Multi-dimensional, comprehensive sample extraction combined with LC-GC/MS analysis for complex biological samples: application in the metabolomics study of acute pancreatitis

Abstract

An analytical strategy employing multi-dimensional sample extraction together with optimal analytical platforms was established. The proposed workflow was based on the fact that most biological samples are complex in terms of molecular species as well as their properties, e.g. hydrophobicity. A few new methods based on methyl tert-butyl ether (MTBE) extraction, which was developed for lipid extraction and used perform a lipidomic study on the upper layer fraction of the MTBE extraction system and a metabolomics analysis by mixing the upper and lower fractions from the same sample, have been developed to acquire more information from samples, and in the present work we established a similar method employing the MTBE procedure to separate samples into polar and apolar fractions, then associated each with an adapted subsequent analysis technique. The polar fraction was silylanized for GC/MS analysis, and apolar lipids were analyzed by reverse phase liquid chromatography (RPLC) on a T3 column. The proposed strategy was applied to investigate acute human pancreatitis and was compared with results obtained by conventional LC/MS-only and GC/MS-only analysis. Features obtained in our LC-GC/MS data outnumbered either the LC/MS-only or GC/MS-only data. Furthermore, with the aid of multivariate analysis, differential metabolites from pancreatitis vs. normal pancreas, mainly amino acids and phospholipids, were identified. This work demonstrates that the proposed analytical strategy is a promising tool to perform metabolomics. In addition, it improved our understanding of the pathogenesis of acute pancreatitis and provided potential biomarkers for AP diagnosis.

---

Introduction

Researchers seek to obtain unbiased and comprehensive information,¹ especially in “omic” studies, to gain insights into biological systems. This highlights the importance of data acquisition procedures in daily scientific work, particularly sample preparation and analysis.

In metabolomics, polar metabolites^2–4 are as important as lipids⁵ in elucidating the physiological status of organisms. Although many well-accepted standard operational procedures of metabolomics have been published,^6–9 few of them meet the demands of comprehensive information extraction. Taking urine and serum studies as an example, samples are generally diluted or deproteinized before being directly analyzed by RPLC or hydrophilic interaction liquid chromatography (HILIC), or silylanized for GC analysis. However, though most of the metabolites in these samples are polar compounds, there are apolar ones. Therefore, neither dilution/deprotein pretreatment nor any single chromatography technique, namely RPLC, HILIC or GC, is sufficient to accomplish a comprehensive analysis. To improve the chromatographic performance, multidimensional chromatography, in online or offline mode, is being developed. For example, Chen et al.¹⁰ used RPLC and HILIC to analyse the urine of people with liver cancer, and t’Kindt et al.¹¹ employed both GC-MS and LC-MS to get comprehensive information in the field of plant metabolomics. As for the issue of sample preparation, a few variations of the method developed by Matyash et al.¹² (referred as the MTBE method in this paper) have been adopted as Patterson et al.¹³ said in a recent paper that the MTBE method is “the best choice for an untargeted biphasic extraction for metabolomics and lipidomics in blood plasma”. Whiley et al.^14,15 reported an IVDE method which allows one to prepare a fluid sample by a one-step extraction in vials and facilitates direct injection from the same vial! Both upper layer lipids and lower layer polar metabolites were injected for subsequent chromatographic analysis from the same vial with a proper adjustment of the needle height (higher for lipids and lower for polar compounds). This method avoided the tedious work involved in conventional sample preparation while retaining more information about the sample. And recently, Chen et al.¹⁶ reported a sample preparation method for both metabolomic and lipidomic studies based on the same MTBE method.¹² The great advantage of the strategy was that it allowed greater coverage of metabolites without increasing the amount of sample required. However, in these reports both fractions, hydrophobic and hydrophilic, were analyzed by RPLC, which usually fails to provide optimal separation of hydrophilic compounds. An elegant way to improve this issue is to associate the sample preparation with the subsequent analysis techniques. Villaseñor et al.¹⁷ proposed a single-phase extraction method for multiplatform analysis and applied it to breast milk profiling using both GC/MS and LC/MS. Be cause they specialize in hydrophilic and hydrophobic compounds, respectively, GC/MS and LC/MS can offer complementary information to each other. However, as the same sample was analyzed twice by different techniques, information redundancy was inevitable.

To get comprehensive but concise information from a sample, we employed the MTBE method for multi-dimensional sample extraction and analyzed the hydrophobic and hydrophilic fractions by different platforms. As GC is a more suitable method for the analysis of polar metabolites such as amino acids¹⁸ and RPLC is dedicated to hydrophobic analytes, in the present work, enriched polar and non-polar metabolites from the same samples were analyzed separately by GC/MS and LC/MS. This strategy was used to explore metabolic differences in the plasma of people with or without acute pancreatitis (AP), a rapidly developing inflammatory disease of the pancreas that can be lethal.

Currently, the etiology of AP is not clear¹⁹ and the diagnosis relies primarily on the presence of certain clinical symptoms such as abdominal pain or on standard laboratory testing of serum lipase activity.²⁰ A more sensitive and specific diagnosis of the disease is needed. A few metabolomic studies employing NMR^21–26 or MS with^27–29 or without³⁰ chromatographic analysis were carried out to address pancreas related disease. Li et al.²⁵ performed ¹H NMR-based metabolomics using serum from AP and healthy rats to identify potential biomarkers of AP, with the goal of identifying diagnostic metabolites for AP. Sakai et al.³¹ investigated the metabolic differences in serum and pancreatic tissue from healthy, cerulein- and L-arginine-induced AP mice by GC/MS. Results from these studies were promising; however, they were based on animal models, and their relevance to human AP should be approached with caution. Lusczek et al.²² presented their metabolomics results of human urine at the 42nd Annual Meeting of the American Pancreatic Association in the year 2006, and demonstrated that there were metabolomic differences between healthy control, acute pancreatitis and chronic pancreatitis patients. In this paper, we attempt to illustrate the advantages of the new proposed method, and present information that could be useful in the diagnosis of AP and in our understanding of its pathophysiology.

Materials and methods

Chemicals and reagents

HPLC-grade acetonitrile (ACN), methanol (MeOH) and methyl tert-butyl ether (MTBE) were purchased from Merck (Darmstadt, Germany) and Tedia (Fairfield, OH, USA), respectively. Methoxyamine hydrochloride, N-methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA), analytical grade formic acid and authentic compounds used for metabolite identification or as internal standard were purchased from Sigma-Aldrich (St. Louis, MO, USA). Ultrapure water of 18.2 MΩ cm was prepared using the Milli-Q system (Millipore, Bedford, MA).

