Predictive metabolite signatures for risk of progression to active TB from QuantiFERON supernatants of household contacts of TB patients

Abstract

The identification of individuals with the greatest risk of progression to active tuberculosis (TB) disease from the huge reservoir of Mycobacterium tuberculosis (Mtb) infected individuals continues to remain an arduous ascent in the global effort to control TB. In a two-year prospective study, we analysed metabolic profiles in the unstimulated and TB antigen stimulated QuantiFERON supernatants of 14 healthy household contacts (HHCs) who progressed to TB disease (Progressors) and 14 HHCs who remained healthy (Non-Progressors). We identified 21 significantly dysregulated metabolites in the TB antigen-stimulated QuantiFERON supernatants of Progressors, of which the combination of Malic acid and N-Arachidonoylglycine had maximum AUC of 0.99. Eighteen significantly dysregulated metabolites were identified in the unstimulated QuantiFERON supernatants of Progressors, among which the combination of Orotic acid and the phosphatidylcholines PC (O-34:1), PC (O-18:1(9Z)/16:0), PC (O-18:1(11Z)/16:0) had the maximum AUC of 0.98. Most of the dysregulated metabolites belonged to the pathways of fatty acid metabolism, lipid metabolism and nitric oxide metabolism. Validation of these metabolic signatures in large, diverse cohorts would pave way for the development of a robust test that can identify individuals at high risk of TB for targetted intervention of TB disease.

Keywords: Metabolites; QuantiFERON supernatant; biomarkers; diagnosis; metabolite signatures; progression; tuberculosis.

Figure 1.

Participant selection from two sites of the C-TRIUMPh cohort study. A total of 1051 adults and children were recruited in the C-TRIUMPh study. Participants were classified based on their baseline Mtb-infection status as positive for QFT (≥0.35 IU/ml) and/or positive for TST (induration diameter ≥5 mm) or negative for both. Among the enrolled participants, those who went on to develop TB during follow-up were identified as Progressors, and those who remained healthy were defined as Non-progressors. Progressors were matched to Non-progressors for age and gender.

Figure 2.

Overrepresentation Analysis of the Significantly Dysregulated Metabolites. Overrepresentation analysis of the significantly altered metabolites in (A) Stimulated supernatants and (B) Unstimulated supernatants. Red horizontal bars represent pathways which are significantly impacted (p value <0.05).

Figure 3.

Biomarker prediction by Multivariate ROC Analysis of significantly altered metabolites in the stimulated supernatants. (A) Overview of all ROC curves created by MetaboAnalyst 6.0 from 6 different biomarker models derived from stimulated QuantiFERON supernatants in the positive mode considering different number of features (2, 3, 5, 7, 10, and 17) with their corresponding AUC values and confidence intervals. (B) Graph presenting the predictive accuracies of the 6 different biomarker models. The red dot specifies the highest accuracy for the 17-feature panel of model 6. (C) Overview of all ROC curves created by MetaboAnalyst 6.0 from 3 different biomarker models derived from stimulated QuantiFERON supernatants in the negative mode considering different number of features (2, 3 and 4) with their corresponding AUC values and confidence intervals. (D) Graph presenting the predictive accuracies of the 3 different biomarker models. The red dot specifies the highest accuracy for the 2-feature panel of model 1.

Figure 4.

Biomarker prediction by Multivariate ROC Analysis of significantly dysregulated metabolites in the unstimulated supernatants. (A) Overview of all ROC curves created by MetaboAnalyst 6.0 from 6 different biomarker models derived from unstimulated QuantiFERON supernatants in the positive mode considering different number of features (2, 3, 4, 5, 6, 7, 8, 9 and 10) with their corresponding AUC values and confidence intervals. (B) Graph presenting the predictive accuracies of the 9 different biomarker models. The red dot specifies the highest accuracy for the 8-feature panel of model 7. (C) Overview of all ROC curves created by MetaboAnalyst 6.0 from 3 different biomarker models derived from unstimulated QuantiFERON supernatants in the negative mode considering different number of features (2 and 3) with their corresponding AUC values and confidence intervals. (D) Graph presenting the predictive accuracies of the 2 different biomarker models. The red dot specifies the highest accuracy for the 2-feature panel of model 1.