Common Variants in the Glycerol Kinase Gene Reduce Tuberculosis Drug Efficacy

Abstract

Despite the administration of multiple drugs that are highly effective in vitro, tuberculosis (TB) treatment requires prolonged drug administration and is confounded by the emergence of drug-resistant strains. To understand the mechanisms that limit antibiotic efficacy, we performed a comprehensive genetic study to identify Mycobacterium tuberculosis genes that alter the rate of bacterial clearance in drug-treated mice. Several functionally distinct bacterial genes were found to alter bacterial clearance, and prominent among these was the glpK gene that encodes the glycerol-3-kinase enzyme that is necessary for glycerol catabolism. Growth on glycerol generally increased the sensitivity of M. tuberculosis to antibiotics in vitro, and glpK-deficient bacteria persisted during antibiotic treatment in vivo, particularly during exposure to pyrazinamide-containing regimens. Frameshift mutations in a hypervariable homopolymeric region of the glpK gene were found to be a specific marker of multidrug resistance in clinical M. tuberculosis isolates, and these loss-of-function alleles were also enriched in extensively drug-resistant clones. These data indicate that frequently observed variation in the glpK coding sequence produces a drug-tolerant phenotype that can reduce antibiotic efficacy and may contribute to the evolution of resistance.IMPORTANCE TB control is limited in part by the length of antibiotic treatment needed to prevent recurrent disease. To probe mechanisms underlying survival under antibiotic pressure, we performed a genetic screen for M. tuberculosis mutants with altered susceptibility to treatment using the mouse model of TB. We identified multiple genes involved in a range of functions which alter sensitivity to antibiotics. In particular, we found glycerol catabolism mutants were less susceptible to treatment and that common variation in a homopolymeric region in the glpK gene was associated with drug resistance in clinical isolates. These studies indicate that reversible high-frequency variation in carbon metabolic pathways can produce phenotypically drug-tolerant clones and have a role in the development of resistance.

Keywords: Mycobacterium tuberculosis; antibiotic resistance; genetics.

FIG 1

Genetic strategy to define bacterial functions that limit drug efficacy. (A, left) Spleen CFU from BALB/c mice infected with transposon mutant library both untreated (circles) and after HRZE treatment (squares). Antibiotic treatment was started at 14 dpi (indicated by gray arrow). Plotted means from 3 biological replicates with standard deviations are shown. (Right) Change in CFU after treatment with the indicated antibiotic for 5 weeks. The change in CFU between pretreatment and posttreatment samples is presented. Significance was calculated using unpaired t test: *, P = 0.03; **, P = 0.002; ***, P = 0.0002; ****, P < 0.0001. (B) Relative abundance of individual mutants, measured by log₂ fold change, in untreated mice (x axis) and HRZE-treated mice (y axis). Significantly altered mutants after treatment are indicated in black. (C) Functional classes of mutants with altered susceptibility in vivo. Classification from Mycobrowser. (D to G) Normalized abundance of mutations in pretreatment (black) and after HRZE treatment (red) at individual TA dinucleotide insertion sites in pncA (D), ppe50-ppe51 (E), glgA-glgC (F), and glpK (G). Shown are the average numbers of unique sequence reads (y axis) plotted versus TA sites (x axis). (H) Log₂ fold change of individual mutants (gray dots) 1 week posttreatment compared to pretreatment. Significantly altered mutants are indicated by black circles.

FIG 2

Glycerol metabolism broadly increases drug efficacy in vitro. (A, top left) Growth of H37Rv on cholesterol and treated with moxifloxacin at the indicated concentrations. Growth was measured by yellow fluorescent protein (YFP) fluorescence. (Top right) GR₅₀ for moxifloxacin (MOX) in medium containing either glycerol, valerate, or cholesterol. (Bottom) GR₅₀ ratios for INH (circles), RIF (squares), and MOX (triangles) grown on different carbon sources. Shown are valerate/glycerol (left) and cholesterol/glycerol (right). Significance was calculated using one-sample t test with a theoretical mean value of 0: *, P = 0.05; **, P = 0.01. (B) Growth kinetics of H37Rv (circles), △glpK (triangles), and complement (squares) strains on glycerol (top) or glucose (bottom). Plotted means from 3 biological replicates with standard deviations are shown. (C) Growth of H37Rv (black bars), ΔglpK (gray bars), and complement glpK (striped bars) strains after treatment with INH or RIF in media containing glycerol, butyrate, or pyruvate and PZA in media containing glycerol or dextrose at pH 5.8. Growth was assessed by the growth constant, k, normalized to no-antibiotic controls and plotted as ratios (treated/untreated), where 1 is the growth constant without antibiotic (indicated by a dotted line). Antibiotic concentrations started at 2 μg/ml, 1 μg/ml, and 400 μg/ml for INH, RIF, and PZA, respectively, and were serially diluted 2-fold for 6 dilutions. Significance was calculated using an unpaired t test with Benjamini-Hochberg multiple-testing correction. *, P = 0.03; **, P = 0.002; ***, P = 0.0002; ****, P < 0.0001.

