In the present study, non-targeted metabolomics technology was used to identify the potential biomarkers of infant sepsis at an early stage. The metabolic profiles of infants with sepsis were significantly different compared to those without. Six differential metabolites were identified, and three of them had excellent diagnostics capability (AUC > 0.7). Further, bioinformatics results showed that metabolic changes were mainly related to lipid metabolism, cell proliferation, cell differentiation and apoptosis, which provided clues for exploring mechanisms of infant sepsis in the future.
The present study found that the levels of the three metabolites (PA (8:0/14:0), PE (16:0/18:2(9Z,12Z)) and CDP-CHO) were increased in the serum of sepsis infants. PA (8:0/14:0) and PE (16:0/18:2(9Z,12Z)) are glycolipid lipids and critical components of the biofilm lipid bilayer, which are involved in metabolism and cell signaling [14]. Moreover, CDP-CHO has been reported to be an essential intermediate of phosphatidylcholine biosynthesis in membrane structures, especially phosphatidylcholine [15]. Furthermore, Jaroonwitchawan et al. used metabolomics analysis and detected elevated PE (16:0/18:2(9Z,12Z)) levels in LPS-tolerant macrophages, which is consistent with our results [16]. In addition, increased serum levels of PA (8:0/14:0) and PE (16:0/18:2(9Z,12Z)) were also identified in septic mice by lipidomic analysis [17]. But, GB Li et al. [18] study showed opposite results that a decreasing tendency of PE (16:0/18:2(9Z,12Z)) was observed in sepsis patients, similar with this study. The contradiction could be related to the heterogeneity of the subjects, which the age of cases enrolled in GB Li study ranging from 15 days after birth to 13 years old. It speculated that the differences of dietary structure and liver function between neonates and young children may affect the metabolism in physiologic and pathophysiologic conditions.
In the current study, correlation analysis showed a negative relationship between PE (16:0/18:2(9Z,12Z)) and sepsis indicators such as PCT, suggesting that PE (16:0/18:2(9Z,12Z)) may be a potential biomarker to predict severe infection. Langley et al. [19] represented the metabolites of patients died of sepsis differed markedly from those of survival patients, and emphasized the predictive value of lipid metabolites for death in patients with sepsis. In addition, the finding of this study was again reinforced by a study of lipid levels in children with sepsis, presenting a positive relationship between lipid levels and infection severity [20]. The relationship of lipids metabolism to sepsis and the exact molecular mechanism deserves further study.
SM (d18:0/16:1(9Z)) is a sphingolipid substance on the membranes of cells that contributes to the structural integrity of those cells. It is also an essential component of signal transduction modules of critical biological reactions, such as cell division, differentiation, gene expression, and apoptosis [21, 22]. This work found that SM (d18:0/16:1(9Z)) levels decreased significantly in the sepsis group, consistent with the previous study by Winkler et al. [23], concluding that SM (d18:0/16:1(9Z)) is an accurate predictor for septic mortality [24].
Phosphocholine is an essential intermediate in lecithin synthesis and participates in various enzymatic reactions. Hecker et al. [25] found that phosphocholine can effectively control the release of ATP-mediated interleukin-1 β (IL-1β) in human and mouse monocytes, and IL-1β is essential in activating the inflammatory responses in sepsis. In this study, decreased expression of phosphocholine was observed in infants with sepsis. Therefore, we hypothesized that the activation of inflammatory response in infants with sepsis might be related to the decrease of endogenous phosphocholine levels. Lysophosphatidylcholine(LysoPC), generated from phosphatidylcholine(PC), may play a role as regulator of immune function [26]. The level of LysoPC decreased in sepsis group, compared with control group (Table 2). A similar result was also found in Drobnik’s research [27], and they reported that the molar ratio of LysoPC and PC, which was a reflection of enzymatic reaction for the lipid generation, can strongly predict sepsis-related mortality. Thus, there is a possibility that LysoPC-PC ratio might be a promising index to predict the outcome of sepsis, and more critical septic cases should be included in future study. Combined with the results of lipidomic analysis of PA(8:0/14:0), PE(16:0/18:2(9Z,12Z)), CDP-CHO and SM(d18:0/16:1(9Z)), we speculated that the changes of lipid metabolites in serum of septic infants might be related to the destruction of the cell membrane during inflammatory response induced by infection.
Besides these lipid metabolites detected in present study, a few amino acids have also been found to be different between the two groups. Phenylalanine, an essential amino acid, was found to be elevated in sepsis group. Another study conducted in adult septic patients, similarly to our results [28]. Furthermore, it indicated that Phenylalanine was found to have the ability to identify high risk patients and predict the mortality related to sepsis. While the mechanism of this change is still unclear, some researchers noted that the increasing concentration of phenylalanine might be related to insufficient tissue perfusion, increased insulin resistance, dysfunctional energy production and the large consumption of tetrahydrobiopterin [29,30,31]. The serum level of another amino acid, histidine decreased in sepsis group in this study. Histidine is essential for infants and may be able to inhibit the expression of pro-inflammatory cytokines [32]. The present results are consistent with another research carried out in adult patients suffered sepsis [16].
There are some limitations of classical indicators in neonatal sepsis: the WBC is not a very useful diagnostic marker because the specificity is low [33]; there are many conditions that cause PCT to increase falsely, such as nonspecific elevation in healthy newborns, prematurity, intracranial hemorrhage, birth asphyxia and neonatal hypoxemia [34], that showed limited value in the diagnosis of early-onset sepsis in neonates; the specificity of CRP is affected by non-infection conditions including hemolysis, tissue injury, surgery, stressful delivery, etc. [35] Correlation analysis revealed that prolylhydroxyproline, PE(16:0/18:2(9Z,12Z)) and CDP-CHO were associated with CRP or PCT, in addition, ROC analysis demonstrated that the three metabolites together improved the diagnostic accuracy. It is still necessary to conduct more studies with larger samples to verify diagnostic and prognostic values of metabolites.
Further, pathway analysis indicated that the glycerophospholipid metabolism, aminoacyl-tRNA biosynthesis and necroptosis might play an important role during sepsis. Subsequently, metabolite network analysis suggested that changes in metabolites were related to classic stress and inflammatory pathways such as ERK/MRPK, NF-κB, AMPK, and mTOR. Therefore, in addition to identifying potential markers of infant sepsis, the present study provides clues to the development of sepsis in infants.
Most of subjects in this study were neonates, differed from the previous pediatric cohort study [18]. All the subjects were term infants, and their mothers had no metabolic factors, which would influence the metabolite detection. Therefore, the results were more reliable. This study suggested clinical value of metabolomics in neonatal sepsis. In the future, the application and mechanisms of metabolites in neonatal infection need to be investigated.
Metabolomics is a valuable tool for diagnosis of sepsis in adults [36]. Compared with adults, infants with the dominantly milk-based diet and with little co-morbidities, the results of metabolites in those are more stable and reliable. The present report describes a clear clustering of the metabolomes in infants with sepsis, highlight the potential metabolites as early biomarkers for infant sepsis. A further strength of our study lies in study population, which studies selected infants or neonates are still scant.
A limitation of this study concerns the relatively limited number of cases. Compare with the number of cases in previously published study [37] on metabolomics in diagnosis of preterm early onset sepsis, the number of included cases in the present study was similar. Another weakness of this report is lack of evidence for association between metabolites and adverse outcomes. Finally, the results need to be validated and improved by further studies.