Fitness Cost and Compensatory Evolution in Levofloxacin-Resistant Mycobacterium aurum

We isolated spontaneous levofloxacin-resistant strains of Mycobacterium aurum to study the fitness cost and compensatory evolution of fluoroquinolone resistance in mycobacteria. Five of six mutant strains with substantial growth defects showed restored fitness after being serially passaged for 18 growth cycles, along with increased cellular ATP level. Whole-genome sequencing identified putative compensatory mutations in the glgC gene that restored the fitness of the resistant strains, presumably by altering the bacterial energy metabolism.

ABSTRACT

We isolated spontaneous levofloxacin-resistant strains of Mycobacterium aurum to study the fitness cost and compensatory evolution of fluoroquinolone resistance in mycobacteria. Five of six mutant strains with substantial growth defects showed restored fitness after being serially passaged for 18 growth cycles, along with increased cellular ATP level. Whole-genome sequencing identified putative compensatory mutations in the glgC gene that restored the fitness of the resistant strains, presumably by altering the bacterial energy metabolism.

INTRODUCTION

Third-generation fluoroquinolones (FQs) levofloxacin and moxifloxacin play a critical role in the treatment of multidrug-resistant tuberculosis (1), and the development of fluoroquinolone resistance (FQ-R) in Mycobacterium tuberculosis dramatically reduces the rates of treatment success (2). FQs inhibit the action of DNA gyrase, which is composed of two A and two B subunits, encoded by gyrA and gyrB, respectively (3). Mutations in two short regions of gyrA and gyrB, known as the “quinolone-resistance-determining regions” (QRDRs) have been shown to cause FQ-R. In M. tuberculosis, the most common FQ-R mutations alter codons 90 and 94 of gyrA, but substitutions are also found in codons 74, 88, 89, and 91 (4, 5). Drug resistance mutations generally impart a fitness cost manifested as reduced growth rates and lower virulence, which could be ameliorated by second-site compensatory mutations (6–8). Such compensatory evolution could play a pivotal role in the maintenance and spread of drug-resistant bacteria (9). Previous study has demonstrated low clustering rates for FQ-resistant clinical M. tuberculosis strains (10), suggesting that FQ-R mutations may somehow make them less likely to be transmitted, but little is known about the fitness costs of FQ-R and the evolution of possible compensatory mutations. Here, we used Mycobacterium aurum as a model organism for studying the fitness cost and compensatory evolution of FQ-R in M. tuberculosis. Compared to other mycobacteria used as models for tuberculosis research, M. aurum does not aggregate, which facilitates precise colony enumeration (11). Although fast growing and nonpathogenic, M. aurum is like M. tuberculosis for FQ-R, and its QRDRs of gyrA and gyrB show amino acid identities of 95 and 97.56%, respectively, with those of M. tuberculosis.

We generated a total of 144 levofloxacin-resistant M. aurum strains from three rounds of selection with 2-fold increasing concentrations of levofloxacin. As shown in Table 1, 45 single mutants were isolated from the wild-type (WT) strain, while the 92 double and 7 triple mutants were derived from strains with single and double mutations, respectively. Primers for amplifying and sequencing of the QRDRs of gyrA and gyrB are described in Table S1 in the supplemental material. Levofloxacin MIC values were determined using broth microdilution method (12). As shown in Table 1, most levofloxacin resistance mutations occurred in codons 90 and 94 of gyrA, and resistance levels increased with mutation accumulation. To further investigate whether FQ-R influences the bacterial fitness, we analyzed the growth kinetics by recording the optical density at 600 nm (OD₆₀₀) of a WT control and levofloxacin-resistant M. aurum strains with different mutations. Most of the mutant strains (23/29, 79%) had little or no growth defect (Fig. 1a), but six mutant strains showed growth rates substantially less than that of the WT strain (Fig. 1b). The gyrA mutations in these strains were G88C, G88D, D94N+D89G, D94G+D89N, D94Y+D89N, and D94Y+G88C (Table 1). To quantify the relative fitness of these six mutant strains compared to WT strain, we performed an in vitro competition assay, as previously described (13), by inoculating equal volumes of WT and mutant cultures (OD₆₀₀ = 1) at a 1:100 dilution, and counting colonies at baseline (0 h) and endpoint (72 h). All six slow-growing mutants showed significantly lower fitness compared to the WT strain (Fig. 1e), but the strain that was least fit had the gyrA G88C substitution, as also seen in Mycobacterium smegmatis (14). Interestingly, the fitness of this G88C mutant was partially restored by the acquisition of a second gyrA D94Y substitution, suggesting an intragenic compensatory mechanism.

