Atypical Genetic Basis of Pyrazinamide Resistance in Monoresistant Mycobacterium tuberculosis

ABSTRACT Pyrazinamide (PZA) is a widely used antitubercular chemotherapeutic. Typically, PZA resistance (PZA-R) emerges in Mycobacterium tuberculosis strains with existing resistance to isoniazid and rifampin (i.e., multidrug resistance [MDR]) and is conferred by loss-of-function pncA mutations that inhibit conversion to its active form, pyrazinoic acid (POA). PZA-R departing from this canonical scenario is poorly understood. Here, we genotyped pncA and purported alternative PZA-R genes (panD, rpsA, and clpC1) with long-read sequencing of 19 phenotypically PZA-monoresistant isolates collected in Sweden and compared their phylogenetic and genomic characteristics to a large set of MDR PZA-R (MDRPZA-R) isolates. We report the first association of ClpC1 mutations with PZA-R in clinical isolates, in the ClpC1 promoter (clpC1p−138) and the N terminus of ClpC1 (ClpC1Val63Ala). Mutations have emerged in both these regions under POA selection in vitro, and the N-terminal region of ClpC1 has been implicated further, through its POA-dependent efficacy in PanD proteolysis. ClpC1Val63Ala mutants spanned 4 Indo-Oceanic sublineages. Indo-Oceanic isolates invariably harbored ClpC1Val63Ala and were starkly overrepresented (odds ratio [OR] = 22.2, P < 0.00001) among PZA-monoresistant isolates (11/19) compared to MDRPZA-R isolates (5/80). The genetic basis of Indo-Oceanic isolates’ overrepresentation in PZA-monoresistant tuberculosis (TB) remains undetermined, but substantial circumstantial evidence suggests that ClpC1Val63Ala confers low-level PZA resistance. Our findings highlight ClpC1 as potentially clinically relevant for PZA-R and reinforce the importance of genetic background in the trajectory of resistance development.

based on the finding that altered Clp protease degradation of PanD functions as part of mechanism of POA action and that mutation in ClpC1 of the Clp complex (Clp protease plus ClpC1) mitigates POA-driven efficiency of PanD proteolysis. Particularly in cases where pncA mutations are absent, the targets mediating resistance against these proposed modes of action are candidates for conferring phenotypic resistance.
A recent analysis of 26 isolates from Sweden with PZA-resistant (PZA-R) but wildtype (WT) pncA was conducted to understand the nature of this noncanonical genetic basis of PZA-R (12). That study was able to attribute resistance to false resistance in 8/ 26 isolates through replicate drug-susceptibility testing (DST) at two concentrations and resistance in another seven isolates to heteroresistance by pncA genotyping following drug pressure in a Bactec mycobacterial growth indicator tube (MGIT) PZA test. To potentially explain the 11 remaining unexplained isolates, the authors of that study genotyped proposed alternative PZA resistance genes (rpsA and panD) with Ion Torrent sequencing. All 11 appeared to be explained by either mixed populations with naturally PZA-R Mycobacterium avium (n = 3), pncA mutation not captured by their primer (n = 1), or alternative mechanisms involving mutation in panD (n = 4) or rpsA (n = 3). However, they found that 17/400 PZA-S isolates harbored nonsynonymous rpsA mutations, including five with the same mutation seen in two of their three PZA-R rpsA mutants (12).
Here, we used long-read sequencing data to genotype 19 PZA-monoresistant isolates from Sweden, including nine of the WT-pncA PZA-R isolates described above. We contrast the genetic basis of PZA resistance in these isolates with that of a large set of multidrug-resistant (MDR PZA-R ) isolates isolated in Sweden during a similar time period to identify aspects of PZA resistance that distinguish PZA monoresistance from MDR PZA-R . We analyzed pncA and genes suggested or shown to mediate alternative resistance mechanisms (panD, rpsA, and clpC1), focusing on how the prevalence of mutations in these genes differs between PZA-monoresistant isolates and MDR PZA-R isolates.

RESULTS AND DISCUSSION
PZA-monoresistant clinical isolates. Isolates were selected from samples collected in Swedish clinical tuberculosis (TB) labs (from patients in Sweden who were most likely infected primarily in Asia and Africa) and processed at the Public Health Agency of Sweden between 2000 and 2015 on the basis of PZA-R in initial phenotypic drug susceptibility testing (DST) and phenotypic susceptibility to isoniazid, rifampin, ethambutol, and amikacin. Genotypic DST corroborated the phenotypic results and implied susceptibility to fluoroquinolones and second-line injectable drugs, with only established lineage markers observed (Table 1). PZA-monoresistant isolates span three of the four major M. tuberculosis lineages and multiple sublineages within the Indo-Oceanic and Euro-American lineages (Table 1).
