Systematic Review of Mutations in Pyrazinamidase Associated with Pyrazinamide Resistance in Mycobacterium tuberculosis Clinical Isolates

Abstract

Pyrazinamide (PZA) is an important first-line drug in the treatment of tuberculosis (TB) and of significant interest to the HIV-infected community due to the prevalence of TB-HIV coinfection in some regions of the world. The mechanism of resistance to PZA is unlike that of any other anti-TB drug. The gene pncA, encoding pyrazinamidase (PZase), is associated with resistance to PZA. However, because single mutations in PZase have a low prevalence, the individual sensitivities are low. Hundreds of distinct mutations in the enzyme have been associated with resistance, while some only appear in susceptible isolates. This makes interpretation of molecular testing difficult and often leads to the simplification that any PZase mutation causes resistance. This systematic review reports a comprehensive global list of mutations observed in PZase and its promoter region in clinical strains, their phenotypic association, their global frequencies and diversity, the method of phenotypic determination, their MIC values when given, and the method of MIC determination and assesses the strength of the association between mutations and phenotypic resistance to PZA. In this systematic review, we report global statistics for 641 mutations in 171 (of 187) codons from 2,760 resistant strains and 96 mutations from 3,329 susceptible strains reported in 61 studies. For diagnostics, individual mutations (or any subset) were not sufficiently sensitive. Assuming similar error profiles of the 5 phenotyping platforms included in this study, the entire enzyme and its promoter provide a combined estimated sensitivity of 83%. This review highlights the need for identification of an alternative mechanism(s) of resistance, at least for the unexplained 17% of cases.

INTRODUCTION

In 2013, there were an estimated 9 million new cases of tuberculosis (TB) (1). Treatment of new cases consists of a 6-month regimen of four first-line drugs: isoniazid (INH), rifampin (RIF), ethambutol (EMB), and pyrazinamide (PZA) (1). PZA is an important first-line drug because it is active against semidormant Mycobacterium tuberculosis (2–7).

Although PZA has been continuously used to treat TB, the World Health Organization (WHO) does not include PZA in the group of antimycobacterial drugs to be routinely tested for resistance because PZA drug susceptibility testing (DST) is notoriously difficult and often inaccurate (8). The size of the inoculum contributes to inaccurate PZA susceptibility tests; if the inoculum is too large, PZA activity may be reduced, leading to false resistance (9, 10). PZA susceptibility tests also require an acidic environment (11), which generates unreliable results by inhibiting growth of the bacteria (10) and creating difficulties in reproducibility (12). An acidic medium is required due to the mechanism of action of PZA. The mechanism was first discovered when Konno et al. observed a lack of pyrazinamidase (PZase) activity in PZA-resistant (PZA^r) M. tuberculosis strains (5). In an isolate with a wild-type PZase, the enzyme converts PZA into pyrazinoic acid (POA) (5). POA is generated in the neutral cytoplasm as an anion, which has no antimycobacterial activity. POA then is pumped out of the cell via efflux pump (13) and, in acidic environments, is protonated (HPOA). HPOA enters the cell, accumulates, and eventually kills the cell (2).

The WHO recommends the radiometric (Bactec 460TB) or nonradiometric (Bactec MGIT 960) liquid medium to test PZA susceptibility, with a MIC of <100 mg/liter considered susceptible and otherwise resistant (14). Because the Bactec 460TB platform was discontinued due the radioactive components, the Bactec MGIT 960 with PZA kit (medium with a preadjusted pH level required for PZA susceptibility testing) is the only platform recommended by the WHO (14). However, many developing countries lack the resources to test for PZA resistance under the WHO guidelines. Finally, as is the case with any other M. tuberculosis growth-based test, PZA susceptibility testing takes weeks due to the slow growth of M. tuberculosis (15–17). This leaves resistant cases without an appropriate treatment for several more weeks while remaining contagious. They may even die in this period due to immune system complications such as those that occur in TB-HIV-coinfected patients.

The WHO does not recommend an enzyme activity assay for drug susceptibility; however, unlike traditional growth-based PZA susceptibility methods, the Wayne assay does not require an acidic medium (18). Therefore, determining enzyme activity is an attractive alternative to conventional DST methods. However, this assay relies on the growth rates of the bacteria (18), which still requires patients to wait weeks for results, diagnosis, and the proper treatment regimen. The results from the Wayne assay are subject to interpretation by the operator, since an observation of a change in color of each medium is required for determining enzyme activity (18). The Wayne assay also may produce nontrivial false-positive (resistance) rates (2, 9).

