Untargeted metabolomic analysis using LC-TOF/MS and LC-MS/MS for revealing metabolic alterations linked to alcohol-induced hepatic steatosis in rat serum and plasma

Abstract

Alcohol-induced hepatic steatosis (AHS), an early stage of alcoholic liver disease, is characterized by large amounts of fat deposited in hepatocytes. Although some important biomarkers have been identified, there still is no systematic view of AHS-associated biomarker alteration in vivo. Blood matrix choice is a fundamental consideration for biomarker exploration. In the present study, a reliable untargeted metabolomic method based LC-TOF/MS and LC-MS/MS was developed to (i) interrogate overall differences of endogenous metabolites in rat serum and plasma, and investigate which of the two matrices is more suitable for the discovery of endogenous biomarkers, (ii) analyze metabolic alterations linked to incident AHS. As a result, metabolite level differences between serum and plasma in normal rats were mainly related to 7 peptides, 8 lysophosphatidylcholines, 3 phosphatidylcholines, 2 heterocyclic compounds, 7 polyunsaturated fatty acids and their oxidative fatty acids. Notably, 7 characteristic peptides were observed only in serum. All the altered metabolites were associated with clotting cascade reaction, indicating that serum conceals the changes in metabolome that have occurred in vitro. The results demonstrated that plasma which represented the original properties of the blood sample was a more suitable matrix to explore potential biomarkers. In the plasma untargeted metabolomic study, the metabolic alterations linked to AHS were mainly involved in phospholipid metabolism, amino acids metabolism, fatty acid metabolism, cholesterol metabolism and sphingolipid metabolism. These findings provided a systematic view of metabolic alterations linked to AHS, demonstrating the untargeted metabonomic method was a robust method for examining the molecular mechanisms of disease.

---

1. Introduction

Untargeted metabolomics has emerged as a powerful method for the discovery of disease biomarkers, providing fundamental insights into cellular biochemistry and clues related to disease pathogenesis over the last two decades.^1,2 It consists of the global profiling of low-molecular-weight metabolites in organisms and identifying the most meaningful metabolites which play important roles in biological pathways. Mass spectrometry (MS)^3,4 and nuclear magnetic resonance (NMR) spectroscopy^5,6 are two principal analytical platforms used to perform untargeted metabolomics in recent years. Although NMR provides conclusive structural information and high reproducible data about metabolites in vivo, it suffers from limitations in sensitivity and chemical resolution.⁷ In contrast, MS provides less-conclusive structural information, but given its high sensitivity, good specificity and large dynamic range, it allows for the acquisition of more data from each biological sample.⁸ Because of their full-spectrum detection capabilities and high mass accuracy, high-resolution mass spectrometers such as time-of-flight (TOF), Orbitrap, or hybrids mass analyzers such as quadrupole-TOF (QTOF) and Q Orbitrap are adequate to detect many more chemical species and develop untargeted metabolomic approaches in complex biological matrixes.^9,10

Alcoholic liver disease (ALD) is a chronic disorder characterized by excessive ingestion of alcohol and manifested as a broad spectrum of disorders, ranging from hepatic steatosis, steatohepatitis, fibrosis and cirrhosis (Laennec’s cirrhosis).¹¹ The earliest liver injury observed in alcoholics is steatosis which is characterized by large amounts of fat deposited in hepatocytes.¹² Alcohol-induced hepatic steatosis (AHS) is largely asymptomatic and the liver damage is reversible after cessation of alcohol consumption, otherwise it advances to irreversible stages of ALD which associated with increased risk of liver cancer.¹³ Detection of ALD at the stage (AHS) is, therefore, key to ameliorate quality of life, improve therapeutic benefit and reduce healthcare burden.

Although some important biomarkers have been identified,^12,14 there still lacks a systematic view of AHS-associated biological processes in vivo, such as broad metabolite level, which is pivotal to clarify the physiological mechanism of disease. Matrix choice is a fundamental consideration for biomarker exploration. Peripheral blood possesses a wealth of information and has been shown to establish the overall pathophysiology mapping of an individual.¹⁵ It does satisfy the criteria of minimal invasiveness and reasonable cost compared with the collection of tissue samples by biopsy.¹⁶ Therefore, serum and plasma samples are widely applied in metabolomic analysis of mammalian systems.¹⁷ However, few studies have investigated potential differences in serum and plasma metabolomes and explored which matrix could capture optimal metabolite information coverage for disease.

In this study, we aimed to (i) interrogate serum and plasma in order to identify overall differences in the metabolite levels, and investigate which of the two matrixes is more suitable for the discovery of endogenous biomarkers, (ii) analyze metabolic alterations linked to AHS. Toward these aims, an untargeted metabolomic approach based LC-TOF/MS and LC-MS/MS was used to globally profile and identify the low-molecular-weight endogenous metabolites in serum and plasma of normal and AHS rats, respectively. Both univariate statistics and multivariate data analysis (MVDA) were employed to test the analytical reproducibility in terms of detected ion intensities and rapidly screen the interesting metabolites associated with AHS.

2. Experimental

2.1 Chemicals and reagents

Chromatographic-grade methanol and acetic acid were purchased from Merck (Darmstadt, Germany) and Sigma Chemical (St. Louis, USA), respectively. Ultra high purity water was produced by Millipore-Q water purification system (Bedford, USA).