With a single exception, PacBio sequencing calls were concordant with initial Sanger genotyping of pncA and with genotyping of alternative resistance targets by Sanger and Ion Torrent sequencing. The sole exception was an isolate (SEA10470) ( Table 2) initially identified as WT/Ser65Ser (a minor population of Ser65Ser) but had a large deletion spanning the beginning of pncA and its upstream region in the PacBio data (Table 2). In their recent work (which included SEA10470), upon observing that their pncA primer for Sanger sequencing from the MGIT PZA tube did not return anything to sequence, Werngren and colleagues sequenced the PZA-containing MGITderived sample on Ion Torrent (12) and discovered the same large deletion that we found in single-molecule real-time (SMRT) sequencing data following several weeks of growth in Löwenstein-Jensen (LJ) medium, implying that WT-pncA was ultimately outcompeted by the pncA deletion in both the presence (MGIT PZA tube) and absence of PZA (growth in LJ medium preceding SMRT sequencing). Thus, we conclude SEA10470 is a heteroresistant pncA mutant with limited or no fitness cost in LJ medium.
panD, rpsA, and clpC1 mutations in PZA-monoresistant isolates with wild-type pncA. To identify the genetic basis of our PZA-monoresistant isolates, we first checked for pncA mutations. PncA alleles for all 10 pncA mutants had been reported previously Tyr389Tyr, promoter: for clinical isolates (Table 3), substantiating their role in conferring PZA-R to these isolates. After accounting for isolates with pncA mutations ( suggesting that the nature of PZA monoresistance is different from that of PZA resistance in MDR isolates. All eight had at least one nonsynonymous mutation in panD, rpsA, or clpC1. Three of the eight were susceptible upon repeated phenotypic DST, and one exhibited low-level resistance (resistant at 100mg/ml but not 200 mg/ml). All four isolates that remained unexplained, with neither PncA mutations nor inconsistent phenotyping, had multiple mutations in alternative PZA resistance genes. Two had nonsynonymous mutations in panD and the other two in rpsA (Table 2), and all had the missense mutation ClpC1 Val63Ala . However, the rpsA mutations also occurred in many Swedish isolates susceptible to PZA (12), suggesting either that these mutations confer low-level resistance around the critical concentration or that other genetic elements are at play. Novel mutations implicate regions mediating PanD proteolysis by ClpC1. Recent work demonstrated that POA increases PanD proteolysis by the Clp complex and proposed this as its mode of action (11). This proteolysis requires a C-terminal degradation tag (PanD 127-139 ) for recognition by the N terminus of ClpC1. While PZA-R clpC1 mutants have not been previously reported among clinical isolates, they have been isolated following selection under PZA (14) and POA (9) pressure in vitro and POA pressure in an in vivo mouse model (8). Here, we report the first account of clpC1 mutations in PZA-R clinical isolates with WT-pncA and a panD mutation (panD Asp133Ala ) in its C-terminal recognition tag novel to clinical isolates.
The first clpC1 mutation is a deletion in the clpC1 promoter (clpC1p), 138 bp upstream of its start codon (clpC1p 2C138 ). The isolate harboring clpC1p 2C138 (SEA14117) ( Table 2) has a pncA mutation (pncA Ile90Ser ) previously reported in PZA-R clinical isolates. However, pncA Ile90Ser is rare and-unlike most clinically impactful pncA mutations (2)is not enriched following selection under PZA pressure in vitro or in mice, and neither were any other codon 90 pncA mutations. While the possibility that pncA Ile90Ser accounts fully for PZA-R in SEA14117 cannot be dismissed, the irreproducibility of pncA Ile90Ser conferring resistance on its own in vitro leaves open the possibility that clpC1p 2C138 may contribute to resistance either alternatively to pncA Ile90Ser or perhaps additively. While clpC1p 2C138 has not been described previously, other clpC1 promoter mutations have emerged under POA selection pressure in vitro (9), suggesting that clpC1 promoter mutations can confer POA resistance. Viewing this observation through the model of Clp proteolysis of PanD as the POA mode of action (11), clpC1p mutations could confer PZA/POA resistance by weakening the promoter, reducing ClpC1 expression and, in turn, Clp-mediated PanD degradation. The second clpC1 mutant was a missense mutation in the ClpC1 N terminus, Val63Ala (clpC1 Val63Ala ). Eleven of 19 PZA-monoresistant isolates were clpC1 Val63Ala mutants, exclusively and invariably of the Indo-Oceanic lineage. These 11 isolates span four Indo-Oceanic sublineages (each with multiple isolates) and comprise seven distinct sublineage-spoligotype combinations (Table 3). Thus, PZA-monoresistant clpC1 Val63Ala mutants are neither a clonal expansion nor monophyletic. Rather, they are spread throughout the Indo-Oceanic lineage. The effects of the variants clpC1 Val63Ala and panD Asp133Ala converge on the interaction between the ClpC1 N-terminal and PanD 127-139 C-terminal degradation tags (9). Though it is only speculation at present, the functional effects implied by these mutations' location fits well into the mode of POA action proposed recently by Gopal and colleagues (11).
