Targeted transcriptomic analyses of tuberculosis treatment response and outcomes in The Gambia

Background Despite availability of effective treatment regimens for drug-susceptible Tuberculosis (TB), some patients still experience poor treatment outcomes. Currently tools for monitoring treatment outcomes are dependent on detection of mycobacteria in sputum, which are slow, expensive and poor at predicting relapse and failure. This study aims to identify new blood-derived markers for predicting treatment response and outcomes. Methods Whole blood was collected in PAXgene tubes from patients with microbiologically confirmed TB at diagnosis, week 2, and at months 2, 4, and 6. Treatment response and outcomes were determined by culture and gene expression was compared between slow and fast responders; and between patients with good (cured) and poor treatment outcomes (failure and recurrent TB) using targeted RNA gene expression. Gene signatures were developed using random forest classification models. Results Significant changes in gene expression were detected over the course of the TB treatment. Notably, major gene expression differences were observed at diagnosis between subsequently cured patients and patients who experienced poor treatment outcomes while minimal changes were detected between slow and fast responders among cured patients at diagnosis. A 7-gene end of treatment signature distinguished patients with good outcomes from those with poor treatment outcomes with area under the curve (AUC) of 0.91 (95% CI 0.85–0.99), 0.98 (95% CI 0.96–0.99), and 1.0 (95% CI 0.99–1.00), at baseline, month 2 and 6, respectively. Additionally, a 6-gene month 2 signature discriminates slow from fast responders with AUCs of 0.49 (95% CI 0.33–0.64), 0.58 (95% CI 0.07–1.00), and 0.93 (95% CI 0.78–1.00) at diagnosis, week 2 and month 2, respectively. Conclusion The study identified genes signatures associated with TB treatment response and outcomes suggesting potential utility for treatment monitoring.