Pyrazinamide Resistance Is Caused by Two Distinct Mechanisms: Prevention of Coenzyme A Depletion and Loss of Virulence Factor Synthesis

Missense Mutations in a CoA Synthesis Gene and Loss of Function Mutations in PDIM Synthesis Genes Cause POA Resistance

We previously showed that, in contrast to common belief, PZA and POA exert growth inhibitory activity against M. tuberculosis at nearly neutral pH 6.5.⁵ Here, we tested the hypothesis that employing nearly neutral instead of acidic agar, as has been attempted unsuccessfully in the past, would enable the isolation of POA-resistant mutants.

(Note that during the course of this work Zhang et al.¹⁷ and Lanoix et al.³¹ showed the feasibility of this approach. See the Discussion.) To facilitate this work, we used Mycobacterium bovis BCG as a biosafety level 2 compatible surrogate for M. tuberculosis. M. bovis BCG is naturally pyrazinamidase-negative (PncA⁻) because of a point mutation (C169G/His57Asp) in the amidase gene pncA and hence is not susceptible to PZA.⁴ M. bovis BCG exhibits similar susceptibility to POA in vitro under both nearly neutral (pH 6.5) and acidic (pH 5.8) conditions, with MIC₅₀ values of 0.5–1 and 0.25 mM, respectively (1 mM POA corresponds to 124 μg/mL).

Pilot experiments to determine broth MICs using standard (pH 6.5) growth media with and without glycerol surprisingly showed that glycerol hypersensitizes wild-type M. bovis BCG to POA (Supporting Information, Figure S1A). Increasing the glycerol concentration from 0.01 to 0.8% resulted in a reduction of the MIC₉₀ of POA from 4 to 1 mM (Supporting Information, Figure S1B). This shows that the POA susceptibility is influenced by the presence of glycerol, a phenomenon that we and others observed previously with other compound classes during antimycobacterial drug screening.^32,33 On the basis of the observed glycerol effect, we used agar containing 0.5% glycerol or agar without glycerol, i.e., with glucose as the main carbon source, for resistant mutant selection. Four independent BCG cultures were plated on the two agar types supplemented with 1, 2, or 4 mM POA. POA-resistant colonies were observed at spontaneous resistance mutation frequencies of 10⁻⁴ to 10⁻⁵/CFU in the presence of glycerol, but in the absence of glycerol, the frequencies were 10⁻³ to 10⁻⁴/CFU. To confirm the resistance, a total of 26 colonies from both screens were restreaked on agar containing corresponding concentrations of POA. The resistance of the 26 strains was further confirmed by the determination of broth MICs via serial dilution in broth with and without glycerol. The POA-resistant strains could be grouped into four phenotypic classes, POA1 to POA4, according to their POA resistance levels and the effect of glycerol on resistance. (See Table 1 for representative strains and Supporting Information, Table S1, for other strains.)

To determine the molecular mechanisms of resistance, the 26 strains were subjected to whole genome sequencing. Rather than finding polymorphisms in the proposed POA targets fas1 and rpsA encoding the fatty acid synthase FAS I and the ribosomal protein S1, respectively, we found mutations in the aspartate decarboxylase panD involved in CoA biosynthesis^16,17 and in the mycocerosic acid synthase mas as well as the phenolpthiocerol synthesis type-I polyketide synthases ppsA-E (Table 1 and Table S1). Both Mas and PpsA-E are involved in the synthesis of the cell wall lipid phthiocerol dimycocerosate (PDIM), a virulence factor required for in vivo but not in vitro growth.^34,35 We also found frameshift mutations in the glycerol kinase glpK that catalyzes the first step of glycerol catabolism.³⁶ In contrast to mutations in panD and mas or ppsA-E, mutations in glpK were specific to POA-resistant strains isolated on agar containing glycerol and were observed only in combination with mutations in mas or ppsA-E, suggesting that glpK mutations alone may not confer POA resistance.

