected probes was missing but the other was not, affecting the sensitivity and specificity for the by-probe analysis but not the by-strain analysis. Although this information regarding the variables affecting the sensitivity and specificity of individual probes is valuable for assay improvement, it does not neces- sarily compromise assay performance if the purpose of this assay is only to identify likely susceptible or resistant samples. Individual samples will be correctly identified as mutant as long as any probe is missing, regardless of whether miscounting occurs or all probes that cover that particular mutation are lost. Continued sequence-level surveillance of pncA in combination with growth-based testing will be critical for possible future versions of this assay, as strains with silent mutations or susceptible nSNPs that are not accounted for in the current version will likely be discovered. Another limitation of LPAs is that the actual SNP cannot be identified, making LPAs unsuitable for precise SNP-level surveillance. Finally, it should be noted that the correlation between mutations and resistance is not absolute and loss of probe binding may not correlate with resistance to PZA in every instance. Conversely, the presence of all probes may not always correlate with susceptibility, as resistance to PZA may occur through another mechanism. Current methods used for evaluation of PZA susceptibility include pncA sequencing, the PZase enzymatic activity assay, and the growth-based Bactec MGIT 960 PZA assay. Sequencing has been shown to have high sensitivity and specificity, i.e., 87 to 92% and 93%, respectively, for predicting PZA resistance and is relatively rapid but requires highly trained technicians and expensive reagents and equipment (10, 20). Enzymatic activity assays involve growing the organism on a slant, adding a ferrous ammonium sulfate solution, and monitoring the culture for the appearance of a red color change at the medium interface (sensitivity, 89 to 91%; specificity, 97%) (20, 21). The Bactec MGIT 960 PZA assay has a reportedly high rate of false PZA resistance results, although it is used by many laboratories to reliably predict resistance to other first- and second-line TB drugs (21). In this study, an alternative technique for evaluating PZA susceptibility, the Nipro Genoscholar PZA-TB II assay, was evaluated and shown to January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 6 Evaluation of the Nipro Genoscholar PZA-TB II Assay Antimicrobial Agents and Chemotherapy achieve reproducible results at two test sites. Although the sensitivity among individual probes was quite variable and the test will need to be updated as additional mutations that do not lead to PZA resistance are identified, the overall sensitivity for identifying strains with mutations in pncA (93.2%) was high and similar to that of other methods currently used for PZA resistance testing. Unfortunately, the by-strain specificity was lower than hoped for (91.2%). However, it should be emphasized that the number of WT strains was small and our inclusion of all strains that would be expected to be susceptible to PZA, based on Sanger sequencing results, and not just strains with fully WT pncA sequences in the WT group had an effect on specificity. The assay had 100% specificity for detecting strains with true WT pncA sequences. Since the assay looks for the absence of WT sequences in order to predict mutations, three strains with suscep- tible nSNPs (Asp110Gly, Val163Ala, and Ala170Val) were determined to represent false-positive results, based on Sanger sequencing results, leading to a decrease in specificity. While this is an area of concern, there are more strains with WT pncA sequences than strains with susceptible nSNPs; therefore, this would likely be a significant issue only in populations in which those particular SNPs are common. Importantly, since many laboratories are currently preforming LPAs for detection of resistance to other drugs, it may be relatively easy to incorporate this assay into current testing algorithms, providing a rapid, relatively simple alternative for detection of PZA resistance. MATERIALS AND METHODS Strain selection. A total of 270 M. tuberculosis