Sample collection and preparation

Twenty-three subjects, including 13 acute pancreatitis patients and 10 healthy controls, were recruited for this study. All individuals were informed of the study procedure and the clinical protocols were approved by the Local Ethics Committee for Human Studies.

Plasma (stored at −80 °C before use) was thawed at 4 °C for sample pretreatment. The workflow for sample preparation employed in this study is shown in Fig. 1. Briefly, all samples were extracted by three different methods, and the resulting fractions were then analyzed using different platforms. Two conventional pretreatments were used as controls for our proposed analytical strategy, namely deproteinization of plasma (Pre_A) and silylation after deproteinization (Pre_B) as shown in Fig. 1. The third method was a two-phase MTBE extraction, which has been widely accepted for lipid extraction, but in the present work, the lower layer was also collected to retrieve information on polar compounds (Pre_C). Different extracts were analyzed by different methods: LC/MS-only analysis for Pre_A, GC/MS-only analysis for Pre_B and LC-GC/MS analysis, which consisted of both LC/MS (MTBE-LC/MS) and GC/MS (MTBE-GC/MS), for Pre_C, respectively.

Fig. 1 Workflow for sample preparation and analysis. (A) Plasma samples were analyzed by the LC/MS-only method after protein precipitation by an organic solvent; (B) molecules in plasma samples were analyzed by the GC/MS-only method after silylation by MSTFA; (C) plasma samples were partitioned between two phases commonly used for lipid extraction and subsequently analyzed by a specific method, the upper layer of lipids was analyzed by LC/MS (MTBE-LC/MS) and the polar fraction was silylated for GC/MS analysis.

Pre_A procedure. Small metabolites were extracted by a method described elsewhere³² with minor modifications. Briefly, a 40 μL aliquot of each thawed plasma sample was transferred to 1.5 mL microfuge tubes, mixed with 160 μL ACN to remove proteins. After centrifugation at 14 000 rpm for 15 min at 4 °C, 160 μL of the supernatant was pipetted out and lyophilized by a Speedvac Concentrator (Thermo Fisher Scientific Inc., MA, USA). The residue was stored at −20 °C for no more than 24 h and was reconstituted in 80 μL of 80% MeOH in H₂O³³ to acquire most of the metabolites in the extract before analysis.

Pre_B procedure. A 40 μL aliquot of thawed plasma sample was mixed with 160 μL ACN, centrifuged at 14 000 rpm for 15 min at 4 °C to remove the protein. Then 160 μL of the supernatant was pipetted into a new tube which contained 10 μL ribitol (120 ppm in water). The mixture was then lyophilized as previously described. The residue was stored at −20 °C for no more than 24 h or at −80 °C for longer preservation. For derivatization, the residue was resuspended in 30 μL methoxyamine hydrochloride (20 mg mL⁻¹ in anhydrous pyridine) and incubated at 37 °C for 90 min, then 30 μL MSTFA was added and samples were held at 37 °C for another 60 min. After centrifugation, the supernatant was transferred to a 1.5 mL vial for GC/MS analysis.

Pre_C procedure. Plasma samples were extracted by the MTBE method.¹² Briefly, 40 μL serum was mixed with 300 μL MeOH, then 1000 μL MTBE was added, samples were vortexed and incubated at room temperature with continued shaking for 1 h. Finally, 250 μL H₂O was added to introduce a two-phase separation. 10 min later, the mixture was centrifuged at 10 000 rpm for 10 min. 500 μL of the upper layer that contained almost all of the apolar metabolites was collected and lyophilized and resuspended in 50 μL CHCl₃/MeOH (2/1) which was diluted 5 fold by solvent MP (5 mM AcONH₄ in IPA/ACN/H₂O, 9/4/2) for LC/MS analysis. 300 μL of lower layer which held most of the polar components in the sample was transferred to a new tube, mixed with 10 μL ribitol (120 ppm in water), lyophilized and derivatized for GC/MS analysis as in section “Pre_B Procedure”. The workflow is shown as in Fig. 1C.

Chromatographic profiling of plasma metabolites

LC/MS analysis. Thermo Scientific™ Dionex™ UltiMate™ 3000 RSLC systems, including a HPG-3400RS binary pump, a WPS-3000 RS autosampler and a TCC-3000RS column oven, was coupled online with the Q Exactive™ Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific Inc., MA, USA) to acquire the LC/MS data. The Q-Exactive mass spectrometer was calibrated daily using a commercial calibration mixture (caffeine, n-butylamine, MRFA, sodium dodecyl sulfate, sodium taurocholate and Ultramark™ 1621). A heated electrospray ionization (HESI-II) source was operated in both positive and negative mode, with the following parameters: ionization voltage, 3.5 (−2.8) kV; sheath and auxiliary gas, 35 and 10, both were nitrogen and in arbitrary units; probe heat temperature, 250 °C; capillary temperature, 325 °C. Full scan resolution was set as 70 000 FWHM, with an m/z scan range of 80–1000. The AGC target (maximum target capacity of the C-trap) was 1 × 10⁶ ions and the maximum injection time was 200 ms. The LC-MS was controlled by Xcalibur 2.2 SP1.48 (Thermo Fisher Scientific Inc., MA, USA).

For LC/MS-only analysis, chromatographic separation was carried out on a BEH C18 column (100 mm × 2.1 mm, 1.7 μm) (Waters, Milford, MA) which was operated at 0.35 mL min⁻¹ at 35 °C. The mobile phase consisted of solvent A (0.1% formic acid in water) and solvent B (ACN). Gradient elution started at 10% B, held for 2 min then increased linearly to 40% B at 4 min and to 100% B at 22 min; 100% B was kept for another 2 min and returned back to the initial 10% B in 1 min. The column was equilibrated for 3 min before the next injection. The injection volume was 5 μL.

For MTBE-LC/MS analysis, separation of lipids was performed on a HSS T3 column (100 mm × 2.1 mm, 1.7 μm) (Waters, Milford, MA). The mobile phases were 10 mM AcONH₄ in 60/40 ACN/H₂O (A) and 10 mM AcONH₄ in 90/10 IPA/ACN (B). The column was held at 40 °C with a flow rate of 0.25 mL min⁻¹. To get optimal separation, the elution was started at 50% B, holding for 1.5 min and then increased to 97% B by 18 min. 3 min later, the gradient was reduced to 50% B in 0.5 min and equilibrated for 3.5 min. The injection volume was 5 μL.

