The relative transmission fitness of multidrug-resistant Mycobacterium tuberculosis in a drug resistance hotspot

Abstract

Multidrug-resistant tuberculosis (MDR-TB) is among the most frequent causes of death due to antimicrobial resistance. Although only 3% of global TB cases are MDR, geographical hotspots with up to 40% of MDR-TB have been observed in countries of the former Soviet Union. While the quality of TB control and patient-related factors are known contributors to such hotspots, the role of the pathogen remains unclear. Here we show that in the country of Georgia, a known hotspot of MDR-TB, MDR Mycobacterium tuberculosis strains of lineage 4 (L4) transmit less than their drug-susceptible counterparts, whereas most MDR strains of L2 suffer no such defect. Our findings further indicate that the high transmission fitness of these L2 strains results from epistatic interactions between the rifampicin resistance-conferring mutation RpoB S450L, compensatory mutations in the RNA polymerase, and other pre-existing genetic features of L2/Beijing clones that circulate in Georgia. We conclude that the transmission fitness of MDR M. tuberculosis strains is heterogeneous, but can be as high as drug-susceptible forms, and that such highly drug-resistant and transmissible strains contribute to the emergence and maintenance of hotspots of MDR-TB. As these strains successfully overcome the metabolic burden of drug resistance, and given the ongoing rollout of new treatment regimens against MDR-TB, proper surveillance should be implemented to prevent these strains from acquiring resistance to the additional drugs.

Fig. 1. The population structure of M. tuberculosis in the country of Georgia.

Maximum likelihood phylogeny of 2,982 drug-susceptible M. tuberculosis genomes (a) and 980 MDR M. tuberculosis genomes (b) collected between 2014 and 2016 in Georgia. Clades coloured in red correspond to L4 strains and clades coloured in blue correspond to L2 strains. The green outer ring in panel B corresponds to genomes that carry the RpoB S450L mutation and the grey outer ring corresponds to the presence of a compensatory mutation in rpoA/B/C. The phylogeny of drug-susceptible strains was constructed from 63,111 variable nucleotide positions and the phylogeny of MDR strains was constructed from 13,649 variable nucleotide positions. Scale bars indicate substitutions per site.

Fig. 2. Phylogenies of M. tuberculosis genomes from Georgia and other countries.

Maximum likelihood phylogeny of 2,348 L2 MDR M. tuberculosis genomes (including 847 from Georgia) (a) and 2,927 L4 MDR M. tuberculosis genomes (including 133 from Georgia) (b). The outer ring in both panels corresponds to the country/region of isolation of the strains. The L2 phylogeny was constructed from 23,616 variable nucleotide positions and the phylogeny of L4 strains was constructed from 27,066 variable nucleotide positions. Scale bars indicate substitutions per site.

Fig. 3. The transmission fitness of MDR M. tuberculosis compared to drug-susceptible M. tuberculosis by lineage.

Posterior distributions of the transmission rate and effective reproductive number of drug-susceptible and MDR strains, estimated by fitting a two-type birth-death model on a random subset of sequences belonging to L2 (N = 200) (a, b) and L4 (N = 200) (c, d).

Fig. 4. The relative transmission fitness of MDR M. tuberculosis by lineage and RpoB mutation.

Posterior distributions of the transmission rate and effective reproductive number of drug-susceptible strains, MDR strains carrying the RpoB S450L mutation, and MDR strains carrying other rifampicin resistance-conferring mutations, estimated by fitting a three-type birth-death model on a random subset of sequences belonging to L2 (N = 274) (a, b) and L4 (N = 228) (c, d).

Fig. 5. The relative transmission fitness of MDR M. tuberculosis by lineage, RpoB mutation and the presence or absence of compensatory mutations.

Posterior distributions of the transmission rate and effective reproductive number of drug-susceptible strains and MDR strains with or without the RpoB S450L mutation and compensatory mutations, estimated by fitting a five-type birth-death model on a random subset of sequences belonging to L2 (N = 374) (a, b) and L4 (N = 228) (c, d).

Fig. 6. Interactions with additional drug resistance-related mutations and compensatory mutations in MDR M. tuberculosis L2 and L4 strains that carry the RpoB S450L mutation.

Co-occurrence patterns of RpoB S450L with other resistance-conferring and compensatory mutations in the L2 (a) and L4 (b) background. In L2, RpoB S450L interacted with 206 different other resistance-conferring mutations (based on the analysis of 767 genomes). In L4, RpoB S450L interacted with 79 different other resistance-conferring mutations (based on the analysis of 70 genomes).