Metabolomics Reveals Hyperlipidemic Biomarkers ...

Send Orders for Reprints to [email protected] Current Metabolomics, 2016, 4, 000-000

1

Metabolomics Reveals Hyperlipidemic Biomarkers and Antihyperlipidemic Effect of Poria cocos Hua Chen1, Lin Chen1, Dan-Dan Tang1, Dan-Qian Chen1, Hua Miao1,2, Ying-Yong Zhao1,2* and Shuang-Cheng Ma3,* 1

Key Laboratory of Resource Biology and Biotechnology in Western China, Ministry of Education, the College of Life Sciences, Northwest University, No. 229 Taibai North Road, Xi’an, Shaanxi 710069, China; 2Division of Nephrology and Hypertension, School of Medicine, University of California, Irvine, Med Sci 1, C352, UCI Campus, Irvine, CA, 92897, USA; 3National Institutes for Food and Drug Control, State Food and Drug Administration, 2 Tiantan Xili, Beijing 100050, China Abstract: Background: Hyperlipidemia has been highlighted to be one of the most prominent global health threats. Poria cocos (PC), a well-known traditional Chinese medicine, is used for treating hyperlipidemia in China. To evaluate its therapeutic function on hyperlipidemia, urinary metabolomics was performed. Method: Diet-induced hyperlipidemic rat model was produced by high fat food, and then the ethanol extract of PC (250 mg/kg) was used to treat hyperlipidemic rats for 6 weeks. Urine samples were analysed using ultra performance liquid chromatography-high definition mass spectrometry coupled with partial least squares-discriminant analysis. Box plots, fold changes, one-way analysis of variance, Mann-Whitney U-test, false discovery rate correction, heatmap display and receiver operating characteristic analysis were employed for further analysis of the identified metabolites. Additionally, visualization of metabolic pathways was conducted by ingenuity pathway analysis and Metscape. Results: Eighteen metabolites were identified including propionylcarnitine, arginine, trimethyltridecanoic acid, methylhippuric acid, aminoadipic acid, citric acid, etc. The metabolites arginine, aminoadipic acid and citric acid were screened as significant biomarkers by various statistical analysis and receiver operating characteristic curves. The results of quantitative enrichment analysis algorithm and cytoscape indicated that thirty-eight metabolic pathways were perturbed by dietinduced hyperlipidemia. The abnormal levels of these metabolites in model group indicated diet-induced hyperlipidemia mainly disturbed amino acid metabolism, tricarboxylic acid cycle, fatty acid metabolism and nucleic acid metabolism. However, PC partially ameliorated these abnormal metabolisms. Conclusion: PC positively regulated the perturbed metabolisms induced by hyperlipidemia and metabolomics was proven to be suitable for characterizing antihyperlipidemic effect of PC.

Keywords: Hyperlipidemia, Poria cocos, Metabolomics, Biomarker, Ultra-performance liquid chromatography, High definition mass spectrometry. Received: April 13, 2016

Revised: May 11, 2016

1. INTRODUCTION As an underlying risk factor for various cardiovascular diseases, hyperlipidemia is highlighted to be a significant health problem worldwide. Hyperlipidemia, also defined as dyslipidemia, is a metabolic disorder disease which is caused by the imbalance of global lipid metabolisms and characterized as increased levels of triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C) and *Address correspondence to these authors at the Key Laboratory of Resource Biology and Biotechnology in Western China, Ministry of Education, the College of Life Sciences, Northwest University, No. 229 Taibai North Road, Xi’an, Shaanxi 710069, China; Tel: +86 29 88305273; Fax: +86 29 88303572; E-mails: [email protected] or [email protected] National Institutes for Food and Drug Control, State Food and Drug Administration, 2 Tiantan Xili, Beijing 100050, China; Tel: +86 10 67095272; Fax: +86 10 67095887; E-mail: [email protected] 2213-235X/16 $58.00+.00

