Overcoming the Challenges of Pyrazinamide Susceptibility Testing in Clinical Mycobacterium tuberculosis Isolates

ABSTRACT

Pyrazinamide (PZA) is one of the first-line agents used for the treatment of tuberculosis. However, current phenotypic PZA susceptibility testing in the Bactec MGIT 960 system is unreliable, and false resistance is well documented. Rapid identification of resistance-associated mutations can confirm the phenotypic result. This study aimed to investigate the use of genotypic methods in combination with phenotypic susceptibility testing for confirmation of PZA-resistant Mycobacterium tuberculosis isolates. Sanger sequencing and/or whole-genome sequencing were performed to detect mutations in pncA, rpsA, panD, and clpC1. Isolates were screened for heteroresistance, and PZA susceptibility testing was performed using the Bactec MGIT 960 system using a reduced inoculum to investigate false resistance. Overall, 40 phenotypically PZA-resistant isolates were identified. Of these, PZA resistance was confirmed in 22/40 (55%) isolates by detecting mutations in the pncA, rpsA, and panD genes. Of the 40 isolates, 16 (40%) were found to be susceptible using the reduced inoculum method (i.e., false resistance). No mutations were detected in two PZA-resistant isolates. False resistance was observed in isolates with MICs close to the critical concentration. In particular, East African Indian strains (lineage 1) appeared to have an elevated MIC that is close to the critical concentration. While this study illustrates the complexity and challenges associated with PZA susceptibility testing of M. tuberculosis, we conclude that a combination of genotypic and phenotypic drug susceptibility testing methods is required for accurate detection of PZA resistance.

INTRODUCTION

Pyrazinamide (PZA) is an effective first-line agent for the treatment of tuberculosis (TB), especially for multidrug-resistant (MDR) TB cases, which are resistant to at least rifampin and isoniazid. PZA is a prodrug and is converted to an active form by the enzyme pyrazinamidase (PZase), which is encoded by pncA in Mycobacterium tuberculosis (1). The essential role of PZA in the treatment of TB emphasizes the importance of accurate and rapid detection of drug resistance. However, PZA is only active at a lower pH, and susceptibility testing must be performed in acidic conditions. This causes unreliable PZA susceptibility results where M. tuberculosis is required to grow in less favorable conditions (2, 3). False resistance to PZA using the Bactec MGIT 960 system is well documented and is mainly due to a high M. tuberculosis inoculum, which may reduce or impair PZA activity by increasing the pH of the test medium (4). This gives rise to the need for a confirmatory assay where phenotypic resistance to PZA has been identified (5, 6).

Molecular identification of mutations in pncA and its promoter region can offer the potential for rapid detection of PZA resistance in M. tuberculosis. Mutations leading to loss of PZase activity is the major mechanism leading to resistance (1). Although the majority of PZA-resistant cases are related to pncA mutations, there is a large spectrum of resistance variants which are highly diverse and scattered over the full length of the pncA gene, including its promoter (7). The absence of a “hot spot region” comprising the majority of mutations complicates the development of a commercial diagnostic assay; therefore, DNA sequencing of the full pncA gene is considered the most reliable method for detection of PZA resistance (1). PZA resistance in strains without pncA mutations is rare, and mutations in potential targets such as rpsA (encoding the aspartate decarboxylase enzyme involved in coenzyme A biosynthesis) and panD (encoding ribosomal protein S1 involved in trans-translation) may be involved in resistance, but to a lesser extent than pncA (8). More recently, clpC1 (encoding an ATPase involved in protein degradation) has been reported to be associated with PZA resistance (9). As of yet, the WHO has not approved the use of molecular techniques for PZA susceptibility testing, and therefore it is still necessary to perform phenotypic susceptibility testing for PZA (10). This study aimed to investigate the use of Sanger sequencing and whole-genome sequencing (WGS), in combination with phenotypic susceptibility testing using a reduced inoculum, for confirmation of PZA-resistant M. tuberculosis isolates.

RESULTS

Detection of pncA mutations associated with pyrazinamide resistance.

