Discovering lipid phenotypic changes of sepsis-induced lung injury using high-throughput lipidomic analysis

Abstract

Sepsis-induced lung injury (SLI) with high mortality and morbidity remains a leading cause of death in intensive care. There is an urgent need for the identification of SLI biomarkers for effective diagnosis. Studies have suggested the promising use of blood lipids as functional intermediate phenotypes in medical research. Lipid metabolism is critical in disease development, and lipidomics is the comprehensive analysis of molecular lipids. Lipid profiling by mass spectrometry may improve SLI risk prediction. The aim of this study was to use lipidomics to identify lipid molecules that could predict SLI patients. We performed a comprehensive and untargeted lipidomic analysis, using ultra-performance liquid chromatography/mass spectrometry on plasma samples from SLI patients and controls, and multivariate analysis methods were used to identify the lipids associated with SLI status. The serum samples were collected from SLI patients in the ICU. Lipid metabolic profiles were evaluated and compared between SLI and healthy cases. Variable importance in projection values were obtained to identify potential lipid biomarkers. A receiver operating characteristic curve was used to evaluate the power of the diagnostic biomarkers. Multivariate analysis revealed a good predictive model to distinguish their metabolic patterns. Seven lipid signatures were identified as potential biomarkers for SLI patients. These candidate lipids were validated in an additional independent cohort. According to the ROC analysis, the values of area under the curve (AUC) range from 0.813 to 0.997, indicating their potential power to distinguish between SLI and healthy cases. These results supported the concept that mass spectrometry-based lipid phenotype profiling might be a useful tool for the effective diagnosis of SLI.

---

Introduction

Sepsis is characterized by an intense systemic inflammatory response affecting the lungs, causing acute lung injury.¹ Sepsis-related lung injury (SLI) with significant morbidity and mortality is one of the main causes of death among hospitalized patients.² Lung injury is the most common form of organ injury in sepsis.³ It is well known that lung injury can occur simultaneously during severe inflammation. The lung is one of the vital organs affected during the sequential development of multi-organ dysfunction in sepsis. Lung injury is assessed by the presence of lung edema, increased vascular permeability, increased inflammatory cell infiltration and cytokine levels.^4–6 However, specific biomarkers are limited. Early detection and diagnosis is very important for improving SLI outcomes.

Lipids are a broad class of molecules with a variety of structures, and with diverse molecules functionally underpinning energy storage, the structure of cell membranes, and cell signaling.^7–10 The perturbations of lipid metabolism and signaling are implicated in the pathogenesis of the most common and devastating diseases. They also play a vital role in cardiovascular diseases,¹¹ inflammation,¹² Alzheimer's disease,^13,14 and metabolic diseases.¹⁵ Lipidomics is defined as “the full characterization of lipid molecules and their biological functions with respect to the expression of proteins involved in lipid metabolism and function, including gene regulation”, which deals with the detection and identification of lipids, fatty acids and their derivatives.¹⁶ The workhorse of high throughput lipidomics is mass spectrometry (MS), often generating hundreds of lipid identifications in a single experiment. Whole-plasma lipidomics takes a more global view of lipid metabolism and can provide a detailed picture of the abnormalities in lipid metabolism.¹⁷ Moreover, lipidomics analysis can lead us to new therapeutic targets and to novel therapeutic agents.¹⁸

The current increase in patients with lipid-related disorders has intensified the focus on lipid metabolism.¹⁹ To date, lipidomics has been used as a powerful tool from the perspective of disease risk,²⁰ but not for the identification of lipid metabolites associated with SLI. Blood plasma is widely used in lipidomics studies mainly due to its availability.²¹ The recent advance in MS technologies has enabled more comprehensive lipid profiling in biological samples.²² Therefore, lipidomics may contribute towards gaining new insights into the mechanisms underlying human SLI from blood samples. In this paper, we developed a high throughput MS-based lipidomics platform that offers lipidomics to identify the molecular lipid levels from systematic lipidomics profiling of large populations associated with SLI.