Accepted: May 13, 2016

very low-density lipoprotein cholesterol as well as decreased levels of high-density lipoprotein cholesterol (HDL-C). Being a common cardiovascular disease, hyperlipidemia is considered to be a serious risk for public health in China and other Asian, American and European developed countries. Even worse, the increasing number of pediatric dylipidemic patients suggests the incidence age of dyslipidemic patients gradually gets younger [1]. Lipid-lowering agents including stains and non-stain drugs are commonly used as clinical therapies for hyperlipidemia, however, most of these drugs have certain adverse effects or other limitations [2]. The antihyperlipidemic treatment with traditional Chinese medicine shows uniquely good performances due to its multi-target effect and hypotoxicity. Some diuretic Chinese medicines such as Rhizoma alismatis, rhubarb, Poria cocos (PC) and Polyporus umbellatus were validated to possess © 2016 Bentham Science Publishers

2 Current Metabolomics, 2016, Vol. 4, No. 2

good effects on improvement of dyslipidemia [3-6]. Among them, PC, the dried sclerotium of the fungus Poria cocos (schw.) wolf (Polyporaceae), is a well-known and popular medicinal fungus possessing antihyperlipidemic, antiinflammatory as well as hypoglycemic effects [6-9]. Moreover, our previous study proved the surface layer of PC had a good therapeutic effect on hyperlipidemia [10]. With a low requirement for the surviving environment, PC can be generally found growing on some pine tree roots or dead barks, which leads to a worldwide distribution. PC is recognized as a common food in North American and Asian countries and also called “Tuckahoe”, “Indian bread” or “Fu-Ling” in Chinese and meanwhile, the name Wolfiporia is frequently used in North American but the name Poria cocos is mainly used in Asia [11]. This mushroom has a soft texture feature and even looks like a potato with a sweet flavor when used as ingredients to cook soup or make tea. Meanwhile, artificial foods of PC such as “Tuckahoe Pie” and “Fuling cake” can be often found in Chinese supermarkets. For these reasons, a big market of PC has been stimulated and PC is being widely produced in China, Korea and Japan. In the current study, a pharmaco-metabolomic study was performed to validate the therapeutic effect of PC on dietinduced hyperlipidemia and to explore why PC can supply this treatment using the advanced analytical technology ultra-performance liquid chromatography coupled with quadruple time-of-flight high-definition mass spectrometry (UPLC-QTOF/HDMS). According to Nicholson’s article, metabolomics was firstly put forward as one of the “omics” methods with quantitative measurement of dynamic, multiparametric metabolic responses of living systems to pathophysiological stimuli or genetic modifications [12]. With the combination of advanced analytical techniques and robust multivariate statistical analysis, metabolomics has been extensively utilized both in animal experimental studies and clinical trials, which plays a momentous role in interpreting disease pathological progression and developing new drugs. Since metabolomics has lots of advantages in studying living systems, it has been widely used not only in public health service but also in plant varieties and biofuels production [13-16]. Metabolomics is also considered as a forceful tool to provide an instantaneous snapshot of the physiology of a cell and contribute to drug-development process and mechanistic illustration [17]. The techniques 1H-NMR, LC-MS and GC-MS are the commonly employed analytical tools to acquire rich information from the biological samples [18-21]. As is reported, LC-MS is more preferred when analysing ionic, polar metabolites and thermally labile compounds among various analytical platforms, and two-dimensional liquid chromatography are indeed better than one for metabolomics studies [22,23]. More importantly, the combinational metabolomics approach of UPLC-MS coupled with multivariate statistical analysis has been confirmed to be more suitable for biological samples to quickly obtain metabolite profiling, especially for large-scale untargeted metabolomics [24]. As a branch of metabolomics, lipidomics is a powerful tool to investigate the diseases related to dyslipidemia [25-27]. As thus, plasma lipidomics was applied to reveal the antihyperlipidemic effect of PC and the results demonstrated PC had a good lipid-lowing activity [28]. Here, to further validate its therapeutic effect, we employed

Chen et al.

