Author Correction: Comparative metabolomics study on therapeutic mechanism of electro-acupuncture and moxibustion on rats with chronic atrophic gastritis (CAG)

Some studies have proved that both acupuncture and moxibustion are very effective for the treatment of CAG. However, little is known about therapeutic mechanism of electro-acupuncture and moxibustion on CAG as well as the difference between them. On the other hand, metabolomics is a ‘top-down’ approach to understand metabolic changes of organisms caused by disease or interventions in holistic context, which consists with the holistic thinking of electro-acupuncture and moxibustion treatment. In this study, the difference of therapeutic mechanism between electro-acupuncture and moxibustion on CAG rats was investigated by a 1H NMR-based metabolomics analysis of multiple biological samples (serum, stomach, cerebral cortex and medulla) coupled with pathological examination and molecular biological assay. For all sample types, both electro-acupuncture and moxibustion intervention showed beneficial effects by restoring many CAG-induced metabolic changes involved in membrane metabolism, energy metabolism and function of neurotransmitters. Notably, the moxibustion played an important role in CAG treatment mainly by regulating energy metabolism in serum, while main acting site of electro-acupuncture treatment was nervous system in stomach and brain. These findings are helpful to facilitate the therapeutic mechanism elucidating of electro-acupuncture and moxibustion on CAG rats. Metabolomics is promising in mechanisms study for traditional Chinese medicine (TCM).

effects by delivering electrical pulses to needles 11 . Moxibustion is conducted to stimulate acupoints by the heat from moxa burning 12 . Electro-acupuncture and moxibustion are believed to have similar clinical effects, although different in their stimulation 13 . Currently, both acupuncture and moxibustion have been shown to be effective on CAG in several studies 14,15 . And it has been reported that acupuncture can down-regulate the expression of miR-21, NF-γB p65 and miR-155 and up-regulate the expression of miR-146a in rats with CAG 16 . In our previous study, it was found that moxibustion can significantly decrease (epidermal growth factor) EGF, (transforming growth factor-α)TGF-α, (proliferating cell nuclear antigen)PCNA, (vascular endothelial growth factor)VEGF, (Ag-nucleolar organization regions)Ag-NORs in gastric mucosal of rats with CAG 17 . The differential expression of genes leads us to much better understanding of CAG as well as mechanism of acupuncture and moxibustion. But it is only part of the information we need because organisms often respond in complex and unpredictable ways to stimuli that causes disease 18 . To give a more complete picture, the metabolic differences of biological status we need to observe.
Metabolomics is a powerful system biological approach, which can simultaneously monitor and evaluate changes of global metabolic compositions involved in biochemical processes caused by changes in disease and other stimulations in a holistic context 19 . This specific feature is concordant with the holistic view of TCM 20 . NMR is one of most preferred platforms and extensively used in metabolomics studies because of non-destructive and minimal sample preparation as well as nonselective analysis 21 . Recently, the application of the 1 H NMR-based metabolomics has shown value in elucidation of the therapeutic mechanisms of acupuncture but moxibustion is less studied 22,23 . In our preliminary study, 1 H NMR-based metabolomics coupled molecular biology is used to analyze biological samples (stomach, cortex urine, serum and medulla) in rats with CAG and after electro-acupuncture treatment. And it was found that electro-acupuncture treatment could relieve CAG by regulating membrane catabolism, function of neurotransmitter in brain and gut microbiota metabolism 24,25 . In this study, the difference of therapeutic mechanism between electro-acupuncture and moxibustion on rats with CAG is investigated by 1 H NMR-based metabolomics combined with pathological evaluation and molecular biological assay.

