A GC-MS-Based Metabolomics Investigation of the Protective Effect of Liu-Wei-Di-Huang-Wan in Type 2 Diabetes Mellitus Mice

Materials and Methods MKR mice were used for the development of diabetes with high-fat diet feeding. These mice were further injected with streptozocin (STZ) to aggravate kidney failure. Fasting blood glucose (FBG) and urinary albumin-to-creatinine ratio (ACR values) were determined to validate the successful establishment of diabetic models with desired kidney dysfunction. Metabolomics approach coupled with gas chromatography-mass spectrometry (GC-MS) and random forest (RF) algorithm was proposed to discover the metabolic differences among model group and control group as well as to examine the therapeutic efficacy of traditional Chinese medicine, Liu-Wei-Di-Huang-Wan (LWDHW), in diabetes and associated kidney failure. Results Some metabolites such as 3-hydroxybutyric acid, citric acid, hexadecanoic acid, and octadecanoic acid showed significant differences between the control group and model group. Treatment with LWDHW resulted in a significant decrease in FBG and ACR values. These results suggested that LWDHW could have beneficial effects in diabetes-associated renal failure.


Introduction
Type 2 diabetes mellitus (T2DM) is known as a chronic multifactorial disease, characterized by metabolic, hormonal, epigenetic, and oxidative imbalance [1,2]. e main pathological features of T2DM include chronic hyperglycaemia and dyslipidaemia. However, the risk of T2DM is associated with various complications, such as retinopathy, nephropathy, neuropathy, ischemic heart disease, and peripheral vasculopathy [1,2]. Epidemiological evidence suggested that diabetic kidney disease (DKD) is one of the most severe diabetic microvascular diseases and a leading cause of end-stage renal disease [3,4]. Approximately 20-40% of diabetic patients will ultimately develop DKD, Mounting evidences have demonstrated that this formulation is also effective in maintaining neuroendocrine immunomodulation balance, improves cognitive function [12,13], and delays the progression of renal failure [14]. Chen et al. have reported that LWDHW can induce CYP1A2, suppress CYP2A6 and NAT2 activities, and affect caffeine metabolism in in vivo model [15]. Cheng et al. have recently found that LWDHW can also help in controlling the plasma glucose levels [16]. Hsu and coworkers reported the beneficial effects of LWDHW on kidney patients. It is also known that integrating TCM healthcare into diabetes care can be associated with decreased risk of developing kidney failure [17].
Metabolomics is a systematic study of metabolic changes of small molecules in response to the changes of both endogenous and exogenous factors and has many potential applications and advantages for the research of complex systems, such as studies of metabolic diseases and therapeutic effects of medicines [18][19][20][21][22][23]. Diabetes and DKD are typical heterogeneous metabolic disorders, characterized by abnormal metabolism of carbohydrates, lipids, and proteins; thus, metabolomics seems to be a powerful tool for evaluating disease process and for identifying the mechanism of action of currently used therapeutic agents. erefore, in this study, we exploit a metabolomics approach to investigate the effects of TCM formulation, Liu-Wei-Di-Huang-Wan, in controlling of FBG and prevention of renal failure.

Sample Preparation.
e drug production standard was that Mu Dan Pi (Paeonia suffruticosa Andr.) was primarily extracted by using a water-vapor volatile oil extraction unit. e volatile oil was collected and the residue was mixed with other five kinds of herbs, immersed in distilled water (1 : 8, w/v) for 0.5 h, and extracted twice by refluxing with boiling water for 2 h. All of the filtrates were mixed together and vacuum concentrated and then mixed with volatile oil to form pill. 1 g of sample was weighted and dissolved in 25 mL methanol (50%), and then the mixture was filtered with 0.45 µm for HPLC analysis.

HPLC
Analysis. An Agilent-1260 HPLC system (Agilent Technologies, MA, USA) equipped with a quaternary pump and UV detection system along with Agilent Poroshell 120 SB-C18 (250 mm * 4.6 mm, 5 µm) column was used to analyze samples. e detection wavelength was set at 235 nm, with a flow rate of 0.2 ml/min, and the column temperature was 25°C. e mobile phase was composed of acetonitrile (solvent A) and 0.1% formic acid (solvent B). e gradient procedure was set as 2% A at 0-2 min, 2-22% A at 2-25 min, 22-35% A at 25-32 min, and 35% A at 32-40 min. e injection volume was 0.2 µL.