2.2 Animals and treatment

All protocols and care of the rats were in accordance with the Guidelines for the Care and Use of Laboratory Animals and approved by the Animal Ethics Committee of China Pharmaceutical University. Male Sprague-Dawley rats (12–14 weeks) weighing 180–220 g were acclimated at 50 ± 20% humidity and 20 ± 2 °C with a 12 h light/dark cycle in an animal breeding room. Purified water and standard chow were provided ad libitum. After 1 week of acclimatization, the rats were randomly divided into the AHS group and control group (n = 8 for each group). Animals were housed individually in cages and provided with the regular Lieber–DeCarli alcohol or pair-feeding (ethanol was substituted isocalorically with carbohydrate) liquid diet for eight weeks in AHS group and control group, respectively. The amount of ethanol (%, w/v) in the liquid diet was 5.00%, 5.14%, 5.29% and 5.43% for every two weeks, respectively. To achieve equal daily energy intake, the rats in the AHS group were fed ad libitum, while control rats were pair-fed the same amount consumed by model rats during the prior day.¹⁸

2.3 Sample collection

All rats were fasted over night before the study. The rats were anesthetized by intraperitoneal injection of urethane (1.6 g kg⁻¹ body weight), and blood samples were collected from hepatic portal vein into heparin anticoagulation (plasma) and no anticoagulant (serum) tubes. The heparin anticoagulation tubes were gently inverted three times to ensure proper mixing of blood with the heparin anticoagulant. The plasma samples were immediately isolated by centrifugation (2500 × g, 10 min, 4 °C) after sampling. The serum samples were kept at 4 °C for 1 h to clot after blood collection, and then centrifuged (2500 × g, 10 min, 4 °C). Volumes of 200 μL plasma or serum samples were aliquoted separately in 2.0 mL microcentrifuge tubes. A quality control pooled (QCP) sample was prepared by mixing aliquots of 20 μL of each of the experimental samples. All samples were stored at −80 °C until sample preparation was initiated.

2.4 Analytical design

2.4.1 Sample preparation. Frozen samples were thawed at 4 °C for 30 min. Matched 200 μL aliquots of serum and plasma samples from each of the eight subjects, including QCP samples were deproteinized with 600 mL cold methanol. After vortex mixing for 1 min, centrifugation at 12 500 × g was performed for 10 min at 4 °C. A 600 μL aliquot of supernatant was transferred into a fresh microcentrifuge tube and speed-vacuum-dried at room temperature, then reconstituted in 200 μL of 20 : 80 methanol/water followed by centrifugation (12 500 × g, 10 min, 4 °C). A 20 μL aliquot was injected for LC-MS analysis.

2.4.2 LC-TOF/MS analysis. Metabolite extracts from rat blood plasma and serum were analyzed on an Agilent-1260 Series LC system which was coupled with an Agilent-6224 TOF mass spectrometer (Agilent Corp., Santa Clara, CA, USA) equipped with an electrospray interface. Chromatographic analysis was performed on a Phecda C18 column (250 mm × 4.6 mm, 5 μm, Hanbon Science & Technology Co., China). The autosampler and column were maintained at 4 and 35 °C, respectively. The mobile phase was water with 0.1% acetic acid (A) and methanol (B). The elution gradient was optimized as follows: 0 min, 5% B; 10 min, 30% B; 20 min, 65% B; 30 min, 85% B; 45–50 min, 90% B; 51–60 min, 95% B, which was delivered at 1.0 mL min⁻¹. Re-equilibration duration was 10 min between individual runs.

The optimized conditions of TOF/MS were: capillary voltage, 4.0 kV/−3.5 kV; drying gas (N₂) temperature, 350 °C; drying gas (N₂) flow rate, 12.0 L min⁻¹; Oct RFV, 750 V; skimmer voltage, 65 V; fragmentor voltage, 175 V; nebulizer pressure, 35 psi; scan range, 100–1200 Da. To ensure accuracy, the mass-to-charge ratio (m/z) of all ions in the mass spectra were real-time corrected by reference ions (m/z 121.050873 (protonated purine) and 922.009798 (protonated hexakis, (1H,1H,3H-tetrafluoropropoxy)phosphazine (HP-921)) for positive mode, m/z 112.985587 (proton-abstracted ammonium trifluoroacetate (TFA) anion) and 1033.988109 (TFA adduct of HP-0921) for negative mode). The acquisition and analysis of data were controlled by Mass Hunter B.04.00 software.

Each of the samples was analyzed by LC-TOF/MS in both positive and negative ionization mode to obtain metabolite profiles and the analysis order of all test samples was randomized.¹⁹ In order to condition the LC-MS system, the QCP sample was injected five times before initiating the run. The QCP sample was re-injected once at the beginning, every four sample injections, and at the end of the run (total of nine injections) to provide a set of data to monitor instrument stability and evaluate reproducibility.

2.5 Data analysis

2.5.1 Data extraction. All the LC-TOF/MS raw files (.d) were converted to common data format (.mzData) files using Mass Hunter B.04.00 software, and subsequently the converted files were uploaded to the interactive XCMS Online platform (https://xcmsonline.scripps.edu) for peak detection, nonlinear alignment in the time domain, automatic integration, metabolite feature annotation, extraction of the peak intensities.^20,21 Data from positive and negative ionization modes were included in two separate data sets in order to analyze them individually.²² Method parameters for pre-processing samples in interactive XCMS Online platform were shown in ESI Text S1.†

2.5.2 Multivariate data processing and visualization. The generated three-dimensional data matrix, comprising of sample code, retention time (t_R)–m/z pairs and ion intensity was exported into SIMCA-P software (13.0 demo version, Umetrics, Sweden) for MVDA after normalization of the intensity of each ion. Principal components analysis (PCA) and orthogonal projection to latent structures discriminant analysis (OPLS-DA) were employed to examine the three-dimensional data in a multivariate setting.²³ In order to reduce the impact of artifacts and noise in the models, all variables were Pareto-scaled prior to model fitting. OPLS-DA was carried out on data to find most influential discriminant features (t_R–m/z pairs) separating sample groups in PCA. The OPLS-DA model was validated through the quality parameters (R²X, Q²) and analysis of variance of cross-validated predictive residuals (CV-ANOVA) (p < 0.01).^24,25

The S-plots generated by OPLS-DA were used to visualize the relative importance of the different sources of variations and obtain a list of candidates. Variables at VIP (variable importance in the projection) >1 were considered. Features at VIP > 1 with a large jack-knifed confidence interval were excluded¹⁹ and the other values were further subjected to unpaired parametric t-test (Welch’s t-test, assuming unequal variance) to determine the significance of each feature. Only t_R–m/z pairs with VIP > 1 and p < 0.05 were selected as potential chemical markers.

