Novel transcriptional signatures for sputum-independent diagnostics of tuberculosis in children

Abstract

Pediatric tuberculosis (TB) is challenging to diagnose, confirmed by growth of Mycobacterium tuberculosis at best in 40% of cases. The WHO has assigned high priority to the development of non-sputum diagnostic tools. We therefore sought to identify transcriptional signatures in whole blood of Indian children, capable of discriminating intra-thoracic TB disease from other symptomatic illnesses. We investigated the expression of 198 genes in a training set, comprising 47 TB cases (19 definite/28 probable) and 36 asymptomatic household controls, and identified a 7- and a 10-transcript signature, both including NOD2, GBP5, IFITM1/3, KIF1B and TNIP1. The discriminatory abilities of the signatures were evaluated in a test set comprising 24 TB cases (17 definite/7 probable) and 26 symptomatic non-TB cases. In separating TB-cases from symptomatic non-TB cases, both signatures provided an AUC of 0.94 (95%CI, 0.88-1.00), a sensitivity of 91.7% (95%CI, 71.5-98.5) regardless of culture status, and 100% sensitivity for definite TB. The 7-transcript signature provided a specificity of 80.8% (95%CI, 60.0-92.7), and the 10-transcript signature a specificity of 88.5% (95%CI, 68.7-96.9%). Although warranting exploration and validation in other populations, our findings are promising and potentially relevant for future non-sputum based POC diagnostic tools for pediatric TB.

Figure 1

Study flowchart. Selection of participants from study A (Light grey boxes) and study B (Darker grey boxes). Hatched grey boxes: Participants from both studies. Study A was a randomized-controlled trial (rct) of the effect of different micronutrient supplementary as an adjunct to anti-tuberculosis therapy (ATT), carried out from January 2008 to June 2012 in Delhi, India. Study B: A neonatal cohort study (NCS) of BCG-vaccinated neonates randomized to active or passive surveillance for 3 years, in Palamaner Taluk, India, April 2007 to September 2010. ¹Inadequate samples or lost to follow-up. ²Ninety of 692 were either QFT/TST positive, or both, indicating M. tuberculosis (Mtb) infection. ³Criteria for possible TB: ≥1 sign and symptom for TB, and either; response to treatment/documented exposure/immunological evidence of Mtb-infection, or; X-ray consistent with TB. ⁴TB ruled out by clinical, radiological and microbiological examination.

Figure 2

Upper figures: Discriminatory abilities for the identified signatures separating TB cases and asymptomatic HHCs in the training set, shown by: (a) receiver operator characteristics (ROC) curves/area under the curve (AUC), and (b) box-and-whisker plots (5–95 percentiles). Lower figures: Discriminatory abilities for the identified signatures separating TB cases from symptomatic non-TB cases in the test set, shown by: (c) receiver operator characteristics (ROC) curves/area under the curve (AUC), and (d) box-and-whisker plots (5–95 percentiles).