Animal Models.
Mice overexpressing a dominant-negative IGF-1R specifically in skeletal muscle (MKR mice) were first established by Fernandez and coworkers [24]. is model showed insulin resistance and can rapidly develop into overt diabetes. en, this model was successfully used in subsequent studies [25,26]. In the current study, forty MKR mice (12 weeks old) were randomly divided into four groups: MKR mice group, model control group, LWDHW group, and western medicine group (gliquidone), 10 mice for each group. However, ten C57BL/6 mice were selected as controls. MKR mice group, LWDHW group, and western medicine group were injected with 1% streptozocin (STZ) dissolved in citric acid buffer (pH � 4.5) at a dose of 50 mg/ kg/day, for 5 days, and the model control group and the control group were injected with the same dose of citric acid buffer. However, the control and the model control group were fed with a routine diet, and a high-fat diet was given to other groups for 4 weeks. Fasting blood glucose (FBG) values and urinary albumin-to-creatinine ratio (ACR) were tested to determine the development of diabetes in desired groups.
LWDHW and western medicine groups were treated with Liu-Wei-Di-Huang-Wan (1.08 g/kg/d) and gliquidone (7.80 mg/kg/d), respectively (once a day for 28 days). On 29 th day, mice were sacrificed, and blood samples were collected by cardiac puncture. FBG values and urinary ACR were determined. e protocols in this experiment were approved by the Animal Ethical Committee of the Hunan University of Chinese Medicine. All experimental procedures were conducted in accordance with the Guideline for the Care and Use of Laboratory Animals.

Collection of Serum Samples and Pretreatment.
Venous blood was collected from fasting mice in a blank tube without anticoagulant or preservative. e fresh blood was stored at 4°C for 1 h, and serum was collected by centrifugation at 3000 rpm and stored at −80°C until further use. Before analysis, the serum was thawed at 4°C for 30 minutes.
Firstly, 10 µL blood was collected from each sample to conduct quality control (QC) sample and vortex-mixed for 60 s, and each 100 µL blood was extracted to form different QC samples for daily analysis. In this way, QC samples can be used to validate the stability of equipment and the performance of the proposed method.
Secondly, 100 µL blood sample was extracted from each sample including QC samples and vortex-mixed with 300 µL methanol (including 1 mg/mL of heptadecanoic acid/methanol as internal standard) for 15 s, and this mixture was further centrifuged for 15 min (14,800 rpm, 4°C) to remove proteins. e supernatant (370 µL) was freeze-dried and mixed with methoxamine/pyridine (20 mg/mL) for 15 s and incubated for 1 h at 70°C, followed by the addition of 100 µL of BSTFA, and incubated again for 1 h (70°C). e supernatant was used for GC-MS analysis. Finally, after being preprocessed, all the samples were analyzed by using GC-MS at random order, and after the 5 samples were analyzed, QC sample was injected once to validate the stability of the equipment. QC samples analyzed on different days can be used to validate the reproducibility of sample preparation.

Gas Chromatography-Mass Spectrometry Conditions.
Metabolite analysis was performed on a DB-5MS capillary column (30 m × 0.25 mm × 0.25 µm) in Shimadzu GC-MS system. e detailed parameters and conditions of GC-MS analysis were similar as reported previously by our group with slight modifications [27]. e temperature was maintained at 70°C for 4 min and then programmed to rise to 300°C at a rate of 8°C/min and held for 3 min. e temperatures of the front injection port, ion source, and interface were set at 260°C, 280°C, and 230°C, respectively. e flow rate of helium gas was 1.0 mL/min. One microliter of the sample was injected at a ratio of 10 :1 split mode. e mass spectrometer was operated under electron impact (EI) mode at ionization energy of 70 eV and 0.90 kV detector voltage in 0.2 s/scan for full scan. e mass spectrometer was selected with m/z ranging from 55 to 600.

Data Processing and Analysis.
Metabolites in these GC-MS profiles were identified by using their chromatographic and mass spectral characteristics. e identification of metabolites and internal standard was based on the search results in the National Institute of Standards and Technology (NIST 07) mass chromatography library in the GC-MS Postrun Analysis software (Shimadzu). Metabolites were also validated by comparing with references' results with retention time and mass spectra. e internal standard was used for normalization in relative quantitative analysis of these metabolites. e raw data acquired by the analytical instruments were firstly processed with peak detection and alignment. Subsequently, the processed data were imported into Matlab software 8.0.1 (Mathworks, USA), and random forest method was used to discover the metabolite differences among different groups and further to identify the most valuable metabolite for distinguishing different groups. MetaboAnalyst 3.0 was used for biological pathway analysis based on these metabolites' information (http://www. metaboanalyst.ca/) [28].

Qualitative Analysis of Bioactive Compounds in Liu-Wei-Di-Huang-Wan.
e high-performance liquid chromatography-ultraviolet (HPLC-UV) method was used for the quality control of LWDHW. Five major constituents, including loganic acid, loganin, 6′-O-galloylpaeoniflorin, benzoylpaeoniflorin, and paeonol, were identified, as depicted in Figure S1. e contents of these five compounds were 0.18 mg/g, 2.53 mg/g, 0.22 mg/g, 0.65 mg/g, and 1.62 mg/g, respectively, and more detailed information can be found in Tables S1 and S2.