Results
Histological Morphology Examinations. Histological examinations of gastric tissues in rats from six groups were conducted (Fig. 1). The results of pathological examination in the control group showed that epithelial cells and glands arranged in the same size. Additionally, epithelial cells were single-layered columnar with membrane integrity and cytoplasm transparent and there were also no hyperemia and edema in the gland and mucosa ( Fig. 1A and a). On the contrary, in the CAG group, reduced gland cells and thinning gastric mucosa layer were clearly seen. And cells arranged intricately with irregular nuclei and distinct nucleoli ( Fig. 1B and b). Based on these observations, the rat modeling of CAG was a successful replication. The gastric mucosas of CAG rats with electro-acupuncture or moxibustion treatment on the stomach meridian acupoints were improved in different degrees with moderate thickness of gastric mucosa and regular gland arrangement (Fig. 1C,c,E and e). These observations illustrate a significant curative effect of electro-acupuncture and moxibustion on rats with CAG. In the other hand, thinning gastric mucosa layer without obvious improvement, incomplete mucosa gland morphology and irregular cell alignment were shown in CAG rats with electro-acupuncture or moxibustion on non-acupoints (Fig. 1D,d,F and f). It suggests that electro-acupuncture and moxibustion treatment on non-acupoints has no much effect on rats with CAG.
Brain-gut Peptide (BGP) Examinations by ELISA. In this study, the levels of substance P and ghrelin in gastric mucosa of rats were determined by ELISA. And it was found that the low expression levels of the substance P and ghrelin in gastric mucosa of rats in CAG group compared with the controls. In the other hand, the levels of these two factors in gastric mucosa and intestinal mucosa of CAG rats with electro-acupuncture or moxibustion treatment on the stomach meridian acupoints were increased to the normal (Fig. 2).

Figure 1.
Histological examination of gastric mucosa from six groups. (A and a, the control rats; B and b, the CAG rats; C and c, CAG rats with electro-acupuncture treatment on the stomach meridian acupoints; D and d, CAG rats with electro-acupuncture treatment on the stomach meridian acupoints; E and e, CAG rats with moxibustion treatment on the stomach meridian acupoints; F and f, CAG rats with moxibustion on nonacupoints). Scale bars represent 2 μm for the top row and 0.5 μm for the bottom row.
1 H NMR Profiles of Serum and Tissues from Rats. The typical 1 H NMR spectra of serum and extracts from tissues (stomach, cerebral cortex and medulla) are shown in Fig. 3. Additionally, two-dimensional (2D) NMR spectra including 1 H-1 H correlation spectroscopy (COSY) and 1 H-13 C heteronuclear single quantum coherence (HSQC) for the samples were also acquired (Figs S2-9). And the main metabolites from the spectra were identified according to published references, our own developed NMR database and Human Metabolome Database (HMDB: http://www.hmdb.ca/) (in Tables S1-4).
Visual inspection of these 1 H NMR spectra exhibited less obvious difference among these groups because of the complexity of the spectra. So, all of resulting datasets were subsequently analyzed to examine the clustering of each group by OPLS-DA, which can reveal any possible variables contributing to CAG molding and electro-acupuncture or moxibustion treatment. For all sample types, a good separation between the control group and CAG group was initially observed in Fig. 4, indicating a noticeable change in metabolic profile occurred in CAG group. We also performed the corresponding S-plot and t-test for metabolites related CAG modeling (in supplementary Fig. S10), which showed some changes in levels of metabolites occurred in CAG group compared Figure 2. The expression of substance P and ghrelin in gastric mucosa of rats in six groups. (Control, control rats; CAG, chronic atrophic gastritis rats; EA, CAG rats with electro-acupuncture treatment on the stomach meridian acupoints; ENA group, CAG rats with electro-acupuncture on non-acupoints; MA group, CAG rats with moxibustion treatment on the stomach meridian acupoints; MNA group, CAG rats with moxibustion on non-acupoints). ( # Means a statistical significance p < 0.05 when compared with the control group; *means a statistical significance p < 0.05 when compared with the CAG group.)  with the controls as follows: a)the serum levels of glycogen, glucose and acetoacetate were increased, whereas the levels of methionine, lactate and betaine were decreased; b) the levels of glutathione and glutamine were up-regulated, while the level of ethanolamine was down-regulated in stomach tissue; c) the higher levels of inosine, methylmalonate and malonic acid occurred but lower levels of phenylalanine and inositol were presented in medulla tissue; d) the increased levels of hypoxanthine, nicotinamide and glycerol occurred in cerebral cortex tissue.
