Abstract
Isoniazid (INH) is a critical antibiotic used worldwide for the treatment and prophylaxis of tuberculosis. Drug resistance (DR) to INH is the single most common type of DR, mediated by multiple genes/loci, including katG, inhA, mabA, mabA-inhA, and the oxyR-ahpC intergenic region. Over the course of 6 years, we performed a two-phase study of 3,696 Mycobacterium tuberculosis complex (MTBC) strains, aiming to determine the molecular basis of INH resistance and assess whole-genome sequencing (WGS) for predicting resistance. In phase 1, we performed a side-by-side study, including 1,767 strains with paired phenotypic drug susceptibility testing (DST) and genotypic DST. We found WGS capable of accurately predicting INH resistance with a sensitivity of 90.3% and a specificity of 99.8%. The negative predictive value of WGS for INH susceptibility was 98.8%. Based on these findings, we developed a molecular testing algorithm where phenotypic DST (pDST) was reduced and applied this new testing algorithm in phase 2 to 1,929 MTBC strains, resulting in streamlined testing, reduced cost, and decreased turnaround time (TAT). The prevalence of INH resistance among MTBC strains in New York was found to be 10.2%. Of the 3,696 isolates tested, 337 were predicted INH-resistant by WGS. Of 41 additional strains exhibiting phenotypic INH resistance, 38 were found to have mutations in genes known to be associated with INH resistance. This study demonstrates the utility of WGS as a molecular tool for predicting INH DR and shows that the vast majority of INH resistance in MTBC has a molecular basis in known resistance loci. IMPORTANCE: Isoniazid (INH) is one of the two most critical antibiotics used as part of standard treatment of tuberculosis and is also used as preventative therapy for contacts of tuberculosis patients, despite having a higher rate of drug resistance than all other antibiotics used in standard therapy. Furthermore, isoniazid resistance typically precedes rifampin resistance in the development of multidrug-resistant TB. As such, the reliable detection of INH resistance is crucial for case management and to limit the acquisition of additional drug resistance. The present study describes a whole-genome sequencing approach to predicting INH resistance from clinical isolates and models how this technology can be used within a reduced phenotypic drug susceptibility testing algorithm to limit duplicate testing, saving resources and time while maintaining the sensitivity of resistance detection.