An analysis of the whole genome sequencing results revealed that amino acid sequence-altering mutations mapped mostly to the C-terminal mycobacterium-specific putative regulatory domain of PanD.^16,17,37 PDIM biosynthesis genes mas and ppsA-E showed mostly frameshift and therefore likely loss-of-function mutations distributed across different domains. glpK mutations were all frameshift loss-of-function mutations.³⁶ The biochemical reactions catalyzed by the various enzymes and details of the mutations and their localization within the PanD, Mas/PpsA-E, and GlpK proteins are summarized in Figure 1.

The 26 POA-resistant strains fell into 4 genotypic classes, and the genotypic classes correlated with the 4 observed phenotypic classes (Table 1 and Supporting Information, Table S1): Class POA1 carries panD plus mas or ppsA-E double mutations, conferring high-level POA resistance (POA^R++) in both the presence and absence of glycerol. Class POA2 harbors panD mutations alone, conferring medium-level POA resistance (POA^R+) in the presence and absence of glycerol. Class POA3 carries mas or ppsA-E mutations alone, conferring low-level resistance (POA^R) in media without glycerol. In media with glycerol, class POA3 is hypersensitive to POA (POA^S+), similar to the wild type. Class POA4 combines glpK and mas or ppsA-E double mutations conferring low-level resistance (POA^R) in the absence of glycerol (similar to class POA3) and medium-level resistance (POA^R+) in the presence of glycerol (Table 1 and Supporting Information, Table S1).

To simplify subsequent analyses, we selected one representative strain for each mutant class as shown in Table 1: M. bovis BCG POA1 [panD1 mas-1], POA2 [panD2], POA3A [mas-1], POA3B [ppsA1], and POA4 [glpK1 mas-1]. The POA-resistant phenotypes for the five representative strains in media without and with glycerol can be seen in Figure 2A,B, respectively. Glycerol sensitizes all strains to POA irrespective of genotype but to different degrees (Figure 2B). This results in a complete loss of resistance for low-resistance-class POA3 (mas or ppsA-E) mutants whereas the low-resistance-class POA4 mutants, containing an additional glpK mutation, retain their resistance in the presence of glycerol.

Mutant selection and all phenotypic characterizations so far were carried out at neutral pH. Because previous POA mutant selection experiments on acidic pH agar had failed to deliver resistant strains,^9,15 we predicted that the representative POA-resistant strains would exhibit their resistance at neutral pH but not at acidic pH. The results from broth dilution MICs carried out at pH 5.8 (Figure 2C) indeed show that POA resistance mutations do not confer POA resistance under acidic pH, which is consistent with the inability to isolate POA-resistant mutants on acidic pH agar.

POA-Resistant Mutations Also Confer PZA Resistance

In our studies, we directly used bioactive POA for the selection of resistant mutants to avoid isolating pncA mutants that constitute the vast majority of PZA resistance in vitro and in vivo.¹⁵ To determine whether POA resistance mutations confer resistance to prodrug PZA, we introduced a functional copy of the pncA gene from M. tuberculosis H37Rv into the five representative mutants and wild-type M. bovis BCG using an integrative plasmid construct. As expected, the introduction of M. tuberculosis pncA into wild-type BCG rendered the previously PZA-resistant strain PZA-susceptible⁴ (Figure 2D). In contrast, the five representative POA-resistant strains remained resistant to PZA after complementation with functional pncA (Figure 2D). Thus, the POA resistance conferring mutations also confers resistance to PZA. As observed previously with POA at acidic pH, the five representative strains complemented with functional pncA were susceptible to PZA at acidic pH 5.8 (Figure 2E).

To exclude the possibility that the identified POA/PZA-resistant genotypes caused unspecific drug resistance, we measured the susceptibility of the five representative strains to isoniazid, rifampicin, streptomycin, and ciprofloxacin. The five POA/PZA-resistant strains were fully susceptible to the drugs, showing the same MICs as the wild-type strain (data not shown). These results indicate that POA/PZA resistance conferring mutations are drug-specific and do not cause general antibiotic resistance.