GC/MS analysis. All the GC/MS analyses (both GC/MS-only and MTBE-GC/MS) were performed by GCMS-QP2010 Ultra (Shimadzu Co., Tokyo, Shimadzu) on an Rtx-5MS column (0.25 mm × 30 m, 0.25 μm) (Restek International, Bellefonte, PA, USA) using the same temperature program. The oven was initially kept at 70 °C, ramped to 230 °C at a rate of 5 °C min⁻¹ and to 320 °C at 90 °C min⁻¹. The column was kept at 320 °C for 5 min. Helium was used as carrier gas with a constant column flow of 1.19 mL min⁻¹. The temperatures of injection, ion source and interface were 240, 220 and 300 °C, respectively. Full MS scan (m/z 33–600) was started at 5.5 min with a split ratio of 10 : 1 and an injection volume of 1 μL.

A quality control (QC) sample was prepared by mixing together equal aliquots of each sample and was processed exactly as the samples were. All samples were analyzed randomly and QC was injected once about every 10 samples to monitor the stability of the analytical system as well as the reproducibility of sample preparation.³⁴

Data processing, statistical analysis and compound identification

Pretreatment. For LC/MS data, raw data were imported into the SIEVE™ 2.2 software (Thermo Fisher Scientific, MA, USA) for peak detection and alignment. The signal detection algorithm was component extraction. The intensity threshold was set as 5 000 000, minimum peak width was 9 and the signal-to-noise was 10, as recommended for Q-Exactive by the vendor. A peak list containing information of m/z, retention time and ion intensities of every sample was exported as a new data set. This was the prototype matrix that would be subjected to multivariate analysis for further pattern recognition. To remove difference inherent to sample variability, variables (m/z intensity) in the same sample were normalized to a total intensity of one unit.³⁵

For GC/MS data, raw data were converted to netCDF files by GCMSsolution (Ver. 2.72, Shimadzu Co., Tokyo, Japan) and were submitted to online XCMS (https://xcmsonline.scripps.edu/) for feature extraction using default parameters for retention time correction, peak matching and missing values filling. The GC/MS-only analysis gave a dataset containing 2419 features and the MTBE-GC/MS analysis from LC-GC/MS analysis gave 2151 features. These ion features were normalized to the ion of m/z 319 which originated from internal standard ribitol.

Statistical analysis. Variables with missing values more than 2 in the AP group and 2 in the healthy control were removed according to the “80% rule”.³⁶ Subsequently, Student’s t-test analysis (SPSS 20.0, Chicago, IL, USA) was applied to exclude non-significant variables (p > 0.05) in the two groups. Finally, ions with high analytical variation (RSD > 30% in QC samples)^34,37 were deleted. Multivariate pattern recognition analyses were conducted using the SIMCA-P+ software (Ver. 11, Umetrics, Umeå, Sweden). Data were Pareto scaled for partial least squares discriminant analysis (PLS-DA).

Identification. For data obtained by the LC/MS-only analysis, structural assignment of metabolites was accomplished based on data acquired from a high resolution MS and MS/MS scan. Some of the identified metabolites were validated by authentic standards; others were determined by comparison with spectra presented in the literature. For identification of compounds from lipid analysis, identification was mostly based on the interpretation of fragment ions from MS and MS/MS scans. For GC/MS identification, ions were first searched against the NIST and Fiehn libraries, followed by authentic standard validation of those with a commercial standard available.

Results and discussion

Sample pretreatment and data acquisition of plasma samples

In analytical chemistry, sample pretreatment is of primordial importance as it reduces sample complexity, concentrates target components and mitigates the effect of the matrix on the analysis. To this end, various extraction or purification methods, such as liquid–liquid extraction or solid phase (micro-) extraction (SPE or SPME), were developed to retain target compounds or remove unwanted ones or the matrix. These techniques are well accepted in daily analytical work.

However, almost all the extraction methods mentioned above are specified to one class of compounds. When it comes to complex biological samples, information loss is inevitable, which is especially unsuitable for global analysis. Analysts have applied different sample processing methods to obtain a more thorough understanding of a sample. Despite the detailed information given by the multi-processing strategy, sample consumption as well as information redundancy are obstacles to its popularity. In addition, the extraction process is usually both time-consuming and costly. Chen et al.¹⁶ employed the MTBE method for serum extraction and combined the upper and lower layers in a global metabolomic study. There is currently no single chromatographic technique which is sufficient for the analysis of both polar and apolar compounds. As such, when Chen et al. employed reverse-phase LC/MS for analysis, optimal separation of polar compounds was sacrificed.

Sample pretreatment. In the present work, a two-phase extraction method adapted from Matyash et al.¹² was used for sample preparation, and both the upper layer lipid fraction and the lower layer polar fraction were collected to keep as much of the sample information as possible. The upper layer was used for LC/MS analysis of plasma lipids and the lower polar ones was silylanized and applied to GC/MS. In addition, the conventional deprotein or silylation methods were also used for sample preparation, and after data acquisition, these results were compared with each other.

Data acquisition. For LC/MS analyses, data were acquired in both positive (Fig. 2A and C) and negative (Fig. 2B and D) modes. LC/MS-only analysis was performed on a BEH C18 column, and the analysis of the lipid fraction (MTBE-LC/MS) was carried out on a slightly hydrophilic T3 column to reduce the retention of hydrophobic compounds on the C18 column. With respect to the quantity of sample, the chromatogram in Fig. 2C was in fact a part of the chromatogram in Fig. 2A. Yet, the separation in Fig. 2C was comparable, if not better, than Fig. 2A, indicating that, when treated using the Pre_A method, most of the lipids were lost. Differential lipids (or lipid fragments) in the healthy control and AP individuals are listed in Table 1.

Fig. 2 LC/MS base peak chromatogram (BPC) of different preparation: (A) +ESI BPC from LC/MS-only analysis; (B) −ESI BPC from LC/MS-only analysis; (C) +ESI BPC from MTBE-LC/MS analysis; (D) −ESI BPC from MTBE-LC/MS analysis.