UPLC-QTOF/HDMS to analyse the urine samples from control, die-induced hyperlipidemic and PC-treated hyperlipidemic rats. For further analysis, partial least-squaresdiscriminate analysis (PLS-DA), heatmap display, correlation coefficient analysis and receiver operating characteristic (ROC) curve were applied to conduct statistical analysis. 2. MATERIALS AND METHODS 2.1. Chemicals and Reagents Formic acid solution (ref. BCBB6918, purity 50%) was supplied by Sigma Chemical Co., Ltd (Sigma Corp., St. Louis, MO, USA). LC-grade acetonitrile and methanol were both purchased from the Baker Company (Mallinckrodt Baker Inc., Phillipsburg, NJ, USA). Ultra purity water was prepared using a Milli-Q water purification system (Millipore Corp., Billerica, MA, USA). 2.2. Animals, Drug and Oral Administration Firstly, PC was ground to produce powder (screened using a sieve with 20 meshes) by a disintegrator. Two kilograms of PC powder were completely immersed in 95% ethanol (15 L) and then extracted by an ultrasonic cleaner for three times (0.5 h each time). The extracting solution was concentrated and dried. Then, the dried extract was dissolved with water and orally administrated to the rats in experimental group. Twenty-four healthy rats (weight between 190 and 210 g) were randomly divided into three groups (n=8/group): control group, model group and PC group. Common diet was offered to control rats during the whole experimental period. The rats in PC and model groups were fed with high fat diets to reproduce diet-induced hyperlipidemic rat model, as described in detail previously [29]. After hyperlipidemic model was successfully built, the rats in PC group were orally administrated with PC ethanol extract (250 mg/kg) in the last six weeks. All procedures in the present animal experiment were in accordance with the standards set forth in the eighth edition of Guide for the Care and Use of Laboratory Animals published by the National Academy of Sciences, The National Academies Press, Washington DC, United States of America. The care of the rats complied with institutional guidelines for animal use in research, and all efforts were made to minimise suffering. 2.3. Sample Collection and Preparation Blood samples were obtained before sacrificing all the rats. Urine samples from each rat were collected within 24 hours and they were stored at
80 °C. Before UPLC-MS analysis, urine were thawed at room temperature. Then the thawed samples were centrifuged at 13000 rpm for 10 min. The supernatant was separated and diluted with distilled water at a ratio of 3:1. The sample preparation was accorded with the method in our previous study [30]. 2.4. Serum Biochemistry Blood specimens from control group, model group and PC group were processed to acquire serum for biochemical parameters determination. Employing an Olympus AU640