Using the strategy mentioned before, all of rats with electro-acupuncture or moxibustion treated presented a clear separation from CAG rats, illustrating that both the electro-acupuncture and moxibustion had a good effect on CAG (in Figs 5 and 6). And the corresponding S-plots and the following t-test suggested as follows: a) levels of glycogen, methionine, betaine, glucose, glutathione, phenylalanine and hypoxanthine were reversed by both electro-acupuncture and moxibustion treatment; b) the levels of glutamine, ethanolamine methylmalonate, malonic acid, inositol, inosine and nicotinamide were returned to normal levels after electro-acupuncture treatment while lactate, acetoacetate and glycerol were regulated by moxibustion treatment.

Correlation analysis of metabolites and BGPs. The correlation analysis on the relationships between
BCPs and potential metabolites from all sample types was shown in the supplementary Fig. S12. Red bars correspond to positive correlations and blue bars correspond to negative correlations. Bar length reflects the magnitude of the correlation coefficients. The bar plots analysis was performed to gain a possible relationship between metabolites and BGPs. In Fig. S12A, betaine, lactate, ethanolamine, phenylalanine and methionine were positively associated with substance P, while malonic acid, glutathione, nicotinamide, glycogen, glucose, acetoacetate, hypoxanthine, glutamine and inosine were negatively associated with substance P. In Fig. S12B, phenylalanine, methionine, betaine, inositol and lactate were positively associated with ghrelin, whereas inosine, malonic acid, acetoacetate, glycerol, glucose, nicotinamide, methylmalonate, hypoxanthine and glutamine were negatively associated with ghrelin.

Discussion
It is known that functional dyspepsia is closely related to brain-gut peptide (BGP), which is effective in regulating various physiological functions in gastrointestinal tract by both enteric nervous system and central nervous system 26 . Substance P is a member of the BGPs and associated with the generation and signals transmission of gastrointestinal visceral hypersensitivity 27 . Ghrelin is also one of important BGPs closely related to appetite and can control gastric motility 28 . In this study, these two factors were reversed to normal by both electro-acupuncture and moxibustion treatment, indicating that both electro-acupuncture and moxibustion treatment play an important role in CAG by regulating enteric nervous system and central nervous system.
In the other hand, for all sample types, the different levels of changes induced by CAG-modeling were reversed by electro-acupuncture or moxibustion treatment. These results may give rise to characteristic metabolomics For all sample types, a good separation between the control group and CAG group was initially observed, indicating a noticeable change in metabolic profile occurred in CAG group.
profiles of CAG modeling and the difference between electro-acupuncture and moxibustion treatment, which will be discussed in further detail below (Fig. 7).

Metabolic Alterations in Serum.
Serum is easy to obtain non-invasively and can provide a holistic view of the whole systems biology. Therefore, the metabolomics approach in serum is often used for disease diagnosis and drug therapy monitoring. In this study, therapeutic mechanism between electro-acupuncture and moxibustion on CAG rats were compared by a serum 1 H NMR-based metabolomics.
