UHPLC-MS-Based Serum and Urine Metabolomics Reveals the Anti-Diabetic Mechanism of Ginsenoside Re in Type 2 Diabetic Rats

Panax ginseng was employed in the treatment of “Xiao-Ke” symptom, which nowadays known as diabetes mellitus, in traditional Chinese medicine for more than a thousand years. Ginsenoside Re was the major pharmacologic ingredient found abundantly in ginseng. However, the anti-diabetic of Ginsenoside Re and its underlying mechanism in metabolic level are still unclear. Serum and urine metabolomic method was carried out to investigate the anti-diabetic pharmacological effects and the potential mechanism of Ginsenoside Re on high-fat diet combined streptozotocin-induced type 2 diabetes mellitus (T2DM) rats based on ultra-high-performance liquid chromatography coupled with quadrupole exactive orbitrap mass spectrometry (UHPLC-Q-Exactive Orbitrap/MS). Serum and urine samples were collected from the control group (CON), T2DM group, metformin (MET) treatment group, and ginsenoside Re treatment group after intervention. The biochemical parameters of serum were firstly analyzed. The endogenous metabolites in serum and urine were detected by UHPLC-MS. The potential metabolites were screened by multivariate statistical analysis and identified by accurate mass measurement, MS/MS, and metabolite databases. The anti-diabetic-related metabolites were analyzed by KEGG metabolic pathway, and its potential mechanism was discussed. The treatment of ginsenoside Re significantly reduced the blood glucose and serum lipid level improved the oxidative stress caused by T2DM. Biochemical parameters (urea nitrogen, uric acid) showed that ginsenoside Re could improve renal function in T2DM rats. Respective 2 and 6 differential metabolites were found and identified in serum and urine of ginsenoside Re compared with T2DM group and enriched in KEGG pathway. Metabolic pathways analysis indicated that the differential metabolites related to T2DM were mainly involved in arachidonic acid metabolism, Vitamin B6, steroid hormone biosynthesis, and bile secretion metabolic pathways. This study verified the anti-diabetic and anti-oxidation effects of ginsenoside Re, elaborated that ginsenoside Re has a good regulation of the metabolic disorder in T2DM rats, which could promote insulin secretion, stimulated cannabinoid type 1 receptor (CB1), and CaMKK β to activate AMPK signaling pathway, inhibited insulin resistance, and improved blood glucose uptake and diabetic nephropathy, so as to play the role of anti-diabetic.