Molecular diagnosis might bypass difficult, inaccurate, and lengthy DST, return results more quickly, and be cost-effective. Nevertheless, the genotypic-phenotypic relationship is not always straightforward for PZA. Unlike RIF and INH, for which phenotypic resistance has a well-established correlation to a few specific rpoB, katG, and inhA promoter mutations, respectively (19, 20), individual mutations that demonstrate a strong association with PZA resistance still elude us.

For PZA, the primary gene associated with resistance is pncA (21). Scorpio and Zhang determined that PZase is encoded by the gene pncA and found that Mycobacterium bovis contains a single mutation in PZase, His57Asp, which is associated with resistance to PZA in M. bovis. These authors cloned M. tuberculosis pncA and discovered that the PZA^r M. tuberculosis isolates also had a mutation in PZase, but not in the same codon as M. bovis. They discovered several mutations in PZase that were associated with PZA resistance (21). Later, it was discovered that mutations in pncA cause PZase to be defective, allowing the bacterium to survive in the presence of PZA (2). Reports have suggested other areas of the genome that might be associated with PZA resistance outside pncA. Currently, mutations in rpsA and panD have been reported to be correlated with PZA resistance (22, 23). fas1 inhibition by POA has also been suggested to be associated with PZA resistance (24). However, several publications have shown that these mutations are not associated with resistance (25–27). Therefore, this systematic review will only include pncA.

Although there is a 2011 meta-analysis on methods to detect resistance to PZA (9), a search through PubMed using key search terms yielded no review with a comprehensive analysis on the PZase mutations associated with PZA resistance in M. tuberculosis. While differing mechanisms of PZA resistance have been described in the literature, most involve PZase in the mechanism (13, 24, 28). This systematic review aims to characterize the global diversity and frequencies of mutations in PZase and the promoter region of pncA and to assess the strength of the association between mutations and phenotypic resistance to PZA among clinical M. tuberculosis strains. This review evaluates the utility of mutations in PZase and in the promoter region of pncA for the molecular diagnosis of PZA resistance in M. tuberculosis.

MATERIALS AND METHODS

Literature search.

A search in PubMed was conducted on all peer-reviewed publications evaluating mutations in PZase in PZA^r and PZA-susceptible (PZA^s) isolates of M. tuberculosis. The search was limited to studies published between 1995 and 10 December 2014. The search was performed through PubMed using the following key words individually and as an exhaustive combination applying the AND operator: “pyrazinamide,” “resistance,” “Mycobacterium tuberculosis,” “pncA,” “mutations,” and “pza” in various combinations. In addition to a database search, references of the articles included were analyzed for more potential studies. For the complete process of reviewing published articles, see the supplemental material.

Study selection criteria.

Studies were included if (i) original data were published, (ii) they were written in English, (iii) phenotypic DST was performed for PZA and assessed the correlation between DST and mutations in PZase and/or the promoter region of pncA on clinical M. tuberculosis strains (DST being the reference standard), and (iv) the whole pncA gene or whole genome was sequenced; mutations in the putative regulatory region or promoter region were included if available. Studies that performed DST on solid or liquid medium were included, as long as the cutoff concentrations were clearly defined. Publications were excluded (i) if the authors did not perform phenotypic DST, (ii) if PZase activity testing was performed (instead of DST) to determine phenotypic DST, (iii) if sequencing of the whole gene was not performed as a method for determining drug-resistant mutations, (iv) if mutation data were not included, or (v) if the publication contained contradictory/inconsistent mutation data. This procedure was repeated independently at multiple time points in order to ensure the accuracy of the data collected.

Data acquisition.

Data extracted from articles that met the inclusion criteria were as follows: author names, publication year, PubMed identifier, title of the paper, total numbers resistant and susceptible determined by phenotypic DST, geographic location, year of collection, reference strain, DST or MIC determination and the specific media used, cutoff concentrations for DST and all concentrations if MICs were determined, and the method for sequencing pncA. Each paper was examined for individual mutations and combinations of mutations within PZase. Each mutation reported was independent of all other mutations within and between studies (except when multiple mutations were reported within PZase). If there was more than one mutation in PZase and/or the promoter region in an isolate, the combination was considered unique and was associated with resistance to PZA.

We recorded specific mutations in PZase and/or the promoter region with the published sequence (GenBank accession numbers U59967 or NC_000962) as the references. The numbers of clinical resistant and susceptible isolates harboring each mutation and combinations of mutations were noted. MIC values were also noted, where applicable. If PZase activity was determined in the study, this was noted for each specific mutation as “yes” for positive PZase activity and “no” for negative PZase activity. Any test to determine PZase activity was accepted. All information was recorded in Microsoft Excel with version control.

Quality assessment.