A summary of pncA sequencing results is shown in Table 1. No mutations associated with PZA resistance were detected in pncA or the promoter region for all 134 PZA-susceptible M. tuberculosis isolates. Phenotypically PZA-susceptible isolates displayed either a wild-type (WT) pncA or the presence of a synonymous mutation (S65S, V155V, and V180V), indicating no association with resistance (i.e., neutral). L35R was detected in one PZA-susceptible isolate and does not confer resistance.

TABLE 1.

pncA sequencing results of M. tuberculosis clinical isolates (n = 174)

No. of isolates	PZA susceptibilitya	pncA sequencing result	Interpretation of mutationb
130	S	WT	No mutation detected
1	S	S65S	Group E: neutral
1	S	V155V	Group E: neutral
1	S	V180V	Group E: neutral
1	S	L35R	Group E: neutral
1	R	A134V	Group A: associated with resistance
1	R	G108R	Group A: associated with resistance
2	R	L4S	Group A: associated with resistance
2	R	a-11g	Group A: associated with resistance
1	R	53-nucleotide deletion	Group A: associated with resistance
1	R	Ins 391 GG	Group A: associated with resistance
1	R	9-nucleotide deletion	Group B: likely associated with resistance
2	R	H51Y	Group B: likely associated with resistance
1	R	V180A	Group B: likely associated with resistance
1	R	D136Y, del-125, S65S	Group B: likely associated with resistance
2	R	G105V	Group B: likely associated with resistance
1	R	I31S	Group B: likely associated with resistance
1	R	H51D	Group B: likely associated with resistance
1	R	H51P	Group B: likely associated with resistance
22	R	WT	No mutation detected

^aS, susceptible; R, resistant.

^bpncA mutations were interpreted based on five groups (groups A to E) proposed by Köser et al. to classify resistance mutations (11). All synonymous mutations were classified as neutral.

In contrast, mutations associated with PZA resistance were detected in the pncA gene and/or promoter region of 18/40 (45%) PZA-resistant M. tuberculosis isolates. These were classified as associated with resistance (group A) or likely associated with resistance (group B) (11). Discordant phenotypic and genotypic drug susceptibility testing (DST) results were observed for 22/40 (55%) PZA-resistant M. tuberculosis isolates where no mutations were detected in the pncA or promoter region.

Investigation of PZA-resistant and pncA WT isolates for heteroresistance.

PZA susceptibility testing was repeated using a standard inoculum with the Bactec MGIT 960 system for 22 isolates with discrepant phenotypic and genotypic DST results. On repeat, 19/22 were phenotypically resistant and 3/22 were susceptible to PZA. Three PZA-susceptible isolates were excluded from heteroresistance testing, and Sanger sequencing of the pncA gene and promoter region was repeated on growth from the Bactec MGIT tube containing pyrazinamide for 19 PZA-resistant isolates to screen for heteroresistance. No pncA mutations were detected in any of the isolates.

Investigation of other PZA resistance mechanisms using WGS.

WGS was performed on 19 M. tuberculosis isolates that were PZA resistant and pncA WT to screen for mutations in potential target genes (rpsA, panD, clpC1, and their promoters). A summary of WGS results is shown in Table 2. WGS showed that four PZA-resistant isolates harbored nonsynonymous mutations in rpsA and panD. The rpsA mutation V260I was detected in two isolates, and t-1833510c was detected in the rpsA promoter region of one isolate. Similarly, the panD mutation I115T was detected in one PZA-resistant isolate belonging to the LAM lineage. The clpC1 mutation V63A was detected in 15 isolates belonging to the East African-Indian (EAI) sublineage (lineage 1, Indo-Oceanic). Synonymous mutations in clpC1 were also detected in 2/15 EAI strains (L15L and V697V) and one isolate belonging to LAM lineage (N806N). No mutations were detected in rpsA, panD, or clpC1 genes for two isolates.

TABLE 2.