Introduction
The epidemic of diabetes mellitus (DM) and its complications poses a major threat to global health; about 1 out of 11 adults worldwide have DM, which is the ninth major cause of death globally [1]. DM is classified into type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM), and nearly 90% of cases are T2DM, which is caused by insulin resistance and lack of proper response to hyperglycemia [2]. DM can cause serious health problems, which may damage almost all organs in the body, such as retinas, liver, and kidneys, and most T2DM patients have at least one complication [3]. It has been reported gradually. There was no significant difference in the control and T2DM groups (p > 0.05). After 30 days of treatment, the FBG level in the MET and Re groups significantly decreased from 12.75 ± 2.37 mmol/L to 8.48 ± 2.79 mmol/L (p < 0.01) and 16.75 ± 1.35 mmol/L to 13.07 ± 2.25 mmol/L (p < 0.05), the hypoglycemic rate was 21.97%. During the 30 days of intervention, the food and water intake of MET and Re groups showed a decreasing trend, while it was no significant difference in the change of BW (p > 0.05), prompting that the effect of ginsenoside Re on reducing the FBG level was not due to the BW loss of the rats, but due to its pharmacological effect.
Molecules 2021, 26, x 3 of 18 (T2DM with normal saline), MET (T2DM with Metformin), and Re (T2DM with Ginsenoside Re) groups. The FGB and BW of each model rat were different, so the average values of the three groups were also different. Therefore, the starting points in Figure  1A,B were different. The FBG level of metformin and ginsenoside Re treatment group decreased gradually. There was no significant difference in the control and T2DM groups (p > 0.05). After 30 days of treatment, the FBG level in the MET and Re groups significantly decreased from 12.75 ± 2.37 mmol/L to 8.48 ± 2.79 mmol/L (p < 0.01) and 16.75 ± 1.35 mmol/L to 13.07 ± 2.25 mmol/L (p < 0.05), the hypoglycemic rate was 21.97%. During the 30 days of intervention, the food and water intake of MET and Re groups showed a decreasing trend, while it was no significant difference in the change of BW (p > 0.05), prompting that the effect of ginsenoside Re on reducing the FBG level was not due to the BW loss of the rats, but due to its pharmacological effect. As shown in Table 1, ginsenoside Re treatment significantly reduced the level of LDL-c and TG (p < 0.05). However, there are no significant differences in TC (p = 0.07), HDL-c (p = 0.42), and NEFA (p = 0.07) among the groups, suggesting that treatment of ginsenoside Re could reduce the blood lipid level of T2DM rats. Ginsenoside Re treatment significantly reduced the level of BUN (p < 0.05), and the level of UA also showed a tendency to decrease, but the differences (p = 0.26) were not statistically significant. It is suggested that ginsenoside Re could improve the renal function of T2DM rats and has the potential to treat diabetic nephropathy. Ginsenoside Re significantly increased the activities of SOD (p < 0.05) and decreased the level of MDA (p < 0.05) in T2DM rats, and the activities of CAT also tended to increase, but the difference was not statistically significant, indicating that ginsenoside Re could mitigate the oxidative stress in T2DM rats. The purpose of determining the biochemical indexes of rats was to observe the biological effects after drug intervention, that was, whether drugs can improve and treat diseases. After determining that the drug had positive biological effects, metabolomics was used to explore the metabolic pathways involved in regulation in this process, and then, explored the potential mechanism of the drug's pharmacological effect. As shown in Table 1, ginsenoside Re treatment significantly reduced the level of LDL-c and TG (p < 0.05). However, there are no significant differences in TC (p = 0.07), HDL-c (p = 0.42), and NEFA (p = 0.07) among the groups, suggesting that treatment of ginsenoside Re could reduce the blood lipid level of T2DM rats. Ginsenoside Re treatment significantly reduced the level of BUN (p < 0.05), and the level of UA also showed a tendency to decrease, but the differences (p = 0.26) were not statistically significant. It is suggested that ginsenoside Re could improve the renal function of T2DM rats and has the potential to treat diabetic nephropathy. Ginsenoside Re significantly increased the activities of SOD (p < 0.05) and decreased the level of MDA (p < 0.05) in T2DM rats, and the activities of CAT also tended to increase, but the difference was not statistically significant, indicating that ginsenoside Re could mitigate the oxidative stress in T2DM rats. The purpose of determining the biochemical indexes of rats was to observe the biological effects after drug intervention, that was, whether drugs can improve and treat diseases. After determining that the drug had positive biological effects, metabolomics was used to explore the metabolic pathways involved in regulation in this process, and then, explored the potential mechanism of the drug's pharmacological effect. Data are expressed as the mean ± SD. * p < 0.05 vs. T2DM group; ** p < 0.01 vs. T2DM group.

Multivariate Statistical Analysis Based on UHPLC-MS
In order to ensure the reliability and stability of the system and data, QC sample was detected in every six samples over the acquisition process, and the PCA model was performed on all the samples from serum and urine, as shown in Figure 2, indicating that the method showed good stability and low variation caused by the instrumental error. Total ion chromatograms (TIC) of the serum and urine samples from CON, T2DM, MET and Re groups were shown in Figures 3 and 4, respectively. Suitable chromatographic and MS conditions were selected for the measurement of the samples in this study. UHPLC-MS metabolic profiling analysis was performed.