2.5.3 Marker identification. In order to elucidate the structures of the potential chemical markers, accurate m/z were first matched to metabolites from online databases (METLIN,²⁰ HMDB,²⁶ and LIPID MAPS²⁷) indicating tentative assignment. After an assessment of retention time and isotopic pattern, in-source fragmentation (LC-TOF/MS) and targeted MS/MS (LC-QqQ/MS) was employed to support ion fragments for structural elucidation. Metabolite structural elucidation according to tandem MS fragmentation pattern was dependent on the ability to obtain acceptable chromatographic resolution and peak intensity.⁸

3. Results and discussion

3.1 Technical reproducibility of the analytical approach

In order to reduce ion suppression induced by co-eluting metabolites of the same class, different mobile phase compositions were screened to obtain LC chromatograms with better peak shape and separation. It was found that 0.1% acetic acid–methanol was the most suitable eluting solvent system. The analytical quality in metabolomics study is of utmost importance as a guarantor for reliable results and accordingly for the success of the entire project. To ensure the robustness of analytical instrumentation and the reproducibility of data, the periodic analysis of QCP samples in the same batch was adopted as a quality assurance strategy in metabolic profiling. The QCP sample is applied for two reasons. The first reason is that the metabolic and sample matrix compositions of QCP samples are closest to the biological samples under study. The second reason is to provide data assessing technical reproducibility of the analytical approach.⁴

Using the 9 QCP samples interspersed throughout the run, both univariate statistics and MVDA were employed to test the analytical reproducibility in terms of detected ion intensities: (1) by calculating the RSD% of detected ion intensities in the QCP samples as a univariate way²⁸ and (2) using PCA as a MVDA method to assess reproducibility. After data extraction, the number of features in positive and negative mode was 537 and 674 for the QCP sample, respectively. After excluding features with 80%-rule²⁹ and %RSD higher than 60%, more than 75% of QCP sample features in positive mode and 71% of QCP sample features in negative mode with %RSD less than 30% were subjected to MVDA (Fig. S1†). The PCA demonstrated a tight grouping of the QCP samples and provided pattern distinction for normal rat plasma, normal rat serum, AHS rat plasma and AHS rat serum samples in positive (Fig. 1A) as well as negative ionization modes (Fig. 1C), implying their high analytical reproducibility across the run and confirming that the t_R–m/z pairs observed between the samples were nonsystematic, but biologically related. To further inspect data quality and run order effect on the test samples and QCP samples in positive (Fig. 1A) and negative-ion modes (Fig. 1C), corresponding Hotelling’s T2Range plot (denotes 95 and 99% significance limits, values larger than the yellow limit are suspect (0.05 level), and values larger than the red limit (0.01 level) can be considered serious outliers) was used to catch outliers in data and observe effects of run order in Fig. 1B and D. It is explicit that there was no outlier and no time-related effect on the detected ions. The high reproducibility of the data demonstrated that the quality assurance procedures can deliver the reliability required by an untargeted metabolomic study.

Fig. 1 (A) PCA score plot of all analyzed samples in positive-ion mode with the statistical parameters (R²X = 0.601, Q² = 0.768), and (B) corresponding Hotelling’s T2Range plot showing sample type to detect any trends in QCP samples and test samples. (C) PCA score plot of all analyzed samples in negative-ion mode with the statistical parameters (R²X = 0.722, Q² = 0.567), and (D) corresponding Hotelling’s T2Range plot. Normal rat plasma (green dots), normal rat serum (blue squares), AHS rat plasma (brown triangles) and AHS rat serum (yellow inverted triangles) samples as well as QCP samples (turquoise diamonds) are shown.

3.2 Differences in metabolite profiles between serum and plasma samples

The differences between the two specimens have been studied by Roche Modular P analyzer and Vitros 950 analyzer on a biochemical level³⁰ and have been analyzed for proteomics or metabolomics applications by LC-MS,³¹ NMR³² and GC-MS.³³ To better investigate and visualize the origin of specimen type differences, both Welch’s t-test and MVDA (PCA and OPLS-DA) were established as described below. After data extraction and quality assurance procedures, supervised OPLS-DA was employed to differentiate the t_R–m/z pairs between the heparin plasma and serum samples. In the score plots of OPLS-DA, it was found that heparin plasma samples (green dots) and serum samples (blue squares) cluster into two discrete groups (Fig. 2A and C), implying that potential discriminant t_R–m/z pairs contributed to the good separation. The discriminant t_R–m/z pairs were visualized and filtered by the S-plot and VIP value from OPLS-DA, respectively. As seen from Fig. 2B and D, the blue 4-point star graph is the S-plot, where the further the distance of the t_R–m/z pairs from the origin point, the higher the confidence level of the t_R–m/z pairs contributed to the clustering observed in the score plots of OPLS-DA. Furthermore, the t_R–m/z pairs with VIP > 1.0 selected from S-plot were further subjected to Welch’s t-test to determine the significance of each variable. Finally, 16 and 15 interesting t_R–m/z pairs (in each respective ion mode) were extracted and their tentative assignments in the online available databases are shown in Table S1.† The metabolites included 7 peptides, 8 lysophosphatidylcholines (lysoPCs), 3 phosphatidylcholines (PCs), 2 heterocyclic compounds, 7 polyunsaturated fatty acids (PUFAs) and their oxidative fatty acids (oxFAs). Among these, phe–phe, lysoPC(18 : 1), lysoPC(16 : 0) and eicosapentaenoic acid were found in both positive- and negative-ion modes.