Table 1 Candidate metabolite ions of acute pancreatitis from MTBE-LC/MS analysis

No.	tR	m/z	M1. VIP[2]	Identification	Fold (AP/Ctrl)
a Ion detected in negative mode.
1	1.58	518.325	2.50	LPC (18 : 3)	0.70
2	2.11	496.341	11.65	LPC (16 : 0)	0.71
3	2.97	525.375	1.27	Isotope of LPC (18 : 0)	0.73
4	2.98	524.373	4.17	LPC (18 : 0)	0.73
5	8.55	756.555	0.99	PC (34 : 3)	1.32
6	9.34	703.577	2.74	SM (18 : 1–16 : 0)	0.90
7	9.57	782.572	4.63	PC (36 : 4)	0.97
8	9.9	782.568	4.71	PC (36 : 4)	0.97
9	9.91	785.59	3.28	Isotope of PC (36 : 3)	1.24
10	9.92	784.587	6.51	PC (36 : 3)	1.24
11	9.93	786.592	0.97	PC (36 : 2)	1.26
12	10.14	785.59	3.38	Isotope of PC (36 : 3)	1.25
13	10.14	784.587	6.74	PC (36 : 3)	1.25
14	10.56	784.588	2.22	PC (36 : 3)	1.32
15	10.8	761.59	4.02	Isotope of PC (34 : 1)	1.18
16	10.8	760.588	7.83	PC (34 : 1)	1.17
17	10.92	812.618	2.68	PC (38 : 3)	1.35
18	10.92	813.62	1.55	Isotope of PC (38 : 3)	1.33
19	17.47	896.771	1.84	TAG (18 : 2–16 : 0–18 : 2)	1.19
20	18.53	878.817	4.12	TAG (18 : 0–16 : 0–18 : 1)	1.83
21	18.53	881.757	1.00	Isotope of TAG (18 : 0–16 : 0–18 : 0)	1.17
22	18.53	876.804	13.28	TAG (18 : 1–16 : 0–18 : 1)	1.36
23	18.59	902.819	4.21	TAG (18 : 1–18 : 1–18 : 1)	1.28
24	18.76	369.352	3.17	Fragment of CE	0.82
25	18.78	902.819	3.88	TAG (18 : 0–18 : 1–18 : 2)	1.28
26	19.00	904.834	2.56	TAG (18 : 0–18 : 1–18 : 1)	1.74
27	19.05	369.352	7.39	Fragment of CE (18 : 2)	0.88
28	19.05	666.619	2.78	CE (18 : 2)	0.89
29	19.63	369.352	2.00	Fragment of CE	1.00
30	2.56^a	303.234	1.56	FA (20 : 4)	0.49

---

For GC/MS analysis, both the chromatographic and MS conditions were the same and the results are shown in Fig. 3A for the GC/MS-only method and in Fig. 3B for MTBE-GC/MS analysis. Chromatographically, there were no obvious differences (Fig. 3A and B), except at the high retention time zone. Chromatographic peaks were searched against the NIST and Fiehn library, and the results are listed in Table 2. Among all the identified items, 3 were missing in MTBE-GC/MS. When subjected to XCMS for feature extraction, the GC/MS-only method gave 2419 ion fragmentations, and the LC-GC/MS method gave 2151. This is what one might expect as in the LC-GC/MS method hydrophobic compounds were extracted from the organic layer of the MTBE system. For example, metabolites such as cholesterol (t_R 36.5 min in Fig. 3B) decreased or disappeared in the chromatogram from the LC-GC/MS method. It is a demonstration that when aiming for comprehensive information on a sample, data from the LC-GC/MS method could reduce data redundancy which is inherent to parallel LC/MS and GC/MS analyses. If we are aiming for partial information, either MTBE-LC/MS or MTBE-GC/MS analysis can reduce the matrix effect, although information loss might occur as in the present study, the LC-GC/MS method gave more candidate biomarkers (206 ions) than the GC/MS-only method (169 ions) in the multivariate analysis.

Fig. 3 GC/MS total ion chromatogram (TIC) of different preparations: (A) TIC from GC/MS-only analysis; (B) TIC from MTBE-GC/MS analysis.

Table 2 Candidate metabolite ions of acute pancreatitis from GC/MS analyses, including GC/MS-only and MTBE-GC/MS analysis^a