Indo-Oceanic strains in this study (11/19), which selected isolates on the basis of PZA monoresistance, were far more prevalent (P , 0.000001, odds ratio = 22.2, 95% CI = 6.05 to 121, two-tailed Fisher's exact test) than PZA-resistant Indo-Oceanic isolates from Sweden in a study (35) (5/80) in which isolates were selected on the basis of MDR phenotype. The frequency of ClpC1 mutants that we observed here stands in contrast to a recent study of 269 PZA-resistant isolates in China where no isolates harbored a nonsynonymous ClpC1 mutation (16). However, that study had mostly Beijing lineage isolates (lineage was not indicated for the remaining isolates). This overrepresentation of Indo-Oceanic isolates in monoresistant PZA-R strains relative to MDR PZA-R strains has been observed previously in the United States (17) and Iran (18), suggesting that it is a phenomenon not isolated to samples derived from patients in Sweden. This repeatedly observed trend of Indo-Oceanic isolates developing PZA monoresistance, when considered alongside the mechanistic plausibility of ClpC1 N-terminal mutations affecting POA resistance (8,9,11,14) and the vacillating of these ClpC1 mutants around the resistance threshold ( Table 2), suggests that clpC1 Val63Ala may confer low-level PZA resistance to Indo-Oceanic isolates. The missense mutation clpC1 Val63Ala is specific to and invariably present in Indo-Oceanic isolates (15,19) and thus would represent a lineagespecific resistance phenotype akin to the intrinsic PZA resistance of the ancestral mycobacteria M. canettii (13) and Mycobacterium bovis (1), but at a lower level. However, confirmatory mutagenesis is needed to verify this possibility.
How might one reconcile the proposed low-level PZA resistance conferred by the clpC1 Val63Ala allele with the observation that Indo-Oceanic isolates often register as phenotypically susceptible to PZA? It is important to consider that the Bactec MGIT 960 critical concentration (100 mg/ml) has been calibrated to detect pncA-conferred resistance, which is severalfold higher than the 2-to 8-fold MIC increases observed for POAresistant mutants harboring panD and clpC1 mutations (8,9). This could explain the variable and discrepant phenotypic results we observe in Indo-Oceanic isolates upon replicating DST (Table 2; Fig. 1). This scenario has been reported for PZA/POA-resistant panD mutants isolated following selection under POA pressure; despite a 2-to 8-fold MIC increase for G139* panD mutants relative to the wild type, seven of eight clinical isolates that Gopal and colleagues found with the mutation were classified as susceptible to PZA (8). Similarly, rpsA D438Ala and rpsA Asp123Ala mutations were recently shown to confer 2-fold increases in MIC relative to H37Rv (20). Low-level resistance conferred by the clpC1 Val63Ala allele would simultaneously explain the prevalence of Indo-Oceanic lineage and WT-PncA PZA-R among PZA-monoresistant isolates (17,18) as well as the variable phenotypic results for these isolates. Epistasis (Fig. 1A) and resistance near the critical concentration vacillating above it due to nongenetic factors (Fig. 1B) could also produce the bias toward Indo-Oceanic lineage isolates and inconsistent DST results among monoresistant PZA isolates resulting from clpC1 Val63Ala . In either case, together, these considerations suggest clpC1 Val63Ala low-level resistance, though confirmation is required. Future work with a larger set of PZA-monoresistant and MDR PZA-R isolates and susceptibility testing across a range of lower MICs would more precisely delineate the role of clpC1 Val63Ala in PZA resistance.