Glycogen is a highly-branched polymer of about 30,000 glucose residues and principal storage form of glucose in animal and human cells. Glucose is synthesized from intermediates such as pyruvate and glycerol by a process known as gluconeogenesis and often degrades by glycolysis. And in the process of glycolysis, metabolic energy is captured 8 . Lactate is the end-product of degradation of glucose under anaerobic conditions and can be considered for the assessment of the severity of the supply/demand imbalance 29 . In the other hand, acetoacetate is one of important intermediate for synthesis and degradation of ketone bodies, which can provide acetoacetyl-CoA and acetyl-CoA for glycolysis or gluconeogenesis 30 . Functioning as a methyl donor, betaine often carries and donates methyl functional groups to facilitate necessary chemical processes such as the synthesis of dimethylglycine 31 . Particularly, it methylates homocysteine to methionine, which is considered as the metabolic precursor for cysteine 32,33 . Additionally, cysteine is important in energy metabolism 34 . So, the levels of glucose, glycogen, betaine and methionine in CAG rats were return to normal after either electro-acupuncture or moxibustion treatment, suggesting that regulating energy metabolism is important in both electro-acupuncture and moxibustion treatment on CAG rats. Only moxibustion treatment can regulate the levels of lactate and acetoacetate CAG rats, indicating that moxibustion may play a more important role in regulating energy metabolism than electro-acupuncture treatment on CAG.

Metabolic Alterations in Extracted Stomach
Tissues. Stomach was accepted as a CAG-targeted organ. As mentioned above, status of gastric mucosa of CAG rats had a different change in both histopathology and BGPs. Herein, a 1 H NMR-based metabolomics was used to compare the therapeutic mechanism between electro-acupuncture and moxibustion on CAG rats.
Ethanolamine is a component of lecithin, which is an important part of cell membranes. Glutathione, synthesized from cysteine, glutamate and glycine, is a coenzyme in various enzymatic reactions 35 . The most important of these reactions are redox reactions, where the thiol grouping on the cysteine portion of cell membranes can protect against peroxidation 36 . In the other hand, glutathione can be converted to glutamic acid, which is a key molecule in cellular metabolism and the most abundant fast excitatory neurotransmitter in the nervous system 37 . Generally, glutamate and glutamine are inter-converted between neurons and astrocytes. In this study, the level of glutathione was reversed to normal by either electro-acupuncture or moxibustion treatment, indicating both electro-acupuncture and moxibustion treatment show beneficial effects in membrane metabolism of CAG rats. In addition, the levels of ethanolamine and glutamine could be regulate by only electro-acupuncture treatment, indicating that electro-acupuncture treatment on CAG rats may play an more important role in membrane metabolism than moxibustion treatment and can regulate neurotransmitter in nervous system.

Metabolic Alterations in Extracted Brain
Tissues. It is accepted that there is a dense neuronal network linking with stomach and central nervous system, where numerous neurotransmitters and peptides play an important role in contributing to gastric mucosal integrity 38,39 . In this study, metabolic alterations in cortex and medulla were investigated by technique of 1 H NMR.
Malonic acid is also one of vital intermediates in the tricaboxylic acid (TCA) cycle and its derivative called methylmalonic acid can lead a block in the enzymatic conversion of methylmalonyl coenzyme A (CoA) to succinyl CoA, which plays a major role in the metabolism of fat and protein 40 . Glycerol is a three-carbon substance that forms the backbone of fatty acids in fats. When the body uses stored fat as a source of energy, glycerol and fatty acids are released. And it is not surprising that glycerol can be converted to glucose and provide energy for cellular metabolism 41 . In addition, niacinamide functions as a component of the coenzyme (nicotinamide adenine Metabolites in red, green and blue represent metabolites regulated by electro-acupuncture, moxibustion and both them, respectively. Both electro-acupuncture and moxibustion intervention showed beneficial effects by restoring many CAG-induced metabolic changes involved in membrane metabolism, energy metabolism and function of neurotransmitters. Notably, the moxibustion played an important role in CAG treatment mainly by regulating energy metabolism in serum, while main acting site of electro-acupuncture treatment is nervous system in stomach and brain. (1, Synthesis and degradation of ketone bodies; 2, Phenylalanine, tyrosine and tryptophan biosynthesis; 3, Phenylalanine metabolism; 4, Glutathione metabolism; 5, Glycerolipid metabolism; 6, Nicotinate and nicotinamide metabolism; 7, Alanine, aspartate and glutamate metabolism; 8, Butanoate metabolism; 9, Cysteine and methionine metabolism).