Assessment of MKR Model and erapeutic Effects of LWDHW.
ree animal groups were established, i.e., normal control group (healthy mice), model control group (diabetic mice), and MKR group (diabetic mice with kidney injury). As MKR mice can develop into overt diabetes by feeding with a relatively high-fat diet, MKR group was further injected with STZ to induce kidney injury. FBG values and urinary albumin-to-creatinine ratio (ACR) values of each group were determined to estimate the animal model, as listed in Table 1. As can be seen from Table 2, the FBG and ACR values for the model control group (diabetic mice) are higher than those of the normal control group (healthy mice), while the FBG and ACR values of the MKR group (diabetic mice with kidney injury) are higher than those of the model control group (diabetic mice). is means the FBG and ACR values can exhibit the states of disease.
After treatment, the FBG value for the LWDHW group was decreased from 14.12 to 9.509; similar results were observed in the western medicine group, which showed a decrease in FBG value from 14.12 to 8.718. e FBG values for the model group (MKR group) did not show a significant decrease. ese results indicated that both western medicine and traditional Chinese medicine could reverse the symptoms of diabetes in mice as well as reduce the impact of a high-fat diet. e urinary albumin-to-creatinine ratio (ACR) was also calculated for each group. ACR values for the LWDHW group were decreased from 150.8 to 115.8, while a decrease in ACR values from 151.2 to 112.4 was also observed in the western medicine group. ese results were consistent with the FBG values, and these results showed the drugs had prevention effects to the kidney.

Reliability of GC-MS-Based Metabolomics Method.
Internal standards and QC samples were used to assess the stability and repeatability of the current method. In the daily analysis process, three QC samples were initially injected for calibrating GC-MS equipment followed by injection of two internal standard samples. Response from both internal standard and QC samples was used to validate the stability of equipment at different analysis times. QC samples were also injected at different intervals during the experiment (after five samples). e results of these QC samples were used to monitor the stability of the whole analytical system as well as the reproducibility of sample preparation. e variations of internal standards and QC samples (except the first three QC) were calculated. RSD values for metabolites ranged from 1.32% to 12.30%. Moreover, the RSD values of two internal standards, 2-isopropylmalic acid and heptadecanoic acid, were 2.3% and 2.9%, respectively.

Endogenous Metabolites Identified by GC-MS.
A total of 30 endogenous metabolites were identified, and the representative total ion chromatogram of each group is presented in Figure S2. Detailed information about the identified metabolites, including compound name, retention time, and quantitative ions, is listed in Table 2.

Metabolites' Analysis between MKR Group and Model
Control Group. In this section, we aimed to find the differences among the MKR group, model control group, and normal control group from their metabolites' aspects and further identify some potential biomarkers for distinguishing different groups that can help in the diagnosis of DKD disease.
All normalized metabolites' profiles of the MKR group, model control group, and normal control group were analyzed by random forest method. Random forest method had been used previously in our research group and has showed its advantages in complex metabolic data analysis [29,30]. It could effectively extract characteristic features from these metabolites' data and visually reveal the relationships among different groups. e classification plot of three groups is shown in Figure 1. ree groups were well separated, and the differences among these groups were obviously observed. ese results showed that metabolites' profiles can reveal the metabolic disturbances in DKD progress, which can also be used for the diagnosis of DKD. erefore, RF algorithm was employed to classify samples from the MKR group and normal control group. e prediction accuracy, sensitivity, and specificity for the current method were 94.33%, 88.96%, and 91.57%, respectively.
Furthermore, in the classification model establishing process, the contribution of each metabolite for distinguishing was calculated. As can be seen from Figure 2, some metabolites such as 3-hydroxybutyric acid, citric acid, hexadecanoic acid, and octadecanoic acid have higher contributions in distinguishing the MKR group from the normal control group. ese metabolites with higher contributions were considered as potential biomarkers for the diagnosis of DKD.
Metabolic pathways of alanine, aspartate, and glutamate metabolism and linoleic acid metabolism play important roles in the progress of DKD (Figure 3). MetaboAnalyst 3.0 software was used to further evaluate the metabolic pathways associated with the identified metabolites. Metab-oAnalyst 3.0 is one of the most popular software used in metabolomics data analysis, which used the high-quality KEGG pathway database as backend knowledgebase. eir impact indexes are 0.7056 and 0.6562, respectively. e detailed metabolic pathways informatics is listed in supporting information Table S4 and Figures S3 and S4, respectively.

