One-day phenotypic drug susceptibility testing for Mycobacterium tuberculosis variant bovis BCG using single-cell imaging and a deep neural network

Buu Minh Tran,Jimmy Larsson,Anastasia Grip,Praneeth Karempudi,Johan Elf
DOI: https://doi.org/10.1101/2024.05.20.594971
2024-05-20
Abstract:Drug-resistant tuberculosis (TB) kills approximately 200,000 people every year. A contributing factor is the slow turnaround time associated with anti-tuberculosis drug susceptibility diagnostics. The prevailing gold standard for phenotypic drug susceptibility testing (pDST) takes at least two weeks. In this study, we used Mycobacterium tuberculosis variant bovis BCG (M. bovis BCG) and Mycobacterium smegmatis as models for tuberculous and nontuberculous pathogens. The bacteria were loaded into a microfluidic chip, trapping them in microchambers, and allowing simultaneous tracking of single-cell growth with and without antibiotic exposure. A deep neural network image-segmentation algorithm was employed to quantify the growth rate over time and determine how the strains responded to the drugs compared to the untreated reference. We determined that the response time of the susceptible strains to isoniazid (INH), ethambutol (EMB), and linezolid (LZD) at MIC was within 3 hours and 1.5 hours for M. bovis BCG and M. smegmatis, respectively. Resistant strains of M. smegmatis were identifiable within 3 hours, suggesting that growth-based pDST can be conducted in less than 12 hours for slow-growing M. bovis BCG. The results obtained for M. bovis BCG are most likely comparable to what we expect for M. tuberculosis as these strains share 99.96% genetic identity.
Microbiology
What problem does this paper attempt to address?