No.	Identification	tR	Fragments	GC/MS-only	MTBE-GC/MS	Fold AP/Ctrl (GC/MS-only)	Fold AP/Ctrl (MTBE-GC/MS)
a *: metabolites validated by authentic standard compounds, **: differential metabolites between healthy control and AP patients in either GC/MS-only analysis or MTBE-GC/MS analysis, UN: unknown, metabolites that cannot be assigned an unambiguous structure, —: metabolites not detected.
1	2-Hydroxypyridine*	5.750	152, 166, 153	✓	✓	1.01	1.08
2	Pyruvic acid*	6.067	174, 74, 89	✓**	✓**	0.31	0.25
3	Acetamide	6.217	134, 130, 77	✓	✓	1.05	1.08
4	D-(−)-Lactic acid*	6.317	117, 191, 190	✓**	✓**	0.82	0.77
5	Glycolic acid*	6.608	43, 69, 57	✓	✓	1.09	1.06
6	L-Valine 1*	6.833	72, 55, 75	✓	✓	0.89	0.99
7	UN1	6.950	71, 57, 43	✓	✓	1.03	1.12
8	L-Alanine 1*	7.225	116, 117, 45	✓	✓**	1.01	0.77
9	3-Methyl-2-ketobutyric acid	7.425	89, 267, 341	✓	✓	0.67	0.68
10	Isoserine	7.608	102, 103, 204	✓	✓	1.02	0.82
11	2-Hydroxyisobutyric acid	7.825	131, 132, 148	✓	✓	0.50	0.46
12	N.A.2 oxalic acid	7.942	148, 66, 45	✓	✓	1.04	0.97
13	3-Hydroxypyridine	7.992	152, 167, 153	✓	✓	1.00	1.08
14	Oxalic acid*	8.15	133, 59, 220	✓	✓**	1.03	1.11
15	L-Leucine 1*	8.425	86, 44, 75	✓	✓	0.78	0.79
16	(R)-3-Hydroxybutyric acid	8.592	117, 191, 148	✓**	✓	0.23	0.19
17	α-Hydroxyisovaleric acid	8.733	145, 75, 146	✓	✓	0.53	0.5
18	DL-Isoleucine 1*	8.942	86, 69, 87	✓	✓	0.84	0.86
19	Hydroxylamine	9.275	132, 73, 89	✓**	✓	0.82	0.81
20	L-Valine 2*	10.017	144, 218, 145	✓**	✓**	0.86	0.72
21	Urea*	10.719	189, 73, 148	✓**	✓**	1.31	1.28
22	L-Serine 1*	11.050	116, 132, 57	✓	✓	0.77	0.85
23	L-Leucine 2*	11.450	158, 159, 102	✓	✓	0.73	0.68
24	Phosphate*	11.592	299, 147, 205	✓	✓**	0.87	0.72
25	L-Isoleucine*	12.025	158, 73, 218	✓	✓	0.85	0.83
26	L-Proline 2*	12.058	142, 143, 144	✓	✓	1.15	1.15
27	Glycine*	12.308	174, 86, 248	✓**	✓**	0.88	0.89
28	Glyceric acid*	13.058	189, 292, 103	✓	✓	1.24	1.26
29	Nonanoic acid	13.533	117, 75, 215	✓	✓	1.01	1.18
30	UN3	13.643		✓	✓	1.16	1.38
31	UN4	13.677		✓	✓	0.63	0.70
32	L-Serine 2*	13.800	204, 73, 218	✓	✓	0.86	0.76
33	Glycylglycine*	14.492	218, 219, 117	✓	✓	0.96	0.91
34	L-Aspartic acid 1*	15.242	160, 130, 116	✓	✓	0.86	0.83
35	β-Alanine*	15.343	248, 174, 86	✓**	✓**	12.35	11.96
36	2,2′-Bipyridine	15.585	156, 155, 128	✓	✓	1.00	1.08
37	UN5	16.125	174, 248, 86	✓	✓	1.01	0.96
38	Aminomalonic acid	16.525	218, 320, 174	✓	✓	0.97	0.99
39	D-Malic acid	17.000	233, 84, 57	✓	✓	0.65	0.55
40	L-Threitol*	17.617	217, 103, 205	✓	✓	1.33	1.50
41	L-Proline*	17.692	156, 157, 258	✓	✓**	0.90	0.80
42	Phenylalanine*	18.135	120, 146, 130	✓	✓	0.76	0.83
43	L-Threonic acid	18.483	115, 292, 143	✓	✓	1.00	1.05
44	L-Threonic acid	18.900	292, 220, 205	✓	✓	0.35	0.37
45	Hexadecane	19.208	57, 71, 43	✓	✓	0.98	1.14
46	Glutamic acid*	20.025	246, 128, 247	✓	✓	0.70	0.60
47	L-Phenylalanine	20.100	218, 192, 100	✓**	✓	0.69	0.66
48	D-(−)-Lyxose	21.650	103, 217, 307	✓**	✓**	2.94	3.05
49	3-Isoquinoline acetic acid, perhydro-, methyl ester	22.475	138, 139, 57	✓	✓	1.02	1.07
51	Ribitol*	22.651		✓	✓	1.00	1.00
52	DL-Ornithine	22.833	174, 186, 142	✓	✓	0.79	0.88
53	L-(+)-Threose	23.158	205, 73, 147	✓	✓	1.37	1.48
54	L-Glutamine*	23.383	156, 155, 245	✓	✓	1.17	1.29
55	9H-Purine	24.000	265, 73, 280	✓	✓**	1.56	1.33
56	L-Ornithine 2*	24.317	142, 174, 143	✓**	✓**	0.78	0.76
57	Citric acid*	24.517	273, 347	✓	✓**	1.17	1.34
58	Tetradecanoic acid*	24.633	117, 73, 285	✓	✓	1.21	1.18
59	Cadaverine	24.9	174, 73, 175	✓	✓	1.04	0.94
60	1,5-Anhydro-D-sorbitol	25.158	217, 218, 191	✓**	✓**	1.22	1.25
61	Tyrosine 1*	25.558	179, 73, 219	✓	✓	1.03	1.15
62	D-Fructose*	25.775	217, 103, 307	✓	✓	0.79	0.82
63	UN6	26.058	319, 205, 160	✓	✓	1.07	1.31
64	Sugar 1	26.108	204, 191, 192	✓	✓	4.37	1.75
65	D-Glucose 1*	26.317	319, 205, 160	✓**	✓**	1.3	1.39
66	L-Lysine 2*	26.392	156, 174, 73	✓	✓	0.93	1.00
67	D-Glucose 2*	26.642	319, 205, 147	✓**	✓**	1.27	1.43
68	L-Tyrosine*	26.750	218, 219, 100	✓	✓	0.90	1.01
69	D-Sorbitol*	26.883	319, 205, 217	✓	✓	1.90	2.79
70	Sugar 2	27.358	217, 361, 103	✓	✓	1.23	1.40
71	Sugar 3	27.892	217, 204, 147	✓	✓**	1.61	1.72
72	cis-9-Hexadecenoic acid*	28.092	73, 117, 75	✓	✓	1.24	1.17
73	D-Gluconic acid 2*	28.283	333, 292, 217	✓	✓	1.72	1.67
74	Hexadecanoic acid*	28.467	117, 313, 73	✓	✓**	1.03	1.07
75	α-D-Mannopyranoside	28.850	204, 73, 319	✓	✓**	1.48	1.47
76	1H-Indole-3-propanoic acid	29.358	202, 73, 333	✓	✓	0.86	0.63
77	Myo-inositol*	29.858	305, 217, 318	✓	✓**	1.18	1.27
78	Uric acid 1*	30.058	441, 456, 382	✓	✓	1.14	1.15
79	9,12-Octadecadienoic acid (Z,Z)-*	31.450	73, 75, 67	✓	✓	0.90	0.75
80	trans-9-Octadecenoic acid*	31.533	117, 73, 75	✓	✓	1.01	0.98
81	11-trans-Octadecenoic acid*	31.650	117, 75, 129	✓	✓	1.07	0.97
82	L-Tryptophan 2*	31.908	202, 73, 203	✓	✓	0.84	0.84
83	Octadecanoic acid	31.975	117, 341, 73	✓	✓	1.03	1.04
84	N,N-Dimethyldodecanamide	32.167	87, 100, 45	✓	✓	0.97	1.01
85	Ritalinic acid ditms	32.242	156, 73, 157	✓	✓	0.98	1.10
86	Octanoic acid, 2-dimethylaminoethyl ester	32.425	58, 71, 72	✓	—	1.19	—
87	Arachidonic acid*	32.983	73, 79, 217	✓	✓	1.19	1.39
88	Linolenic acid*	33.092	73, 75, 67	✓	✓	1.03	1.28
89	cis-4,7,10,13,16,19-Docosahexaenoic acid*	33.783	79, 91, 73	✓	—	1.02	—
90	Inosine*	33.925	217, 230, 245	✓**	✓**	4.09	5.15
91	α-Tocopherol*	36.108	502, 237, 73	✓	—	1.17	—
92	Cholesterol*	36.342	129, 329, 368	✓	✓	1.08	0.92