Antihyperlipidemic Effect of Poria cocos by Metabolomics

automatic analyser, the concentrations of serum TC, TG, HDL-C and LDL-C were obtained for further assessment. 2.5. UPLC-HDMS Analysis By a Waters ACQUITY UPLC system (Waters, USA), UPLC-HDMS analysis was performed on a HSS T3 column (2.1 mm  100 mm, 1.8
m). The mixture of water (A) and acetonitrile (B) was used as the mobile phase and each contained 0.1% formic acid. Gradient elution was applied and showed as follows: 0–0.5 min, 1% B; 0.5–12.0 min, 1–30% B; 12.0–15.0 min, 30–99% B; 15.0–16.0 min, 99% B and 16.0–20.0 min, 99.0–1.0% B. Every prepared urine sample (2
L) was injected for each run. The autosampler and the column were stored at 4 °C and 55°C, rspectively. The flow rate was set at 0.45 mL/min. The corresponding mass spectrometry data of isolated compounds were collected on a quadrupole and orthogonal acceleration time-of-flight tandem mass spectrometer with the scan range from 50 to 1200 m/z. The capillary and cone voltage were 2.5 kV and 45 V, respectively. At a temperature of 550 °C, the desolvation gas was controlled at 900 L/h. The flow rate of cone gas was set at 50 L/h and the source temperature was maintained at 120 °C. All the data acquisition and analysis were performed by Waters MassLynx v4.1 and MakerLynx software. 2.6. Data Processing, Biomarker Identification and Metabolic Pathway Analysis The UPLC-MS data were extracted and introduced to Markerlynx XS to conduct the operations of peak detection and alignment. Completing data normalisation, the summed total ion intensity was output to EZinfo 2.0 software for PLS-DA analysis. For further discriminant analysis, some compounds were selected and identified according to the values of variable in project (VIP). Box plots of relative intensity were analysed by GraphPad Parism 5.0. Fold changes (FC), One-way analysis of variance (ANOVA), MannWhitney U-test and false discovery rate (FDR) correction (Benjamini Hochberg method) were performed by SPSS 19.0. Heatmap display and ROC curves were produced by Metaboanalyst 3.0 and Medcalc 12.7. All the p values lower than 0.05 represented the relationship between these two subjects showed significant discrepancy. Additionally, ingenuity pathway analysis (IPA) was employed based on metabolomics pathway analysis (MetPA) to visualize metabolomics. For further investigation, hyperlipidemia-related metabolic pathway clarification was presented by the quantitative enrichment analysis (QEA) algorithm which was conducted in the metabolite set enrichment analysis (MSEA). Visualization of metabolic pathways was acquried by Metscape which was operated on Cytoscape 3.3. 3. RESULTS AND DISCUSSION 3.1. Serum Biochemistry Results The concentrations of TC, TG, HDL-C, and LDL-C in control group were 2.64±0.53, 0.59±0.06, 0.96±0.10 and 1.71±0.12 mmol/L, respectively. In model group, remarkably elevated concentrations of TC (4.60±0.41 mmol/L), TG (1.13±0.15 mmol/L) and LDL-C (3.02±0.24 mmol/L) were observed whereas HDL-C (0.59±0.08 mmol/L) showed a

Current Metabolomics, 2016, Vol. 4, No. 2

3

lower level compared with the control rats. After making a comparison with model group, TC (3.77±0.31 mmol/L), TG (0.81±0.14 mmol/L) and LDL-C (2.45±0.17 mmol/L) concentrations in PC group were significantly lower than those in model group. However, HDL-C level (0.79±0.05 mmol/L) in PC group was obviously higher than in model group. As thus, the serum biochemistry results demonstrated the hyperlipidemic rat model was successfully reconstructed and PC could regulate the abnormal concentrations of these four serum parameters to be closer to normal levels. 3.2. Metabolite Identification and Statistical Analysis The base peak intensity (BPI) chromatograms in positive and negative ESI ion modes were presented in Fig. (S1). To validate the reproducibility, six replicated analysis of the same urine sample were conducted. Ten peaks (5.36_476.3059, 21. 2.37_154.048, 3.40_281.0979, 4.65_362.2159, 5.09_432.2791, 7.10_475.2486, 11.49_391.2861, 2.79_367.1489, 3.81_281.0976, 1.43_169.0336) were selected and extracted to perform the methodology anatomy. The values of relative standard deviations of retention time and peak area were lower than 0.72% and 2.92%, respectively. MarkerLynx 4.1 was employed to conduct the discriminant analysis based on PLS-DA model. From the collected data, 5098 and 5789 variables in positive and negative ion mode were processed, as shown in Fig. (1A) and (1B). The identified variables were marked with red hollow squares in the PLS-DA loading plots. The results indicated these metabolites had higher VIP values and contributed greatly to the profiles. Additionally, scores plots among control group, model group and PC group were acquired by PLS-DA model. Fig. (2A) and (2C) displayed 2D of PLS-DA scores plots in positive and negative ion mode, respectively. Fig. (2B) and (2D) displayed 3D of PLS-DA scores plots in positive and negative ion mode, respectively. The results suggested the test specimens in each group clustered well, which obviously validated a high predictability and adaptability. Fifty variables with VIP values more than 0.70 in positive ion mode and 50 variables with VIP values more than 0.67 in negative ion mode were screened for identifying structures. To continue selecting and analysing endogenous metabolites, xenobiotics and inappropriate fragment ions from the same metabolites were discarded. Eighteen metabolites including seven in positive ion mode and eleven in negative ion mode were identified (Table 1). Box plots of relative intensities of these metabolites were made and exhibited in Fig. (3), indicating 7 metabolites including tyrol, arginine, methylhippuric acid, aminoadipic acid, 3hydroxysebacic acid, citric acid and creatininie in the PC group showed a consistency with the control group but an opposite tendency with the model group. PLS analysis of five components (Fig. 1C) and correlation analysis (Fig. 1D) were conducted by Metaboanlyst 3.0 based on the relative intensities of 18 metabolites. The metabolites in PLS analysis were correspondent with the variables indicated in global PLS-DA loadings. Furthermore, dendrograms (Fig. 2E) and scores plot of PLS analysis of 18 metabolites (Fig. 2F) were also generated according to the data from the differential groups. The dendrograms indicated the samples could be