SCienTifiC RepoRTS | 7: 14362 | DOI:10.1038/s41598-017-13195-5 dinucleotide) NAD, which is used extensively in glycolysis and the TCA cycle of cellular respiration 42 . So, being reversed to normal in levels of malonic acid, methylmalonic acid, niacinamide by electro-acupuncture treatment is another indication about both electro-acupuncture and moxibustion treatment on CAG by regulating energy metabolism. And the level of glycerol was regulated by only moxibustion treatment, suggesting moxibustion treatment may be more important in regulating energy metabolism than electro-acupuncture treatment.
Phenylalanine is an essential amino acid highly concentrated in brain and precursor for the neurotransmitter called catecholamines 43 . Adenosine, composed of adenine and d-ribose, plays a role in signal transduction as a neurotransmitter 44 . Inosine is a purine nucleoside that is an intermediate in the degradation of purines and purine nucleosides like hypoxanthine and also occurs in the anticodon of certain transfer RNA molecules. And adenosine to inosine (A-to-I) RNA editing plays a vital role in neurotransmission in brain 45 . As a second messenger in a cell, inositol plays an important role in the form of inositol phosphates, which are important in signal transduction. In current study, the levels of these four metabolites were altered to normal by electro-acupuncture treatment while moxibustion treatment could regulate only two of them, which suggests that electro-acupuncture treatment on CAG plays a more important role in regulating nervous system than moxibustion treatment.
Together, these findings suggests that both electro-acupuncture and moxibustion treatment on CAG rats are related to regulating membrane metabolism, energy metabolism and function of neurotransmitter. From the global point of view, the main acting site of moxibustion treatment on CAG rats is blood by regulating energy metabolism, whereas electro-acupuncture treatment can ameliorate CAG mainly by repairing nervous system in stomach and brain. This is consistent with previous study about the warming effect of moxibustion can promote gastrointestinal blood circulation and enhance gastrointestinal motility energy while that clinical efficacy of acupuncture relies on the mediation of the central nervous system 46,47 .
In conclusion, a 1 H NMR-based metabolomics analysis of multiple biological samples (serum, tomach, cortex and medulla) coupled with pathological examination and BGPs examination was used to investigate the difference of therapeutic mechanism between electro-acupuncture and moxibustion on CAG rats in this study. It was found that for all sample types in rats of CAG group, a noticeable change in metabolic profile occurred, involved in membrane metabolism, energy metabolism and function of neurotransmitter. In the other hand, both electro-acupuncture and moxibustion intervention showed beneficial effects by restoring many CAG-induced metabolic changes. Notably, the moxibustion played an important role in CAG treatment mainly by regulating energy metabolism in serum, while main acting site of electro-acupuncture treatment is nervous system in stomach and brain of CAG rats. Further studies on electro-acupuncture and moxibustion interventions by gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS)-based metabolomics are also required to validate our findings. The current work will be beneficial for better understanding difference of biological mechanism between electro-acupuncture and moxibustion treatment on CAG. Animals. 36 healthy male S.D. rats (170-210 g weight) used in this study were maintained in a temperature-controlled environment (22 ± 1 °C) and a 12: 12 h light/dark cycle (lights on at 8:00 a.m.) with free access to water and food.

Ethical
After adaptation for one week, all animals were randomly divided into six groups (n = 6 per group): (1) control group, (2) CAG rats, (3) CAG rats with electro-acupuncture treatment on the stomach meridian acupoints, (4) CAG rats with electro-acupuncture on non-acupoints, (5) CAG rats with moxibustion treatment on the stomach meridian acupoints and (6) CAG rats with moxibustion on non-acupoints. The rats in control group were freely accessed to water clean water and standard rat diet. According to our previous study 25 , all of rats except ones in control group were subjected to CAG modeling including compulsive sporting, irregular fasting as well as mixture of 30% alcohol and 2% sodium salicylate poured down the throats for 12 weeks.