To control for errors in the reported data within a study and errors in extracting data, every reference amino acid reported was compared to the published H37rv sequence (GenBank accession number NC_000962.2). Mutations corresponding to discrepant reference amino acids were excluded from the analysis, while those with the correct amino acid reference (from the same article) were included in analysis (29–34).

Statistical analysis.

With the aim of a potential predictive value for resistance-related mutations for molecular diagnostic testing, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for each unique mutation. Sensitivities and specificities were also calculated for the Zhang and Mitchison (2)-reported hot spot regions, codons 3 to 17 (5 double mutants, one mutation in the hot spot region, another outside the hot spot region), 61 to 85 (5 double mutants), and 132 to 142 (34 double mutants: 21 from 1 study, 13 from another), using the following formulas:

$${\text{sensitivity}}_{\text{h}otspot}=\hspace{0.17em}\frac{n_{\text{hotspot}}^{\text{r}}}{N_{\text{total}}^{\text{r}}}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}{\text{specificity}}_{\text{hotspot}}=\hspace{0.17em}1-\hspace{0.17em}\frac{n_{\text{hotspot}}^{\text{s}}}{N_{\text{total}}^{\text{s}}}$$

where n_hotspot^r is the total number of resistant isolates with mutations within a hot spot region and n_hotspot^s is the number of susceptible isolates with a mutation within the hot spot region. N_hotspot^r and N_hotspot^s refer to the total number of resistant and susceptible isolates included in this study, respectively. Sensitivity and specificity were also calculated for two separate groups: mutations in isolates that had the promoter region as well as the coding region sequenced and mutations in isolates that only had the coding region sequenced.

RESULTS

Description of included studies.

A search through PubMed (Medline) revealed 248 articles published between 1975 and 10 December 2014. Out of 248 potential studies, 61 studies met all of the inclusion criteria (8, 17, 21, 25, 29–85). A PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram in Fig. 1 illustrates the breakdown of the number of articles excluded because of the specified criteria. Of the 61 studies, the earliest study was from 1996, while 20 (33%) were published in the last 3 years (8, 25, 30, 31, 35, 41, 43, 47, 60, 70, 73, 75, 77, 81–85). In total, there were 6,089 isolates, 2,760 phenotypically PZA^r and 3,329 phenotypically PZA^s, in the 61 articles that had at least the pncA coding region sequenced. The geographic locations of these isolates were very diverse, encompassing 42 different countries and 4 continents. Table S1 in the supplemental material provides an overview of the geographic origins of the isolates included in this study.

FIG 1.

PRISMA flow diagram. The search terms used were “Mycobacterium tuberculosis” and “pyrazinamide.”

In total, 641 unique PZase and promoter region mutations were discovered in PZA^r isolates. If there were multiple mutations in PZase or its promoter, the combination was considered unique and the combination (rather than the individual mutations) was associated with PZA resistance.

Because DST is the standard resistance-determining test, we noted the type of medium and the cutoff concentration used. If MICs were determined, we noted all levels of concentrations as well as the medium used. Although we recorded MIC information for each mutation, where applicable, here we only report the MIC data collected and have not performed any analysis due to the incomparability of the results from various media (see Table S2 in the supplemental material). The medium suggested by the WHO is a liquid medium (Bactec 460TB and Bactec MGIT 960 only) (14). There were also insufficient data for a statistical analysis of the relative platform bias and trends. Out of 61 studies, 42 studies performed DST or determined the MIC on either Bactec 460TB or Bactec MGIT 960 medium. A full list of the types of media and the cutoffs (if available) used in each study is presented in Table S3 in the supplemental material.

PZase activity is believed to be an important measure of resistance to PZA (2, 9). We, therefore, noted PZase activity whenever it was reported by the study. In addition to sequencing pncA and performing phenotypic DST, 39 studies (64%) determined PZase activity (17, 21, 31, 34, 35, 37–42, 45, 46, 49, 53, 54, 56–69, 74, 77–81, 83, 84). All studies used the assay developed by Wayne to ascertain PZase activity (18).

A PRISMA checklist has been included to ensure the integrity of the systematic review (see Table S4 in the supplemental material).

Mutations in PZase associated with PZA resistance.