strains from the CDC (Atlanta, GA) (n ⫽ 144) and the PHAS (n ⫽ 126), for which the pncA genotype had been determined previously by Sanger sequencing, were tested using the Nipro Genoscholar PZA-TB II assay; the group included a subset of strains with WT pncA (n ⫽ 21). Strains with a wide variety of mutations within pncA and its promoter were selected from laboratory culture collections based on the availability of pncA sequence data. Strains harbored muta- tions within the pncA coding region (n ⫽ 235) or its promoter (n ⫽ 9) or both (n ⫽ 5) and had either one (n ⫽ 198) or more (n ⫽ 7) SNPs, an indel (n ⫽ 39), or both a SNP and an indel (n ⫽ 5). In this sample set, 189 strains harbored a single nSNP, including three strains containing nSNPs that have been reported not to lead to PZA resistance (Glu37Val, Val163Ala, and Ala170Val) (18). Nine strains harbored only a sSNP (Leu35Leu, Ala38Ala, and Ser65Ser). Twelve strains contained various combinations of indel, nSNP, and sSNP. The strain collection contained 218 total SNPs in 210 strains, with 142 unique substitutions in 87 (of 187) different codons. Strains represented a wide geographical distribution. DNA isolation. For samples tested at the CDC, the DNA used either had been isolated from strains in a previous study (22) or was freshly prepared from a 30-␮l aliquot of a frozen stock that was heated for 30 min at 97°C. DNA was diluted 1:10 and used as the template for PCRs. For samples tested at the PHAS, two 1-␮l loops of bacterial growth were transferred from a Lowenstein-Jensen slant into 200 ␮l of Tris-EDTA buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0). The suspensions were heated at 95°C for 30 min and centrifuged at 13,000 rpm for 2 min, and the resulting supernatant was used as the template for PCRs. Nipro Genoscholar PZA-TB II assay. This assay is based on 48 overlapping 15- to 21-nucleotide WT probes that encompass the entire pncA open reading frame and an additional 18 nucleotides upstream of the gene. These probes are bound in sequential order to a nitrocellulose strip. Several sSNPs in pncA have been identified in clinical isolates, and probes for three mutant sequences not associated with resistance to PZA (Gly60Gly, Ser65Ser, and Thr142Thr) are included on the membrane in combination with the WT probe for that region of pncA, thus ensuring probe binding in the presence of either the WT sequence or any of the three susceptible sSNPs. The assay was performed according to the package insert, using 5 to 10 ␮l of template and the Nipro Multiblot NS-4800 system. Briefly, pncA was amplified in a 50-␮l reaction mixture containing 30 ␮l amplification solution Z, 1 ␮l DNA polymerase, the template, and water. Reactions were denatured at 94°C for 2 min and then cycled 55 times at 98°C for 10 s and 67°C for 1 min, followed by a final 2-min extension at 65°C. Five microliters of each resulting amplicon was mixed with 5 ␮l of denaturation solution and loaded into the Multiblot NS-4800 hybridization instrument (Nipro) along with the test strips, prewarmed hybridization solution, rinse solution, conjugate solution, and substrate solution (all supplied in the Genoscholar PZA-TB II kit), and strips were processed as recommended by the manufacturer. WT amplicons bind to complementary probes, whereas the presence of a mutation within the amplified pncA sequence precludes hybridization to that probe. For strains containing one of the three susceptible sSNPs accounted for in this assay (Gly60Gly, Ser65Ser, and Thr142Thr), amplicons with either the WT sequence or the sSNP hybridize to the LPA strips. Bound amplicons are visualized by colorimetric development. Strips were read by eye in a blinded fashion, and the absence of probes was noted. Sequencing of pncA from discordant strains. All strains with discordance between genotype and LPA results were subjected to repeat sequencing of pncA and its promoter region, to confirm the previous sequencing results. The pncA gene was amplified from extracted DNA using primers pncA1 R January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 7 Willby et al. Antimicrobial Agents and Chemotherapy (GGTCATGTTCGCGATCGTCG) and pncA2 F (ACAGTTCATCCCGGTTCGGC) (CDC) or pncA F3 (AAGGCCGC