Electro-acupuncture and moxibustion treatment. In this study, two acupoints in the stomach meridian including Liangmen (ST 21, located 5 mm horizontally to the spot above 2cm from navel) and Zusanli (ST 36, located 5 mm below the fibular head and lateral to the anterior tubercle of the tibia) were selected according to "The Veterinary Acupuncture of China" and Government Channel and Points Standard GB12346-90 of China (Fig. S13). According to previous studies 25 , for rats in the electro-acupuncture treatment group, Liangmen (ST 21) and Zusanli (ST 36) were treated with electro-acupuncture (a 30-minute session per day) for 2 weeks after CAG modeling. While for CAG rats with electro-acupuncture on non-acupoints, the corresponding non-acupoints located 5 mm away from each of the two acupoints were treated with electro-acupuncture as mentioned above (Fig. S3). And these non-acupoints have nothing to do with the stomach meridian and do not lie on any other known acupoints. In the other hand, for CAG rats with moxibustion on acupoints and non-acupoints, the above-mentioned acupoints and non-acupoints were selected respectively and subjected to moxibustion (a 30-minute session per day) for two weeks after CAG modeling.
Histopathology. The gastric mucosa samples in rats were collected and immersed in the phosphate-buffered 10% formalin. After sample dehydration, the biopsies were embedded in wax and then sectioned at 5μm. At last, a histopathological examination by light microscopy was conducted after being stained with hematoxylin and eosin.
Enzyme linked immunosorbent assay (ELISA) assessment. After two weeks of electro-acupuncture or moxibustion treatment, all rats were anesthetized and the gastric glands (1 cm ×1 cm) were collected. After being cut into pieces, each tissue sample was mixed with ice-cooled phosphate buffered saline (PBS) (0.1 M, 1:9 (w/v)). And then the mixture was transferred to 2 mL tube and centrifuged for 15 min (3000 rpm, 4°C) after being homogenated at 10 000 rpm on ice. Finally, the upper supernatants were collected in clean tubes and stored at −80 °C prior to analysis.
The levels of substance P and ghrelin were determined using an ELISA kit purchased from Cusabio (Wuhan, Hubei, China). The ELISA-based method was conducted according to protocol provided by manufacturer.
Biological sample collection and 1 H NMR experiments. After the experimental procedure, all animals were sacrificed anesthetized with isoflurane. Then the abdominal aortic blood and tissue samples were collected. The serum samples were obtained by centrifuging at 10000 rpm for 10 min at 4 °C after standing for 1h at room temperature. The cerebral cortex, medulla and stomach tissue samples were excised and snap-frozen in liquid nitrogen immediately for tissue extraction. All of samples were stored at −80 °C prior to analysis.
According to our preliminary study 25 , each serum sample of 400 μL was added to 200 μL phosphate buffer solution (90 mM K 2 HPO 4 /NaH 2 PO 4 , pH 7.4, 99.9% D 2 O) and centrifuged at 10 000 rpm for 10 min at 4 °C. And then 500 μL of supernatant was transferred into 5 mm NMR tube for NMR experiment. The 1 H NMR spectra of serum samples were obtained by a 600 MHz Bruker spectrometer at 298 K. A standard one-dimensional (1D) Carr-Purcell-Merboom-Gill (CPMG, RD-90 (τcp-180-τcp)-acquisition) with water suppression was used to captured 1 H NMR spectra. For every sample, 64 FIDs were collected into 20K data points over a spectral width of 12 000 Hz with a relaxation delay of 0.23 ms.
