We performed a cross-sectional study examining pncA polymorphisms in M. tuberculosis isolates from a convenience sample of patients with MDR or XDR TB who were prospectively enrolled in studies from KwaZulu-Natal, South Africa (n = 451, diagnosed 2011–2014), and Georgia (n = 103, diagnosed November 2011–April 2012) (Technical Appendix). Cultures and DST were performed at the provincial TB reference laboratory in Durban, South Africa, and the National Reference Laboratory in Tbilisi, Georgia. Samples underwent PCR amplification followed by standard capillary sequencing of the pncA promoter and coding DNA sequence at the Public Health Research Institute in Newark, New Jersey, USA, as previously described (9). Polymorphisms were identified by alignment of nucleotide sequences to the H37Rv reference strain by using ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/). We calculated the frequency of each pncA polymorphism and classified mutations as synonymous or nonsynonymous. We compared polymorphisms with those reported in the literature to identify any that are known to be associated with phenotypic susceptibility and to determine the proportion that are likely to confer phenotypic resistance (10–15).

To determine the effect of clonal expansion of MDR and XDR TB on the diversity of pncA mutations, we also compared IS6110-based restriction fragment-length polymorphism (RFLP) patterns with pncA mutations for the isolates from South Africa (conducted at the Public Health Research Institute). The distribution of RFLP patterns among isolates with identical pncA mutations was examined to determine if the pncA mutation arose de novo or may have been transmitted.

We completed targeted pncA gene sequencing for 554 unique patient-isolates [1 isolate/patient], 167 MDR TB and 387 XDR TB; of these, 99 (59%) of MDR TB and 215 (56%) of XDR TB patients had previously received treatment for TB. A pncA polymorphism was found in 117 (70%) MDR TB and 370 (96%) XDR TB isolates (Tables 1, 2). The proportion of MDR TB and XDR TB isolates with pncA polymorphisms did not differ significantly between those from South Africa and Georgia (MDR, 68% vs. 72%, p = 0.74; XDR, 96% vs. 90%, p = 0.73, respectively).

A total of 69 distinct pncA polymorphisms were identified (Tables 1 and 2). Of these, 12 were insertions (313 patient-isolates), 5 were deletions (6 patient-isolates), and 52 were single-nucleotide polymorphisms (SNPs; 168 patient-isolates); all but 2 of these were nonsynonymous. No polymorphism was found in common between the isolates from South Africa and Georgia. Among the pncA SNPs identified, only 6 (9 patient-isolates) have previously been associated with phenotypic pyrazinamide susceptibility (10–12); 40 SNPs have been associated with phenotypic pyrazinamide resistance (146 patient-isolates), and 7 SNPs were not previously reported (23 patient-isolates) (10–15).

There were 34 polymorphisms identified from South Africa, of which 14 (41%, constituting 388 patient-isolates) were present in >1 patient (Table 1). We found that, for 382 (98%) of 388 patients, the RFLP pattern was identical to that of at least 1 other patient with the same pncA mutation (Table 1). Moreover, each pncA polymorphism was associated with only 1 RFLP pattern in 10 of the 14 polymorphisms. By comparison, 13 RFLP patterns were seen among the 40 patients with a wild-type pncA sequence.

In this study, we found that 70% of MDR TB and 96% XDR TB patient-isolates had pncA polymorphisms. Given the high likelihood of frameshift mutations resulting in resistance and the high specificity (94%–98%) of pncA SNPs for pyrazinamide resistance (13,14), we estimate that at least 56%–66% of MDR TB and 90%–95% of XDR TB cases from these settings are likely to be resistant to pyrazinamide. Only a small number of mutations were synonymous, previously associated with pyrazinamide susceptibility, or had an SNP for which phenotypic susceptibility has not been previously tested. This finding has implications regarding the effectiveness of empiric use of pyrazinamide for drug-resistant TB or novel treatment regimens. Further studies are needed to fully determine the association of pncA mutations with treatment outcomes.

A diversity of pncA mutations—69 distinct polymorphisms—were observed among MDR and XDR TB patients, of which none were shared in common between the isolates from South Africa and Georgia. Most pncA polymorphisms were unique to individual patients. When the same pncA polymorphism was seen in >1 patient, the IS6110 RFLP pattern was nearly always similar, suggesting that the pncA mutation was acquired before transmission. The diversity of polymorphisms underscores previous findings that there is no clear hotspot for pncA mutations (12), unlike resistance-conferring regions for other TB drugs (e.g., rpoB, katG) (11). Development of rapid molecular tests for pyrazinamide susceptibility may be hampered by the lack of a hotspot for mutations; 1 assay has been developed to detect the full wildtype pncA sequence, but its diagnostic accuracy has not yet been adequately tested (7).

A limitation of our study is that phenotypic pyrazinamide susceptibility testing was not performed on the sequenced isolates. Nonetheless, correlation with phenotypic testing has been previously reported in the literature for most polymorphisms, enabling us to estimate the proportion likely to be pyrazinamide resistant (10–15). In addition, the studies that provided these isolates were not specifically designed to be representative of all diagnosed cases of drug-resistant TB; nonetheless, the study populations were carefully selected to provide a high level of generalizability to the broader population. National drug resistance surveys that include pyrazinamide genotypic and phenotypic susceptibility should be designed to confirm these findings.

The high prevalence of pncA polymorphisms from geographically disparate countries suggests that guidelines to empirically use pyrazinamide in drug-resistant TB regimens, including shorter MDR TB regimens (4), should be reconsidered. Simplified assays to test pyrazinamide susceptibility are needed, although they may be difficult to develop given the genotypic or phenotypic complexities. Considering the potential synergy of pyrazinamide with new TB drugs, routine assessment of pyrazinamide will be increasingly necessary and useful.