GATGACACCTCT) and pncA R4 (GTGTCGTAGAAGCGGCCGAT) (PHAS). Amplicons were purified, and Sanger sequencing was undertaken using the amplification primers along with primers MtbpncA304 F (CGTACAGCGGCTTCGAAGGA) and MtbpncA304 R (TCCTTCGAAGCCGCTGTACG) (CDC) or the amplifica- tion primers along with primers P3 F (ATCAGCGACTACCTGGCCGA) and P4 R (GATTGCCGACGTGTCCA GAC) (PHAS). Sequence data were aligned with the H37Rv pncA reference sequence (GenBank accession number NC_000962.3) using Lasergene SeqMan 12 (CDC) or Vector NTI Advance v9 (InfoMax, Inc.) (PHAS) software, and SNPs were identified using the reference sequence. Data analysis. Data were analyzed in two ways. In the by-strain analysis, the correlation between a strain harboring a mutation (as identified previously by Sanger sequencing), other than one that has been reported not to be associated with resistance, and loss of binding of any probe was assessed. In the by-probe analysis, the accuracy of each probe in detecting mutations within that probe’s binding region was assessed. Sensitivity and specificity were calculated for each method. CIs were calculated using Wilson’s score interval. ACKNOWLEDGMENTS We thank Samuel Schumacher and Sophia Georghiou at FIND for their critical reading of the manuscript. Financial support was provided to the Public Health Agency of Sweden from FIND. The Nipro Corporation provided reagents and loaner Nipro Multiblot NS-4800 instru- mentation for completion of this study. Nipro did not contribute in any way to study design, data analysis, or manuscript preparation. The use of trade names is for identification only and does not constitute endorse- ment by the U.S. Department of Health and Human Services, the U.S. Public Health Service, or the CDC. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the funding agency. REFERENCES 1. World Health Organization. 2010. Treatment of tuberculosis. World rium tuberculosis clinical isolates. Antimicrob Agents Chemother 59: Health Organization, Geneva, Switzerland. 5267–5277. https://doi.org/10.1128/AAC.00204-15. 2. McDermott W, Tompsett R. 1954. Activation of pyrazinamide and nico- 11. Miotto P, Cabibbe AM, Feuerriegel S, Casali N, Drobniewski F, Rodionova tinamide in acidic environments in vitro. Am Rev Tuberc 70:748 –754. Y, Bakonyte D, Stakenas P, Pimkina E, Augustynowicz-Kopec E, Degano 3. Konno K, Feldmann FM, McDermott W. 1967. Pyrazinamide susceptibility M, Ambrosi A, Hoffner S, Mansjo M, Werngren J, Rusch-Gerdes S, Ni- and amidase activity of tubercle bacilli. Am Rev Respir Dis 95:461– 469. emann S, Cirillo DM. 2014. Mycobacterium tuberculosis pyrazinamide 4. Gonzalo X, Drobniewski F, Hoffner S, Werngren J. 2014. Evaluation of a resistance determinants: a multicenter study. mBio 5:e01819-14. https:// biphasic media assay for pyrazinamide drug susceptibility testing of doi.org/10.1128/mBio.01819-14. Mycobacterium tuberculosis. J Antimicrob Chemother 69:3001–3005. 12. Maningi NE, Daum LT, Rodriguez JD, Mphahlele M, Peters RP, Fischer https://doi.org/10.1093/jac/dku230. GW, Chambers JP, Fourie PB. 2015. Improved detection by next- 5. Zhang Y, Permar S, Sun Z. 2002. Conditions that may affect the results of generation sequencing of pyrazinamide resistance in Mycobacterium susceptibility testing of Mycobacterium tuberculosis to pyrazinamide. J tuberculosis isolates. J Clin Microbiol 53:3779 –3783. https://doi.org/10 Med Microbiol 51:42– 49. https://doi.org/10.1099/0022-1317-51-1-42..1128/JCM.01179-15. 6. Diacon AH, Dawson R, von Groote-Bidlingmaier F, Symons G, Venter A, 13. Jureen P, Werngren J, Toro JC, Hoffner S. 2008. Pyrazinamide resis- Donald PR, van Niekerk C, Everitt D, Hutchings J, Burger DA, Schall R, tance and pncA gene mutations in Mycobacterium tuberculosis. Anti- Mendel CM. 2015. Bactericidal activity of pyrazinamide and clofazimine microb Agents Chemother 52:1852–1854. https://doi.org/10.1128/ alone and in combinations with pretomanid and bedaquiline. Am J AAC.00110-08. Respir Crit Care Med 191:943–953. https://doi.org/10.1164/rccm.201410 14. Sekiguchi J, Nakamura T, Miyoshi-Akiyama T, Kirikae F, Kobayashi I, -1801OC. Augustynowicz-Kopec E, Zwolska Z, Morita K, Suetake T, Yoshida H, Kato 7. Diacon AH, Dawson R, von Groote-Bidlingmaier F, Symons G, Venter A, S, Mori T, Kirikae T. 2007. Development and evaluation of a line probe Donald PR, van Niekerk C, Everitt D, Winter H, Becker P, Mendel CM, assay for rapid identification of pncA mutations in pyrazinamide- Spigelman MK. 2012. 