Mining the literature included in this study and compiling the list of all mutations reported in PZA^r isolates implicated 171 PZase amino acids (91% of the gene) in drug resistance. Of 2,760 resistant isolates, 2,277 isolates (83%) harbored at least one mutation in PZase and/or its promoter region, 2,187 (79%) isolates had mutations only in PZase, 78 (3%) isolates had mutations only in the promoter region, and 12 (0.4%) had a mutation in the promoter region and one in the coding region. Furthermore, of the 3,329 PZA^s isolates, 300 (9%) had a mutation in PZase (Table 1). Sensitivity and specificity were also calculated based on 4 different categories, similar to Miotto et al. (84). The criteria and resulting sensitivities and specificities are located in Table S5 in the supplemental material. In addition, calculation of the sensitivity and specificity based solely on mutations from isolates that were tested with Bactec 460TB or Bactec MGIT 960 resulted in 80% sensitivity and 91% specificity.

TABLE 1.

Stratification of isolates included in this study into phenotypic groups

Group	No. of isolates	% per phenotype
All resistant isolates (with at least the pncA coding region sequenced)	2,760	100
Resistant isolates with PZase and its promoter sequenced	2,565	93
Resistant isolates only with mutations in PZase	2,187	79
Resistant isolates only with mutations in the promoter region of pncA	78	3
Resistant isolates with mutations in both the promoter and the open reading frame	12	0.4
Resistant isolates without a mutation in PZase	483	18a
All susceptible isolates (with at least the pncA coding region sequenced)	3,329	100
Susceptible isolates with pncA and its promoter sequenced	3,207	96
Susceptible isolates only with mutations in PZase	307	6
Susceptible isolates only with mutations in the promoter region of pncA	14	0.4
Susceptible isolates with mutations in both the promoter and the open reading frame	0	0
Susceptible isolates without a mutation in PZase or the promoter region	3,022	91

^aIt is important to recognize that not all the isolates in this group are considered “unexplained resistance” cases. These percent values have been calculated with respect to all isolates with a pncA sequence of the same phenotype. Because not all isolates included an analysis of the promoter region, some of these isolates may harbor a mutation in the promoter region and therefore are not in the unexplained resistance group.

The distribution of mutations (and their frequencies) observed in PZA^r isolates along PZase and its promoter region is demonstrated in Fig. 2. It is important to note that this figure combines the frequencies of the different mutations that are observed in each codon. See Table S5 in the supplemental material for a complete list of all mutations and their individual frequencies. The hot spot mutations suggested by Zhang and Mitchison (3 to 17, 61 to 85, and 132 to 142) (2) are located between the brackets on the graph in Fig. 2. Many mutations do not fall within these regions, and those that do only account for a small percentage of the study's PZA^r population (7% of PZA^r). Sensitivities for the three hot spot regions were 10%, 13%, and 1%, respectively. Six mutations were excluded from certain studies because the reference amino acid did not match H37rv (29, 31–33). The mutations that were excluded, along with the article information from which the mutations were reported, can be found in Table S6 in the supplemental material.

FIG 2.

The distribution and frequency of mutations observed in pyrazinamide-resistant Mycobacterium tuberculosis clinical isolates in pncA and its promoter region. The black brackets indicate hot spot regions (3 to 17, 61 to 85, and 132 to 142).

In addition to single nucleotide polymorphisms (SNPs), frequently single nucleotide insertions or deletions (indel) were also reported. In many cases, an indel co-occurred with an SNP in the same isolate, while sometimes it was reported as the only variant in the gene; 369 isolates (16% of PZA^r with a mutation) had only an indel. In one study, the deletion of the entire gene was reported (74). A single amino acid substitution in PZase was the most common type of mutation, occurring in 1,804 isolates (79%). A distribution of the types of mutations is demonstrated in Fig. S1 in the supplemental material. The most common mutation was L85P, with 41 isolates harboring the mutation, although 13 of the isolates with this mutation came from one study, which might indicate clonal expansion (71). One hundred six isolates harbored more than one mutation (usually a substitution and a deletion) (see Fig. S1 in the supplemental material). A full list of all of the mutations associated with PZA resistance found in the published literature as well as sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) is located in Table S6 in the supplemental material.

Statistical analysis of mutations in PZase.

The standard metrics of sensitivity, specificity, PPV, and NPV were calculated for each unique mutation (Table S6 in the supplemental material) to determine if any mutation can be used in a molecular diagnostic test. The highest sensitivity was 1.48% at mutation L85P. This mutation also had a specificity of 98%, PPV of 100%, and NPV of 55%. As expected, all mutations had very low sensitivities (the majority had <1%),and most had specificities and PPVs of 100%. NPVs for the majority of mutations were around 55%. To determine the gene-wide metrics, we performed two sets of calculations. In the first set, we only considered mutations in the coding region and then we added the promoter region mutations in order to calculate the combined coding-promoter region metrics. The sensitivity and specificity for the coding region were 79% and 90%, respectively, while those of the combined coding-promoter regions of pncA were calculated to be 83% and 90%, respectively. NPVs and PPVs for both sets are shown in Table S7 in the supplemental material.