Concluding remarks. By analyzing monoresistant PZA isolates collected in Sweden, we found that monoresistant isolates have mutational profiles distinct from those of MDR PZA-R isolates and identified clpC1 Val63Ala as a potential low-level resistance marker common to Indo-Oceanic isolates. Despite this mutation's being common to Indo-Oceanic isolates (19) and POA/PZA-resistant ClpC1 mutants isolated from in vitro screens (8,9), this is, to our knowledge, the first report of PZA-resistant ClpC1 mutants in clinical isolates. Further study of the hypothesized link between clpC1 Val63Ala and lowlevel resistance warrants mechanistic study, as does evaluation of its effect on PZA/ POA resistance in vitro and in the context of infection. The functional effects implicated by clpC1 Val63Ala and panD Asp133Ala converge on the interaction between the ClpC1 N-terminal and PanD 127-139 C-terminal degradation tags (9), consistent with the emerging model of POA mode of action where PanD degradation by Clp proteolysis is enhanced by POA (11). Functional studies assessing the effect of these two mutations on PanD degradation natively and in the presence of POA would be of great interest. More broadly, the bias of PZA monoresistance toward Indo-Oceanic isolates underlines the relevance of genetic background in evolution of resistance (21)(22)(23), a factor that is often precluded from analysis in molecular epidemiology and genome-wide association studies on the assumption that they do not influence resistance. Our findings suggest otherwise.

MATERIALS AND METHODS
Isolate selection. PZA-monoresistant isolates were selected from TB patients seen in Sweden between 2003 and 2015, irrespective of lineage, TB presentation, or geographic origin. The geographic origin of the 19 TB strains isolated in Sweden was typical of the general TB epidemiology in Sweden, with roughly half of the patients coming from Africa, a fifth from Asia, and the rest from Sweden or other parts of the globe (24).
Culturing and DNA extraction. M. tuberculosis samples were prepared and extracted at the World Health Organization Supranational Reference Laboratory in Stockholm, Sweden. Isolates growing on LJ medium were received from clinical TB labs and subsequently Sanger sequenced. The genotypic DST result (pncA-Sanger) was then compared to the phenotypic DST result (PZA Bactec 460 until 2008, MGIT 960 thereafter) from the clinical lab, and discordant cases were retested in PZA MGIT. In cases of verified PZA resistance, Sanger sequencing was repeated on the growth from the PZA MGIT tube. Samples still lacking a pncA mutation were then subjected to whole-genome sequencing (WGS). Preparation and extraction were performed as previously described (12). Epistasis. In the epistatic scenario, clpC1 Val63Ala alone confers low-level resistance, invariably below the 100-mg/liter critical concentration, but some isolates (reds) harbor additional mutations that confer higher-level PZA-R through additive or synergistic interaction with clpC1 Val63Ala . (B) Near-critical concentration MIC. In this scenario, low-level PZA resistance conferred by clpC1 Val63Ala alone sometimes exceeds the 100-mg/liter cutoff, due to nongenetic factors contributing to MIC variance (30,31), such as inoculum size (32), growth medium, or interlaboratory variation (33). The phenomena depicted in panels A and B are not mutually exclusive and could both be at work.
SMRT sequencing. DNA sequencing was performed at the Institute for Genomic Medicine at the University of California, San Diego, CA. DNA libraries for PacBio (Pacific Biosciences, Menlo Park, CA) were prepared using PacBio's DNA template prep kit with no follow-up PCR amplification. Briefly, sheared DNA was end repaired, and hairpin adapters were ligated using T4 DNA ligase. Incompletely formed SMRTbell templates were degraded with a combination of exonuclease III and exonuclease VII. The resulting DNA templates were purified using solid-phase reversible-immobilization (SPRI) magnetic beads (AMPure; Agencourt Bioscience, Beverly, MA) and annealed to a 2-fold molar excess of a sequencing primer that specifically bound to the single-stranded loop region of the hairpin adapters. SMRTbell templates were subjected to standard SMRT sequencing using an engineered phi29 DNA polymerase on the PacBio RS II system according to the manufacturer's protocol.
Sanger and Ion Torrent sequencing. DNA sequencing with Sanger and Ion Torrent was carried out at World Health Organization Supranational Reference Laboratory in Stockholm, Sweden, as previously described (12).
Drug susceptibility testing. For cases with discordance between initial Sanger sequencing of pncA and the PZA Bactec 460/MGIT result from the clinical lab, PZA DST was performed in the Bactec MGIT system (Becton Dickinson) according to instructions from the manufacturer. The DST inoculum was prepared from bacterial growth on Löwenstein-Jensen egg medium at 37°C. Briefly, two 1-ml loops of bacteria were suspended in 3 ml phosphate-buffered saline (PBS) in a small glass tube with glass beads. The bacterial suspension was homogenized using vortexing or an ultrasound water bath to disperse any clumps. The suspension was then left to sediment for 20 min, and the upper 2 ml was transferred to a new tube and left to sediment for another 15 min. Before inoculation of the MGIT PZA medium culture tubes (pH 5.9), the bacterial suspension was adjusted to a McFarland turbidity of 0.5 and diluted in PBS per the manufacturer's PZA test protocol.
Statistical tests. Two-tailed Fisher's exact test was used to assess independence and implemented in R statistical programming language (29).