Mutations Affecting PDIM Synthesis but Not Mutations Affecting CoA Synthesis Confer Cross Resistance to POA Analogs

Nicotinic acid and benzoic acid are two structural analogs of POA, and both show antimycobacterial activity,⁷ albeit with a lower potency than for POA (Figure 3B). Previous attempts to isolate mutants against these analogs on acidic agar were also unsuccessful.⁷ The genetic data above suggest that resistance to POA may be due to two distinct mechanisms involving either panD or mas/ppsA-E mutations, the latter causing a lower level of resistance. To determine whether the analogs shared either of the two mechanisms of action with POA, cross-resistance studies were carried out. The growth of wild-type BCG was inhibited by nicotinic and benzoic acid as expected (Figure 3A). POA-resistant strains containing mas or ppsA-E mutations were also resistant to the analogs, whereas panD mutations alone (POA2 mutant class) did not confer cross-resistance (Figure 3A). The results indicate that the three carboxylic acids share a PDIM synthesis-related mechanism of action whereas the CoA-related mechanism of action is unique to POA. It is interesting that glycerol in the medium hypersensitized M. bovis BCG to inhibition by benzoic acid and more weakly to nicotinic acid (Figure 3B), as was observed for POA.

To confirm the role of mutations in PDIM synthesis, genes in POA, and nicotinic/benzoic acid resistance, a functional copy of the mas gene, encoding the 2111 amino acid mycocerosic acid synthase, was introduced into POA-resistant strain M. bovis BCG POA3A [mas-1]. Upon complementation with functional mas, BCG POA3A [mas-1] lost its resistance to POA and reverted to wild-type susceptibility (Supporting Information, Figure S2A). Figure S2B shows that mas complementation in BCG POA3A also restored susceptibility to nicotinic acid and benzoic acid.

POA but Not POA Analogs Deplete CoA, and panD Mutations Prevent Such Depletion

What is the metabolic mechanism of the resistance mediated by panD mutations? On the basis of the nature and the location of missense mutations clustered in the C-terminal domain of PanD, we hypothesized that POA could bind to and inhibit PanD (as suggested by Zhang and colleagues^16,17), causing the depletion of CoA, the final product of the biosynthetic pathway and essential acyl carrier cofactor. To test our hypothesis, we first measured the effect of POA on the cellular levels of CoA in wild-type M. bovis BCG using a fluorometric assay. Growth and CoA concentrations were measured over time in BCG cultures exposed to POA. Growth remained unaffected for 1 day after the addition of POA, indicating that POA is a slow-acting drug (Figure 4A). CoA levels dropped significantly between 12 and 24 h after POA treatment, suggesting that POA causes growth arrest via depletion of CoA. Figure 4C shows that in contrast to the response of wild-type bacteria to POA, the treatment of strains containing panD mutations (M. bovis BCG POA1 [panD1 mas-1] and POA2 [panD2]) did not result in the depletion of CoA. These results are consistent with the inhibition of PanD by POA, causing the depletion of CoA and thus growth termination of the bacilli. Mutations in the C-terminal domain of PanD may block POA binding and cause resistance by preventing CoA depletion.¹⁷ To demonstrate that POA inhibits the CoA biosynthesis pathway as opposed to modulating CoA levels through a different mechanism, we measured the levels of the CoA precursor pantothenate located downstream of PanD in the pathway using LC-MS. Pantothenate displayed a CoA-like pattern: POA treatment caused the depletion of pantothenate in wild-type bacteria, whereas panD mutations prevented the POA-induced collapse of pantothenate levels (Figure 4D).

Our genetic and chemical probing studies suggest that the PDIM-related mechanism of action of POA is distinct from its CoA-related mechanism of action. Therefore, we hypothesized that mas or ppsA-E mutations would not prevent POA-induced CoA depletion. As predicted, strains harboring mas or ppsA-E mutations without panD comutations exhibited POA-induced CoA and pantothenate depletion similar to those of wild-type bacteria (Figure 4C,D). In addition, the POA analogs did not cause the depletion of CoA, consistent with their PDIM-specific mechanism and with the cross-resistance results (Figure 4B). Furthermore, the lack of CoA depletion at growth inhibitory concentrations of nicotinic acid and benzoic acid implies that the observed CoA depletion is POA-specific and not a general phenomenon associated with drug-induced growth inhibition.