Mutations in PZase of PZA-susceptible isolates.

Of 3,329 PZA^s isolates, 300 harbored mutations in PZase (9% of PZA^s isolates) (17, 29–31, 35, 41, 55, 58, 66, 70, 73, 82, 84). In total, 96 distinct PZase mutations were found in PZA^s isolates; while 32 of those were unique to susceptible isolates, none were found in the resistant set. Table S8 in the supplemental material depicts mutations found in PZA^s isolates and compares the frequencies of mutations found in susceptible isolates to the frequencies of those same mutations among resistant isolates.

Pyrazinamidase activity.

PZase activity has been reported to be absent in PZA^r isolates (5). To confirm this, we noted negative or positive PZase activity for each PZase mutation in PZA^r and PZA^s isolates whenever it was reported. Of the studies that determined PZase activity (39 studies, 1,979 PZA^r isolates), 1,125 PZA^r isolates tested negative for PZase activity. However, there were 247 PZA^r isolates (12%) that tested positive for PZase activity. A comparison of the enzyme activity, DST, and mutation report stratifies PZA^r isolates into three discrepant groups: those with positive PZase activity that harbored a mutation(s) in PZase (120 isolates), those with positive PZase activity that did not harbor a mutation(s) in PZase (127 isolates), and those with no detectable PZase activity and no mutations in PZase (69 isolates). Percentages of the discrepant groups are presented in Table S9 in the supplemental material. Further breakdown of PZA^s isolates is demonstrated in Fig. S2 in the supplemental material. The first group of isolates with positive PZase activity and a mutation(s) in PZase is shown in Table S10 in the supplemental material. Miotto et al. reported the greatest number of resistant isolates that harbored a mutation in PZase and tested positive for PZase activity (68 isolates, 0.689%) (84). Table S10 in the supplemental material lists the number of PZA^r isolates with PZase activity and the location of mutations in PZase compared to those of resistant isolates with the same mutations and negative PZase activity.

DISCUSSION

As a whole, the combination of the coding region of pncA and its promoter is an acceptable marker for determining PZA resistance in M. tuberculosis, with 83% sensitivity and 90% specificity. PZase activity determination using the Wayne assay is also a fair diagnostic test for resistance, although it has noticeably lower specificity (9). The WHO recommends performing DST using Bactec MGIT 960 or Bactec 460TB medium for PZA susceptibility testing if the organization can afford the suggested platforms. These platforms are expensive and are not available in many developing countries. Therefore, a reliable and cost-effective platform for diagnosing PZA resistance in M. tuberculosis is still an unmet need that the private sector can explore.

Molecular diagnostics of PZA resistance.

Figure 2 demonstrates the diversity of mutations in PZase in PZA^r isolates with a random distribution pattern. This distribution is highly unusual and unlike those observed for markers of resistance for other drugs (e.g., aminoglycoside resistance, where mutations in a single position, 1401, in rrs can produce a global testing sensitivity of 78%) (86). An explanation for this diversity still eludes us. Among the more popular hypotheses, the cost of mutations in this gene and the nonessential role of the product can be mentioned. In brief, since pncA is believed to be nonessential (40), mutations may confer low to no fitness cost, resulting in no selective pressure on the functional protein, which in turn can result in mutations occurring anywhere in the gene without a loss of fitness. However, it has been shown previously that although a gene may be considered nonessential in vitro, it may be essential in vivo (87, 88). From the protein science point of view, it is difficult to imagine how mutations scattered all over the enzyme and away from the active site can render the enzyme dysfunctional.

Considering mutations that are observed in resistant isolates as associated with PZA resistance provided an estimated sensitivity of 83% and specificity of 91%. We developed four additional categories similar to the work of Miotto et al. (84) in order to analyze the strength of the association of variants to PZA resistance (Table S5 in the supplemental material). Although specificity increases by excluding mutations that also exist in susceptible isolates, sensitivity suffers. With a more systematic sequencing effort, we believe more resistance-associated mutations will be observed in susceptible isolates. This phenomenon has already been seen through this systematic study where one study reports a mutation only in resistant isolates and another has observed it in several susceptible isolates. In the case of dismissal of such mutations, resistant isolates that harbor them would be considered “unexplained,” hence reducing sensitivity.

Several included studies did not use the WHO-recommended Bactec platforms for PZA susceptibility testing. Using isolates that were tested on Bactec platforms only, we estimated the sensitivity and specificity of pncA and its promoter combined. The results were 80% and 91%, respectively. These statistics are very similar to the overall statistics (83% sensitivity and 91% specificity) calculated using isolates tested on all platforms.

