Characterising tuberculosis treatment success and failure using metabolomics
Tuberculosis (TB) is one of the deadliest infectious diseases of our time, with 1.4 million deaths globally, recorded in 2010 (3800 deaths a day) by the World Health Organization (WHO). Currently, South Africa ranks third on the 2011 list of 22 high-burden TB countries in the world and it was estimated that each active-TB person could potentially infect 10–15 people annually. The WHO additionally reported that in the year 2009, 87% of all TB patients worldwide were successfully treated, with a treatment success rate of 74% reported for South Africa. Despite this however, non-adherence to anti-TB treatment is still a major issue, due to it resulting in a global increased prevalence of drug resistant TB and subsequently TB treatment failure. Treatment failure is thought to be caused by a number of factors, however, it still remains largely misunderstood. One aspect of this, that isn't clearly addressed in the literature, is the underlying variation in each patient, resulting in his/her varying reaction to the drug regimen, and hence it’s varying efficacy from one patient to the next. Furthermore, little is known about the underlying variation of the host to the primary TB infection or response to the TB disease state, and how some patients have more effective mechanisms for eliminating the infection, or recovering from the disease. Considering this, a metabolomics research study using GC×GC-TOFMS was conducted, in order to identify potential metabolite markers which may be used to better characterise the underlining mechanisms associated with poor treatment outcomes (treatment failure). The first aim was to evaluate the accuracy and efficiency of the methodology used, as well as to determine the capability and accuracy of the analyst to perform these methods. In order to evaluate the GCxGC-TOFMS analytical repeatability, one QC sample was extracted and injected repeatedly (6 times) onto the GC×GC-TOFMS. Similarly, the analyst's repeatability for performing the organic acid extraction and analyses was also determined, using 10 identical QC samples, which were extracted and injected separately. CV values were subsequently calculated from the collected and processed data as a measure of this. Of all the compounds detected from the 6 QC sample repeats used for GCxGC-TOFMS repeatability, 95.59% fell below a 50% CV value, and 93,7% of all the compounds analysed for analyst repeatability had a CV < 50. Subsequently, using the above metabolomics approach, in addition to a wide variety of univariate and multivariate statistical methods, two patient outcome groups were compared. A sample group cured from TB after 6 months of treatment was compared vs a sample group where treatment failed after the 6 month period. Using urine collected from these two patient groups at various time points, the following metabolomics comparisons where made: 1) at time of diagnosis, before any anti-TB treatment was administrated, 2) during the course of treatment, in order to determine any variance in these groups due to a varying response to the anti-TB drugs, 3) over the duration of the entire 6 months treatment regimen, in order to determine if differences exist between the two groups over time. A clear natural differentiation between the cured and failed outcome groups were obtained at time of diagnosis, and a total of 39 metabolites markers were subsequently identified. These metabolites were classified according to their various origins, and included (1) those associated with the presence of M. tuberculosis bacteria, (2) those resulting from an altered host metabolism due to the TB infection, and (3) metabolites of various exogenous origins. The detailed interpretation of these metabolites suggests that a possible underlying RCD or some sort of mitochondrial dysfunction may be present in the treatment failure group, which may also be induced through an external stimulus, such as alcohol consumption. We hypothesise that this may possibly result in a far greater severity to M. tuberculosis infection in this group, subsequently causing a reduced capacity for a successful treatment outcome, also considering the critical role of the mitochondria in the metabolism of anti-TB drugs. Furthermore, 20 metabolite markers were identified when comparing the two outcome groups during the treatment phase of this metabolomics investigation. A vast majority of these 20 metabolites were also identified as markers for time 0 (time of diagnosis). Additionally, metabolites associated with anti-TB drug induced side effects, were also found to be comparatively increased in the treatment failure group, indicative of more pronounced liver damage, accompanied by metabolites characteristic of a MADD metabolite profile, due to a deficient electron transport flavoprotein, confirming previous experiments done in rats. These side effects have also previously been implicated as a major contributor of poor treatment compliance, and ultimately treatment failure. Lastly, 35 metabolite markers were identified by time dependent statistical analysis and represented those metabolites best describing the variation between the treatment outcome groups over the entire study duration (from diagnosis, to week 26). This time dependent statistical analysis identified markers, using an alternative statistical approach, and confirmed previous findings and added in a better characterisation of treatment failure. Considering the above, we successfully applied a metabolomics approach for identifying metabolites which could ultimately aid in the prediction and monitoring of treatment outcomes. This additionally led to a better understanding and or characterisation of the phenomenon known as treatment failure, as well as the underlying mechanisms related to this occurrence.