Analysis of 19 M. tuberculosis isolates selected as PZA resistant and pncA wild type^a

Isolate no.	Original DST result for:				WGS result for:				Reduced inoculum DST result (PZA)		MIRU VNTR lineage
	RIF	INH	EMB	PZA	pncA	rpsA	panD	clpC1b	50 μg/ml	100 μg/mld
Alternative mechanism (n = 4)
16IE38	S	S	S	R	WT	V260I	WT	V63A, L15L	R	R	EAI
19IE45	S	S	S	R	WT	V260I	WT	V63A	R	R (GU 213)	EAI
17IE30	S	S	S	R	WT	WT	I115T	N806N	R	R	LAM
18IE32c	S	S	S	R	WT	Promoter t-1833510c	WT	WT	NA	NA	S
Unknown mechanism (n = 2)
19IE132	S	S	S	R	WT	WT	WT	V63A	R	R (GU 204)	EAI
16IE36	S	S	S	R	WT	WT	WT	V63A	R	R	EAI
False resistance (n = 13)
16IE46	S	S	S	R	WT	WT	WT	WT	R	S	Euro American
16IE45	S	S	S	R	WT	WT	WT	WT	R	S	Euro American
16IE07	S	S	S	R	WT	WT	WT	V63A	R	S	EAI
16IE37	S	S	S	R	WT	WT	WT	V63A	R	S	EAI
17IE29	S	S	S	R	WT	WT	WT	V63A	R	S	EAI
20IE83	S	S	S	R	WT	WT	WT	V63A	R	S	EAI
18IE42	R	R	S	R	WT	WT	WT	V63A	R	S (GU 65)	EAI
18IE34	S	S	S	R	WT	WT	WT	V63A, V697V	R	S	EAI
18IE41	S	S	S	R	WT	WT	WT	V63A	R	S	EAI
18IE33	S	S	S	R	WT	WT	WT	V63A	R	S	EAI
20IE85	S	S	S	R	WT	WT	WT	V63A	S	S	EAI
18IE31	S	S	S	R	WT	WT	WT	V63A	S	S	EAI
20IE80	S	S	S	R	WT	WT	WT	V63A	S	S	EAI

^aAbbreviations: S, susceptible; R, resistant; GU, growth unit; RIF, rifampin; INH, isoniazid; EMB, ethambutol; PZA, pyrazinamide; ND, not determined.

^bV63A in clpC1 is a phylogenetic mutation linked with East African Indian (EAI) strains.

^c18IE32 failed to grow subsequently for PZA susceptibility testing using a reduced inoculum.

^d100 μg/ml PZA represents the critical concentration for drug susceptibility testing of M. tuberculosis.

In addition, whole-genome single-nucleotide polymorphism (SNP) analysis was performed to determine the genetic relatedness of isolates sharing the same phenotype and genotype (Table 2). None of these isolates were clonal, except for two which were designated Euro-American lineage (data not shown). The finding of V63A in clpC1 among unrelated EAI strains suggests a phylogenetic link. To further investigate the role of V63A in clpC1, WGS data of 128 PZA-susceptible isolates were analyzed (see Table S1 in the supplemental material). V63A in clpC1 was found in all 22 PZA-susceptible isolates belonging to EAI. Furthermore, M432T, N75G, and R212R mutations in rpsA were also detected in PZA-susceptible isolates.

Pyrazinamide susceptibility testing using a reduced inoculum.

PZA susceptibility testing was performed using a reduced inoculum to investigate false resistance associated with the Bactec MGIT 960 system on PZA-resistant and pncA WT isolates (Table 2). Three PZA-resistant M. tuberculosis isolates with mutations in rpsA and panD were included as resistant controls. Three PZA-resistant isolates with rpsA or panD mutations were resistant at 50 μg/ml and 100 μg/ml PZA, while one isolate (18IE32) failed to culture and could not be tested. Two isolates belonging to the EAI lineage without mutations associated with PZA resistance remained resistant at 50 μg/ml and 100 μg/ml PZA. In contrast, 10 isolates retested as susceptible to 100 μg/ml PZA but resistant to 50 μg/ml PZA when using a reduced inoculum (among them, 1 MDR isolate [18IE42] did not show clear susceptibility when tested at 100 μg/ml PZA [growth unit (GU) = 65]). Three isolates were also found to be susceptible at 50 μg/ml and 100 μg/ml PZA. Interestingly, false resistance was observed in 11 isolates belonging to EAI, and among them, 8/11 had elevated MICs close to the critical concentration (i.e., resistant at 50 μg/ml and susceptible at 100 μg/ml PZA) when tested using a reduced inoculum method.

