Host transcriptional biosignatures for TB progression in Indian household contacts

Achieving the goal of the End TB Strategy depends on new tools, preferably blood-based point-of-care (POC) tests, to identify individuals at risk of progression from M. tuberculosis infection to subclinical or clinical tuberculosis (TB). A decade of signature discovery and validation has resulted in several transcriptional signatures with promising capacity for predicting TB progression within the next 3–6 months, but evaluation of signature performance in Asian populations is lacking. Nested within a prospective observational cohort study of Indian household contacts (HHC study), we adapted the RISK6 and Sweeney3 along with our locally derived INDIA11 signature to the microfluidic RT-qPCR platform (Fluidigm) and evaluated their capacity to predict TB progression during a 12-month follow-up in a head-to-head comparison. As readily available in our dataset, the recently published single-gene signatures GBP2, FCGR1B, and SERPING1 were also assessed. Of the 525 recruited household contacts, 12 (6 cases in children aged 5–14 years) progressed to TB (5 at 6 months, 7 at 12 months) as defined by growth of M. tuberculosis in respiratory specimens. RISK6 and Sweeney3 were successfully adapted to the Fluidigm platform. One gene in the INDIA11 assay failed, resulting in the INDIA10 signature with an overall failure rate of 5.7%. RISK6, Sweeney3, and INDIA10 demonstrated comparable but statistically non-significant predictive performance for TB progression, with AUCs: RISK6 0.61 (95%CI 0.41–0.82), Sweeney3 0.61 (95%CI 0.41–0.80), INDIA10 0.59 (95%CI 0.46–0.72). Applying a fixed value of sensitivity 75% corresponding to the WHO minimum target for TB prediction resulted in the corresponding specificities RISK6 0.49 (95%CI 0.41–0.56), Sweeney3 0.32 (95%CI 0.25–0.39), and INDIA10 0.42 (95%CI 0.35–0.50). Of single-gene signatures, the AUCs of FCGR1B and GBP2 were significant with specificities of FCGR1B 0.56 (95%CL 0.49–0.64) and GBP 0.50 (95%CL 0.43–0.58). The expression of interferon-gamma–inducible gene signatures were able to discriminate between TB progressors and non-progressors in an hitherto underexplored Indian population, supporting the generalizability of RNA technology to this setting. Given the limited sensitivity and specificity of current biosignatures in all populations, exploration of integrated diagnostic algorithms relying on established and novel tools in a multidisciplinary approach is warranted.