Experimental methods

Reagents and materials

HPLC grade acetonitrile was purchased from Merck (Germany). Leucine enkephalin was purchased from Sigma-Aldrich (St. Louis, MO, USA). Water was produced by a Milli-Q ultra-pure water system (Millipore, Billerica, USA). Both methanol and formic acid (HPLC grade) were purchased from Tedia (USA). The standards were purchased from Sigma-Aldrich, Inc. (St. Louis, MO, USA).

Patients

This study was performed in the Intensive Care Unit (ICU) at the Heilongjiang University of Chinese Medicine, in Harbin, the capital metropolitan city of Heilongjiang Province in north-eastern China, between May 2014 and August 2015. This study was approved by the Ethics Committee of Heilongjiang University of Chinese Medicine (HUCM2014-03791), and performed according to the Declaration of Helsinki. All participants were from the Chinese Han population and informed consent forms were signed by all of the subjects. The diagnosis of each patient was made independently by at least 3 experienced psychiatrists after evaluating potential participants according to the criteria of Consensus Conference 2001,¹⁷ without cardiopulmonary resuscitation, emergency origin, non-surgical, non-pregnant and non-chronic kidney disease. The healthy individuals had no disease symptoms and did not use any medication. Detailed characteristics for the patients are shown in ESI Table 1.† The plasma samples from subjects on week 1 of enrollment were selected for lipid profiling analysis. We collected whole plasma (5 mL) samples from each participant in the morning (7:30–8:30 am) following at least 10 hours of fasting. The plasma was stored at −80 °C until measurements were made.

Sample preparation and processing

Sample preparation for the lipidomics analysis was conducted as follows: 20 μL of plasma was added to a glass HPLC vial containing a 400 μL glass insert. Ten microlitres of high purity water and 40 μL of MS-grade methanol were added to each sample, followed by a 2 min vortex mix to precipitate the proteins. Then, 200 μL of methyl t-butyl ether was added, and the samples were mixed via vortex at room temperature for 1 h. After the addition of 50 μL of high purity water, a final sample mixing was performed before centrifugation at 3000 g for 10 min. The upper, lipid-containing, methyl t-butyl ether phase was then injected onto the LC-MS system directly from the vial by adjustment of the instrument needle height.

UPLC analysis

Chromatographic separation was achieved using an Acquity ultra-performance liquid chromatography system (Waters Inc., Milford, USA) with a 2.1 μm × 100 μm × 1.7 μm BEH C₈ column (Waters Inc., Milford, USA). The column temperature was maintained at 40 °C and the injection room temperature was 4 °C. The gradient used 0.1% formic acid in water as mobile phase A and 0.1% formic acid in acetonitrile as mobile phase B. The gradient consisted of 0 min (5% B), 2 min (40% B), 6 min (80% B), 8 min (100% B), 10 min (100% B). The solvent was delivered at a flow rate of 1 mL min⁻¹ and the injection volume was 2 μL. QC samples that were prepared by pooling all the analysed samples were injected at the beginning, middle and end of the run to check the system's performance.

MS analysis

Lipid analysis was carried out by positive-ion electrospray ionization mass spectrometry (ESI-MS) on a QTOF mass spectrometer (Waters Inc., Milford, USA). MS parameters were set up as follows: the electrospray voltage was 5.0 kV, the capillary temperature was 250 °C, the sheath gas flow rate was 20 units, and the desolvation temperature was 250 °C, with a desolvation gas flow of 400 L h⁻¹, and an extraction cone voltage of 3.0 V. A precursor ion isolation width of 0.1 m/z units was used, with a 10 ms activation time for MS/MS experiments. Full scan MS spectra and MS/MS spectra were acquired with a maximum ionization time of 10 ms and 50 ms, respectively. The collision energy varied between 10 eV and 20 eV for MS/MS. Accurate mass data and isotopic distributions for the precursor and product ions were analyzed and compared to the spectral data from lipidmaps (https://www.lipidmaps.org), MassBank (https://www.massbank.jp) and METLIN (http://www.metlin.scripps.edu). The tolerance for MS and MSMS identification was set to 5 ppm in the scans. Finally, the lipids were confirmed by comparison with authentic standards.

