Optimizing Multi-drug Mycobacterium tuberculosis Therapy for Rapid Sterilization and Resistance Suppression

Project Summary/Abstract In P01 AI123036, we were able to generate an algorithm that ranked single agents for Mycobacterium tuberculosis (MTB), identified promising 2-drug combinations and, with a completely novel mathematical approach, identified 3-drug regimens predicted to be significantly better than 2-drug regimens. These predictions were prospectively validated in a BALB/c model (H37Rv) and in a Non-Human Primate model of MTB (Erdman strain). In this proposal, we will extend our previous work. There is a large number of new MTB agents, many with novel mechanisms of action. We have 4 Specific Aims (SA) that, when complete, will allow us to identify multi-drug combinations that will optimize rate of kill for organisms in 3 different metabolic states and will suppress resistance emergence. In the Hollow Fiber Infection Model [HFIM] (SA#1), we will be able to rank new agents on the bases of potency and physicochemical properties. The HFIM provides insight into the drug’s exposure-response for kill and resistance suppression. We identified a near optimal 3-drug regimen (PMD/MFX/BDQ). With new single agents, we can examine substituting a new agent for an older agent AND we can expand the regimens to identify a near- optimal 4-drug regimen. This will be particularly important for patients with high bacterial burdens. In SA #2, we will test regimens from SA#1 in two murine models (BALB/c & C3HeB/FeJ mice). These will give somewhat different information. Both give information regarding kill and resistance suppression. Kramnik mice have pathology more closely resembling that in humans. We will use Matrix-Assisted Laser Desorption Ionization-MS Imaging and Laser Capture Microdissection LCMS. This allows identification of spatial distribution and quantification of drugs. A question regarding cure is how long to wait to sacrifice animals to document eradication. Some agents (BDQ) have long tissue half-lives. We will document rates of ingress/egress of drugs into the infection site, allowing determination when animal cohorts may be sacrificed to document eradication. In SA #3, we will document mechanisms of antimicrobial effect quantitatively. We have generated a first-of-a- kind dynamic model for PBP-binding in MTB, and will link this to rates of cell kill. We have also developed AMP/ADP/ATP intracellular assays. These will be employed for agents like diarylquinolines (e.g. BDQ) and PMD that act as energy poisons (for PMD, this occurs under anaerobic/non-replicative conditions. We will measure intracellular (MTB) drug concentrations, linking them to effect alone and in combination therapy experiments. Proposal success rests on modeling of the data. In SA #4, we have written code to extend earlier analyses, going from 3- to 4-drug regimens. For these high dimensional models, we developed several approaches to speed up analysis making them computationally tractable. At proposal end, we shall develop a 4-drug algorithm allowing rapid identification of near optimal regimens that work for both susceptible and less-susceptible organisms. The algorithm will be general. It will work well for today’s agents but also for agents as discovered.