Discussion
Improvements in systems biology theory and the development of metabolomics technologies have led to the recognition of the significance of metabolomics to clinical work in the field of critical care medicine.6 ,13 Many previous metabolomics experiments in animals have shown that metabolite profiles change dramatically and that small molecule metabolites can be used to identify infections and assess prognosis.14–17 However, clinical trials of metabolomics in sepsis have been limited. Our study used LC-MS/MS methods to screen serum biomarkers of the different stages of sepsis. We found that metabolite profiling was altered significantly in patients with SIRS in comparison with healthy control patients (figure 2A). The metabolomics approach could reflect changes in the metabolic state of the different stages of sepsis (figures 2B–D). Based on the bioinformatics analyses, 12 metabolites were ultimately identified as potential biomarkers of sepsis, disease severity and death.
An nuclear MR-based metabolomics study explored systemic metabolic changes in traumatic, critically ill patients and found that carbohydrate and amino acid levels were changed in the early SIRS stage.18 A targeted metabolomics study used LC-MS/MS to reveal that lipid compounds (acylcarnitines and glycerophosphatidylcholines) may be helpful for differentiating infection, and that identified lipid compounds may become promising candidate biomarkers.19 Stringer et al20 found four metabolites that could distinguish sepsis-induced acute lung injury from healthy controls, and these metabolites showed significant correlations with acute physiology scores. We found a sharp increase in S-(3-methylbutanoyl)-dihydrolipoamide-E and N-nonanoyl-glycine and significant decreases in lactitol dehydrate and S-phenyl-d-cysteine (table 2). S-(3-methylbutanoyl)-dihydrolipoamide-E participates in the degradation of valine, leucine and isoleucine and in the synthesis of branched chain fatty acids. Valine, leucine and isoleucine are classified as branched-chain amino acids (BCAAs), and they play regulatory roles in protein metabolism. Both animal experiments and clinical studies have shown that the concentrations of BCAAs are reduced during sepsis.21 ,22 We conclude that S-(3-methylbutanoyl)-dihydrolipoamide-E may play an important role in this pathway process. Moreover, it has been reported that severe disorders in amino acid metabolism commonly occur in critically ill patients.23 ,24 Although we were not able to obtain sufficient previous literature about N-nonanoyl-glycine and S-phenyl-d-cysteine, the trends in the upregulation and downregulation of their expression indirectly indicate metabolic changes in glycine and cysteine. In addition, a recent study has suggested that hydrophilic metabolites may provide an early signature of sepsis even prior to the onset of clinical symptoms.25 Lactitol dehydrate identified here is a hydrophilic carbohydrate metabolite, which supported this viewpoint. ROC curves were employed to evaluate the diagnostic efficiency of four selected biomarkers. Lactitol dehydrate showed the highest sensitivity, while S-(3-methylbutanoyl)-dihydrolipoamide-E had the highest specificity.
When the body is subjected to various pathological or physiological stimuli, changes in metabolic profiles may indicate the state of a disease or its severity.26–28 All these factors, including elevated body temperature, tissue hypoxia, changes in the internal and external environment, inflammatory cytokine production and abnormal levels of reactive oxygen species and hormones significantly affect the expression levels and activities of metabolic enzymes, which contribute to metabolic disorders in carbohydrates, proteins, fatty acids and nucleic acids. We speculate that some of the metabolites may reflect the severity of sepsis. In the present study, we found four metabolites that showed a significant downward trend. Glyceryl-phosphoryl-ethanolamine is a characteristic property of anticoagulation activity.29 We hypothesise that it was reduced due to coagulation disorders and vascular endothelial dysfunction during severe sepsis. Ne, Ne dimethyllysine mediated the response to oxidative stress30 and participated in the histone modification and gene regulation.31 Based on the human metabolome database (HMDB), 2-phenylacetamide is an intermediate in phenylalanine metabolism. Changes in 2-phenylacetamide and d-cysteine indicated the existence of an amino acid metabolism disorder. Based on ROC analyses, the sensitivities of glyceryl-phosphoryl-ethanolamine, 2-phenylacetamide and d-cysteine were higher.
Similarly, the metabolomics approach can also be used to evaluate the prognosis of sepsis. Mickiewicz et al32 screened a list of potentially important metabolites that could be used for diagnosis and the prediction of mortality in septic shock in a paediatric population. Seymour et al30 found that the global metabolomic profile is broadly different between survivors and non-survivors who were hospitalised with pneumonia and sepsis. It has also been reported that significant disturbances in fat metabolism are found in multiple organ dysfunction syndrome (MODS) and in cases with a poor prognosis.3 ,18 We identified four metabolites that were elevated within the 48 h prior to death. S-(3-methylbutanoyl)-dihydrolipoamide-E appeared again during this stage, which suggested severe metabolic disorders of BCAA and fatty acid synthesis. The elevated serum levels of PG (22:2(13Z, 16Z)/0:0) may be attributed to high concentrations of lipoproteins.19 The HMDB indicates that glycerophosphocholine is involved in signal transduction, glycerophospholipid metabolism, prostaglandin and leukotriene metabolism, and energy storage and consumption. However, its specific mechanism in sepsis is not known. The increased S-succinyl-glutathione concentration most likely resulted from oxidative stress. Our findings have some similarity with the recent study by Langley et al33, which also demonstrated that the different profiles of metabolites clustered into fatty acid transport and β-oxidation, gluconeogenesis and the citric acid cycle. According to the ROC curve, S-(3-methylbutanoyl)-dihydrolipoamide-E and PG (22:2(13Z, 16Z)/0:0) had higher sensitivities, while glycerophosphocholine had the highest specificity in the assessment of prognosis. These compounds may become potential metabolic biomarkers of sepsis.
This study has certain limitations. The limited sample size of this study included only 35 cases of patients with sepsis. In accordance with different prognoses, we found that the sulfur-containing amino acid concentrations of the non-survivor group were significantly lower than those of the survivor group at all time points, but the differences were not statistically significant at certain time points. One explanation for the lack of statistical significance could be the heterogeneity among the patients with sepsis, which then led to many different metabolic disorders. It would be necessary to use a larger sample size to validate our conclusions, particularly in view of the multiple comparisons performed in our study. Additionally, the study of sepsis metabolomics in patients with diabetes and the inclusion of lactate as a biomarker would each be worthy of future research.