Comparative Metabolites' Analysis between MKR Group and Treated
Group. Changes in metabolic, physiological, and pathological conditions induced after administration of Liu-Wei-Di-Huang-Wan and gliquidone were further analyzed. e cluster analysis results for the model group, LWDHW group, western medicine group, and normal control group are shown in Figure 4.
As shown in Figure 4, the normal control group is located in the right of the cluster analysis plot, while the MKR group is located in the left part of cluster plot. ese results indicated an obvious difference in the metabolic patterns of International Journal of Analytical Chemistry 5 Both the LWDHW group and western medicine group were closer to the normal control, which indicated a comparative therapeutic efficacy of LWDHW with gliquidone. Besides, the authors also found that the intensities of some metabolites were different among LWDHW and western medicine groups. LWDHW had better regulation effects on metabolites, such as oxalic acid, butyric acid, serine, D-galactose, D-glucose, D-arabinose, and linoleic acid, while gliquidone had better regulation effects on L-proline, glycine, L-threonine, erythronic acid, N-acetyl-D-glucosamine, D-mannose, L-lysine, and D-turanose. ese results might be caused by the different therapeutic pathways of two medicines. More comprehensive and deep study would be implemented based on these metabolites' information. us, the advantage of metabolomics was that it can provide more information for the mechanism research.

Discussion
It is generally considered that TCMs exert a synergistic action to achieve the therapeutic benefit; the synergies may contain a series of complex interactions due to the coexistence of multiple components. It is difficult to explain all the effects caused by these components. Metabolomics approach provided a suitable perspective to monitor the comprehensive physiological and pathological changes caused by disease and treatment. Metabolites are the final products of various actions from genomics to proteomes and so on.
us, metabolomics is a research frontier of systems biology, which provides a new perspective to understand diseases and drug actions by identifying the signature metabolites that represent the global biochemical changes in living systems.
Previous studies have reported the presence of unusually high lipid deposition in the CKD and DKD patients, which suggest that excessive fatty acid (FA) accumulation and lipotoxicity can be a potential cause of kidney dysfunction. FAs are carboxylic acids with saturated or unsaturated aliphatic tails obtained from de novo synthesis or hydrolysis of triglycerides. ey are primary substrates in energy metabolism, synthesis of membrane phospholipids, and bioactive compounds [31,32]. An increased level of urinary FAs was found in DN patients, and these high levels were found to be in positive correlation with renal tubulointerstitial injury [33,34].
3-Hydroxybutyric acid is synthesized in the liver from acetyl-CoA and can be used as an energy source by the brain when the blood glucose level is low. Blood levels of 3hydroxybutyric acid may be monitored in diabetic patients as an indication of diabetic ketoacidosis. One of the functions of 3-hydroxybutyric acid is to provide acetoacetyl-CoA for the synthesis of cholesterol, fatty acids, and complex lipids [35]. In this study, we observed elevated levels of hexadecanoic acid and octadecanoic acid in the DKD progress. In the pathological status, the increased albuminbound FAs are reabsorbed by proximal tubular epithelial cells (PTECs) and promote FA deposition in the kidney [36,37].

Conclusion
In the present study, we employed metabolomics method to accurately differentiate the model group and control group and further represent the metabolic profile changes treated by TCM with high classification accuracy. ese results demonstrated that metabolomics could be effectively used to mine the information hidden in the complex metabolomics data and hence can pave the way for the interpretation of these data. erefore, metabolomics is an efficient and attractive alternative technique for diseases' diagnosis, pathogenesis, and pharmacodynamics research.

International Journal of Analytical Chemistry
Data Availability e quality control and metabolomics data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare no competing financial interests.

Authors' Contributions
Jian-hua Huang and Dan He contributed equally to this work. Jian-hua Huang and Rong Yu designed the study and drafted the manuscript. Dan He, Du Qing, Chen Lin, and You Qin participated in the animal experiments and interpretation of the results; Shui-han Zhang, Cai Ping, and Yu-hui Qin revised the manuscript. All authors read and approved the final manuscript. Table S1: the linear range for these five compounds. Table S2: identification information of the selected peaks by using HPLC-Q-TOF. Table S3: metabolite information of QC groups. Table S4: result from key metabolic pathway analysis with MetaboAnalyst 3.0. Figure S1: the HPLC chromatography of LWDHW samples and five standards: (1) loganic acid, (2) loganin, (3) 6′-O-galloylpaeoniflorin, (4) benzoylpaeoniflorin, and (5) paeonol. Figure S2: the TIC plots of different groups: blue line: normal control group, red line: MKR model, green line: LWDHW group, and purple line: western medicine group. Figure S3: alanine, aspartate, and glutamate metabolism (impact: 0.7056). Figure S4: linoleic acid metabolism (impact: 0.6562). (Supplementary Materials)