Fig. 2 Score plots of OPLS-DA in positive-ion mode (A) with the statistical parameters (R²X = 0.645, R²Y = 0.998, Q² = 0.977, p[CV-ANOVA] = 7.3 × 10⁻⁷) and in negative-ion mode (C) with the statistical parameters (R²X = 0.723, R²Y = 0.991, Q² = 0.961, p[CV-ANOVA] = 7.9 × 10⁻⁶) between the serum (blue squares) and heparin plasma (green dots). The corresponding S-plot from OPLS-DA model of the two groups in positive-ion mode (B) and in negative-ion mode (D).

Serum was characterized by the presence of some dominating singly and doubly charged peptides. From the trend plots of S-plot, the t_R–m/z pairs of peptides that only existed in the serum constitute the major difference between serum and heparin plasma samples. Taking the t_R–m/z 9.77–120.0811 (a singly charged peptide) and t_R–m/z 31.21–453.1721 (a doubly charged peptide) in positive-ion dataset for example, the interesting t_R–m/z pairs were extracted easily at the top of the S-plot (Fig. 3A and D). Then, ±20 ppm narrow mass window extracted ion chromatograms (EICs) (Fig. 3B and E) and corresponding in-source fragmentations (Fig. 3C and F) were used to further confirm. The characteristic peptides which are only present in serum and not detectable in plasma are believed to emerge from peptides derived from cellular components or the clot, and the release of platelet-derived peptides during the coagulation process. This ex vivo generation of multiple peptides have been speculated to potentially affect proteome profiling³¹ and could also lead to a potential misinterpretation of serum metabolomics data.

Fig. 3 Intensity trend plots, EICs and corresponding in-source fragmentations of a singly charged peptide (t_R–m/z 9.77–120.0811, fragmentation of phenylalanine–phenylalanine) (A–C) and a doubly charged peptide (t_R–m/z 31.21–453.1721) (D–F) in positive-ion mode (X-axis represents serum (blue squares) and heparin plasma (green dots) samples and Y-axis represents response in (A) and (D)).

8 lysoPCs(lysoPC(15 : 0), lysoPC(16 : 0), lysoPC(16 : 1), lysoPC(17 : 0), lysoPC(18 : 0), lysoPC(18 : 1), lysoPC(18 : 2), lysoPC(20 : 4)), 3 PCs (PC(34 : 2), PC(36 : 4), PC(40 : 5)) and 5 PUFAs and their oxFAs (eicosapentaenoic acid, arachidonic acid, 12-hydroxyheptadecatrienoic acid, 12-hydroxyeicosapentaenoic acid and 5,6-dihydroxyeicosatrienoic acid) were markedly higher in serum than in heparin plasma. PCs are a class of phosphoglycerolipids (PLs) that incorporate choline as a headgroup. They are the major components of biomembranes. LysoPCs, as a major bioactive phospholipid component of oxidized low-density lipoprotein, are formed by hydrolysis of sn-2 fatty acyl bond of phospholipids by phospholipase A2 (PLA2).³⁴ The specimen-associated differences among the levels of 8 lysoPCs, 3 PCs, 5 free PUFAs and their oxFAs were also related to the clotting process in vitro. The coagulation process results in thrombin-stimulated platelets releasing PLs, lyso-PLs and PUFAs from biomembrane by activated PLA2 via inositol 1,4,5-phosphate signaling.¹⁶ Whereas two oxFAs metabolites (8-hydroxyeicosatetraenoic acid, 5,15-dihydroxyeicosatetraenoic acid) were significantly lower in serum than in heparin plasma. Degradation of the two lipid metabolites in serum may occur during the coagulation process. Another difference between serum and heparin plasma was related to two heterocyclic compounds (hypoxanthine and its metabolite xanthine). The higher level of hypoxanthine and xanthine in serum had been attributed to presence of platelets and erythrocytes, which could continue to release these oxypurines after the blood specimen was drawn.³⁵ Therefore, the levels of these metabolites associated with clotting cascade reaction in serum are unable to reflect real biological processes in vivo.

3.3 Differences in metabolite profiles between normal and AHS rat samples

After data extraction and quality assurance procedures, MVDA was applied to generate a comparison of normal rat serum and AHS rat serum samples, normal rat plasma and AHS rat plasma samples. Score plots of OPLS-DA and the corresponding S-plot between the normal rat serum and AHS rat serum samples are shown in Fig. S2,† the normal rat plasma and AHS rat plasma samples are shown in Fig. S3.† Based on VIP >1.0 of OPLS-DA and p < 0.05 of Welch’s t-test, 35 differential metabolites between the normal rat serum and AHS rat serum, 34 differential metabolites between the normal rat plasma and AHS rat plasma were identified (Table 1). The typical total ion chromatograms of normal rat plasma and AHS rat plasma in positive-ion mode (A, B) and in negative-ion mode (C, D) are shown in Fig. S4.† Metabolite identification was conducted with accurate mass and fragmentation behavior, as well as online database analyses. To illustrate the identification of metabolites, we took the t_R–m/z 50.63–524.3713 in positive ion dataset and t_R–m/z 55.71–363.2583 in negative ion dataset for example (Fig. 4). In positive ion ESI-MS, two ions at m/z 524.3713 and 546.3529 were extracted at 50.63 min (Fig. 4A). We inferred that the quasi-molecular ions were 524.3713 [M + H]⁺ and 546.3529 [M + Na]⁺, respectively. The molecular formula of the metabolite was established as C₂₆H₅₄NO₇P. According to the HMDB database and mass fragments supplemented by MS/MS scan (Fig. 4B), the differential metabolite was tentatively assigned as lysoPC(18 : 0). In negative ion ESI-MS, two ions at m/z 363.2553 [M + CH₃COO]⁻ and 339.2029 [M + Cl]⁻ were found at 55.71 min (Fig. 4C) and molecular formula of the metabolite was established as C₂₀H₃₂O₂. The fragment ions at 259.24 [M − H − CO₂]⁻, 205.20 [M − H − CO₂ − C₄H₆]⁻ and 163.14 [M − H − CO₂ − C₄H₆ − C₃H₆]⁻ were produced using target ESI-MS/MS spectra (Fig. 4D). According to the accurate mass and its fragment information, the metabolite was tentatively identified as arachidonic acid.

