Evolution of drug resistance in tuberculosis: recent progress and implications for diagnosis and therapy

Abstract

Drug-resistant tuberculosis is a growing threat to global public health. Recent efforts to understand the evolution of drug resistance have shown that changes in drug-target interactions are only the first step in a longer adaptive process. The emergence of transmissible drug-resistant Mycobacterium tuberculosis is the result of a multitude of additional genetic mutations, many of which interact, a phenomenon known as epistasis. The varied effects of these epistatic interactions include compensating for the reduction of the biological cost associated with the development of drug resistance, increasing the level of resistance, and possibly accommodating broader changes in the physiology of resistant bacteria. Knowledge of these processes and our ability to detect them as they happen informs the development of diagnostic tools and better control strategies. In particular, the use of whole genome sequencing combined with surveillance efforts in the field could provide a powerful instrument to prevent future epidemics of drug-resistant tuberculosis.

Fig. 1

A web of epistasis mediates drug resistance in M. tuberculosis. Key genes in M. tuberculosis drug resistance have been plotted, taking into account their approximate position in the genome. Genes in bold are known to be directly involved in antibiotic resistance. Lines denote putative epistatic interactions; connecting genes involved in the physiology of a drug as well as more broad/indirect mechanisms referred to as ‘ancillary to drug resistance’. This categorization is meant to include factors mediating complex aspects of cell physiology, such as cell permeability and mutation-induced physiological changes. Bold lines connecting rpoB to rpoC and embB to Rv3972 refer to in vitro validated compensatory mechanisms. Specific mutation pairs using M. tuberculosis numbering are shown where known. Figure based on information from the following references: [, , , , , , , , –94]. AG aminoglycosides, EMB ethambutol, ETH ethionamide, FQ fluoroquinolones, INH isoniazid, PAS para-aminosalicylic acid, PZA pyrazinamide, RIF rifampicin, STR streptomycin. Many ancillary genes were omitted for the sake of clarity—see original papers for a more comprehensive list [7, 54]

Fig. 2

The healthcare continuum. A diagrammatic representation of the current healthcare continuum as described in Wells et al. [71], showing different healthcare levels with their distance from the patient. The resources dimension encompasses a breadth of parameters, from access to infrastructure and apparatus, to technical proficiency of staff and financial resources that are available. Different diagnostic and DST technologies are shown as bars with the arrows indicating the levels at which we would ideally deploy them in the future. WGS whole genome sequencing, LPA line probe assay, MGIT mycobacterial growth indicator tube: phenotypic DST using the Bactec MGIT 960 instrument, automated NAT automated nucleic acid amplification technology

Fig. 3

Evolutionary trajectories of epistasis-driven resistance. The relative fitness of strains carrying key drug-resistance mutations grown in the absence of drug was plotted against their contribution to MIC to illustrate the relationship between genotype and phenotype for important drug-resistance mutations. Lines connecting individual mutations denote strains carrying two mutations, while arrows denote estimates of the fitness of double/triple-mutants (three different types of arrows are used to illustrate different evolutionary trajectories). Reproductive number (R ₀) defined as the number of secondary cases caused by an infected individual is roughly equal to the product of an organism’s transmissibility (t), number of contacts (c), and the duration of infection (d). Fitness estimates were summarized from Gagneux et al. [45] (RIF, M. tuberculosis), Borrell et al. [46] (FQ, Msm), Safi et al. [47] (EMB, M. tuberculosis), and Huitric et al. [85] (BDQ, M. tuberculosis). We were unable to find true relative fitness measurements for KatG S315T and GyrA A90V; these were estimated from Pym et al. [95] and Poissy et al. [96]. Fold increases in MIC shown are averages of values obtained from Sougakoff et al. [–99], Pang et al. [–99], and Anthony et al. [–99] for RIF; Pym et al. [95] for INH; Safi et al. [47, 100, 101], Plinke et al. [47, 100, 101], and Starks et al. [47, 100, 101] for EMB; Aubry et al. [, –105], Matrat et al. [, –105], Cheng et al. [, –105], Duong et al. [, –105], and Malik et al. [, –105] for FQ; and Huitric et al. [85] for BDQ. BDQ bedaquiline, EMB ethambutol, FQ fluoroquinolones, INH isoniazid, MIC minimum inhibitory concentration, RIF rifampicin