Figure 2 also demonstrates that a single (or even a few) mutation(s) in PZase is not sufficient for diagnostics and that the whole gene and its promoter need to be considered. The most common mutation was observed in 41 isolates, which only provides 1% sensitivity. The current molecular diagnostic test used to determine RIF resistance, Gene Xpert MTB/RIF, provides 92.2% sensitivity for culture-positive patients (89). This test only analyzes a small region of the rpoB gene, the RIF resistance-determining region (RRDR), which is 81 bp long (89). A similar platform would only provide, at most, 13% sensitivity for PZA resistance.

Furthermore, 483 resistant isolates (18% of all resistant isolates) had a wild-type PZase. This means that 483 patients would be falsely labeled as having a PZA^s M. tuberculosis infection if only pncA sequencing was performed as a diagnostic test. The common explanations for such discrepant cases in the literature have been the unreliability of the DST and the low MIC cutoff (2, 9, 40, 77). There have been reports of false resistance with many platforms used to perform phenotypic DST (61, 65, 70, 90, 91). Raising the MIC cutoff to 128 mg/liter (77), 200 mg/liter (2, 9), or 300 mg/liter (40) has been suggested in order to reduce the number of such unexplained isolates. It is important to note that while raising the cutoff may classify some of the isolates with no PZase mutations as susceptible and thereby improve the specificity, the approach would also classify isolates with borderline MIC levels and a PZase mutation as susceptible. This, in turn, will reduce the current diagnostic sensitivity of pncA as a whole. A second argument against raising the MIC cutoff is the consideration for the clinical applicability of the bacteriological discoveries. High concentrations of PZA are toxic in humans (92); consequently, clinicians can only prescribe 25 mg/kg to a patient for a daily dosage or 35 mg/kg for three times weekly dosing (93), which makes a large dose of PZA for treatment unrealistic and therefore susceptibility of the isolate at higher doses clinically irrelevant. Moreover, raising the concentration for treatment is not useful as the bactericidal activity of PZA only slightly increases with higher concentrations (92). For these reasons, the clinical benefits of raising the in vitro susceptibility cutoff are unclear.

Due to the complications in PZA susceptibility testing, molecular diagnostics becomes an attractive alternative. Because of the lack of hot spots in PZase, a PZA molecular diagnostic platform must analyze the entire gene and its promoter. A PCR-based two-step DNA synthesis (PTDS) method is one such diagnostic method due to its ability to analyze relatively larger segments of the genome (94).

Another solution is whole-genome sequencing (WGS). WGS provides unique, but unrealized, diagnostic opportunities for drug-resistant TB. It offers entire gene and promoter-based diagnostics not only for PZA but also for all other first- and second-line drugs. It, therefore, offers a “complete” genotypic profile. This includes estimation of the phenotyping for all first- and second-line drugs concurrently (95), lineage and fitness analysis (96), heterogeneity through metagenomic analysis (97), and more. These capabilities are becoming important as the incidences of multidrug-resistant TB (MDR-TB) and extensively drug-resistant TB (XDR-TB) are on the rise (1). The cost of WGS is high compared to that of current molecular diagnostic tests but is rapidly decreasing (98). The time to determine the resistance profile is also rapidly decreasing, currently taking up to 11 days (98). These aspects are ideal for developing countries with scarce resources but high incidences of MDR- and XDR-TB which require multiple phenotypic tests. Currently, 75% of MDR cases go undetected globally, which therefore results in ineffective treatment regimens and further extends the infectious period of these complex cases (1). Although current WGS approaches rely on growth and extraction of DNA material from a culture, which again takes several weeks, direct WGS of biological samples is an important developmental next step. Genotyping using mycobacterial interspersed repetitive-unit–variable-number tandem-repeat (MIRU-VNTR) analysis, spoligotyping, or any form of phylogenetic processing is also possible using WGS. This is important in outbreak situations and for determining the origin of infection. It is important to note here that MIRU-VNTR analysis currently cannot be done by WGS platforms that use PCR amplification, such as Ion Torrent (Thermo Fisher Scientific) or the MiSeq system (Illumina, Inc.); only the Pacific Biosciences of California RS platform is able to do this for M. tuberculosis.

Pyrazinamide-susceptible isolates harboring mutations in PZase.

