Title: Tracking clock variations, contrasted demographies and selection signatures in Mycobacterium tuberculosis lineages: From simulations to the field

Authors: Thierry Wirth^1,2, Emilie Rousseau^1,2,3, Nour Gharbi^1,2, Matthias Merker³, Stefan Niemann^4,5

Affiliations: ¹EPHE, PSL University, Paris, France; ²Institut de Systématique, Évolution, Biodiversité, ISYEB, Muséum national d’Histoire naturelle, CNRS, Sorbonne Université, EPHE, Université des Antilles, Paris, France; ³Evolution of the Resistome, Research Center Borstel, Borstel, Germany; ⁴Molecular and Experimental Mycobacteriology, Research Center Borstel, Borstel, Germany; ⁵German Center for Infection Research, Partner site Hamburg-Lübeck-Borstel-Riems, Germany

Abstract: The M. tuberculosis lineage 2 is one of the most widely dispersed lineages globally and is particularly associated with the spread of multi-resistant strains in Eurasia. Yet, the reasons behind the success of Lineage 2 are only partially understood. In the present study, we wanted to develop an integrative approach that combines predictive simulations, sequencing data and laboratory fluctuation assays to answer this question.

One of our first objectives was to determine whether Bayesian phylogenomic approaches (BEAST algorithm) were able to distinguish contrasting mutation rates between lineages, or even to detect a potential acceleration of mutation rates along branches during an outbreak, based on the mutational parameters presented by Ford et al. (2015). According to our predictive simulations on SLIM, mimicking population sizes and demographic parameters sticking to recent outbreaks observed on the L2 and L4 lines, Bayesian inferences are able to detect relatively fine differences in mutation rates at the population level (around 10%). Furthermore, forward simulations and random sampling of sharply expanding M. tuberculosis populations under a constant mutation rate of 1 x 10^-7 substitutions/nucleotide/year did not show any signal of accelerated mutation rates under the Beast analyses (tip-dated trees), excluding therefore a potential confounding effect of the demographic component. Once reassured on this point, we compared two recent clones responsible for outbreaks belonging respectively to the L2 and L4 lineage (W148 and Haarlem/Hamburg) and presenting the characteristics of measurably evolving populations. Interestingly, the inferred mutation rate estimates confirmed that L2 outbreak strains evolve faster (P < 0.01) than L4 outbreak strains. Even more surprising, was the fact that the L2 strains displayed a significant pattern of mutation rate acceleration along the branches, and this at a scale of only 30 years.

Last but not least, the same trend was observed at larger scales, where Modern lineages (L2, L3 and L4) displayed slightly different mutational regimens on their house-keeping genes depending on the codon positions (1^st, 2^nd and 3^rd base), arguing for undergoing diversifying selection processes. This trend was even more marked within the modern lineage for the genes involved in antibiotic resistance, where the 2^nd base turned out to be the one with the most significant rate of evolution.