FIG 3

Loss of glycerol kinase increases survival under PZA treatment in vivo. (A) Lung CFU of H37Rv (circles) and △glpK (triangles) strains from BALB/c mice after aerosol infection with a dose of 500 to 700 CFU/mouse. Shown are plotted means from 4 biological replicates with standard deviations. (B) Spleen CFU from BALB/c mice after intravenous infection with pooled mutant strains both untreated (black circles) and treated with the indicated antibiotic. Plotted means from 4 biological replicates with standard deviations are shown. (C) Relative abundance of ΔglpK (top) and Δppe51 (bottom) strains compared to that of the wild type in vivo after antibiotic treatment. Treatment times were 14 days for PZA and MIX, 28 days for RIF, and 35 days for INH and EMB. Individual points are biological replicates normalized to day 0 ratios. Significance was calculated using unpaired t test with Benjamini-Hochberg multiple testing correction: *, P = 0.03; **, P = 0.002; ***, P = 0.0002; ****, P < 0.0001. (D) Lung CFU of H37Rv, ΔglpK, and complement strains from BALB/c mice after aerosol infection and treatment with PZA. Data represent two competition infections: 1:1 H37Rv and ΔglpK (black and blue circles, respectively) strains, dose of 700 to 1,000 CFU/mouse, and 1:1 H37Rv and complement (black and gray squares, respectively) strains, dose of 300 to 500 CFU/mouse. Treatment with PZA was started at 21 dpi and continued to 35 dpi. Shown are plotted means and standard deviations, and individual points are biological replicates. Limits of quantification are indicated by dotted red lines. Significance was calculated using unpaired t test with Benjamini-Hochberg multiple testing correction: *, P = 0.03; **, P = 0.002; ***, P = 0.0002; ****, P < 0.0001; ns, not significant.

FIG 4

Glycerol catabolic mutations associated with XDR strains. (A) Growth kinetics of drug-susceptible (DS) and drug-resistant (DR) clinical isolates on glycerol- or butyrate-containing media. Shown are plotted means from 3 biological replicates with standard deviations. (B) Sanger sequencing of glpK from H37Rv and clinical isolate KT0149-1. The homopolymer region is in the first domain of the protein. One-bp insertion changes downstream amino acid sequence and introduces a premature stop codon at amino acid 252. (C) Phylogenetic tree of M. tuberculosis isolates from Korea with various drug susceptibility profiles: DS (orange); DR, 1 to 4 antibiotics (blue); DR, 5 to 7 antibiotics (green); and DR >7 antibiotics (red). Mutations in glpK gene are indicated: frameshift mutations, purple stars; missense mutations, purple circles.

FIG 5

GlpK mutations associated with drug resistance in clinical isolates from Peru. (A) Phylogenetic tree of M. tuberculosis isolates from Peru. GlpK mutations are indicated (red circles). (B) Representation of GlpK mutations in different lineages: lineage distribution of 1,031 GWAS samples (outer); distribution of 68 glpK mutations (middle); distribution of 45 single-base expansions, T57GT, of the glpK homopolymer (inner). (C) Association between glpK mutations and drug resistance. Statistical significance (*) based on Bonferroni correction with a type 1 error rate of 0.01. INH, isoniazid; RIF, rifampin; RBU, rifabutin; EMB, ethambutol; PZA, pyrazinamide; STR, streptomycin; LIN, linezolid; MOX, moxifloxacin; AMK, amikacin; KAN, kanamycin; CAP, capreomycin; ETH, ethionamide.