Although mutations in PZase are frequently interpreted as phenotypic resistance, some PZA^s isolates also harbor mutations in PZase. Table S8 in the supplemental material presents a summary of all mutations reported in PZA^s isolates as well as the number of PZA^r isolates with the corresponding mutation. Of 3,329 susceptible isolates, 300 had a mutation in PZase (9%). Since DST errors that would cause a resistant isolate to be labeled as susceptible are exceedingly infrequent, a proper explanation of this phenomenon still eludes us. In comparison, for example, the A1401G mutation in rrs is observed in 78% of amikacin-resistant and 0% of amikacin-susceptible isolates and in 56% of kanamycin-resistant and 0% of kanamycin-susceptible isolates (86). For both drugs, no susceptible isolates have mutations at the 1401 position and a large majority of resistant isolates harbor this mutation. PZA resistance in M. bovis is monoclonal and associated with a single amino acid substitution in PZase (21). Yet, PZA resistance in M. tuberculosis does not follow this pattern and seems to be quite different. Among the potential explanations for PZA^s isolates harboring PZase mutations, the incorrect inoculum size (9, 10) or incorrect modification of the pH in the medium (9) has been mentioned.

While the frequencies are not high enough to be conclusive, the mutations that occurred equally among resistant and susceptible strains, such as Cys14Gly and Thr47Ala, might be phylogenetic markers because they occur regardless of phenotype. Phylogenetic markers have been suggested before in other resistance-associated genes, such as codon 463 in katG (INH resistance) and codon 95 in gyrA (fluoroquinolone resistance) (99). It is important to note that resistant isolates with only such potentially phylogenetic markers in PZase or its promoter have been reported in the literature (31, 68, 73). As a result, if these mutations turned out to be phylogenetic markers, more resistant isolates would end up as having no genetic explanation for resistance, slightly decreasing the sensitivity of the gene in diagnostics.

Pyrazinamidase activity in pyrazinamide-resistant isolates.

PZase activity is an important measurement to determine the susceptibility to PZA because PZase converts PZA into POA, which is protonated and becomes active against the mycobacterium (2). A nonfunctional PZase allows the mycobacterium to survive in the presence of PZA (100–102). This concept is applied in tests for PZase activity.

The three tests (phenotypic, genotypic, and enzymatic) considered in this review stratified the isolates for which all three types of results were reported into eight groups. Groups 2 and 7 in Fig. S2 in the supplemental material hold isolates that have concordant results across all three tests. The remaining six groups have some type of discrepancy in the result of one of the three tests compared with those of the other two. The figure depicts the typical diagnostic conclusion that is made given the results. While the typical tendency is to side with the two tests that are concordant, we do not believe that the discordant results are always errors. Here, some discrepancies indicate a high probability of a test error but some might hold the potential for alternative explanations which, if proven, might significantly further our understanding of the mechanism of resistance to PZA.

The first interesting discrepant group is group 1 (Fig. S2 in the supplemental material), containing resistant isolates with detectable PZase activity and a mutation in PZase. This does not fit the currently accepted mechanism of PZA resistance. PZase activity is an indication of susceptibility. In terms of location on the gene, there is not a clustering of certain mutations that might allude to why these isolates still have PZase activity (see Table S10 in the supplemental material). One possibility is that the PZase assay (89% sensitivity) may have been incorrect or that the assay may have been misinterpreted (9). Another possibility is that these mutations may not affect the catalytic site and only lower PZase activity rather than completely eliminating it (103). Additionally, a possible reason for the phenotypic-genotypic discordance is heteroresistance (mixed populations). Among the more interesting explanations, though, we note the possibility of another element in the genome being involved and regulating the production (and possibly compensating for the reduced activity) of PZase.

The second discrepant group (group 3 in the figure) includes resistant isolates with detectable PZase activity and no mutations in PZase, which points to a possible DST inaccuracy. This group fits the commonly assumed error pattern of PZA DST, namely, false resistance.

Group 4 is the group of isolates that holds the strongest evidence for a mechanism of resistance other than pncA. This is the group of PZA^r isolates that do not have detectable PZase activity and also do not harbor any mutations in PZase or its promoter (2). While error in molecular testing is a possibility, this error rate has been significantly reduced with the introduction of new sequencing technologies and repeated molecular testing. An element elsewhere in the genome (or transcriptome) that regulates the pncA activity would be a good explanation and a very exciting discovery. Other possible explanations include false phenotypic resistance, sample contamination with other naturally PZA^r nontuberculous mycobacteria, and heteroresistance (mixed populations of PZA^r and PZA^s strains) (2). A repeat DST is the best first step in these cases in order to rule out the erroneous DST results.

Groups 5 and 6 include isolates that have been classified as susceptible by DST, but the genotypic results indicate a mutant PZase. Since false sensitivity is a much rarer event in PZA DST, isolates in these two groups potentially point to an alternative mechanism of resistance. Isolates belonging to these two groups are rare; thus, the two groups have much smaller populations than group 4.