Each of pre-weighed stomach, cerebral cortex, medulla samples (200 mg) were homogenated in CH 3 OH (600 mL) and H 2 O (300 mL) and then vortexed for 1min. The samples were centrifuged for 10 min (10000 rpm, 4 °C) after partitioning on ice for 10 min. And then the upper supernatant from each sample were lyophilized and mixed with 600 μL D 2 O containing sodium 3-trimethylsilyl-(2, 2, 3, 3-d4)-1-propionate (TSP, 1 mM). Finally, a 500 μL of mixture was piped into 5 mm NMR tube for NMR analysis. 1 H NMR spectra of these samples were also obtained using a Bruker 600 MHz spectrometer at 298 K. Standard 1D 1 H spectra were captured with a Nuclear Overhauser Effect Spectroscopy (NOESY, RD-901-t1-90°-tm-90° -acquire) pulse sequence. For all sample, 64 FIDs were collected into 64K data points over a spectral width of 12 000 Hz with a relaxation delay of 6.5 μs.
For good metabolite identification, two-dimensional (2D) NMR spectra for the samples were also obtained by spectrometer mentioned above, including 1 H-1 H correlation spectroscopy (COSY) and 1 H-13 C heteronuclear single quantum coherence (HSQC) as our previous study 48 . The COSY spectra were acquired with a relaxation delay of 1.5s, and 48 transients were collected into 1024 data points with a spectral width of 12 ppm for both dimensions. For the 2D 1 H-13 C HSQC spectra, a 1.5 s relaxation delay was used for water suppression by presaturation. For each 2D spectrum, 2048×128 data points were collected using 8 scans per increment. The spectral widths were set to 12 and 240 ppm in the proton and carbon dimensions, respectively The metabolites in NMR spectra from all samples were assigned according to published literature 21,24,25 , our own developed NMR database and HMDB database (http://www.hm db.ca/). Data Preprocessing and multivariate statistical analysis. The all of 1 H NMR spectra from samples were phased and baseline corrected using MestReNova v9.0.1 software (Mestrelab Research, Santiago de Compostella, Spain). The 1 H NMR spectra were referenced to a double peak of CH 3 from the internal lactate at 1.33 ppm (for serum samples) or a single peak from TSP at 0.0 ppm (for tissue samples). All spectra were also peak-aligned to overcome peak-shift problem. After the chemical shift range of δ 4.70-5.2 ppm were removed to eliminate the influence of water, the spectra were segmented at δ 0.01 intervals across the region of 0.5 to 8.5 ppm(for serum samples) and 0.5-9.0 ppm (for tissue samples). To compensate for significant concentration differences between the samples, the integral values from each spectrum were normalized to a sum of all of integrals in a spectrum, and then the date matrices were formed for further multivariate analysis.
The 1 H NMR spectral data was imported into SIMCA-P14.1 (Umetrics, Sweden) for multivariate analysis. Prior to multivariate analysis, the Pareto-scaling was used to decrease the effect of noise and artefacts in the models. Firstly, a principal component analysis (PCA) was used for a natural separation among the all groups by visual inspection of the score plots in this study. And then a supervised orthogonal projection to latent structures discriminant analysis (OPLS-DA) was carried out to distinguish different groups. Additionally, parameters for model fitness (R2) and predictive ability (Q2) were used to assess quality of OPLS-DA model. And the p-value from cross-validated analysis of variance (CV-ANOVA) was also calculated to indicate the level of significance for group separation in OPLS-DA. To discover the potential variables for differentiation, the corresponding S-plot of OPLS-DA model was conducted. The potential biomarkers were extracted according to the variable importance in the project (VIP) of the established OPLS-DA model (VIP≥1.00) and an independent-sample t-test (p <0.05) using (OriginProver.8.1). In addition, a systematically statistical correlation analysis was used to investigate the relationships between these metabolites and BCPs with online data tool Metaboanalyst 3.0 (http://www. Metaboana lyst.ca).