Furthermore, 10 PZA-susceptible isolates harboring V63A in clpC1 were tested at 50 μg/ml and 100 μg/ml PZA using a reduced inoculum to gain more insight into this particular genotype (Table S2). Overall, 6/10 isolates were resistant to 50 μg/ml and susceptible to 100 μg/ml PZA (among them, 2 isolates did not show clear susceptibility at 100 μg/ml). On the other hand, 4/10 isolates were susceptible to both 50 μg/ml and 100 μg/ml PZA (2 which did not show clear susceptibility when tested at 50 μg/ml).

DISCUSSION

The Bactec MGIT 960 system is the method recommended by the WHO for PZA susceptibility testing, although it is associated with a relatively high rate of false resistance in some cases (6, 12). DNA sequencing of the pncA gene is considered the most reliable genotypic method for detection of pyrazinamide resistance, but evidence of non-pncA mutations associated with PZA resistance is emerging (8, 13). PZA resistance in M. tuberculosis is most commonly associated with mutations in the pncA gene or promoter region, which leads to reduced PZase activity. In this study, mutations were detected in 18/40 (45%) PZA-resistant isolates by Sanger sequencing of the pncA gene and promoter region. A method for interpreting the association between mutations and phenotypic drug resistance in M. tuberculosis was standardized by Miotto et al., who recently proposed five groups of pncA mutations to inform the use of PZA (11, 14). While most pncA mutations are categorized as group A, associated with resistance, or group B, likely associated with resistance, knowledge gaps still exist for novel pncA mutations with inconclusive evidence for resistance and resistance caused by mutations in genes such as rpsA, panD, and clpC1 (3, 7, 14–16). Köser et al. also proposed that pncA mutations should be assumed to cause resistance unless disproven (11). Likewise, frameshift and in-frame mutations in the pncA gene can, respectively, be classified as associated or likely associated with resistance, as they are predicted with confidence to cause a loss of functional phenotype (11, 14, 17).

Phenotypic PZA-resistant isolates without pncA mutations can be attributable to false resistance, other resistance mechanisms, and/or heteroresistance (18). Although PZA-resistant strains without pncA mutations are rare, we found that 22/40 (55%) PZA-resistant isolates had no mutations in the pncA gene or its promoter region. PZA-resistant isolates carrying a wild-type pncA were mainly mono-PZA resistant (n = 18), with single isolates being associated with multidrug resistance or streptomycin or isoniazid mono-resistance. Sanger sequencing of the pncA and promoter region confirmed PZA resistance in 16/17 (94%) MDR TB isolates identified in the Irish Mycobacteria Reference Laboratory (IMRL). Of note, it has been estimated that 60.5% of MDR M. tuberculosis strains are likely to be PZA resistant (19). Although we did not detect any heteroresistance, Werngren et al. showed that 8/26 (30.7%) PZA-resistant and pncA WT isolates were associated with heteroresistance, as low-frequency mutations could be missed by Sanger sequencing (8).

Mutations in target genes such rpsA and panD have been reported in PZA-resistant isolates but to a lesser extent than in pncA, and evidence of mutations linked to clpC1 is emerging (20, 21). The role of rpsA in PZA resistance is widely debated in the literature, with different publications supporting and dismissing the role of this gene (22–26). While rpsA is thought not to be involved in PZA resistance in some studies, it remains unclear how rpsA mutations are associated with PZA resistance in clinical M. tuberculosis isolates, and further research is needed to determine this relationship (13, 27). The rpsA mutation V260I was detected in two of our PZA-resistant isolates belonging to the EAI lineage, and this is consistent with a previous study (8). V260I in rpsA has been suggested to be a phylogenetic marker for the EAI lineage and was associated with low-level resistance (8). A mutation in the rpsA promoter was also detected in one PZA-resistant isolate which had not been previously reported. Likewise, we detected the panD mutation I115T in one PZA-resistant isolate belonging to LAM lineage, and this mutation was also previously reported in two PZA-resistant isolates (8). More recently, phylogenetic mutations in genes implicated in antibiotic resistance in M. tuberculosis have been studied to generate a comprehensive database for interpreting genotypic DST results, although clpC1 was not investigated (28). We also identified some mutations in rpsA and clpC1 in PZA-susceptible isolates which appeared to be phylogenetically linked and do not correlate with PZA resistance. Further research is required to characterize mutations in rpsA, panD, and clpC1.