Table 1 OPLS-DA markers (t_R–m/z) of normal and AHS rat plasma samples in positive- and negative-ion mode^a

Features (tR–m/z)	Adduct	Formula	MW/Da	Diff./ppm	Metabolite	Related pathway	VIP	Intensity response
a Abbreviations: MW, monoisotopic weight; VIP, variable importance on projection; LysoPC, lysophosphatidylcholine; PG, prostaglandin; Phe, phenylalanine; GCDCA, glycochenodeoxycholic acid; PC, phosphatidylcholine; GDCA, glycodeoxycholic acid. *: response compared to control group samples.
Positive mode
50.63–524.3713	[M + H]^+	C26H55NO7P	524.3716	−0.6	LysoPC(18 : 0)	Phospholipid metabolism	2.01	↓*
3.21–138.0541	[M + Na]^+	C5H9NO2Na	138.0531	7.2	L-Proline	Amino acids metabolism	1.94	↑*
55.62–381.2635	[M + H]^+	C22H37O5	381.2641	−1.6	20-Ethyl-PGE2	Fatty acid metabolism	1.88	↓*
6.39–165.0560	[M + H]^+	C9H9O3	165.0552	4.8	Phenylpyruvic acid	Amino acids metabolism	1.75	↑*
45.18–522.3556	[M + H]^+	C26H53NO7P	522.3560	−0.8	LysoPC(18 : 1)	Phospholipid metabolism	1.73	↓*
49.99–351.2521	[M + H]^+	C21H35O4	351.2535	−4.0	PGA1 methyl ester	Fatty acid metabolism	1.62	↑*
50.57–546.3534	[M + H]^+	C28H53NO7P	546.3560	−4.8	LysoPC(20 : 3)	Phospholipid metabolism	1.49	↑*
3.11–132.1019	[M + H]^+	C6H14NO2	132.1025	−4.5	L-Leucine	Amino acids metabolism	1.47	↑*
43.71–518.3235	[M + Na]^+	C24H50NO7PNa	518.3223	2.3	LysoPC(16 : 0)	Phospholipid metabolism	1.41	↓*
50.36–282.2789	[M + H]^+	C18H36NO	282.2797	−2.8	Oleamide	Fatty acid metabolism	1.36	↑*
54.32–306.2781	[M + Na]^+	C18H37NONa	306.2773	2.6	Stearamide	Fatty acid metabolism	1.25	↑*
41.48–542.3242	[M + Na]^+	C26H50NO7PNa	542.3223	3.5	LysoPC(18 : 2)	Phospholipid metabolism	1.22	↑*
3.46–204.0633	[M + Na]^+	C9H11NO3Na	204.0637	−2.0	L-Tyrosine	Amino acids metabolism	1.25	↑*
12.74–188.0720	[M + H]^+	C11H10NO2	188.0712	4.3	Indoleacrylic acid	Amino acids metabolism	1.18	↑*
33.21–299.2577	[M + H]^+	C18H35O3	299.2586	−3.0	3-Keto stearic acid	Fatty acid metabolism	1.12	↑*
41.16–590.3211	[M + Na]^+	C30H50NO7PNa	590.3223	−2.0	LysoPC(22 : 6)	Phospholipid metabolism	1.09	↑*
52.82–414.2992	[M − 2H2O + H]^+	C26H40NO3	414.3008	−3.9	GCDCA	Cholesterol metabolism	1.06	↓*
38.20–468.3079	[M + H]^+	C22H47NO7P	468.3090	−2.3	LysoPC(14 : 0)	Phospholipid metabolism	1.04	↓*
3.02–118.0861	[M + H]^+	C5H12NO2	118.0868	−5.9	L-Valine	Amino acids metabolism	1.02	↑*
41.17–566.3817	[M + H]^+	C28H57NO8P	566.3822	−0.9	PC(20 : 0)	Phospholipid metabolism	1.01	↓*
Negative mode
55.71–363.2553	[M + CH3COO]^−	C22H35O4	363.2535	5.0	Arachidonic acid	Fatty acid metabolism	1.97	↑*
3.21–114.0551	[M − H]^−	C5H8NO2	114.0555	−3.5	L-Proline	Amino acids metabolism	1.94	↑*
38.97–378.2416	[M − H]^−	C18H37NO5P	378.2409	1.9	Sphingosine-1-phosphate	Sphingolipid metabolism	1.87	↓*
20.10–832.5835	[M − H]^−	C48H83NO8P	832.5856	−2.5	PC(40 : 6)	Phospholipid metabolism	1.42	↑*
50.74–280.2649	[M − H]^−	C18H34NO	280.2640	3.2	Oleamide	Fatty acid metabolism	1.41	↑*
32.84–335.2238	[M − H]^−	C20H31O4	335.2222	4.8	Leukotriene B4	Fatty acid metabolism	1.35	↓*
50.37–560.3336	[M − H]^−	C28H51NO8P	560.3352	−2.9	PC(20 : 2)	Phospholipid metabolism	1.31	↑*
50.41–508.3417	[M − H]^−	C25H51NO7P	508.3403	2.8	LysoPC(17 : 0)	Phospholipid metabolism	1.28	↓*
44.95–526.2759	[M + CH3COO]^−	C23H45NO10P	526.2781	−4.2	PC(13 : 0)	Phospholipid metabolism	1.27	↓*
27.81–381.2628	[M − H]^−	C22H37O5	381.2641	−3.4	20-Ethyl PGF2α	Fatty acid metabolism	1.23	↓*
12.58–203.0831	[M − H]^−	C11H11N2O2	203.0821	4.9	Tryptophan	Amino acids metabolism	1.19	↑*
23.45–754.5403	[M − H]^−	C42H77NO8P	754.5387	2.1	PC(34 : 3)	Phospholipid metabolism	1.13	↑*
55.77–282.2793	[M − H]^−	C18H36NO	282.2797	−1.4	Stearamide	Fatty acid metabolism	1.10	↑*
48.22–254.2475	[M − H]^−	C16H32NO	254.2484	−3.5	Palmitic amide	Fatty acid metabolism	1.07	↑*
34.54–391.2851	[M − C2H4NO]^−	C24H39O4	391.2848	0.8	GDCA	Cholesterol metabolism	1.06	↓*
43.55–480.3094	[M − H]^−	C23H47NO7P	480.3090	0.3	LysoPC(15 : 0)	Phospholipid metabolism	1.04	↓*
44.95–556.3043	[M − H]^−	C28H47NO8P	556.3039	0.7	PC(20 : 4)	Phospholipid metabolism	1.03	↑*

