Metabolomics is a powerful new technology that allows for the assessment of global metabolic profiles in easily accessible biofluids and biomarker discovery in order to distinguish between diseased and non-diseased status. We utilized this approach in a pilot study in urine and plasma samples from pneumonia patients from The Gambia. The global metabolic profiling and subsequent multivariate analysis clearly distinguished severe pneumonia patients from matched community controls. Although no common pathogenic factor was identified in all the cases, it is noteworthy that a similar disease manifestation allows for similar metabolic profiles and identification of biomarkers. Six metabolites emerged as markers of key differences between the two groups: uric acid, L-histidine, hypoxanthine, glutamic acid, L-tryptophan, and ADP. These metabolites together are important for the host response to infection through antioxidant, inflammatory and antimicrobial pathways, and energy metabolism.
Our observation of lower levels of urinary uric acid in severe pneumonia patients relative to controls suggests increased tubular reabsorption and renal retention of the analyte perhaps to protect against oxidative stress 
. In contrast, plasma levels of uric acid were elevated in pneumonia cases. In in vitro
studies, uric acid reacted rapidly with ozone and conferred protection of plasma lipids from peroxidation and erythrocyte lysis 
. Additional studies have shown that uric acid released from injured cells constitutes a major endogenous danger signal that activates the NALP3 inflammasome (also called cryopyrin or NLRP3), leading to IL-1b production 
as part of the host response to lung inflammation and fibrosis. Taken together, these studies suggest that uric acid plays a major anti-inflammatory role in pneumonia cases and allows for protection of the host organism from oxidative damage. Additionally, plasma hypoxanthine levels were elevated in patients relative to controls. Hypoxanthine and xanthine are the precursors of uric acid and uric acid was also elevated in the plasma of pneumonia patients, although not statistically significant. Xanthine on the other hand was not identified as a marker through the statistical analysis. High concentrations of hypoxanthine, xanthine, and uric acid have also been shown in patients with bacterial meningitis 
. This may be because sepsis provokes significant alterations in energy metabolism homeostasis with hypoxanthine and uric acid, offering possibly useful surrogate markers of infection 
. Elevation of hypoxanthine has also been reported during septic shock and may reflect early high energy nucleotide failure 
An additional marker with potentially important implications for disease outcome is ADP. The main role of ADP in the blood is the activation of platelets for effective hemostasis and blood aggregation 
. Our results indicated that ADP levels in plasma from pneumonia patients are downregulated. This in turn may lead to decreased platelet activation and decreased formation of aggregates and thromboemboli. Lack of purinergic receptors for ADP is a possible way to protect against aggregate formation, however the reduced plasma levels of ADP in pneumonia patients in our studies could confer a protective mechanism against organ failure 
. Nonetheless, other mechanisms, such as chemokine activation 
, appear to mediate the platelet activation under certain conditions, such as low ADP availability. Further work is required in this direction of low plasma ADP in pneumonia cases and its role in disease outcome and patient survival.
The last three markers identified in this study are amino acids. Hendriksen et al stressed that glutamic acid uptake and synthesis is important for full Streptococcus pneumoniae
fitness and virulence 
. The higher levels of glutamic acid in the plasma in patients relative to controls in our study may indicate cellular injury and protection. Additionally, excess circulation of glutamic acid during the disease state requires special attention in this particular human population as it can be associated with central nervous system damage, which is sometimes associated with severe pneumonia 
. L-tryptophan, on the other hand, was lower in plasma of patients relative to controls. Tryptophan starvation is a recognized antimicrobial defense mechanism through Indoleamine 2,3-dioxygenase (IDO) and mediates immunoregulatory effects 
. It is possible therefore, that tryptophan starvation initially exhibits an antimicrobial effect to aid in fighting the disease status and later contributes in regulating the T-cell response from a possible overstimulation 
. Urinary L-histidine was also lower in patients relative to controls. L-histidine is a precursor for the synthesis of histamine, a major contributor to inflammation, asthma, and potentially pneumonia in both human 
and animal studies 
. The lack on detection of histamine in urines of pneumonia patients and retention of L-histidine is possible indication that histidine is being converted to histamine in the tissues of pneumonia patients and contributing to their inflammatory state and propagation of disease status.
The sex dependent clustering of the pneumonia group, which was not present in the controls, points to differences in the metabolic response to pneumonia between male and female patients. Sex differences have been documented in survival following community-acquired pneumonia and nosocomial infections, which could be explained by differences in immune responses, genetics, or sex hormone levels 
. Additionally, a study by Casimir et al on childhood pneumonia revealed significant differences in inflammatory markers between male and female patients 
, making identification of metabolic differences between male and female patients an attractive candidate for future studies on diagnosis and drug development.
This small-scale preliminary study has clear limitations. It is not possible to say whether the metabolomic profile seen in these children with severe pneumonia is pneumonia-specific or associated with a wider spectrum of illness. Either of these possibilities is potentially diagnostically significant, and further work investigating the specificity of the findings must be done to resolve this question. The literature reporting metabolomic analysis in infectious diseases is limited. The majority of the work has been conducted in meningitis patients and considerable work has been conducted on the assessment of global metabolic profiling of bacteria 
. Pneumonia specific urinary metabolomic studies have concentrated on primarily adult populations with specific aetiology; however, this study is the first to provide pneumonia metabolomic analysis in urine and plasma from a specific pediatric population in parallel. Additionally, the differences in markers identified could be attributed to age and population related differences, overall aetiology of the pneumonia phenotype, and technologies and analytical methods utilized. This study was too small to define organism-specific metabolic responses, which will be useful for diagnosis, and it was not possible using available sensitive molecular techniques to distinguish causative pathogens from asymptomatic ‘DNAemia’. This is a general challenge for the growing field of molecular diagnostics rather than a limitation of this study in particular. The size of this study also means it has likely failed to identify other metabolites that will be important for diagnostics in the future.
The ability of the methods used in this study to clearly distinguish the children with severe pneumonia from their controls points to the considerable potential of metabolomics to improve diagnosis in sick children and to advance the knowledge of disease mechanisms. This preliminary work's importance is further emphasized by the fact that specific markers were identified in an outbred human population with genetic variability, no clear common causative agent, and simply a shared clinical syndrome. Metabolomics may provide an effective means to overcome the inability of current molecular pathogen detection techniques to distinguish causative pathogens from organisms that are ‘innocent bystanders’. Larger scale studies are now needed to determine the extent of its potential and to identify markers for different causative agents and for other potentially important aspects of disease such as illness severity, key comorbidities, and response to treatment. Once a panel of key biomarkers has been established there is the potential to take metabolomics closer to the bedside through its incorporation into point-of-care devices, which it is hoped will deliver breakthroughs in care in high mortality settings, and the evolution of which will likely be rapid in the next few years.