---

Metabolomic results

To get insights into the differences between plasma samples from healthy controls and individuals with acute pancreatitis (Ctrl and AP), as well as to compare information obtained by the LC-GC/MS analysis and LC/MS- or GC/MS-only analysis, multivariate analyses of data from different sample fractions or by different analytical methods are performed. A data matrix consisting of information on retention time, fragment intensity and fragment (m/z), and dataset derived from MTBE-LC/MS analysis was used to exemplify the multivariate analysis. First, an orthogonal signal corrected partial least square discriminant analysis (OSC PLS-DA) model was conducted. As we can tell from Fig. 4A, in the score plot of the PLS-DA model, AP and Ctrl samples were well separated from each other. The model was validated by a permutation test run 999 times (Fig. 4B). The two main parameters of the model, the cumulative explained variance of the model (R²) and the precision of prediction by the model (Q²), were 0.95 and 0.85 for the current PLS-DA model, and the intercept of R² and Q² were 0.347 and −0.349, respectively. The intercepts indicated the PLS-DA model was not over-fitted; therefore, further information exploration on the model can be carried out. All variables are displayed in Fig. 4D, and the top 100 variables that contributed to the separation of AP and Ctrl are marked by a red square. To narrow down this differential list of candidate variables, an S-plot (Fig. 4C) was graphed to view the importance of these variables in pattern recognition. Variables located near the center of the plot have little contribution in discriminant analysis, and were excluded, as well as those with a large jack-knifing confidence (Fig. 4E).

Fig. 4 Multivariate statistical analysis results of the MTBE-LC/MS analysis from the LC-GC/MS method. (A) OSC PLS-DA score plot; (B) cross validation plot of the model with a R² and Q² of 0.95 and 0.85 and the corresponding intercept of 0.347 and −0.349, respectively; (C) S-plot, top 100 ions on the VIP list are marked with red square; (D) OSC PLS-DA loading plot, top 100 ions on the VIP list are marked with red square; (E) jack-knifing confidence of selected differential variables.

MTBE-GC/MS analysis and LC/MS- and GC/MS-only analysis data were processed as mentioned and results of LC/MS-only analysis are shown in the ESI (Fig. S1†). Differential metabolites from LC/MS-only analysis, which also is a conventional approach for metabolomics studies, could be classified to categories: the polar compounds represented by amino acid and its derivates and apolar lipids. The newly proposed strategy in the present work, the LC-GC/MS method, gave more detailed information on each class of compounds. When comparing MTBE-GC/MS data with the polar fraction of LC/MS-only data, more amino acids as well as carbohydrates and other compounds were obtained, and the same conclusion can be draw for comparison of MTBE-LC/MS lipids with that obtained using LC/MS-only method, as shown in Tables 1, 2 and S1.† And as presented in Table 2 and Fig. 3, the MTBE-GC/MS method is almost as good as the GC/MS-only method. In Fig. 3, the chromatograms were zoomed to the same scale to unfold any difference between the two methods and they were nearly the same except that a few peaks at the end of the temperature program were faded out or disappeared. These peaks were mainly more hydrophobic compounds such as cholesterol, which were more suitable for LC/MS analysis. And as shown in Table 1, the MTBE LC/MS method did profile a few cholesterol derived compounds while the LC/MS-only method failed to do so. Thus, the proposed LC-GC/MS strategy outperformed either the LC/MS- or GC/MS-only method in information retrieving and might provide more comprehensive results for metabolomics studies.

Potential biologically important metabolites in acute pancreatitis

Metabolomic investigations concerned with the dysfunction of pancreas have not been performed as often as with other organ-related diseases.³⁸ These studies covered three aspects: pancreatic cancer, chronic and acute pancreatitis, and both MS^28,30,31 and NMR^24,25,39 techniques were applied. From the pathophysiologic point of view, pancreatic failure, either in situ or transferred, can trigger hypercatabolism, increase demands of energy and/or promote protein catabolism.⁴⁰ This is coincident with our results as candidate metabolites in the present study are mainly amino acids, carbohydrates, small organic acids, lipids and other polar molecules which are involved in multiple metabolic pathways such as amino acid metabolism, glucose metabolism and lipid metabolism. Generally, most of the amino acids were decreased in AP, and so were some energy related compounds such as pyruvic acid, lactic acid, 3-hydroxybutyric acid; while carbohydrates including lyxose, 1,5-anhydro-sorbitol, glucose, α-mannopyranoside as well as the metabolic intermediates oxalic acid and citric acid were increased under the AP condition. The result gave us a glimpse of the potential metabolic state of AP and might shed some light on the pathogenesis of AP at a molecular level; from these data, information on early diagnosis of AP could be derived. In the following part of this paper the potential biological functionality in AP of the metabolites that are worth further investigation are discussed.

Amino acid metabolism and related metabolites. Plasma/serum amino acids have long been of concern in pancreas related disease^41,42 and our work revealed a few amino acids which were disturbed by acute pancreatitis and most of them were decreased in the AP group. The down regulated amino acids are alanine, valine, glycine, proline, phenylalanine, ornithine and the up regulated one is β-alanine. Adrych et al.⁴³ found some essential and aromatic amino acids were decreased in the serum in people with chronic pancreatitis and they attributed the decrease to decreased exocrine function. This fact may hold true in the case of acute pancreatitis and it is possible that in the present work, certain metabolite such as valine and proline are decreased under disease conditions as a result of dysfunction of the organism. Besides, in a recent review,⁴⁴ the author sorted out the factors of influence in AP and found that elevated serum concentration of alanine aminotransferase which might lead to the decrease of alanine “may be suggestive of gallstone aetiology”. Apart from studies that were consistent with our work, there have been publications with different opinions. Li et al.²⁵ performed a serum ¹H NMR metabolomics study on rats with experimental AP and found lactate, valine, 3-hydroxybutyric acid increased in AP and glucose, glycine and phosphatidylcholine (PC) decreased. We shared the results that glycine and PC were decreased but contradicted the change of lactate, valine and glucose in AP. Considering that Li’s work was based on experimental rat AP; the contradiction might arise from a metabolism difference between species or suggest a different effect of the incidence of AP. In either case, further studies are necessary to validate the results and assumptions.

