Transcriptomic Signatures of Progression to Tuberculosis Disease Among Close Contacts in Brazil

Abstract

Background: Approximately 5% of people infected with Mycobacterium tuberculosis progress to tuberculosis (TB) disease without preventive therapy. There is a need for a prognostic test to identify those at highest risk of incident TB so that therapy can be targeted. We evaluated host blood transcriptomic signatures for progression to TB disease.

Methods: Close contacts (≥4 hours of exposure per week) of adult patients with culture-confirmed pulmonary TB were enrolled in Brazil. Investigation for incident, microbiologically confirmed, or clinically diagnosed pulmonary or extrapulmonary TB disease through 24 months of follow-up was symptom triggered. Twenty previously validated blood TB transcriptomic signatures were measured at baseline by real-time quantitative polymerase chain reaction. Prognostic performance for incident TB was tested by receiver operating characteristic curve analysis at 6, 9, 12, and 24 months of follow-up.

Results: Between June 2015 and June 2019, 1854 close contacts were enrolled. Twenty-five progressed to incident TB, of whom 13 had microbiologically confirmed disease. Baseline transcriptomic signature scores were measured in 1789 close contacts. Prognostic performance for all signatures was best within 6 months of diagnosis. Seven signatures (Gliddon4, Suliman4, Roe3, Roe1, Penn-Nicholson6, Francisco2, and Rajan5) met the minimum World Health Organization target product profile for a prognostic test through 6 months and 3 signatures (Gliddon4, Rajan5, and Duffy9) through 9 months. None met the target product profile threshold through ≥12 months of follow-up.

Conclusions: Blood transcriptomic signatures may be useful for predicting TB risk within 9 months of measurement among TB-exposed contacts to target preventive therapy administration.

Keywords: biomarkers; blood; prognostic; transcriptomic; tuberculosis.

Figure 1.

Study flow diagram. Mtb, Mycobacterium tuberculosis; qPCR, quantitative polymerase chain reaction; TB, tuberculosis.

Figure 2.

Prognostic performance of transcriptomic signatures. Representative (A) box-and-whisker plots, (B) scatter plots, and (C) ROC curves of the transcriptomic signatures with the best prognostic performance through 6 months (Gliddon4 and Kaforou22) and 9 through 24 months (Duffy9) of follow-up. A, The box-and-whisker plots depict signature score distribution at enrollment by TB status; each dot represents a participant. Red crosses represent IGRA-negative (IGRA−) and IGRA-positive (IGRA+) nonprogressors; blue and green dots, pulmonary and extrapulmonary incident TB cases, respectively; and triangles and circles, microbiologically confirmed and clinically diagnosed TB, respectively. P values for comparison of median signature scores between groups in the box-and-whisker plots were calculated with the Mann-Whitney U test. Box, IQR; midline, median; whiskers, IQR ± (1.5 × IQR). Numbers of participants are included in Table 3. B, The scatter plots depict signature score distribution at enrollment and month 6 of follow-up by months to incident TB diagnosis. Each dot represents a participant with incident TB; blue and green dots, pulmonary and extrapulmonary TB cases, respectively; and triangles and circles, microbiologically confirmed and clinically diagnosed TB, respectively. LOESS curve, the local polynomial regression; shaded area, 95% CI on the LOESS regression. Spearman rank order correlation coefficient (ρ), demonstrating the association between time to incident TB diagnosis and signature score, and P values are shown for each signature. C, The ROC curves depict the prognostic performance (AUC with 95% CI) of signatures for differentiating progressors (incident TB disease) vs nonprogressors (healthy close contacts) through 6, 9, 12, and 24 months of follow-up. Numbers of participants are included in Table 3. The solid box depicts the optimal criteria (90%, sensitivity; 90%, specificity) and the dashed box the minimal criteria (75%, sensitivity; 75%, specificity) as set out in the World Health Organization’s target product profile for an incipient TB test [16]. (D) Summary of signature prognostic performance in the order of AUC estimates through 6 months of follow-up. The prognostic AUC estimates through 9, 12, and 24 months are also shown. Midline, the AUC estimate; error bars, 95% CI. AUC, area under the curve; IGRA, interferon-γ release assay; LOESS, locally estimated scatterplot smoothing; ROC, receiver operating characteristic; TB, tuberculosis.