---

Fig. 4 (A) EICs for lysoPC(18 : 0) ionized in positive ion mode with two different adducts: [M + H]⁺ and [M + Na]⁺. (B) MS/MS spectra of lysoPC(18 : 0) and its fragmentation behavior. (C) EICs for arachidonic acid ionized in negative ion mode with two different adducts: [M + CH₃COO]⁻ and [M + Cl]⁻. (D) MS/MS spectra of arachidonic acid and its fragmentation behavior.

The metabolites profiling altered by AHS was found to be similar between serum and plasma. However, we observed that the increased phe–phe levels in AHS rat serum were therefore unlikely to reflect AHS-associated biological processes in vivo. In the plasma untargeted metabolomics study, the metabolites altered by AHS mainly involved in phospholipid metabolism (increased levels of PC(13 : 0), PC(20 : 0), PC(20 : 2), PC(20 : 4), PC(40 : 6); decreased levels of lysoPC(14 : 0), lysoPC(15 : 0), lysoPC(16 : 0), lysoPC(17 : 0), lysoPC(18 : 0), lysoPC(18 : 1) and increased levels of lysoPC(18 : 2), lysoPC(20 : 3), lysoPC(22 : 6)), amino acids metabolism (increased levels of L-proline, tryptophan, L-valine, L-tyrosine, L-leucine, indoleacrylic acid, phenylpyruvic acid and phenylalanine), fatty acid metabolism (increased levels of 3-keto stearic acid, oleamide, stearamide, palmitic amide, arachidonic acid and decreased levels of prostaglandin A1 (PGA1) methyl ester, 20-ethyl PGE2, 20-ethyl PGF2α, leukotriene B4), cholesterol metabolism (decreased levels of GDCA, GCDCA) and sphingolipid metabolism (decreased levels of sphingosine-1-phosphate).

3.4 Analysis of potential metabolite biomarkers of AHS

3.4.1 Phospholipid metabolism. Phosphoglycerolipid metabolites (PCs and lysoPCs) were significantly altered in the plasma of AHS rats compared to the normal rats. It has been reported that heavy alcohol consumption can stimulate PCs synthesis,³⁶ and inhibit the activity of lecithin cholesterol acyltransferase and enzymes of the lipolytic transformation of lipoproteins.³⁷ The increase of PCs implied that alcohol led to alterations in the ingredients and structures of biomembrane in AHS rats. Polyunsaturated lysoPCs were also increased in AHS rat plasma, whereas saturated or monounsaturated lysoPCs were decreased. The activity of PLA2 is increased in alcohol-injured liver.³⁸ This may explain the increase of polyunsaturated lysoPCs. The decrease of saturated or monounsaturated lysoPCs may attributed to enhanced conversion of these lysoPCs to lysophosphatidic acids by lysophospholipase D.³⁹

3.4.2 Amino acids metabolism. Oxidative stress plays an important role in the mechanisms of AHS. Chronic alcohol intake can increase the production of reactive oxygen species, reduce cellular antioxidant levels, and aggravate oxidative stress in many tissues, especially the liver, leading to the cell membrane permeability change.⁴⁰ The elevated levels of L-proline, L-valine, L-tyrosine, L-leucine, tryptophan and its metabolite (indoleacrylic acid), phenylalanine and its metabolite (phenylpyruvic acid) suggested that alcohol altered the membrane permeability and made amino acids and their metabolites release into plasma from damaged liver cells. Among these, L-proline can be catalyzed into hydroxyproline by proline hydroxylase to synthesize collagen, which can stimulate liver injury progression.⁴¹

3.4.3 Fatty acid metabolism. The increased levels of 3-keto stearic acid, oleamide, stearamide, palmitic amide in AHS rat plasma suggested that alcohol consumption could lead to the inhibition of fatty acid oxidation and enhancement of lipogenesis. Arachidonic acid is a polyunsaturated omega-6 fatty acid that presents in the membrane phospholipids of mammalian cells. It is released from PCs and phosphatidylethanolamine by PLA2, or from phosphatidylinositol by phospholipase C (PLC), or from diacylglycerol (DAG) by DAG lipase.⁴² The increased level of arachidonic acid and the decreased of its metabolites such as PGA1 methyl ester, 20-ethyl PGE2, leukotriene B4, 20-ethyl PGF2α in AHS rat plasma implied that alcohol consumption may enhance the activity of PLA2, PLC or DAG lipase, inhibit metabolism and transformation of arachidonic acid.

3.4.4 Cholesterol metabolism and sphingolipid metabolism. Cholesterol is an important raw material for synthesizing bile acids of GDCA and GCDCA in vivo. The decreased levels of GDCA and GCDCA in AHS rat plasma indicated that alcohol inhibited cholesterol metabolism. Cholesterol metabolism disorder can cause the increase of blood cholesterol, which is an important risk factor for the formation of fatty liver disease.⁴³ Sphingosine-1-phosphate is a pleiotropic lysophospho-lipid mediator which is converted primarily from sphingosine by sphingosine kinase and regulates a diverse range of cellular processes, including cell proliferation, differentiation and motility. Sphingosine-1-phosphate and its receptors are also involved in liver regeneration and oxidative injury.^44,45 The decreased levels of sphingosine-1-phosphate indicated that alcohol consumption inhibited sphingosine kinase and aggravated hepatic oxidative injury.

