RESUMO
BACKGROUND: High proportions of Mycobacterium tuberculosis cells in sputum containing triacylglycerol-rich lipid bodies have been shown to be associated with treatment failure or relapse following antituberculous chemotherapy. Although lipid body determination is a potential biomarker for supporting clinical trial and treatment decisions, factors influencing variability in sputum frequencies of lipid body-positive (%LB+) M tuberculosis in patients are unknown. We aimed to test our hypothesis that exposure to host-generated NO and M tuberculosis strains are factors associated with differences in sputum %LB+. METHODS: In this observational study, we determined %LB+ frequencies before treatment by microscopy in patients with smear-positive tuberculosis from two separate prospective observational study settings (Gondar, Ethiopia, recruited between May 1, 2010, and April 30, 2011, and Fajara, The Gambia, who provided sputum samples before treatment between May 5, 2010, and Dec 22, 2011). In Ethiopia, fractional exhaled nitric oxide (FeNO) was measured as a biomarker of host NO, and M tuberculosis strain differences were determined by spoligotyping. Treatment response was assessed by percentage weight change after 7 months. In The Gambia, treatment responses were assessed as change in BMI and radiographic burden of disease after 6 months. Sputum M tuberculosis isolates were studied in vitro for their %LB+ and triacylglycerol synthase 1 (tgs1) mRNA responses to NO exposure. Propidium iodide staining was used as a measure of NO strain toxicity. Correlation between in vitro %LB+ frequencies following NO exposure and those of the same strain in sputum was examined with linear regression and Dunnett's multiple comparison test. FINDINGS: In Ethiopia, 73 patients who were smear positive for pulmonary tuberculosis were recruited (43 [59%] were male and 30 [41%] were female). Of these, the %LB+ in the sputum of 59 patients showed linear correlation with log10 FeNO (r2=0·28; p<0·0001) and an association with strain spoligotype was suggested. Seven M tuberculosis strains from The Gambia showed different dose-responses to NO in vitro, demonstrated by changing lipid body content, tgs1 transcription, and bacterial toxicity. In sputum %LB+ frequencies correlated with in vitro %LB+ responses to NO of the corresponding isolate. In a subset of 34 patients across both cohorts, higher sputum %LB+ frequencies before treatment were associated with weaker responses to treatment than lower sputum %LB+ frequencies. INTERPRETATION: M tuberculosis strain and exposure to host-generated NO are associated with sputum %LB+. Our results support the use of M tuberculosis strain-dependent sputum %LB+ as a predictive biomarker of treatment response. FUNDING: The Medical Research Council, the University of Leicester, and the University of Gondar.