Analysis of Genome-Wide Mutational Dependence in Naturally Evolving Mycobacterium tuberculosis Populations

Abstract

Pathogenic microorganisms are in a perpetual struggle for survival in changing host environments, where host pressures necessitate changes in pathogen virulence, antibiotic resistance, or transmissibility. The genetic basis of phenotypic adaptation by pathogens is difficult to study in vivo. In this work, we develop a phylogenetic method to detect genetic dependencies that promote pathogen adaptation using 31,428 in vivo sampled Mycobacterium tuberculosis genomes, a globally prevalent bacterial pathogen with increasing levels of antibiotic resistance. We find that dependencies between mutations are enriched in antigenic and antibiotic resistance functions and discover 23 mutations that potentiate the development of antibiotic resistance. Between 11% and 92% of resistant strains harbor a dependent mutation acquired after a resistance-conferring variant. We demonstrate the pervasiveness of genetic dependency in adaptation of naturally evolving populations and the utility of the proposed computational approach.

Keywords: computational biology; microbial evolution; microbial genomics; microbiology; mycobacterium.

Fig. 1.

Patterns leading to detected evolutionary dependency. A simple framework classifying observed types of evolutionary dependencies in antibiotic resistance development. (A) Dependencies can potentiate resistance development. Potentiating mutations may amplify resistance, that is, directly influence the inhibitory concentration of the drug, or they may instead have a general effect on growth, virulence, and metabolism that increase the probability of acquisition of directly causal drug resistance mutations. (B) After initial resistance evolution, consequential mutations (i.e., arising as a consequence of resistance) are observed and manifest through multiple mechanisms. (C) Consequential mutations may restore fitness lost with the acquisition of resistance variants. The latter can be mediated through direct physical interactions or pathway-mediated changes in related genes. (D) Lastly, consequential mutations can causally amplify resistance, either through individual effects or epistatic effects such that the combination of the two variant effects is different than the sum of the individual effects.

Fig. 2.

Computational workflow for finding dependencies between mutations. (A) We found 1,184,177 pairs of SNPs across 4,743 sites that co-occur either sequentially or simultaneously at least once. We began with a data set of 31,428 isolate genomes and performed phylogeny and ancestral sequence reconstruction. We called each SNP as ancestral or derived relative to the pan-susceptible M. tuberculosis ancestral sequence (H37Rv) and then enumerated all SNPs that arise at least five times independently, dividing them into pairs that appear at least once sequentially or simultaneously. (B) For sequentially occurring pairs, we determine whether the probability of mutation a is affected by the presence of mutation b by inferring the distribution of the probability of mutation a in the context of b using a beta distribution, and then comparing it with the expected probability of mutation a not in the context of b. (C) For simultaneously occurring mutations, we determine whether the probability of observing mutations a and b simultaneously is higher than expected based on the product of the individual probabilities of mutation a and b—that is, assuming the two events are independent.

Fig. 3.

Sequential and simultaneous mutation pairs are enriched in functional categories. We determine the identity of the top 100 pairs of significant hits for (A) sequential mutation pairs and (C) simultaneous mutation pairs. We categorize mutation pairs as those where a known resistance mutation occurs first, known resistance mutation occurs second, both mutations are known resistance mutations, one mutation is in a known antigen protein, both mutations are in a known antigen protein, or other category (not any of the above). For simultaneous mutations, we compute the categories for the top 100 hits found within 100 bp on the genome and for the top 100 hits found outside 100 bp. The genomic distance in megabases of all pairs of significant dependent mutations for (B) sequential mutations and (D) simultaneous mutations are shown.

Fig. 4.

Dependent mutations within resistance-associated genes. We measured the identity and prevalence of significant dependent mutations occurring after initial resistance evolution. (A) Mutations that occur after mutations in known antibiotic resistance genes, visualized on the genome using pyCircos (github.com/ponnhide/pyCircos), with colors corresponding to the antibiotics in B. Antibiotics with shared genetic basis of resistance are shown in the same color. Only mutations that happen sequentially at least five times are shown. (B) Fraction of resistant strains that display one or more pairs of sequential dependent mutations. (C) Example of pairs of dependent mutations within the kanamycin resistance pathway, shown on a per-gene basis. Kanamycin's inhibition of the ribosome is blunted by ribosomal RNA mutations, while cellular kanamycin levels are reduced by increased levels of Eis, putatively caused by both mutations in the eis promoter region and mutations in the regulatory proteins WhiB7 and WhiB6. Dependencies between these mutations demonstrate multistep resistance evolution.

Fig. 5.

Dependencies between antibiotics. The detected significant dependent mutations between resistance-conferring mutations follow a particular order that mirrors the usage of different antibiotics. For each antibiotic, we took the top dependent pair between known resistance-conferring genes and other genes and between known resistance genes for different antibiotics. We display pairs and links where mutation a occurs sequentially or simultaneously with mutation b at least ten times. Link intensity corresponds to the number of occurrences. The prefix “p_” before a gene name indicates that the mutations are found in the upstream region of that gene. The drug para-aminosalicylic acid (PAS) is not included in the WHO catalog, but folC is a candidate resistance gene for this drug (Wei et al. 2019).