Rapid and accurate methods to monitor tuberculosis (TB) treatment response do not currently exist. Efforts to improve outcomes have focused on early identification of rifampicin susceptibility followed by prompt treatment initiation and adherence monitoring. The rapid molecular susceptibility tests most often used give dichotomous cutoffs. Recent studies though show that minimum inhibitory concentrations (MICs) just below these breakpoints also predict poor outcomes. Even if a patient takes most of their therapy, clinical response can still vary substantially. Delays in sputum clearance (culture conversion from growth to no growth) can range from a few days to 5 months and failure or relapse rates can be as high as 20% in drug-susceptible TB. During the weeks to months of human infection and antibiotic treatment, in host Mtb populations experience substantial measurable genetic changes. These changes may be neutral or allow pathogen adaption to immune, antibiotic or metabolic pressure, e.g. low iron or cobalamin levels that may result in heritable drug tolerance and resistance phenotypes. Here we propose to study in host longitudinal pathogen dynamics including changes in population diversity over time and identify genes under selection to shed light on host-pathogen interactions. The study of in host pathogen dynamics can improve our understanding of cure from infection and pave the way for the use of whole genome sequencing for monitoring treatment response, circumventing the delays and biohazards of traditional culture-based approaches. We additionally propose the development of a genome- based predictor of MIC and to assess if MIC predictions are associated with delays in culture conversion and poor clinical response. We will systematically study pathogen samples from a well characterized TB treatment patient cohort (NIAID TRUST TB cohort in Worcester, South Africa -PI Dr. Jacobson) combining long and deep short-read sequencing to resolve full genome assemblies and variants at low allele frequency. We have strong preliminary data that long-read sequencing unmasks more Mtb genetic diversity than detectable by short-read sequencing alone and have previously characterized directional selection in a subset of genes including resistance loci, the B12 biosynthesis pathway, and PPE genes known to interact with host innate defense. The proposed work is enabled by our methodological expertise in population genetics, machine learning and resistance prediction for clonal bacteria like Mtb and will allow, for the first time, the study of directional and diversifying selection on the full repertoire of Mtb genetic variation. It will also allow the training of an MIC prediction model on a large ~17,000 isolate dataset curated across studies and geographies. Both study aims promise to inform our understanding of how pathogen genetic variation affects Mtb survival in host and the response to treatment.
Patients with tuberculosis (TB) can benefit greatly if we could identify signs of poor response earlier and adjust therapy intensity accordingly. Innovations in DNA isolation and sequencing technologies now enable the study of TB pathogen populations in an individual patient with very high resolution. Here, we propose to use these technologies and a highly well-characterized cohort of South African TB patients to investigate how pathogen population genetic changes link to pathogen survival and treatment response, laying the foundation for improvements in clinical treatment monitoring using DNA sequencing.