Group 8 is the group of isolates that have been classified as susceptible by DST and molecular testing, but the enzymatic assay shows no enzyme activity. Like group 4, this group of isolates may indicate a complementary mechanism of resistance that regulates the function of pncA.

Alternative or complementary mechanism of resistance.

Eighteen percent of the resistant isolates did not have a mutation in PZase or its promoter. As mentioned above, while an incorrect phenotype might be the source of this inconsistency, there is a possibility that a complementary mechanism of resistance regulates the activity of pncA. Other inconsistencies such as cases with enzymatic activity but a resistant phenotype are harder to explain with a complementary mechanism and point more to an entirely different (alternative) mechanism of resistance. Of course, erroneous DST, in particular the case of false positives, might be the source of many of these inconsistencies. The current recommended solution to this issue is to raise the DST cutoff, which would make many borderline resistant cases susceptible. However, due to, at times, a wide range of MICs associated with certain mutations (see Table S10 in the supplemental material), raising the phenotypic cutoff might eliminate markers that serve as an explanation for some borderline resistant cases but also might eliminate some that have the same marker but a high MIC level (e.g., Trp34Asp reported in separate isolates with MIC levels of 100 mg/liter or >400 mg/liter) (57, 67). In such a case, the explanation for resistance (e.g., W34D) might no longer serve as an explanation of resistance for the higher MIC, and therefore an alternative mechanism of resistance would become prominent.

Further justification from the protein science point of view lies in the effect of the observed variations on the protein. For instance, an insertion of a C is observed in nucleotide 16 of the gene in a susceptible isolate (47). This variation causes a frameshift that results in the emergence of two stop codons in the middle of the gene, breaking the gene into three segments. It is difficult to understand how, after such a destruction of the protein, the enzymatic function remains undisturbed and the isolate still remains susceptible.

Alternative mechanisms of resistance have previously been suggested. Among these, fas1, rpsA, efflux pumps, and panD can be mentioned (13, 23, 24, 28). While some of these (e.g., rpsA) have since been refuted (25, 27), others, once confirmed, can serve as complementary targets for molecular diagnostics.

Limitations and future research.

Although inclusion and exclusion criteria were strictly adhered to, there are some limitations to this review. Most platforms used to test for PZA susceptibility are known to have inaccuracies (8). Statistics for regional populations may differ from those of their global counterparts presented in this study. Selection bias in studies included in this review might contribute to skewed statistics. For instance, some authors may have only chosen isolates from patients in whom treatment failed or may have chosen isolates based on a specific resistance profile, such as MDR- or XDR-TB. Several studies did not indicate the isolate selection process, which makes the assessment of selection bias impossible (8, 25, 51, 65, 67, 71, 75, 83, 84). More focused studies are needed in order to determine whether specific cohorts, such as PZA monoresistant strains, have a unique mechanism of PZA resistance.

The patient may have been infected with multiple strains, but the authors may have isolated only one strain for testing (104), or, if the authors did not perform species determination tests, they may have isolated nontuberculous mycobacteria (105). Moreover, several studies used different assays to determine PZA susceptibility. Although the cutoff concentrations to determine susceptibility should be comparable across different types of media, there may be discrepancies. All studies included in this review performed PCR to amplify pncA. Amplification is associated with a bias to GC-rich regions (106). This bias might have caused errors in the sequencing process.

Future studies should include the promoter region as well as the coding region when pncA is sequenced for PZA susceptibility determination. This improves the sensitivity of molecular-based diagnostic testing (see Table 1). Although pncA is considered nonessential, compensatory mutations should be studied. Compensatory mutations have been reported in RIF resistance in rpoC and might also be present in PZA resistance (107, 108).

Concluding remarks.

PZA is an important first-line drug in treating active and latent TB (109). PZA is also important in treating HIV-positive patients because the drug does not exhibit drug-drug interactions with antiretroviral therapy (110). However, detection of PZA resistance is challenging and threatens early intervention and treatment of complex TB cases. Questions about existing PZA susceptibility testing methods suggest a need for improved or new diagnostics. In addition, mutations in PZase are proposed to serve as markers for PZA resistance. Nonetheless, individual mutations in PZase alone are not enough to diagnose resistance, and molecular-based platforms need to consider the entire gene and its promoter, as suggested previously (9). More than 13% of the resistant isolates in this review did not harbor any mutations in PZase or its promoter. While the accuracy of the susceptibility testing and sampling bias may increase or decrease this percentage, it is highly probable that, as with all other drugs, pncA and its promoter will not explain all PZA-resistant cases. The search for an alternative or complementary mechanism is therefore important for the development of a highly sensitive diagnostic platform.