The antibiotic resistance arrow of time: efflux pump induction is a general first step in the evolution of mycobacterial drug resistance

Abstract

We hypothesize that low-level efflux pump expression is the first step in the development of high-level drug resistance in mycobacteria. We performed 28-day azithromycin dose-effect and dose-scheduling studies in our hollow-fiber model of disseminated Mycobacterium avium-M. intracellulare complex. Both microbial kill and resistance emergence were most closely linked to the within-macrophage area under the concentration-time curve (AUC)/MIC ratio. Quantitative PCR revealed that subtherapeutic azithromycin exposures over 3 days led to a 56-fold increase in expression of MAV_3306, which encodes a putative ABC transporter, and MAV_1406, which encodes a putative major facilitator superfamily pump, in M. avium. By day 7, a subpopulation of M. avium with low-level resistance was encountered and exhibited the classic inverted U curve versus AUC/MIC ratios. The resistance was abolished by an efflux pump inhibitor. While the maximal microbial kill started to decrease after day 7, a population with high-level azithromycin resistance appeared at day 28. This resistance could not be reversed by efflux pump inhibitors. Orthologs of pumps encoded by MAV_3306 and MAV_1406 were identified in Mycobacterium tuberculosis, Mycobacterium leprae, Mycobacterium marinum, Mycobacterium abscessus, and Mycobacterium ulcerans. All had highly conserved protein secondary structures. We propose that induction of several efflux pumps is the first step in a general pathway to drug resistance that eventually leads to high-level chromosomal-mutation-related resistance in mycobacteria as ordered events in an "antibiotic resistance arrow of time."

Fig 1

Ratio of the intracellular azithromycin concentration to that in the central compartment. At each of the time points when simultaneous drug concentrations were measured, azithromycin concentrations were much higher inside macrophages for all systems. A log₁₀ scale is used on the y axis; even with that scale, it can be seen that the ratios are not similar at all time points, and the relationships between intracellular and extracellular concentrations are thus nonlinear. The error bars indicate SD.

Fig 2

Azithromycin dose-response curves in the hollow-fiber system. Given that extracellular concentrations did not exceed the MIC in most cases but the drug nevertheless could kill MAC, we chose to use the intracellular concentrations for the dose-response curve.

Fig 3

Relationship between the azithromycin-resistant subpopulation and azithromycin exposure on day 10 of the dose-effect study. The relationship between the size of the low-level-resistant subpopulation and drug exposures (AUC/MIC) was described by the familiar inverted U shape.

Fig 4

Effect of the dosing schedule on the total microbial population. There was virtually no difference in microbial kill by dosing schedule for each drug exposure examined, consistent with an AUC/MIC-linked effect.

Fig 5

Emergence of high-level drug resistance follows low-level resistance mediated by efflux pumps. (A) On day 7, a low-level drug-resistant subpopulation was encountered. The subpopulation is a phenotype that grows on agar supplemented with 3× MIC (96 mg/liter) but not on agar supplemented with 256 mg/liter azithromycin. Since some of the population did not grow on plates supplemented with the efflux pump inhibitor thioridazine at 1 mg/liter (which has no effect on the microbial kill on its own), this means that efflux pumps accounted for a substantial portion of the resistance. The scale for the bacterial burden is a log scale, so what looks like small differences on this scale are relatively large on a linear scale. (B) By day 28, this low-level-resistant subpopulation had increased substantially, as had the proportion that could be inhibited by the efflux pump inhibitor. (C) A high-level-resistant subpopulation (which grew on 256 mg/liter azithromycin) had emerged by day 28 in some HFSs. Thioridazine had no effect on the size of that population. Neither a low-level- nor a high-level-resistant azithromycin subpopulation was encountered in untreated controls (not shown). The error bars indicate SD.

Fig 6

Real-time PCR of three putative azithromycin efflux pumps. MAC was incubated with 16 mg/liter azithromycin. (A) MAV_1659 demonstrated no significant upregulation during the first 72 h of exposure to azithromycin. (B) MAV_1406 demonstrated no significant change at 1 h (1.36- ± 0.30- versus 1.01- ± 0.17-fold) but was upregulated at 24 h (56.05- ± 9.99- versus 1.01- ± 0.17-fold) compared to baseline. There was no further change in induction beyond 24 h. (C) MAV_3306 induction was stepwise, starting at 1 h (1.68- ± 0.46- versus 1.00- ± 0.46-fold) to 50.60- ± 9.41- versus 1.01- ± 0.11-fold at 72 h. The error bars indicate SD.

Fig 7

Antibiotic resistance arrow of time. The diagram depicts the proposed steps in the evolution of high-level antibiotic resistance in mycobacteria.

Fig 8

Predicted protein secondary structure of the putative efflux pump encoded by MAV_3306 and its orthologs. Each red arrow represents an α-helix, while each blue arrow represents a β-strand. Highly conserved residues are shown in blue, while relatively conserved residues are shown in black. The degree of similarity among all proteins is represented by the bar diagram below the consensus sequence.

Fig 9

Predicted protein secondary structure of the putative efflux pump encoded by MAV_1406 and its orthologs. Each red arrow represents an α-helix, while each blue arrow represents a β-strand. Highly conserved residues are shown in blue, while relatively conserved residues are shown in black. The degree of similarity among all proteins is represented by the bar diagram below the consensus sequence.