4. Conclusions

In the present work, a reliable untargeted metabolomic method based on LC-TOF/MS and LC-MS/MS was proposed and applied to rapidly analyze metabolic alterations linked to incident AHS and interrogated serum and plasma in order to identify overall differences in the metabolites levels, investigated which of the two matrices was more suitable for the discovery of endogenous biomarkers. As a result, metabolite level differences between serum and plasma samples in normal rats were mainly related to 8 LysoPCs, 3 PCs, 2 heterocyclic compounds, 7 PUFAs and their oxFAs. Notably, 7 characteristic peptides were observed only in serum. All the altered metabolites were associated with clotting cascade reaction, indicating that serum conceals the changes in metabolome that have occurred in vitro, thus obscuring variations between specimens. The results demonstrated that, compared with serum, heparin plasma which represented the original properties of the blood sample was a more suitable matrix to explore potential chemical markers in vivo. In the plasma untargeted metabolomics study, the metabolic alterations linked to AHS were mainly involved in phospholipid metabolism, amino acids metabolism, fatty acid metabolism, cholesterol metabolism and sphingolipid metabolism. These findings provide a systematic view of metabolic alterations linked to AHS, demonstrating the untargeted metabonomic method was a robust method for examining the molecular mechanisms of disease.

Acknowledgements

We would like to express our gratitude to National Natural Science Foundation of China (No. 81274063), Technology Innovation Program for post-graduate of Jiangsu province (No. KYLX_0619) and a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions. We also wish to thank Kaiwen Luo, Juanjuan Wang for assistance with the experiments.