Ten isolates were classified as PZA susceptible at the critical concentration (100 μg/ml) when using the reduced inoculum method, while being resistant to 50 μg/ml of PZA. This suggests that isolates with MICs close to the critical concentration can yield inconsistent results, making PZA susceptibility testing challenging. Our results demonstrate that a reduced inoculum method could be used to confirm true PZA-resistant results obtained with standard Bactec MGIT DST. Piersimoni et al. suggested that isolates tested as PZA resistant using a standard inoculum should undergo repeat testing using a reduced inoculum and that pncA sequencing should be performed in cases of resistance (5). In this study, we showed that testing at both 50 μg/ml and 100 μg/ml with a reduced inoculum is useful to identify if isolates have an MIC close to the critical concentration, thereby predicting possible false resistance. The identification of isolates which have an MIC below but close to the critical concentration (i.e., resistant to 50 μg/ml and susceptible to 100 μg/ml) may have clinical implications, as some pncA mutations can lead to false PZA susceptibility results (i.e., T47A and I31T in pncA are likely to have an MIC below or at the critical concentration) (11). Köser et al. suggested that PZA susceptibility testing at 50 μg/ml and 100 μg/ml PZA can be used to refine the classification of pncA mutations (11). Previously, it was suggested that the critical concentration should be reassessed based on a single study in Sweden which attempted to reevaluate the critical concentration of PZA by testing 46 clinical isolates characterized by pncA mutations (29). However, it is unclear whether the current critical concentration of PZA (100 μg/ml) corresponds to the epidemiological cutoff value, as this has not been rigorously assessed using a reliable susceptibility testing method (11). Nevertheless, we suggest that retesting at 100 μg/ml with a reduced inoculum can mitigate random false resistance in routine PZA susceptibility testing based on current knowledge.

Likewise, it was recently reported that lineage 1 was overrepresented in a Swedish collection of mono-PZA-resistant isolates compared to PZA-resistant isolates in a previous study where isolates were selected based on MDR phenotype (8, 30). Modlin et al. also reported the identification of V63A in clpC1 among their PZA-monoresistant isolates, suggesting its association with low-level PZA resistance based on testing 100 μg/ml and 200 μg/ml PZA using the standard Bactec MGIT testing method (30). This is similar to our finding that EAI strains (lineage 1) appear to have an elevated background MIC when tested using a reduced inoculum method. However, further research is needed to ascertain the role of V63A in clpC1 and its association with low-level PZA resistance, given that all EAI isolates harbored this mutation and 72% (18/25) had elevated MICs (i.e., resistant to 50 μg/ml and susceptible to 100 μg/ml PZA) in our study. In addition, the clinical implication of this finding is unclear. More clinical outcome data are needed to determine whether the current dose of PZA is effective for strains with elevated MIC or if a higher dose may compensate for low-level resistance (11). Of note, two isolates belonging to EAI were tested as resistant to 50 μg/ml and 100 μg/ml PZA when using the reduced inoculum method and only harbor V63A in clpC1. Further research is required to determine the range of resistance in these isolates. However, other unknown mechanisms conferring PZA resistance may still exist, which would warrant continued phenotypic testing.

In conclusion, monoresistance to PZA was observed mostly in EAI strains, and those with MICs close to the critical concentration were prone to false resistance. The sensitivity was 55% (22/40) and specificity was 100% (134/134) for sequencing pncA, panD, and rpsA to confirm PZA resistance obtained using a standard inoculum in the Bactec MGIT system. Following the investigation of PZA-resistant isolates using a reduced inoculum, the sensitivity of sequencing pncA, panD, and rpsA increased to 91.6% (22/24), and specificity was 100% (150/150). More recently, WGS of 10,209 M. tuberculosis isolates showed potential for predicting pyrazinamide susceptibility with 91.36% sensitivity and 96.8% specificity based on pncA sequencing (31). However, WGS is not yet implemented routinely in most diagnostic laboratories, and there is a need to correlate PZA mutation data with levels of phenotypic drug resistance in order to inform clinical decision-making (10).