14-day bactericidal activity of PA-824, bedaquiline, resistant Mycobacterium tuberculosis strains. J Clin Microbiol 45: pyrazinamide, and moxifloxacin combinations: a randomised trial. Lan- 2802–2807. https://doi.org/10.1128/JCM.00352-07. cet 380:986 –993. https://doi.org/10.1016/S0140-6736(12)61080-0. 15. World Health Organization. 2016. The use of molecular line probe assays 8. Dawson R, Diacon AH, Everitt D, van Niekerk C, Donald PR, Burger DA, for the detection of resistance to second-line anti-tuberculosis drugs: Schall R, Spigelman M, Conradie A, Eisenach K, Venter A, Ive P, Page- policy guidance. World Health Organization, Geneva, Switzerland. http:// Shipp L, Variava E, Reither K, Ntinginya NE, Pym A, von Groote- www.who.int/tb/WHOPolicyStatementSLLPA.pdf. Bidlingmaier F, Mendel CM. 2015. Efficiency and safety of the combina- 16. World Health Organization. 2015. Global tuberculosis report 2015. tion of moxifloxacin, pretomanid (PA-824), and pyrazinamide during the World Health Organization, Geneva, Switzerland. http://apps.who.int/ first 8 weeks of antituberculosis treatment: a phase 2b, open-label, partly medicinedocs/documents/s22199en/s22199en.pdf. randomised trial in patients with drug-susceptible or drug-resistant 17. Nathavitharana RR, Hillemann D, Schumacher SG, Schlueter B, Ismail N, pulmonary tuberculosis. Lancet 385:1738 –1747. https://doi.org/10.1016/ Omar SV, Sikhondze W, Havumaki J, Valli E, Boehme C, Denkinger CM. S0140-6736(14)62002-X. 2016. Multicenter noninferiority evaluation of Hain GenoType MTBDR- 9. Scorpio A, Zhang Y. 1996. Mutations in pncA, a gene encoding plus version 2 and Nipro NTM⫹MDRTB line probe assays for detection of pyrazinamidase/nicotinamidase, cause resistance to the antitubercu- rifampin and isoniazid resistance. J Clin Microbiol 54:1624 –1630. https:// lous drug pyrazinamide in tubercle bacillus. Nat Med 2:662– 667. doi.org/10.1128/JCM.00251-16. https://doi.org/10.1038/nm0696-662. 18. Whitfield MG, Warren RM, Streicher EM, Sampson SL, Sirgel FA, van 10. Ramirez-Busby SM, Valafar F. 2015. Systematic review of mutations in Helden PD, Mercante A, Willby M, Hughes K, Birkness K, Morlock G, van pyrazinamidase associated with pyrazinamide resistance in Mycobacte- Rie A, Posey JE. 2015. Mycobacterium tuberculosis pncA polymorphisms January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 8 Evaluation of the Nipro Genoscholar PZA-TB II Assay Antimicrobial Agents and Chemotherapy that do not confer pyrazinamide resistance at a breakpoint concentra- 21. Simons SO, van Ingen J, van der Laan T, Mulder A, Dekhuijzen PN, tion of 100 micrograms per milliliter in MGIT. J Clin Microbiol 53: Boeree MJ, van Soolingen D. 2012. Validation of pncA gene sequenc- 3633–3635. https://doi.org/10.1128/JCM.01001-15. ing in combination with the mycobacterial growth indicator tube 19. Driesen M, Kondo Y, de Jong BC, Torrea G, Asnong S, Desmaretz C, method to test susceptibility of Mycobacterium tuberculosis to pyr- Mostofa KSM, Tahseen S, Whitfield MG, Cirillo DM, Miotto P, Cabibbe AM, azinamide. J Clin Microbiol 50:428 – 434. https://doi.org/10.1128/JCM Rigouts L. 2017. Evaluation of a novel line probe assay to detect resis-.05435-11. tance to pyrazinamide, a key drug used for tuberculosis treatment. Clin 22. Campbell PJ, Morlock GP, Sikes RD, Dalton TL, Metchock B, Starks AM, Microbiol Infect https://doi.org/10.1016/j.cmi.2017.05.026. Hooks DP, Cowan LS, Plikaytis BB, Posey JE. 2011. Molecular detection of 20. Chang KC, Yew WW, Zhang Y. 2011. Pyrazinamide susceptibility testing mutations associated with first- and second-line drug resistance com- in Mycobacterium tuberculosis: a systematic review with meta-analyses. pared with conventional drug susceptibility testing of Mycobacterium Antimicrob Agents Chemother 55:4499 – 4505. https://doi.org/10.1128/ tuberculosis. Antimicrob Agents Chemother 55:2032–2041. https://doi AAC.00630-11..org/10.1128/AAC.01550-10. January 2018 Volume 62 Issue 1 e01871-17 aac.asm.org 9