References

1. G. J. Patti, O. Yanes and G. Siuzdak, Nat. Rev. Mol. Cell Biol., 2012, 13, 263–269.
2. N. Zamboni, A. Saghatelian and G. J. Patti, Mol. Cell, 2015, 58, 699–706.
3. K. Contrepois, L. H. Jiang and M. Snyder, Mol. Cell. Proteomics, 2015, 14, 1684–1695.
4. W. B. Dunn, D. Broadhurst, P. Begley, E. Zelena, S. Francis-McIntyre, N. Anderson, M. Brown, J. D. Knowles, A. Halsall, J. N. Haselden, A. W. Nicholls, I. D. Wilson, D. B. Kell, R. Goodacre and The Human Serum Metabolome (HUSERMET) Consortium, Nat. Protoc., 2011, 6, 1060–1083.
5. O. Beckonert, H. C. Keun, T. M. D. Ebbels, J. G. Bundy, E. Holmes, J. C. Lindon and J. K. Nicholson, Nat. Protoc., 2007, 2, 2692–2703.
6. L. An, Q. S. Shi and F. Feng, RSC Adv., 2015, 5, 84048–84055.
7. E. Holmes, A. Wijeyesekera, S. D. Taylor-Robinson and J. K. Nicholson, Nat. Rev. Gastroenterol. Hepatol., 2015, 12, 458–471.
8. Z. J. Zhu, A. W. Schultz, J. H. Wang, C. H. Johnson, S. M. Yannone, G. J. Patti and G. Siuzdak, Nat. Protoc., 2013, 8, 451–460.
9. M. Raro, M. Ibáñez, R. Gil, A. Fabregat, E. Tudela, K. Deventer, R. Ventura, J. Segura, J. Marcos, A. Kotronoulas, J. Joglar, M. Farré, S. Yang, Y. Xing, P. Van Eenoo, E. Pitarch, F. Hernández, J. V. Sancho and Ó. J. Pozo, Anal. Chem., 2015, 87, 8373–8380.
10. X. Y. Yu, J. Luo, L. J. Chen, C. X. Zhang, R. T. Zhang, Q. Hu, S. L. Qiao and L. Li, RSC Adv., 2015, 5, 69800–69812.
11. G. Szabo, Gastroenterology, 2015, 148, 30–36.
12. H. Fernando, K. K. Bhopale, S. Kondraganti, B. S. Kaphalia and G. A. S. Ansari, Toxicol. Appl. Pharmacol., 2011, 255, 127–137.
13. S. Manna, M. Thompson and F. Gonzalez, in Biological Basis of Alcohol-Induced Cancer, ed. V. Vasiliou, S. Zakhari, H. K. Seitz and J. B. Hoek, Springer International Publishing, 2015, vol. 815, ch. 13, pp. 217–238.
14. M. Jaremek, Z. Yu, M. Mangino, K. Mittelstrass, C. Prehn, P. Singmann, T. Xu, N. Dahmen, K. M. Weinberger, K. Suhre, A. Peters, A. Doring, H. Hauner, J. Adamski, T. Illig, T. D. Spector and R. Wang-Sattler, Transl. Psychiatry, 2013, 3, e276.
15. H. J. Issaq, Z. Xiao and T. D. Veenstra, Chem. Rev., 2007, 107, 3601–3620.
16. M. Ishikawa, K. Maekawa, K. Saito, Y. Senoo, M. Urata, M. Murayama, Y. Tajima, Y. Kumagai and Y. Saito, PLoS One, 2014, 9, e91806.
17. D. C. Wedge, J. W. Allwood, W. Dunn, A. A. Vaughan, K. Simpson, M. Brown, L. Priest, F. H. Blackhall, A. D. Whetton, C. Dive and R. Goodacre, Anal. Chem., 2011, 83, 6689–6697.
18. W. Zhong, W. L. Zhang, Q. Li, G. X. Xie, Q. Sun, X. H. Sun, X. B. Tan, X. G. Sun, W. Jia and Z. X. Zhou, J. Hepatol., 2015, 62, 1375–1381.
19. T. Barri and L. O. Dragsted, Anal. Chim. Acta, 2013, 768, 118–128.
20. R. Tautenhahn, K. Cho, W. Uritboonthai, Z. J. Zhu, G.J. Patti and G. Siuzdak, Nat. Biotechnol., 2012, 30, 826–828.
21. H. Gowda, J. Ivanisevic, C. H. Johnson, M. E. Kurczy, H. P. Benton, D. Rinehart, T. Nguyen, J. Ray, J. Kuehl, B. Arevalo, P. D. Westenskow, J. H. Wang, A. P. Arkin, A. M. Deutschbauer, G. J. Patti and G. Siuzdak, Anal. Chem., 2014, 86, 6931–6939.
22. M. Garcia-Aloy, R. Llorach, M. Urpi-Sarda, O. Jáuregui, D. Corella, M. Ruiz-Canela, J. Salas-Salvadó, M. Fitó, E. Ros, R. Estruch and C. Andres-Lacueva, Mol. Nutr. Food Res., 2015, 59, 212–220.
23. P. A. Vorkas, G. Isaac, M. A. Anwar, A. H. Davies, E. J. Want, J. K. Nicholson and E. Holmes, Anal. Chem., 2015, 87, 4184–4193.
24. A. M. Wheelock and C. E. Wheelock, Mol. BioSyst., 2013, 9, 2589–2596.
25. H. Wu, X. Li, X. Yan, L. An, K. Luo, M. Shao, Y. Jiang, R. Xie and F. Feng, J. Pharm. Biomed. Anal., 2015, 115, 315–322.
26. D. S. Wishart, T. Jewison, A. C. Guo, M. Wilson, C. Knox, Y. F. Liu, Y. Djoumbou, R. Mandal, F. Aziat, E. Dong, S. Bouatra, I. Sinelnikov, D. Arndt, J. G. Xia, P. Liu, F. Yallou, T. Bjorndahl, R. Perez-Pineiro, R. Eisner, F. Allen, V. Neveu, R. Greiner and A. Scalbert, Nucleic Acids Res., 2013, 41, D801–D807.
27. E. Fahy, M. Sud, D. Cotter and S. Subramaniam, Nucleic Acids Res., 2007, 35, W606–W612.
28. R. Dallmann, A. U. Viola, L. Tarokh, C. Cajochen and S. A. Brown, Proc. Natl. Acad. Sci. U. S. A., 2012, 109, 2625–2629.
29. S. Bijlsma, L. Bobeldijk, E. R. Verheij, R. Ramaker, S. Kochhar, I. A. Macdonald, B. van Ommen and A. K. Smilde, Anal. Chem., 2006, 78, 567–574.
30. R. R. Miles, R. F. Roberts, A. R. Putnam and W. L. Roberts, Clin. Chem., 2004, 50, 1704–1706.
31. H. Tammen, L. Schulte, R. Hess, C. Menzel, M. Kellmann, T. Mohring and P. Schulz-Knappe, Proteomics, 2005, 5, 3414–3422.
32. O. Teahan, S. Gamble, E. Holmes, J. Waxman, J. K. Nicholson, C. Bevan and H. C. Keun, Anal. Chem., 2006, 78, 4307–4318.
33. L. S. Liu, J. Y. Aa, G. J. Wang, B. Yan, Y. Zhang, X. W. Wang, C. Y. Zhao, B. Cao, J. A. Shi, M. J. Li, T. A. Zheng, Y. T. Zheng, G. Hao, F. Zhou, J. G. Sun and Z. M. Wu, Anal. Biochem., 2010, 406, 105–112.
34. J. Aoki, A. Taira, Y. Takanezawa, Y. Kishi, K. Hama, T. Kishimoto, K. Mizuno, K. Saku, R. Taguchi and H. Arai, J. Biol. Chem., 2002, 277, 48737–48744.
35. W. E. Wung and S. B. Howell, Clin. Chem., 1980, 26, 1704–1708.
36. C. S. Lieber, S. J. Robins, J. J. Li, L. M. Decarli, K. M. Mak, J. M. Fasulo and M. A. Leo, Gastroenterology, 1994, 106, 152–159.
37. A. A. Chirkin, N. Y. Konevalova, I. N. Grebennikov, V. A. Kulikov, Y. V. Saraev, V. U. Buko, I. A. Chirkina, E. O. Danchenko and K. J. Gundermann, Addict. Biol., 1998, 3, 65–70.
38. J. Chang, J. H. Musser and H. McGregor, Biochem. Pharmacol., 1987, 36, 2429–2436.
39. S. Pyne, K. C. Kong and P. I. Darroch, Semin. Cell Dev. Biol., 2004, 15, 491–501.
40. A. I. Cederbaum, Y. K. Lu and D. F. Wu, Arch. Toxicol., 2009, 83, 519–548.
41. C. S. Lieber, J. Hepatol., 2000, 32, 113–128.
42. X. Tang, E. M. Edwards, B. B. Holmes, J. R. Falck and W. B. Campbell, Am. J. Physiol.: Heart Circ. Physiol., 2005, 290, H37–H45.
43. H. K. Min, A. Kapoor, M. Fuchs, F. Mirshahi, H. P. Zhou, J. Maher, J. Kellum, R. Warnick, M. J. Contos and A. J. Sanyal, Cell Metab., 2012, 15, 665–674.
44. Y. Liu, S. Y. Saiyan, T. Y. Men, H. Y. Gao, C. Wen, Y. Liu, X. Zhou, C. T. Wu, L. S. Wang and C. P. Cui, J. Pathol., 2013, 230, 365–376.
45. H. Ikeda, N. Watanabe, I. Ishii, T. Shimosawa, Y. Kume, T. Tomiya, Y. Inoue, T. Nishikawa, N. Ohtomo, Y. Tanoue, S. Iitsuka, R. Fujita, M. Omata, J. Chun and Y. Yatomi, J. Lipid Res., 2008, 50, 556–564.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra27910k

---