Model-Based Prediction of Functional SNPs Suggests Factors for Metabolic Diversity and Drug Resistance across Human-Associated Mycobacterium Tuberculosis

The Mycobacterium tuberculosis complex (MTBC) is a group of closely related pathogenic bacteria that can cause tuberculosis (TB) upon infection. TB is a leading cause of human mortality worldwide and threatens to remain so for the foreseeable future due to the emergence of multidrug-resistant MTBC strains. A characteristic feature of the MTBC is that it harbors comparatively little genetic variation, but a large share of its single-nucleotide polymorphisms (SNPs) are non-synonymous and experimental evidence is challenging the long-held assumption that phenotypic diversity in the MTBC is negligible. Bridging the genotype-phenotype gap in the MTBC could therefore provide valuable insights into the evolution of phenotypes such as drug resistance.

To address this problem, we built a single constraint-based model that integrates the exometabolomes and genomes of 18 MTBC strains from six lineages native to different parts of the world with strain-specific genome-scale metabolic models. We used the model to predict the metabolic effects of non-synonymous SNPs in enzyme-encoding genes via a three-step optimization approach, aiming to explain as much of the observed metabolic variation as possible by as few SNPs as possible while obtaining consistent flux distributions. Using our predicted SNP effects, we classified 88 SNPs (15%) as functional. These functional SNPs affect 67 unique enzymes across most metabolic pathways and include a SNP in pyruvate kinase previously shown to be functional in Mycobacterium bovis. In addition, we predicted three functional SNPs in enzymes involved in folate metabolism and we suggest a possible explanation for differential sensitivity to para-aminosalicylic acid, one of the antibiotics currently used to treat multidrug-resistant TB. Concluding, our method is capable of predicting the metabolic effects of genetic variation in microbes and allowed us to connect genetic and metabolic diversity in the MTBC.