Abstract
Active tuberculosis (aTB) remains a major killer from infectious disease, partially due to delayed diagnosis and hence treatment. Classical microbiological methods are slow and lack sensitivity, molecular techniques are costly and often unavailable. Moreover, available immuno-diagnostic tests lack sensitivity and do not differentiate between aTB and latent TB infection (LTBI). Here, we evaluated the performance of the combined measurement of different chemokines/cytokines induced by two different stage-specific mycobacterial antigens, Early-secreted-antigenic target-6 (ESAT-6) and Heparin-binding-haemagglutinin (HBHA), after a short in vitro incubation of either peripheral blood mononuclear cells (PBMC) or whole blood (WB). Blood samples were collected from a training cohort comprising 22 aTB patients, 22 LTBI subjects and 17 non-infected controls. The concentrations of 13 cytokines were measured in the supernatants. Random forest analysis identified the best markers to differentiate M. tuberculosis-infected from non-infected subjects, and the most appropriate markers to differentiate aTB from LTBI. Logistic regression defined predictive abilities of selected combinations of cytokines, first on the training and then on a validation cohort (17 aTB, 27 LTBI, 25 controls). Combining HBHA- and ESAT-6-induced IFN-γ concentrations produced by PBMC was optimal to differentiate infected from non-infected individuals in the training cohort (100% correct classification), but 2/16 (13%) patients with aTB were misclassified in the validation cohort. ESAT-6-induced-IP-10 combined with HBHA-induced-IFN-γ concentrations was selected to differentiate aTB from LTBI, and correctly classified 82%/77% of infected subjects as aTB or LTBI in the training/validation cohorts, respectively. Results obtained on WB also selected ESAT-6- and HBHA-induced IFN-γ concentrations to provided discrimination between infected and non-infected subjects (89%/90% correct classification in the training/validation cohorts). Further identification of aTB patients among infected subjects was best achieved by combining ESAT-6-induced IP-10 with HBHA-induced IL-2 and GM-CSF. Among infected subjects, 90%/93% of the aTB patients were correctly identified in the training/validation cohorts. We therefore propose a two steps strategy performed on 1 mL WB for a rapid identification of patients with aTB. After elimination of most non-infected subjects by combining ESAT-6 and HBHA-induced IFN-γ, the combination of IP-10, IL-2 and GM-CSF released by either ESAT-6 or HBHA correctly identifies most patients with aTB.