Multivariate analysis

The data were acquired and the results were analysed using the MarkerLynx System 4.1 (V4.1, Waters Inc., Milford, USA) for peak detection and alignment. Data visualization was performed on EZinfo Data System 2.0 (Waters Inc., Milford, USA). Data post processing and normalization were performed using an in-house developed data management system.²³ Multivariate analyses were performed by principal component analysis (PCA) and Orthogonal Projection to Least Squares Discriminant Analysis (OPLS-DA). Principal components analysis was then used for the detection of outliers. In addition, the value of variable importance for the projection (VIP) of OPLS-DA was applied to evaluate the variable contribution and find out the potential biomarkers. Receiver operating characteristic curve (ROC) was used to quantify the diagnostic power of the candidates. The prediction ability of these candidates was assessed by area under curve, accuracy, sensitivity and specificity. Statistical analyses were carried out using SPSS 18 software. Differences with p values of less than 0.05 performed by student′ test were considered statistically significant.

Results and discussion

Lipidomics multivariate analysis

Lipidomic analyses were performed for plasma samples collected from SLI and healthy subjects. In this study, UPLC/MS was applied to collect the lipid metabolites. Typical examples of lipid metabolite fingerprint profiles in human plasma samples from patients and healthy subjects are shown in Fig. 1. An OPLS-DA model (Fig. 2) from SLI subjects and healthy samples was built to discover the relevant lipid features. In PCA, R2Y (cum) and Q2 (cum) parameters were used for the evaluation of the models, indicating the fitness and prediction ability, respectively. In this study, the model's figures of merit were R2X = 0.734, R2Y = 0.895 and Q2 = 0.64, and suggested its reliability. The score plot of SLI and healthy cases could be separated into two distinct clusters, which indicated that the state of lipid metabolites was different between the SLI and healthy types.

Fig. 1 UPLC/MS base peak chromatograms of healthy cases (up) and SLI patients (down) from the ICU.

Fig. 2 PLS-DA score plot of lipid profiles data for the discrimination of SLI patients (black) and healthy controls (red). t[1]P and t[2]O represent the principal components.

Lipidomics signatures

To screen the relevant lipid biomarkers for SLI, we thus first examined the differences in lipid metabolite profiles of the plasma in all the groups. VIP values from the OPLS-DA model were obtained. The key lipid metabolites were selected from the VIP plot (Fig. 3) which was used to rank the contribution of the lipid metabolites. A total of 7 potential metabolites met the criterion of VIP >10, and also showed significant differences (p < 0.05) between the SLI and controls, of which 5 metabolites were higher in the SLI group, whereas 2 metabolites were higher in the healthy group. The fold changes of the lipid metabolite abundances ranged from 2.85 to 46.72. The lipid metabolites included PE(P-19:1(12Z)/0:0), PE(22:2(13Z,16Z)/15:0), PC(17:0/0:0), LysoPC(P-16:0), PE(20:3(8Z,11Z,14Z)/0:0), LysoPE(16:0/0:0) and PC(17:1(10Z)/0:0).

Fig. 3 VIP plot of PLS-DA model for the interesting lipid metabolites of SLI patients. Red depicts the lipids with marked changes.

Prediction of selected lipids for SLI risk

The prediction ability of the selected lipids was evaluated by the ROC curve. Table 1 shows the predictive power of the 7 selected lipids to discriminate SLI patients and healthy cases. The values of AUC range from 0.813 to 0.997, indicating the potential capacity of these differential lipids to distinguish between SLI patients and controls. Notably, PE(P-19:1(12Z)/0:0) had a sensitivity of 98.1% and a specificity of 97.3%, and the AUC value was 0.997, exhibiting good diagnostic performance. Three lipids (PE(P-19:1(12Z)/0:0), PE(22:2(13Z,16Z)/15:0), PC(17:0/0:0)) were selected to form a biomarker group (AUC > 0.95) to improve the risk discrimination between the SLI patients and healthy cases. This suggests that lipidomics appears to be the most promising method for determining the metabolic phenotype of SLI patients in clinical settings.