TABLE 1.

Mutation and levofloxacin MICs of levofloxacin-resistant M. aurum strains

No. of mutations	Prior mutationa	Secondary mutationa		Selective concn (μg/ml)b	MIC (μg/ml)	No. of strainsc
		gyrA	gyrB
Single mutation	None	G88C	WT	0.2	0.4	3*#
		G88D	WT	0.2–0.4	6.4	9*
		G89D	WT	0.4	3.2	2*
		S90L	WT	0.05–0.4	0.8	10*
		D94N	WT	0.05–0.4	0.8	11*#
		D94G	WT	0.2	0.8	2*#
		D94Y	WT	0.2	0.4	4*#
		WT	R485L	0.2	0.4	4*
Double mutations	gyrA G88C	D94Y	WT	1.6	3.2	1
	gyrA S90L	D94Y	WT	0.8–6.4	25.6	2*
		D94A	WT	1.6–3.2	12.8	3*
		D94G	WT	0.8–3.2	25.6	7*
		D94N	WT	0.8–6.4	25.6	8
		WT	D500N	1.6	6.4	1*
	gyrA D94N	G88C	WT	0.8	1.6	1
		G88D	WT	0.4–3.2	6.4	2*
		G89N	WT	0.8	3.2	1
		D89G	WT	0.8	1.6	1*
		S90L	WT	0.8–3.2	25.6	17
		S90A	WT	0.8	25.6	3
	gyrA D94G	A74S	WT	0.8	1.6	2
		G88D	WT	0.8	3.2	1
		D89N	WT	0.8	3.2	2*
		S90L	WT	0.8	12.8	7*
		S90A	WT	0.8	12.8	1*
		S91P	WT	0.8	3.2	1#
		WT	D500N	0.8	3.2	1*
		WT	L538F	0.8	3.2	2*
	gyrA D94Y	G88C	WT	0.8–1.6	3.2	2*
		D89N	WT	0.8	3.2	3*
		D89G	WT	0.8	3.2	1
		S90L	WT	0.8	25.6	3
		S91P	WT	0.8	3.2	1
		WT	R485H	0.8	3.2	2*
	gyrB R485L	G88D	WT	0.8–1.6	3.2	13
		S90L	WT	0.8	3.2	1*
		D94G	WT	0.8	1.6	1*
		D94Y	WT	0.8	3.2	1*
Triple mutations	gyrA S90L + gyrA D94A	WT	L538F	12.8	51.2	1*
		WT	D500H	25.6	51.2	2*
	gyrA S90L + gyrB D500N	WT	D94N	25.6	51.2	1*
		WT	D94Y	12.8–25.6	51.2	3*

^aAll gyrA and gyrB mutations were transformed to corresponding codon substitutions in M. tuberculosis.

^bThe concentrations used to select levofloxacin-resistant M. aurum mutants increased twofold.

^cAsterisks (*) indicate strains subjected to growth rate measurements in a preliminary screen for levofloxacin-resistant strains with decreased fitness, and pound signs (#) indicate the mutations in the indicated strains were observed in clinical M. tuberculosis strains.

FIG 1.