Exogenous Pantothenate Causes Resistance to POA but Not to POA Analogs

Under our proposed model, POA inhibits PanD, which causes the depletion of CoA and therefore growth termination. Hence, we predicted that exogenous addition of the CoA precursor pantothenate would phenocopy the panD mutations. Figure 5A shows that the addition of pantothenate to M. bovis BCG wild type indeed mimicked the BCG POA2 [panD2] resistance phenotype whereas the addition of exogenous pantothenate to BCG POA3A [mas-1] and BCG POA3B [ppsA1] mimicked the resistance phenotype of BCG POA1 [panD1 mas-1] (Figure 5B). Pantothenate prevents POA-mediated CoA depletion in M. bovis BCG wild type as expected^17,18 (Figure 5C). Finally, consistent with our hypothesis that POA shares a CoA-unrelated, PDIM-associated mechanism of action and resistance with nicotinic acid and benzoic acid, we observed that the addition of pantothenate did not confer resistance to these POA analogs (Figure 5D).

CoA and PDIM Resistance Pathways Are Recapitulated in M. tuberculosis

To confirm that POA resistance conferring mutations in panD and mas or ppsA-E can also be observed in virulent M. tuberculosis upon selection in neutral pH media, we plated M. tuberculosis H37Rv on POA-containing neutral agar. Whole genome sequencing of 10 confirmed POA-resistant strains revealed missense mutations in the C-terminal domain of PanD or frameshift mutations in PDIM synthesis genes, as observed in M. bovis BCG (Table 2). This suggests that the results obtained with M. bovis BCG apply to its virulent cousin. Interestingly, one POA-resistant M. tuberculosis strain carried wild-type panD and mas/ppsA-E but contained polymorphisms in dlaT, Rna60, Rv0907, and hemZ. In addition, some of the strains containing panD or ppsA-E mutations also contained mutations in genes including Rv3626c, PPE47, mmpL2, Rv1178, and mqo (Table 2). Because of the fact that only single mutants were obtained for these genes, it is not possible to conclude whether they are involved in POA resistance. However, these findings together with the isolation of one POA-resistant M. tuberculosis strain with wild-type panD and mas/ppsA-E alleles (Table 2) suggest that there might be additional mechanisms of resistance to be uncovered.

Bacterial Strains, Culture Medium, and Chemicals

M. bovis BCG (ATCC 35734) and M. tuberculosis H37Rv (ATCC 27294) strains were maintained in complete Middlebrook 7H9 medium (BD Difco) supplemented with 0.05% (v/v) Tween 80 (Sigma), 0.5% (v/v) glycerol (Fisher Scientific), and 10% (v/v) Middlebrook albumin-dextrose-catalase (BD Difco) at 37 °C with agitation at 80 rpm. Pyrazinamide, pyrazinoic acid, benzoic acid, and nicotinic acid were purchased from Sigma-Aldrich and were freshly dissolved in 90% DMSO to a concentration of 0.5 M and sterilized using 0.2 μm PTFE membrane filters (Acrodisc PALL). Kanamycin and sodium pantothenate were obtained from Sigma-Aldrich, dissolved in deionized water, and sterilized using 0.2 μm Minisart high-flow syringe filters (Sartorius) prior to adding to specific growth media under aseptic conditions. The pH of broth and liquid media after the addition of POA/NA/BA was checked using pH paper strips (Merck) and verified to be around 6.5.

