Deconvoluting drug interactions based on M. tuberculosis physiologic processes:Transcriptional disaggregation of the BPaL regimen in vivo

Abstract

Identification of optimal antibiotic combination treatments for tuberculosis (TB) in preclinical studies is impeded by the limited information conventional pharmacodynamic (PD) markers provide about drug interactions. Measurement of individual drug activity based on colony forming units (CFU) does not reliably predict the activity of drug combinations, potentially because one drug may affect the physiology of Mycobacterium tuberculosis (Mtb) in a way that either favors or disfavors the activity of a second drug. SEARCH-TB is a novel candidate PD approach which uses targeted in vivo transcriptional profiling to evaluate the effects of drugs on Mtb physiology. To test the capacity of SEARCH-TB to elucidate drug interactions, we deconstructed the BPaL (bedaquiline, pretomanid, linezolid) regimen in the BALB/c high-dose aerosol mouse infection model, measuring the effect of 2, 7, and 14-day treatment with drugs in monotherapy, pairwise combinations, and as a 3-drug combination. Monotherapy rapidly induced drug-specific Mtb transcriptional responses by day 2 with continued evolution over 14 days. Bedaquiline dominated pairwise combinations with both pretomanid and linezolid. The pretomanid-linezolid combination gave a blended response, inducing transcriptional profiles "intermediate" between either drug. In the 3-drug BPaL regimen, the addition of both pretomanid and linezolid to bedaquiline yielded a greater transcriptional response than expected based on pairwise results. This work demonstrates that physiologic perturbations induced by a single drug may be modified in complex ways when drugs are combined. This establishes proof of concept that SEARCH-TB is a highly granular readout of drug interactions in vivo, providing information distinct from CFU burden and suggesting a future in which regimen selection is informed by in vivo molecular measures of Mtb physiology.