Characterising tuberculosis and treatment failure thereof using metabolomics

Abstract

Tuberculosis (TB), a highly contagious bacterial disease caused by Mycobacterium tuberculosis, is considered the leading cause of death globally from a single bacterial pathogen. The latest reports indicate 10.4 million new TB cases are diagnosed globally per annum, of which approximately only 5.7 million actually receive treatment, resulting in an estimated 1.8 million deaths. This high TB prevalence may be ascribed to a number of factors, including, amongst others, untimely and inaccurate diagnostics, inadequate treatment regimens, drug-resistant M. tuberculosis strains, human immunodeficiency virus (HIV) co-infection, and inadequate knowledge of the TB disease in general. This study is novel in the sense that it used a metabolomics research approach to identify new metabolite markers from patient-collected urine, for better characterising/understanding the TB disease state and treatment failure thereof. Using a validated urinary organic acid extraction and a two-dimensional gas chromatography time-of-flight mass spectrometry (GCxGC-TOFMS) metabolomics approach, we were able to differentiate a culture-confirmed active TB-positive group and TB-negative healthy control group, based on their detected metabolite differences, utilising a variety of multi- and univariate statistical methods. We identified the most significant urinary TB metabolite markers contributing to these differences, which shed light on previously unknown mechanisms/adaptations of the host in response to M. tuberculosis and other host–pathogen interactions. The most significant of these were the TB-induced changes resulting in an abnormal host fatty acid and amino acid metabolism, mediated through changes in interferon gamma and possibly insulin. This also explains some of the symptoms associated with TB and provides clues to better treatment approaches. Thereafter, the same approach was used to compare TB-positive patients with an unsuccessful treatment outcome to those successfully treated. Differentiation of the groups was achieved using the urine samples collected from these patients at time of diagnosis, i.e. before any treatment was administered. The identified urinary biomarkers were then used to better understand the underlying biology related to TB treatment failure. The most significant observations were the elevated levels of those metabolites associated with a gut microbiome imbalance, which has been shown to alter an individual’s response to anti-TB drugs and also negatively influence their immune function, contributing to an unsuccessful treatment outcome. Another interesting observation was those metabolites traditionally used for diagnosing inborn abnormalities in any of the three enzymes of the mitochondrial trifunctional protein complex in the treatment failure group. Since L-carnitine and various short-chain fatty acids are also reduced in these individuals, and are well-known for their antimycobacterial properties, this metabolic profile may explain an additional mechanism responsible for these individuals having an increased disease severity and/or a poor response to TB treatment. Considering the possible significance of these findings from a diagnostic perspective, the GCxGC-TOFMS data generated from the aforementioned treatment outcome experiment was reanalysed using a univariate statistical approach, in order to find possible diagnostic markers for predicting treatment outcome, utilising urine collected at time of diagnosis. Using a logistic regression model, two predictors, i.e. 3,5-dihydroxybenzoic acid and 3-(4-hydroxy- 3-methoxyphenyl)propionic acid, displayed the capacity to predict an unsuccessful treatment with an area under the receiver operating characteristic curve value of 0.94, and a leave-oneout cross-validation value of 0.89, indicating high sensitivity and specificity. Furthermore, these two identified predictors are also associated with an imbalance in gut microbiota, confirming the previously proposed mechanisms related to treatment failure in these individuals. Considering the results, this study not only proved the capability of a metabolomics research approach to identify new metabolite markers which could be used towards better understanding TB, and treatment failure thereof, but also possibly diagnostically for predicting treatment outcome of first-line anti-TB drugs at time of diagnosis. Furthermore, the fact that these markers can be detected from patient-collected urine, as opposed to sputum, has additional benefits for both research and diagnostic applications, considering the ease by which such samples can be obtained, with very little discomfort to the individual.