4 Current Metabolomics, 2016, Vol. 4, No. 2

Chen et al.

Fig. (1). PLS analysis and correlation analysis among control group, model group and PC group. PLS-DA loading plot of global metabolic profiles in positive ion mode (A) and negative ion mode (B). The identified metabolites are located in the loading plots and the corresponding spots are marked with red boxes. (C) PLS analysis of five components of 18 identified metabolites in control group, model group and PC group, respectively. The red, green and blue symbols represent control group, model group and PC group. (D) Correlation analysis of the differential metabolites are shown. The color of each section is proportional to the significance of the change in the metabolites (red, upregulated; blue, down-regulated). The variables labeled with compound names are the identified metabolites with significant changes among control group, model group and PC group.

Antihyperlipidemic Effect of Poria cocos by Metabolomics

Current Metabolomics, 2016, Vol. 4, No. 2

5

Fig. (2). PLS-DA model results for control group, model group and PC group. (A) and (C) show the 2D of PLS-DA model in positive and negative ion mode, respectively. (B) and (D) show the 3D of PLS-DA model in positive and negative ion mode, respectively. Dendrograms (E) of hierarchical clustering of different groups are classified by PLS-DA model with the significantly changed metabolites. PLS analysis of two components of 18 metabolites is performed and shown in scores plot (F). The red, green and blue symbols represent control group, model group and PC group. Table 1. No.

Identified metabolites, fold change (FC) and p-values among control, model and PC groups. Metabolites

Model vs. Control

PC vs. Contol

PC vs. Model

VIPa

FCb

p-valuec

p-valued

FDRe

FCb

p-valuec

p-valued

FDRe

FCb

p-valuec

p-valued

FDRe

1

Tyrosol

1.94

1.46

1.01
10-08

5.64
10-09

6.06
10-08

0.84

2.52
10-02

2.91
10-01

3.24
10-02

0.58

1.16
10-14

4.47
10-08

2.09
10-13

2

Indoxyl sulfate

1.67

0.89

6.00
10-05

2.20
10-03

1.08
10-04

0.89

8.22
10-05

2.21
10-03

1.85
10-04

1.00

8.71
10-02

9.73
10-01

1.21
10-01

3

Propionylcarnitine

1.28

0.32

4.81
10-12

5.47
10-07

8.66
10-11

0.32

4.38
10-12

5.47
10-07

7.88
10-11

0.99

9.75
10-01

8.90
10-01

9.75
10-01

4

Arginine

1.14

1.69

2.80
10-09

5.64
10-09

2.52
10-08

1.09

3.77
10-01

5.80
10-02

3.99
10-01

0.64

1.37
10-09

3.51
10-07

1.23
10-08

5

3-Methyldioxyindole

1.11

0.78

6.87
10-05

1.77
10-02

1.12
10-04

0.68

6.32
10-08

1.87
10-05

2.84
10-07

0.87

2.36
10-02

1.87
10-04

3.86
10-02

6 Current Metabolomics, 2016, Vol. 4, No. 2

Chen et al.

Table (1) contd….

No.

