Analysis of the Plasma Metabolome after Trauma, Novel Circulating Sphingolipid Signatures, and In-Hospital Outcomes

Abstract

Background: Trauma is the leading cause of death and disability for individuals under age 55. Many severely injured trauma patients experience complicated clinical courses despite appropriate initial therapy. We sought to identify novel circulating metabolomic signatures associated with clinical outcomes following trauma.

Study design: Untargeted metabolomics and circulating plasma immune mediator analysis was performed on plasma collected during 3 post-injury time periods (<6 hours [h], 6 h-24h, day 2-day 5) in critically ill trauma patients enrolled between April 2004 and May 2013 at UPMC Presbyterian Hospital in Pittsburgh, PA. Inclusion criteria were age ≥ 18 years, blunt mechanism, ICU admission, and expected survival ≥ 24 h. Exclusion criteria were isolated head injury, spinal cord injury, and pregnancy. Exploratory endpoints included length of stay (overall and ICU), ventilator requirements, nosocomial infection, and Marshall organ dysfunction (MOD) score. The top 50 metabolites were isolated using repeated measures ANOVA and multivariate empirical Bayesian analysis for further study.

Results: Eighty-six patients were included for analysis. Sphingolipids were enriched significantly (chi-square, p < 10^-6) among the top 50 metabolites. Clustering of sphingolipid patterns identified 3 patient subclasses: nonresponders (no time-dependent change in sphingolipids, n = 41), sphingosine/sphinganine-enhanced (n = 24), and glycosphingolipid-enhanced (n = 21). Compared with the sphingolipid-enhanced subclasses, nonresponders had longer mean length of stay, more ventilator days, higher MOD scores, and higher circulating levels of proinflammatory immune mediators IL-6, IL-8, IL-10, MCP1/CCL2, IP10/CXCL10, and MIG/CXCL9 (all p < 0.05), despite similar Injury Severity Scores (p = 0.12).

Conclusions: Metabolomic analysis identified broad alterations in circulating plasma sphingolipids after blunt trauma. Circulating sphingolipid signatures and their association with both clinical outcomes and circulating inflammatory mediators suggest a possible link between sphingolipid metabolism and the immune response to trauma.

Figure 1:

Patient selection and broad structure of the metabolomic dataset. (A) Graphical projection of the first 2 principle component dimensions highlighting that, using an unsupervised statistical approach, there are differences between timepoint 1 (<6 h) and timepoint 3 (D2-D5), with timepoint 2 (6 h-24 h) intermediate. Percentages listed after the dimensions represent the percentage of the total dataset variance encompassed by each dimension. (B) Using a supervised statistical method to maximize covariance and identify group discrimination (partial least squares-discriminant analysis, [PLS-DA]), a more pronounced group separation is identifiable. As before, the percentages listed after the dimensions represent the percentage of the total dataset variance encompassed by each dimension. (C) Comparison of metabolite superfamily representation among the initial metabolite dataset (n = 1000 metabolites) and the top 50 metabolites, as identified by consensus results of repeated-measures ANOVA and multivariate empirical Bayesian analysis to identify the metabolites that varied the most over time. Chi-square testing was used to compare the distribution of metabolite identities among the top 50 with the distribution within the original 1000 metabolites. (D) Heatmap summarizing distribution of top 50 metabolites across all 3 timepoints, organized according to metabolite superfamily and response profile. (E) Subfamiliy breakdown of the lipid metabolite class both in the original dataset (n = 488 lipids overall) and among the top 50 overall metabolites (n = 33 lipids represented). The distribution of lipids among the top 50 metabolites was statistically significantly different than the original dataset lipid distribution by chi-square analysis (p <0.005).

Figure 2:

Evaluation of the ceramide-sphingomyelin axis in blunt trauma patients. (A) Schematic breakdown of the relationship between identifiable subclasses of ceramide and sphingomyelin derivatives among the entire dataset. Heatmap data represent the group averages across timepoints demonstrating broad patterns over time within these metabolic subclasses. (B) Detailed heatmap of the data represented in (A). A notable pattern develops among the sphinganines, sphingosines, and glycosphingolipids by the D2-D5 timepoint, whereby possible subgroups of individual responders become apparent.

Figure 3:

Metabolomic identification of sphingolipid response subgroups in a blunt trauma cohort. (A) Isolation of the D2-D5 timepoint of the sphinganines, sphingosines, and glycosphingolipid subcategories followed by Pearson clustering demonstrates 3 distinct classes of patients: those who do not enrich in any sphingolipid signal, those who enrich the sphinganines and sphingosines, and those who enrich the glycosphingolipids. (B) Individual plots of representative biomolecules from each of the categories, demonstrating distinct patterns of enrichment (either a distinct upregulation by D2-D5, or an upregulation across all timepoints). Here, the dotted lines represent the average values of the noted metabolites taken from a separate metabolomic profile of 2000 individual patients from the Heart SCORE study, while the shaded box represents the range encompassing one standard deviation from the mean of the associated HeartSCORE data. Due to differences between the metabolomic studies in both patient selection and sample collection, no statistics were performed comparing trauma patients to the “control” group; average values provided purely for observational purposes. Labels at individual graph corners correspond to significance testing with repeated-measures ANOVA: G = Sphingolipid group significance p < 0.05, T = Time significance p < 0.05, I = Interaction significance p < 0.05.

Figure 4:

Cytokine profiles over time from patients in the blunt trauma cohort stratified by sphingolipid response group. Circulating cytokine and chemokine levels over time between times <6 h to day 5 are shown by sphingolipid enrichment group. Novel patterns emerge demonstrating clear differences in cytokine and chemokine profiles when patients are grouped according to sphingolipid responses, as described in the text. Labels at individual graph corners correspond to significance testing with 2-way (not repeated measures) ANOVA: G = Sphingolipid group significance p < 0.05, T = Time significance p < 0.05, I = Interaction significance p < 0.05.