Lysophospholipids Are Associated With Outcomes in Hospitalized Patients With Mild Traumatic Brain Injury

Abstract

Mild traumatic brain injury (mTBI) accounts for 70-90% of all TBI cases. Lipid metabolites have important roles in plasma membrane biogenesis, function, and cell signaling. As TBI can compromise plasma membrane integrity and alter brain cell function, we sought to identify circulating phospholipid alterations after mTBI, and determine if these changes were associated with clinical outcomes. Patients with mTBI (Glasgow Coma Score [GCS] ≥13 and loss of consciousness <30 min) were recruited. A total of 84 mTBI subjects were enrolled after admission to a level I trauma center, with the majority having evidence of traumatic intracranial hemorrhage on brain computed tomography (CT). Plasma samples were collected within 24 h of injury with 32 mTBI subjects returning at 3 months after injury for a second plasma sample to be collected. Thirty-five healthy volunteers were enrolled as controls and had a one-time blood draw. Lipid metabolomics was performed on plasma samples from each subject. Fold change of selected lipid metabolites was determined. Multivariable regression models were created to test associations between lipid metabolites and discharge and 6-month Glasgow Outcomes Scale-Extended (GOSE) outcomes (dichotomized between "good" [GOSE ≥7] and "bad" [GOSE ≤6] functional outcomes). Plasma levels of 31 lipid metabolites were significantly associated with discharge GOSE using univariate models; three of these metabolites were significantly increased, while 14 were significantly decreased in subjects with good outcomes compared with subjects with poor outcomes. In multivariable logistic regression models, higher circulating levels of the lysophospholipids (LPL) 1-linoleoyl-glycerophosphocholine (GPC) (18:2), 1-linoleoyl-GPE (18:2), and 1-linolenoyl-GPC (18:3) were associated with both good discharge GOSE (odds ratio [OR] 12.2 [95% CI 3.35, 58.3], p = 5.23 × 10^-4; OR 9.43 [95% CI 2.87, 39.6], p = 7.26 × 10^-4; and OR 5.26 [95% CI 1.99, 16.7], p = 2.04 × 10^-3, respectively) and 6-month (OR 4.67 [95% CI 1.49, 17.7], p = 0.013; OR 2.93 [95% CI 1.11, 8.87], p = 0.039; and OR 2.57 [95% CI 1.08, 7.11], p = 0.046, respectively). Compared with healthy volunteers, circulating levels of these three LPLs were decreased early after injury and had normalized by 3 months after injury. Logistic regression models to predict functional outcomes were created by adding each of the described three LPLs to a baseline model that included age and sex. Including 1-linoleoyl-GPC (18:2) (8.20% improvement, p = 0.009), 1-linoleoyl-GPE (18:2) (8.85% improvement, p = 0.021), or 1-linolenoyl-GPC (18:3) (7.68% improvement, p = 0.012), significantly improved the area under the curve (AUC) for predicting discharge outcomes compared with the baseline model. Models including 1-linoleoyl-GPC (18:2) significantly improved AUC for predicting 6-month outcomes (9.35% improvement, p = 0.034). Models including principal components derived from 25 LPLs significantly improved AUC for prediction of 6-month outcomes (16.0% improvement, p = 0.020). Our results demonstrate that higher plasma levels of LPLs (1-linoleoyl-GPC, 1-linoleoyl-GPE, and 1-linolenoyl-GPC) after mTBI are associated with better functional outcomes at discharge and 6 months after injury. This class of phospholipids may represent a potential therapeutic target.

Keywords: biomarkers; lysophospholipids; mild traumatic brain injury; outcomes.

FIG. 1.

Circulating levels of lipid metabolites in relation to mild traumatic brain injury (mTBI) outcomes. mTBI patients were dichotomized by good or bad discharge Glasgow Outcome Scale Extended (GOSE), and plasma lipid metabolite fold changes were calculated for statistical comparison. Volcano plots demonstrating plasma lipid metabolite fold changes within 24 h (A) and 3 months (B) after mTBI. All p values are false discovery rate (FDR) corrected. Fold changes were considered to be significant at log₂ (fold change) >0.6 or < -0.6, and corrected p < 0.05.

FIG. 2.

Changes in lipid metabolites over time after injury. Volcano plots of 24 h (A) and 3-month (B) changes after mild traumatic brain injury (mTBI). Fold changes are depicted comparing mTBI to controls. Metabolites found to be significantly higher in subjects with good outcomes (1-linoleoyl-GPC [18:2], 1-linoleoyl-GPE [18:2], and 1-linolenoyl-GPC [18:3]) are shown in red. Levels of each metabolite are shown in C–E. Levels are shown early (within 24 h) and late (3 months) after injury for those with bad (Glasgow Outcome Scale [GOSE] ≤6, red) and good (GOSE ≥7, blue) outcomes. *p < 0.001.

FIG. 3.

Associations between plasma choline and phospholipids. Each pixel represents a correlation coefficient corresponding to the phospholipid in that column and choline. Red, blue, and white coloring indicate positive, negative, or near-zero correlations. Correlations are shown between lysophosphlipids (LPL) and choline (A) and phosphatidylcholines (PC) and choline (B). *p < 0.05 in controls, ^#p < 0.05 early after mild traumatic brain injury (mTBI), ^†p < 0.05 late after mTBI.

FIG. 4.

Receiver operating curve (ROC) analysis. Machine learning models were developed in order to determine the prognostic value of plasma lipid metabolites to predict outcomes after mild traumatic brain injury (mTBI). Baseline models included age and sex. Models were developed adding each of the three metabolites found to be upregulated in subjects with good discharge outcomes and to have significant associations with outcomes in multivariable models. Areas under the curve (AUCs) relative to the baseline model are shown for predicting discharge outcomes (A). A model was also created including all three metabolites added to the baseline model to predict discharge outcomes (B). Only one metabolite (1-linoleoyl-GPC [18:2]) when added to the baseline model individually significantly increased the AUC to predict 6-month outcomes (C). Principal components (3) generated from 25 detected lysophosphlipid (LPL) metabolites were added to the baseline model to predict 6-month outcomes (D).

FIG. 5.

Schematic overview of pathways involved. A diagram of the Kennedy and Lands Cycles is shown depicting the biosynthesis pathways of membrane phosphatidylcholine. Phospholipases remove an acyl group from phosphatidylcholine to produce lysophospholipids and free fatty acids (FFAs). Lysophospholipids can be combined with diverse acyl groups to generate phosphatidylcholine for membrane repair. FFAs, such as polyunsaturated fatty acids (PUFAs), removed from the sn-2 position of phosphatidylcholine may be used in downstream signaling pathways. An analogous pathway exists for phosphatidylethanolamine. Abbreviations: correlation (cor), lysophospholipid (LPL), phosphatidylcholine (PC), diacylglycerol (DAG), phospholipase A₂ (PLA₂), lysophospholipid acyltransferase (LPLAT).