Metabolites

Model vs. Control VIP

a

FC

b

c

d

PC vs. Contol

p-value

p-value

FDR

e

FC

b

c

PC vs. Model d

p-value

p-value

FDR

e

FC

b

p-valuec

p-valued

FDRe

Cytidine

1.04

0.38

2.49
10-07

8.75
10-06

8.96
10-07

0.32

2.85
10-08

1.53
10-06

2.57
10-07

0.84

4.86
10-02

1.34
10-02

7.29
10-02

7

Trimethyltridecanoic acid

1.03

0.71

8.96
10-03

7.36
10-02

9.49
10-03

0.36

2.18
10-07

9.18
10-07

7.85
10-07

0.50

1.68
10-03

7.03
10-07

3.78
10-03

8

3-Hydroxyanthranilic acid

0.97

0.78

3.52
10-02

3.63
10-02

3.52
10-02

1.28

3.45
10-05

7.20
10-04

8.87
10-05

1.63

7.54
10-06

4.67
10-04

3.39
10-05

9

Xanthine

0.90

1.27

3.04
10-05

5.89
10-05

6.08
10-05

1.21

6.43
10-04

2.39
10-04

1.16
10-03

0.96

2.75
10-01

4.14
10-01

3.09
10-01

10

Methylhippuric acid

0.89

1.51

7.20
10-08

6.69
10-06

3.24
10-07

1.30

5.41
10-04

7.29
10-05

1.08
10-03

0.86

3.67
10-03

1.32
10-02

7.34
10-03

11

Aminoadipic acid

0.84

1.43

3.91
10-04

9.86
10-04

5.41
10-04

0.99

9.60
10-01

8.32
10-01

9.60
10-01

0.70

3.96
10-04

1.58
10-04

1.19
10-03

12

3-Methyluridine

0.84

1.26

6.82
10-06

1.55
10-04

1.53
10-05

1.24

2.54
10-05

9.53
10-05

7.62
10-05

0.99

7.57
10-01

7.08
10-01

8.02
10-01

13

N-Acetylneuraminic acid

0.82

0.57

3.41
10-04

4.90
10-04

5.12
10-04

0.56

8.43
10-03

4.41
10-02

1.26
10-02

0.97

2.03
10-01

2.97
10-01

2.44
10-01

14

3-Hydroxysebacic acid

0.79

1.29

1.21
10-03

1.97
10-02

1.56
10-03

0.80

2.33
10-02

1.10
10-04

3.23
10-02

0.62

1.29
10-08

3.07
10-07

7.74
10-08

15

Citric acid

0.77

1.22

3.34
10-03

2.44
10-02

3.76
10-03

0.93

3.63
10-01

4.49
10-01

4.08
10-01

0.76

2.10
10-05

2.78
10-04

7.56
10-05

16

Creatinine

0.72

1.30

3.56
10-07

1.98
10-05

1.07
10-06

1.17

1.51
10-03

5.62
10-05

2.47
10-03

0.90

1.91
10-02

4.13
10-02

3.44
10-02

17

4-Aminohippuric acid

0.71

0.90

1.29
10-03

4.30
10-03

1.55
10-03

1.06

1.45
10-01

6.49
10-02

1.74
10-01

1.17

1.51
10-03

8.70
10-03

3.88
10-03

18

Uric acid

0.70

1.30

4.48
10-06

7.42
10-06

1.15
10-05

1.38

4.17
10-08

7.03
10-07

2.50
10-07

1.06

1.71
10-01

1.18
10-01

2.20
10-01

a

VIP value was extracted through PLS-DA model. b The FC was calculated based on binary logarithm for model vs. control, PC vs. control or PC vs. model. FC with a value greater than 1.00 indicates a higher intensity of the metabolite between model vs. control, PC vs. control or PC vs. model, while a FC value less than 1.00 indicates a lower intensity of the metabolite between model vs. control, PC vs. control or PC vs. model. c p-values were calculated using one-way ANOVA analysis. d p-values were calculated from nonparametric test Mann–Whitney U-test. e The FDR value was obtained from the adjusted p-value of FDR correction according to Benjamini–Hochberg method.

Fig. (3). Box plots of relative intensity of 18 identified metabolites among the control group, model group and PC group. The statistical significance between two groups is indicated. Model group vs. control group *p