Tuberculosis (TB) takes longer to cure than almost any other bacterial infection. Subpopulations of Mycobacterium tuberculosis (Mtb) become phenotypically drug-tolerant in association with host-imposed stresses that put Mtb into a slowly-replicating or non-replicating state (here called NR for simplicity). We hypothesize that NR Mtb represent a major population of the "persisters" that survive the first months of chemotherapy. One subset of NR Mtb is termed "viable but nonculturable" (VBNC) because they elude detection as colony forming units (CFU) on agar. The nonculturability of VBNC Mtb has left us with almost no knowledge of their essential genes, metabolic pathways or susceptibilities to specific drugs and drug candidates. We have now improved a limiting dilution assay for VBNC Mtb and used it to confirm that an average of 87% of the viable Mtb detected in sputum from 33% of the untreated TB patients studied were VBNC forms missed both by CFU assays and by liquid culture in the BACTEC MGIT assay. Further, we developed a reproducible way to generate VBNC Mtb in vitro by prolonged starvation in axenic culture. In close collaboration with Project 5 and the Clinical Core, we will further improve and apply the resuscitation assay to characterize VBNC Mtb by chemical genomics in vitro, by genetics in the mouse, and by their presence in sputum from patients before and after standard therapy. We will deliver: compounds that eradicate VBNC Mtb in vitro and identification of the targets of these compounds; a mouse model where the efficacy of such inhibitors can be tested in vivo; and an assay that can be used in future clinical trials to complement the standard early bactericidal activity test on sputum so as to identify drugs that impact VBNC Mtb in the human host.