Fitness costs and compensation of mutant M. aurum strains. (a and b) Graphs show the growth curves of M. aurum strains in nonantibiotic 7H9 broth. The OD₆₀₀ is plotted as a function of time. Mutations at codons 88, 89, 90, 91, and 94 are located in the gyrA gene, and mutations at codons 500 and 538 are located in the gyrB gene. The WT and single mutants are indicated by a circle, while double and triple mutants are indicated by triangles and asterisks, respectively. (c and d) Growth curves of M. aurum strains after 0 (solid lines) and 18 (dashed lines) growth cycles in nonantibiotic and 0.2 μg/ml of levofloxacin 7H9 broth, respectively. (e and f) Relative fitness (e) and ATP levels (f) of M. aurum strains after 0 (dark shading) and 18 (light shading) growth cycles, respectively. The results show the means ± the standard deviations of three independent experiments, and the exact P values are listed above the pairs of data (paired t test).

We then experimentally evolved the six slow-growing mutants in the laboratory, with the WT strain as a control. Eight independent cultures for each strain were grown in antibiotic-free 7H9 broth for 3 days and then serially passaged by transferring 3 μl of culture into 3 ml of fresh medium to initiate the next 3-day growth cycle. After 18 growth cycles, five of the six evolved mutant strains showed growth rates that were restored to that of the WT and improved fitness compared to their mutant parents (Fig. 1c to e). The only mutant that did not evolve restored fitness after 18 growth cycles was the strain with the single G88C substitution. Illumina whole-genome sequencing (WGS) found that all five evolved fitness-restored strains convergently had frameshift insertions in the glgC gene, which encodes an ADP-glucose pyrophosphorylase (AGPase) (Table S2). The evolved strains maintained the same MICs as their slower growing ancestors, indicating that the compensatory mutations had no impact on drug resistance (Table S2). In addition, WGS identified mutations in 12 other genes in the evolved strains, including some from the passaged WT strain (Table S3), but none of these mutations were found in evolved strains from more than one original mutant.

FQ-R mutations have been reported to alter the enzymatic activity of the ATP-dependent DNA gyrase (15). Because the AGPase, encoded by glgC gene, is involved in the ATP-consuming process of glycogen synthesis (16), frameshift insertions that inactivate the glgC gene might reduce glycogen synthesis and therefore ATP consumption. Consistent with this explanation, cellular ATP concentrations, measured using an ATP assay kit (Beyotime, Shanghai, China) according to the manufacturer’s instructions, were significantly higher in all faster growing evolved strains than in their pre-evolved mutant parents (paired t test, P < 0.01) (Fig. 1f). The higher ATP levels could perhaps increase the activity of the ATP-dependent DNA gyrase, thus compensating for a putatively reduced functional gyrase capacity in the low fitness mutants. Since the GlgC proteins of M. aurum and M. tuberculosis are highly similar (identity, 91.34%), we searched for glgC mutations in a data set of more than 700 clinical FQ-resistant M. tuberculosis strains but identified only one strain, with gyrA D94G plus gyrB A504V substitutions, that had an insertion mutation in the glgC gene. It is possible that the putative compensatory mutations in the glgC gene are not found in clinical M. tuberculosis isolates because low fitness FQ-R mutant strains would not survive long enough in a clinical setting to develop a compensatory mutation. However, occasional clinical strains are found with the fitness-costing mutations we describe, but they may have compensatory mutations that, due to differences in the biology of M. tuberculosis, are distinct from the AGPase inactivating insertions found in M. aurum. The reports that FQ-R strains are less frequently transmitted could indicate that the more common FQ-R mutations have subtle fitness costs that we could not appreciate through in vitro studies of M. aurum but may affect the relative abilities of M. tuberculosis strains to cause disease.

In conclusion, we demonstrated that some levofloxacin-resistant M. aurum strains with mutations in gyrA showed fitness costs that were compensated with the acquisition of insertions into the glgC gene. The fitness compensated strains showed higher ATP concentrations than the WT strain, perhaps due to decreased ATP consumption resulting from the absence of the glgC encoded AGPase. This study represents a first step toward a better understanding of the fitness effects of FQ-R mutations in mycobacteria. Further studies are warranted to investigate the molecular basis of fitness costs and the compensatory mechanisms in FQ-R mycobacteria.

Data availability.

The WGS data have been submitted to the Sequence Read Archive of the National Center for Biotechnology Information as recalibrated BAM files (accession number PRJNA598519).