Glucose metabolism and related metabolites. In the present study, an augmented level of glucose as well as lyxose, 1,5-anhydro-sorbitol, α-mannopyranoside in the AP group was found and this is predictable as hyperglycemia is common in severe illnesses such as acute pancreatitis.⁴⁵ An augmented glucose level was also found in chronic pancreatitis while the opposite effect was seen in pancreatic cancer³⁹ as in all the other carcinogenic processes.⁴⁶ The differentiated consumption of glucose in pancreatitis and pancreatic cancer indicated that in pancreatitis either the glycolysis pathway was partially retardant or there was a retro-complementation of glucose such as gluconeogenesis.⁴⁷ As in our results, pyruvic acid and lactic acid which could be regarded as the endpoints of glycolysis were decreased in pancreatitis and so was the ketone body 3-hydroxybutyric acid, an intermediate of the gluconeogenesis pathway; it was more likely that the glycolysis in pancreatitis was reduced. Besides, as one of the intermediates of tricarboxylic acid (TCA) cycle, citric acid, was found elevated in AP group, the TCA cycle of acute pancreatitis could be enhanced and be the source of energy in acute pancreatitis.

Correlations of some of the polar compounds detected are plotted in Fig. 5 with the data presented as box plots.

Fig. 5 Pathway of some amino acid related compounds; data is depicted by box plots. The up arrow (↑) on the right of the compound name indicates the up regulation of the corresponding metabolite in AP groups and the down arrow (↓) indicates down regulation. Ctrl: healthy control group; AP: acute pancreatitis group.

Lipid metabolism and related metabolites. Another group of compounds that discriminated the AP and healthy control samples are related to lipids. They might reflect the high-energy expenditure metabolic process of AP. A series of lysophosphotidylcholine (LPC), phosphotidylcholine (PC) and triglyceride (TAG) were found to be differential metabolites between the healthy control and AP. This was predictable as the AP progressed with increased energy requirements. Besides, as phospholipids are one of the main components of cell membranes, the change of phospholipids levels in the AP group might reflect the decomposition of tissues. Idegami et al.⁴⁸ studied the changes in lung phospholipids in AP subjects, and found that both the degradation and synthesis of alveolar PC were accelerated in rats with acute pancreatitis. And the increased PC and decreased LPC in the present study may indicate that the synthesis of PC exceeded its degradation. van Minnen et al.⁴⁹ investigated the effects of phospholipids and cholesterol crystals on pancreatitis induced by bile salts, and found that higher concentrations of phosphatidylcholine physiologically mitigated pancreatitis. Therefore, the change of LPC and PC levels observed in the present AP samples may have resulted from the metabolism of PC executing a protective effect. And changes of plasma TAG concentration were feedback of the demand for PC synthesis of the body.

Conclusion

The proposed two-dimensional sample pretreatment procedure along with optimized multiplatform analysis provided a new strategy to obtain more information from samples without increasing the amount of sample required. First, for sample preparation, a conventional two-phase extraction protocol using MTBE was adopted to separate metabolites into two major classes according to their polarity. Then, the two extracts were analyzed by a chromatographic technique optimized for each, i.e. GC chromatography for polar compounds and LC for apolar (lipid) compounds. The small amount of sample required and the comprehensive but non-redundant information obtained are the two principal merits of this approach. Our protocol was employed to study the human AP metabolome. When results were compared with those of conventional protocols, more information was obtained, as the GC analysis of whole samples is similar to that of the polar extract of samples, while the MTBE LC/MS analysis of the plasma lipid gave a cohort of lipids rather than a few lipid compounds eluted during a window of 3–5 min in the LC/MS analysis of whole plasma. Differential metabolites such as amino acids and lipids may be suggested as biomarkers of AP. Though we are fully aware that the results obtained here were based on a small sample size, and it is necessary to evaluate the specificity and sensitivity of the proposed molecular markers with large samples and further elucidate their physiological functions. This method can be used to perform analogous “omic” applications as well as conventional analysis, especially where samples are not easily available or their amounts are small.

Acknowledgements

Financial support was provided by the National Natural Science Foundation of China (No. 31471128, 21505054) and Natural Science Foundation of Jiangsu Province (No. BK20150132).