Selection of Spontaneous Resistant Mutants and Growth and Resistance Determination

CFU (10⁴ to 10⁶) from mid-log-phase cultures of M. bovis BCG (M. tuberculosis) was plated on complete 7H10 agar plates or 7H10 without glycerol containing 1, 2, or 4 mM POA and grown for 4 weeks. Colonies were picked from selection plates and restreaked on agar containing the same concentration of POA for colony purification and to verify drug resistance. Isolated colonies were picked from the restreak plates and expanded in 7H9 to an OD₆₀₀ of 0.7–0.8 and stored in 25% glycerol in 1 mL aliquots at −80 °C. This frozen stock was then used for the subsequent phenotypic and genetic characterization of mutants.

Growth inhibitory activities of compounds against M. bovis BCG were determined as described for M. tuberculosis in ref 5. The MIC₉₀ and MIC₅₀ reported here are those concentrations of drug that inhibit 90 and 50% of growth, respectively, as compared to a drug-free control after 5 days of incubation. All MICs were performed in technical and biological replicates.

Whole Genome Sequencing

Genomic DNA was isolated from M. bovis BCG and M. tuberculosis H37Rv, wild type, and mutants as described.⁵³ Library construction using a NEBNext Ultra DNA library preparation kit (New England Biolabs), whole genome sequencing on the Illumina MiSeq platform, and bioinformatics analyses was performed by AIT Biotech, Singapore. (Details are described in Supporting Information, Supplemental methods).

pncA and mas Complementation Constructs

To restore pyrazinamidase PncA activity in M. bovis BCG, which has a nonfunctional endogenous pncA locus, the integrative vector pMV306 was used.⁵⁴ The M. tuberculosis pncA coding sequence including the putative promoter was PCR amplified using primers pncA_F 5′ATGACACCTCTGTCACCGGACGGA3′ and pncA_R 5′AAGCTTCAGGAGCTGCAAACCAACTCG-A3′ and inserted into pCR2.1TOPO by TOPO TA cloning (Life Technologies). The insert was subsequently excised using EcoRI and integrated into pMV306. The mas gene of M. bovis BCG including its putative promoter was amplified using PCR with specific primers mas_F 5′TGGTACCGGTCGGATGTGA-TGTG3′ and mas_R 5′TATCTAGAATTCGACCGCTAT-GATGCC3′ and inserted into the KpnI/XbaI restriction sites of pMV306. The integrity of the constructs was confirmed by capillary sequencing (AIT Biotech, Singapore). The constructs were electroporated into M. bovis BCG strains and selected on Kanamycin (25 μg/mL) containing 7H10 agar plates at 37 °C. Kanamycin-resistant colonies were picked after 3 weeks, expanded with Kanamycin selection, and frozen at −80 °C as glycerol stock, which was subsequently characterized.

CoA Quantification

M. bovis BCG cultures were grown to mid-log phase in 1 L roller bottles (Corning) at 37 °C and pelleted in 50 mL tubes at 3200 rpm for 10 min. Pellets were resuspended in the specific growth medium after adjusting to an OD₆₀₀ of 0.2 and incubated with or without POA, nicotinic acid, or benzoic acid. At each time point for each treatment, the samples were plated on 7H10 to determine CFU, and an equivalent of 2 × 10⁹ CFU was pelleted by centrifugation. The pellet was resuspended in 500 μL of phosphate-buffered saline (PBS) and homogenized by bead beating (Precellys 24 homogenizer) at 6500 rpm three times for 30 s each. The lysate was pelleted by centrifugation, and the supernatant was subsequently used for analysis by a CoA assay kit (Sigma-Aldrich MAK034) per the manufacturer’s instructions. The fluorescence intensity was read on a Tecan Infinite M200 plate reader (λ_ex = 535 nm/λ_em = 587 nm), and the CoA quantity was determined in nanomoles from a standard curve that was obtained for each experiment. These experiments were performed in technical replicates of independent biological replicates and normalized per CFU.

The lysate supernatant was also used for mass spectrometric quantification of CoA and pantothenate, which was achieved by using liquid chromatography coupled to mass spectrometry (LC-MS). (Details are described in Supporting Information, Supplemental methods.)

Supplementary Materials