Table 1 Information on the lipidomics signatures in plasma detected by ultra-performance liquid chromatography/mass spectrometry

No.	VIP	Rt	[M + H]^+	Formula	Mass error (ppm)	Metabolites^a	HMDB	Max fold change	Trend	AUC	Sensitivity (%)	Specificity (%)
a All compounds were verified with an authentic standard.
1	19.47	1.85	478.3286	C24H48NO6P	−1.19	PE(P-19:1(12Z)/0:0)	LMGP02070003	46.72	↑	0.997	98.1	97.3
2	17.62	2.89	758.5717	C42H80NO8P	1.05	PE(22:2(13Z,16Z)/15:0)	HMDB09549	14.39	↑	0.980	97.2	96.5
3	14.81	4.84	510.3554	C25H52NO7P	0.00	PC(17:0/0:0)	LMGP01050024	6.26	↓	0.980	95.7	97.1
4	14.65	6.19	480.3469	C24H50NO6P	1.21	LysoPC(P-16:0)	HMDB10407	9.16	↑	0.942	92.9	88.1
5	11.53	4.40	504.3102	C25H46NO7P	−1.44	PE(20:3(8Z,11Z,14Z)/0:0)	LMGP02050022	2.85	↑	0.901	80.2	88.7
6	11.17	4.43	454.296	C21H44NO7P	2.08	LysoPE(16:0/0:0)	HMDB11503	3.93	↓	0.820	81.8	89.5
7	10.37	3.20	508.3403	C25H50NO7P	1.12	PC(17:1(10Z)/0:0)	LMGP01050002	4.59	↑	0.813	85.0	82.4

---

In conclusion, high-throughput UPLC/MS was used to characterize sensitive lipid biomarker(s) associated with SLI patients. The score plot of orthogonal partial least squares discriminant analysis showed significant discrimination between the SLI and healthy cases. Lipid profiling showed SLI to be associated with a profound abnormality in metabolic phenotype. The levels of PE(P-19:1(12Z)/0:0), PE(22:2(13Z,16Z)/15:0), PC(17:0/0:0), LysoPC(P-16:0), PE(20:3(8Z,11Z,14Z)/0:0), LysoPE(16:0/0:0), and PC(17:1(10Z)/0:0) in the SLI subjects were significantly different from the control cases. It is interesting to note that the selected lipids PE(P-19:1(12Z)/0:0), PE(22:2(13Z,16Z)/15:0), PC(17:0/0:0) with high AUC values exhibited satisfactory diagnostic performance. Our study suggests that lipidomics signatures may possess great potential for the diagnosis of SLI patients. Undoubtedly, the clinical validation of these selected lipids requires further studies including large series of patients.

Conflict of interest

The authors declare no competing financial interests.

Acknowledgements

This work was supported by grants from the Key Program of Natural Science Foundation of State (Grant No. 81470196, 81302905), Natural Science Foundation of Heilongjiang Province of China (H2015038), and Youth Innovative Talent Program of Heilongjiang Province of China (UNPYSCT-2015118).