References

1. O. Fiehn and T. Kind, in Metabolomics, ed. W. Weckwerth, Humana Press, 2007, vol. 358, ch. 1, pp. 3–17.
2. B. Feng, S. Wu, F. Liu, Y. Gao, F. Dong and L. Wei, BMC Gastroenterol., 2013, 13, 73.
3. H. L. Chuang, Y. T. Huang, C. C. Chiu, C. D. Liao, F. L. Hsu, C. C. Huang and C.C. Hou, J. Nutr. Biochem., 2012, 23, 752–758.
4. C. C. Huang, W. T. Lin, F. L. Hsu, P. W. Tsai and C. C. Hou, Eur. J. Appl. Physiol., 2010, 108, 557–566.
5. K. Ikeda, Y. Oike, T. Shimizu and R. Taguchi, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2009, 877, 2639–2647.
6. E. J. Want, I. D. Wilson, H. Gika, G. Theodoridis, R. S. Plumb, J. Shockcor, E. Holmes and J. K. Nicholson, Nat. Protoc., 2010, 5, 1005–1018.
7. E. C. Y. Chan, K. K. Pasikanti and J. K. Nicholson, Nat. Protoc., 2011, 6, 1483–1499.
8. H. Luan, N. Meng, P. Liu, Q. Feng, S. Lin, J. Fu, R. Davidson, X. Chen, W. Rao, F. Chen, H. Jiang, X. Xu, Z. Cai and J. Wang, J. Proteome Res., 2014, 13, 1527–1536.
9. 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 and R. Goodacre, Nat. Protoc., 2011, 6, 1060–1083.
10. J. Chen, W. Wang, S. Lv, P. Yin, X. Zhao, X. Lu, F. Zhang and G. Xu, Anal. Chim. Acta, 2009, 650, 3–9.
11. R. t’Kindt, K. Morreel, D. Deforce, W. Boerjan and J. Van Bocxlaer, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2009, 877, 3572–3580.
12. V. Matyash, G. Liebisch, T. V. Kurzchalia, A. Shevchenko and D. Schwudke, J. Lipid Res., 2008, 49, 1137–1146.
13. R. E. Patterson, A. J. Ducrocq, D. J. McDougall, T. J. Garrett and R. A. Yost, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2015, 1002, 260–266.
14. L. Whiley, A. Sen, J. Heaton, P. Proitsi, D. García-Gómez, R. Leung, N. Smith, M. Thambisetty, I. Kloszewska, P. Mecocci, H. Soininen, M. Tsolaki, B. Vellas, S. Lovestone and C. Legido-Quigley, Neurobiol. Aging, 2014, 35, 271–278.
15. L. Whiley, J. Godzien, F. J. Ruperez, C. Legido-Quigley and C. Barbas, Anal. Chem., 2012, 84, 5992–5999.
16. S. Chen, M. Hoene, J. Li, Y. Li, X. Zhao, H.-U. Häring, E. D. Schleicher, C. Weigert, G. Xu and R. Lehmann, J. Chromatogr. A, 2013, 1298, 9–16.
17. A. Villaseñor, I. Garcia-Perez, A. Garcia, J. M. Posma, M. Fernández-López, A. J. Nicholas, N. Modi, E. Holmes and C. Barbas, Anal. Chem., 2014, 86, 8245–8252.
18. J. K. Kim, T. Bamba, K. Harada, E. Fukusaki and A. Kobayashi, J. Exp. Bot., 2007, 58, 415–424.
19. S. Tenner, World J. Gastroenterol., 2014, 20, 16529–16534.
20. P. A. Banks, T. L. Bollen, C. Dervenis, H. G. Gooszen, C. D. Johnson, M. G. Sarr, G. G. Tsiotos and S. S. Vege, Gut, 2013, 62, 102–111.
21. N. J. Zyromski, A. Mathur, G. A. N. Gowda, C. Murphy, D. A. Swartz-Basile, T. E. Wade, H. A. Pitt and D. Raftery, Pancreatology, 2009, 9, 410–419.
22. E. R. Lusczek, J. A. Paulo, J. R. Saltzman, V. Kadiyala, J. Rosenblum, P. A. Banks, D. L. Conwell and G. Beilman, Pancreas, 2011, 40, 1335.
23. D. Ouyang, Hepato-Gastroenterology, 2012, 59, 2314–2317.
24. E. R. Lusczek, J. A. Paulo, J. R. Saltzman, V. Kadiyala, P. A. Banks, G. Beilman and D. L. Conwell, J. Pancreas, 2013, 14, 161–170.
25. J. Li, X.-l. Zhao, Y.-x. Liu, X.-h. Peng, S.-f. Zhu, H. Guo, Y.-l. Liu, M.-h. Wan and W.-f. Tang, BMC Gastroenterol., 2014, 14, 115.
26. G. Sun, J. Wang, J. Zhang, C. Ma, C. Shao, J. Hao, J. Zheng, X. Feng and C. Zuo, Mol. BioSyst., 2014, 10, 467–474.
27. J. J. Pesek, M. T. Matyska, J. A. Loo, S. M. Fischer and T. R. Sana, J. Sep. Sci., 2009, 32, 2200–2208.
28. T. Kobayashi, S. Nishiumi, A. Ikeda, T. Yoshie, A. Sakai, A. Matsubara, Y. Izumi, H. Tsumura, M. Tsuda, H. Nishisaki, N. Hayashi, S. Kawano, Y. Fujiwara, H. Minami, T. Takenawa, T. Azuma and M. Yoshida, Cancer Epidemiol., Biomarkers Prev., 2013, 22, 571–579.
29. Y. Zhang, L. Qiu, Y. Wang, X. Qin and Z. Li, Analyst, 2014, 139, 1697–1706.
30. A. Leichtle, U. Ceglarek, P. Weinert, C. Nakas, J. M. Nuoffer, J. Kase, T. Conrad, H. Witzigmann, J. Thiery and G. Fiedler, Metabolomics, 2013, 9, 677–687.
31. A. Sakai, S. Nishiumi, Y. Shiomi, T. Kobayashi, Y. Izumi, H. Kutsumi, T. Hayakumo, T. Azuma and M. Yoshida, Arch. Biochem. Biophys., 2012, 522, 107–120.
32. L. Zhou, L. Ding, P. Yin, X. Lu, X. Wang, J. Niu, P. Gao and G. Xu, J. Proteome Res., 2012, 11, 5433–5442.
33. M. Yuan, S. B. Breitkopf, X. Yang and J. M. Asara, Nat. Protoc., 2012, 7, 872–881.
34. F. Michopoulos, L. Lai, H. Gika, G. Theodoridis and I. Wilson, J. Proteome Res., 2009, 8, 2114–2121.
35. H. M. Parsons, D. R. Ekman, T. W. Collette and M. R. Viant, Analyst, 2009, 134, 478–485.
36. S. Bijlsma, I. Bobeldijk, E. R. Verheij, R. Ramaker, S. Kochhar, I. A. Macdonald, B. van Ommen and A. K. Smilde, Anal. Chem., 2006, 78, 567–574.
37. J. Zhou, L. Zhang, Y. Chang, X. Lu, Z. Zhu and G. Xu, J. Proteome Res., 2012, 11, 4351–4360.
38. J. Zhu, D. Djukovic, L. Deng, H. Gu, F. Himmati, E. G. Chiorean and D. Raftery, J. Proteome Res., 2014, 13, 4120–4130.
39. L. Zhang, H. Jin, X. Guo, Z. Yang, L. Zhao, S. Tang, P. Mo, K. Wu, Y. Nie, Y. Pan and D. Fan, Clin. Biochem., 2012, 45, 1064–1069.
40. M. D. Johnson, Gut, 2005, 54, iii1–iii9.
41. P. Sandstrom, L. Trulsson, T. Gasslander, T. Sundqvist, U. von Dobeln and J. Svanvik, Amino Acids, 2008, 35, 225–231.
42. N. Fukutake, M. Ueno, N. Hiraoka, K. Shimada, K. Shiraishi, N. Saruki, T. Ito, M. Yamakado, N. Ono, A. Imaizumi, S. Kikuchi, H. Yamamoto and K. Katayama, PLoS One, 2015, 10, e0132223.
43. P. Giesbertz, J. Ecker, A. Haag, B. Spanier and H. Daniel, J. Lipid Res., 2015, 56, 2029–2039.
44. S. J. F. Harper and S. Cheslyn-Curtis, Ann. Clin. Biochem., 2011, 48, 23–37.
45. Y. Zuo, Y. Kang, W. Yin, B. Wang and Y. Chen, J. Crit. Care, 2012, 27, 146–152.
46. L. W. Yuan, H. Yamashita and Y. Seto, World J. Gastroenterol., 2016, 22, 2046–2059.
47. S. O. Diaz, J. Pinto, G. Graça, I. F. Duarte, A. S. Barros, E. Galhano, C. Pita, M. d. C. Almeida, B. J. Goodfellow, I. M. Carreira and A. M. Gil, J. Proteome Res., 2011, 10, 3732–3742.
48. K. Idegami, K. Mori, A. Misumi and M. Akagi, Jpn. J. Surg., 1983, 13, 227–235.
49. L. P. van Minnen, N. G. Venneman, J. E. van Dijk, A. Verheem, H. G. Gooszen, L. M. A. Akermans and K. J. van Erpecum, Pancreas, 2006, 32, 369–375.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra26708k

---