This study illustrates the complexities and challenges associated with PZA susceptibility testing of M. tuberculosis. A combination of genotypic and phenotypic DST methods is required for accurate confirmation of PZA resistance. Monoresistance to PZA is uncommon, and determination of the MIC using a reduced inoculum based on the Bactec MGIT 960 method could be performed to aid in the interpretation of genotypic DST data while minimizing false resistance.

MATERIALS AND METHODS

M. tuberculosis clinical isolates.

A total of 174 nonduplicate M. tuberculosis clinical isolates were included in this study. All PZA-resistant isolates (n = 40), including both mono-resistant and MDR-TB received in the Irish Mycobacteria Reference Laboratory (IMRL) between 2013 and 2020 were selected. A subset of PZA-susceptible isolates (n = 134) representing the major M. tuberculosis lineages circulating in Ireland between 2009 and 2020 were also selected. Susceptibility testing was originally performed using the Bactec MGIT 960 system (Becton, Dickinson, New Jersey, United States) as part of routine diagnostics. Mycobacterial interspersed repetitive units variable number tandem repeats (MIRU-VNTR) genotyping was also performed on all isolates as part of public health surveillance in Ireland.

Sanger sequencing of pncA and investigation of heteroresistance.

DNA extraction was performed using the GenoLyse DNA extraction kit (Bruker-Hain Diagnostics, Nehren, Germany) and Sanger sequencing of the pncA, and its promoter was performed to detect mutations associated with PZA resistance. The pncA primer sequences used in this study were provided by the Supranational Reference Laboratory for Tuberculosis in Milan, Italy. PZA susceptibility testing using the Bactec MGIT 960 system was repeated for resistant isolates with no mutations detected in the pncA or its promoter. Sanger sequencing was performed from M. tuberculosis growth in the MGIT culture tube containing PZA, as previously described, to detect heteroresistance (8). pncA mutations were interpreted based on five groups (groups A to E) proposed by Köser et al. to classify PZA resistance mutations (11).

Investigation of other resistance mechanisms using WGS.

WGS was performed on PZA-resistant isolates where no pncA mutations were detected with Sanger sequencing (n = 19) and on PZA-susceptible isolates where no pncA Sanger sequencing had previously been performed (n = 128). Genomic DNA was extracted from heat-inactivated cultures using the QuickGene DNA tissue kit S (Kurabo, Osaka, Japan), and DNA library preparation was performed using the Nextera XT DNA library preparation kit (Illumina, USA). Paired-end sequencing was performed on the Illumina MiniSeq or MiSeq sequencing system. WGS analysis was performed using the MTBseq pipeline v1.04 with default parameters. Mutations in pncA, rpsA, panD, and clpC1 were searched for manually (32). Low-frequency mode was also used to detect low-frequency mutations with one read in both the forward and reverse directions, a phred score of at least 20, and if the allele was indicated by a minimum frequency of 5% in mapped reads.

Pyrazinamide susceptibility testing using a reduced inoculum.

PZA susceptibility testing was performed using a reduced inoculum in the Bactec MGIT 960 system as previously described by Piersimoni et al. (5). Briefly, the drug susceptibility testing (DST) inoculum was prepared by using a positive MGIT liquid culture on day 1 or 2. Homogenization was performed by vortexing with sterile glass beads, and the suspension was left to sediment for 30 min before transferring approximately 2 ml to a new universal container. A reduced inoculum volume (250 μl) of this culture was added to a Bactec MGIT PZA medium culture tube (Becton, Dickinson, New Jersey, USA). Similarly, a 1:10 dilution of the culture was prepared, and 250 μl was used to inoculate the growth control tube. PZA susceptibility testing was performed using drug concentrations of 50 μg/ml and 100 μg/ml. Susceptibility to PZA was defined as a growth unit (GU) level of <100 in the culture tube containing PZA, and resistance was defined as a GU level of ≥100 in the culture tube containing PZA.

Data availability.

Raw reads were submitted to the European Nucleotide Archive under project accession number PRJEB42011.