References

1. N. Mulchandani, et al., Stimulation of Brain AMP-Activated Protein Kinase Attenuates Inflammation and Acute Lung Injury in Sepsis, Mol. Med., 2015, 21(1), 637–644.
2. Q. Liang, et al., High-throughput metabolic profiling for discovering metabolic biomarkers of sepsis-induced acute lung injury, RSC Adv., 2016, 6, 11008–11013.
3. W. C. Lin, et al., Kallistatin protects against sepsis-related acute lung injury via inhibiting inflammation and apoptosis, Sci. Rep., 2015, 5, 12463.
4. M. Acosta-Herrera, et al., Lung Transcriptomics during Protective Ventilatory Support in Sepsis-Induced Acute Lung Injury, PLoS One, 2015, 10(7), e0132296.
5. M. Xue, et al., Diagnostic and prognostic utility of tissue factor for severe sepsis and sepsis-induced acute lung injury, J. Transl. Med., 2015, 13, 172.
6. W. Tang, et al., The growth factor progranulin binds to TNF receptors and is therapeutic against inflammatory arthritis in mice, Science, 2011, 332(6028), 478–484.
7. H. Begum, et al., Discovering and validating between-subject variations in plasma lipids in healthy subjects, Sci. Rep., 2016, 6, 19139.
8. K. M. Sas, et al., Targeted Lipidomic and Transcriptomic Analysis Identifies Dysregulated Renal Ceramide Metabolism in a Mouse Model of Diabetic Kidney Disease, J. Proteomics Bioinf., 2015, 14, 002.
9. M. Wang, et al., Novel advances in shotgun lipidomics for biology and medicine, Prog. Lipid Res., 2016, 61, 83–108.
10. G. Mao, et al., Mitochondrial Redox Opto-Lipidomics Reveals Mono-Oxygenated Cardiolipins as Pro-Apoptotic Death Signals, ACS Chem. Biol., 2016, 11(2), 530–540.
11. X. Shi, et al., A Caenorhabditis elegans model for ether lipid biosynthesis and function, J. Lipid Res., 2016, 57(2), 265–275.
12. T. Hyötyläinen and M. Orešič, Analytical Lipidomics in Metabolic and Clinical Research, Trends Endocrinol. Metab., 2015, 26(12), 671–673.
13. F. Di Gaudio, et al., FragClust and TestClust, two informatics tools for chemical structure hierarchical clustering analysis applied to lipidomics. The example of Alzheimer's disease, Anal. Bioanal. Chem., 2016 , Epub ahead of print.
14. D. Ogasawara, et al., Rapid and profound rewiring of brain lipid signaling networks by acute diacylglycerol lipase inhibition, Proc. Natl. Acad. Sci. U. S. A., 2016, 113(1), 26–33.
15. T. Cajka and O. Fiehn, Toward Merging Untargeted and Targeted Methods in Mass Spectrometry-Based Metabolomics and Lipidomics, Anal. Chem., 2016, 88(1), 524–545.
16. J. Xu, et al., Lipidomics comparing DCD and DBD liver allografts uncovers lysophospholipids elevated in recipients undergoing early allograft dysfunction, Sci. Rep., 2015, 5, 17737.
17. Q. Liang, et al., Lipidomics Analysis Based on Liquid Chromatography Mass Spectrometry for Hepatocellular Carcinoma and Intrahepatic Cholangiocarcinoma, RSC Adv., 2015, 5, 63711–63718.
18. J. K. Prasain, et al., Comparative Lipidomics of Caenorhabditis elegans Metabolic Disease Models by SWATH Non-Targeted Tandem Mass Spectrometry, Metabolites, 2015, 5(4), 677–696.
19. J. Keereetaweep, et al., Lipidomic Analysis of Endocannabinoid Signaling: Targeted Metabolite Identification and Quantification, Neural Plast., 2016, 2016, 2426398.
20. Y. Zhang, et al., High resolution mass spectrometry coupled with multivariate data analysis revealing plasma lipidomic alteration in ovarian cancer in Asian women, Talanta, 2016, 150, 88–96.
21. B. J. Blaise, et al., Metabolic phenotyping of traumatized patients reveals a susceptibility to sepsis, Anal. Chem., 2013, 85, 10850–10855.
22. P. J. Crick, Guan XL.Lipid metabolism in mycobacteria-insights using mass spectrometry-based lipidomics, Biochim. Biophys. Acta, 2016, 1861(1), 60–67.
23. Q. Liang, T. Zhang, Y. Jiang and A. Zhang, Untargeted lipidomics study of coronary artery disease by FUPLC-Q-TOF-MS, Anal. Methods, 2016, 8, 1229–1334.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra03979k

---