Multivariate Statistical Analysis Based on UHPLC-MS
In order to ensure the reliability and stability of the system and data, QC sample was detected in every six samples over the acquisition process, and the PCA model was performed on all the samples from serum and urine, as shown in Figure 2, indicating that the method showed good stability and low variation caused by the instrumental error. Total ion chromatograms (TIC) of the serum and urine samples from CON, T2DM, MET and Re groups were shown in Figures 3 and 4, respectively. Suitable chromatographic and MS conditions were selected for the measurement of the samples in this study. UHPLC-MS metabolic profiling analysis was performed.        PCA was an unsupervised multivariate analysis method used to reflect the overall distribution trend among samples. For each comparison group of serum ( Figure 5A) and urine ( Figure 5B), PCA score plots showed an obvious trend of separation, suggesting that the Pharmacological experiment model was established successfully, and the drug intervention changed the metabolism in rats.
A supervised OPLS-DA model was established to further screen the differential metabolites. The separated clusters were observed between samples of each contrast group. The OPLS-DA model parameters (R2Y, Q2) were obtained through a 7-fold cross-validation to measure the fitness and the prediction ability of the model. R2Y values in serum (R2Y = 0.69) and urine (R2Y = 0.94) were all greater than 0.5, suggesting that the model could meet the predictive ability of the data. Then, the OPLS-DA model was assessed by a permutation to verify the quality of the model, overfitting did not occur according to the results. Both PCA and OPLS-DA score plots showed an obvious separation trend in serum and urine comparison groups, indicating that ginsenoside Re treatment intervened in the metabolic profiles in T2DM rats.
Molecules 2021, 26, x Figure 5. PCA analysis, OPLS-DA analysis, OPLS-DA permutation test and volcano plot of (A) and urine (B) samples from the ginsenoside Re compared to T2DM group, respectively A supervised OPLS-DA model was established to further screen the diff metabolites. The separated clusters were observed between samples of each c group. The OPLS-DA model parameters (R2Y, Q2) were obtained through a cross-validation to measure the fitness and the prediction ability of the model. R ues in serum (R2Y = 0.69) and urine (R2Y = 0.94) were all greater than 0.5, suggesti the model could meet the predictive ability of the data. Then, the OPLS-DA mod assessed by a permutation to verify the quality of the model, overfitting did no according to the results. Both PCA and OPLS-DA score plots showed an obvious tion trend in serum and urine comparison groups, indicating that ginsenoside R ment intervened in the metabolic profiles in T2DM rats.
Univariate statistical analysis was then carried out to visualize the fold chan value and significance test of difference (p-value) through volcanic map analysis 5), red-dots represented significantly up-regulated metabolites, blue-dots repre significantly down-regulated metabolites, and gray-dots represented insignifican Univariate statistical analysis was then carried out to visualize the fold change (FC) value and significance test of difference (p-value) through volcanic map analysis ( Figure 5), red-dots represented significantly up-regulated metabolites, blue-dots represented significantly down-regulated metabolites, and gray-dots represented insignificantly metabolites (p > 0.05). There were 10 (in serum) and 12 (in urine) differential metabolites filtered out in the Re compared to the T2DM group, which are represented in Table 2.

KEGG Pathway Enrichment Analysis
Among the differential metabolites, two compounds (in serum) and six compounds (in urine) were identified which could be enriched in KEGG metabolic pathway (shown in Table 3). Differential metabolites are putatively identified by accurate mass measurement, MS/MS spectra, and database searching. The molecular MS error of all identified metabolites was less than 10 ppm. These differential metabolites were identified by comparing the retention time, accurate mass, and MS/MS spectra with these characteristics of the authentic standards. The three representative compounds involved in the following discussion were taken as examples to explain the metabolite identification. The extracted ion chromatogram (EIC), full scan mass spectrum, and MS/MS spectrum of compounds were shown in Figure 6.  Figure 6B,C. All the identified metabolites were identified based on the accurate mass measurement and MS/MS product ion analysis compared with authentic standards or databases.
Pathway enrichment analysis was performed using KEGG online database to explore the best-fit pathways and potential metabolic mechanism. Two (in serum) and six (in urine) metabolites were enriched in the KEGG pathway, respectively, as shown in Table 3. Pathway enrichment analysis was performed using KEGG online database to explore the best-fit pathways and potential metabolic mechanism. Two (in serum) and six (in urine) metabolites were enriched in the KEGG pathway, respectively, as shown in Table 3.

Discussion
The differential metabolite 12(R)-HETE was detected in serum, with up-regulated expression (log2FC > 0). It is a 12-lipoxygenase (12-LO)-mediated arachidonic acid decomposition product, which is widely associated with the dysfunction of human and mouse pancreatic islet β-cell [23]. The level of 12(R)-HETE in patients with diabetes is controversial, some studies have shown an increase, while others have shown no change

Discussion
The differential metabolite 12(R)-HETE was detected in serum, with up-regulated expression (log 2 FC > 0). It is a 12-lipoxygenase (12-LO)-mediated arachidonic acid decomposition product, which is widely associated with the dysfunction of human and mouse pancreatic islet β-cell [23]. The level of 12(R)-HETE in patients with diabetes is controversial, some studies have shown an increase, while others have shown no change or a decrease in the level. However, it was both an important metabolite associated with diabetes [24]. Arachidonic acid (AA) had potent anti-inflammatory and anti-diabetic effects, and the levels of AA in the liver and plasma were significantly decreased in obese and T2DM rats [25]. The up-regulation showed that ginsenoside Re could activate the AA metabolic pathway and improve diabetes in the diabetic rat model. The compound was also detected during the intervention of total ginsenoside of ginseng berry in T2DM rats, indicating that ginsenoside Re was one of the pharmacodynamic substances of ginseng berry on anti-diabetes [20]. 4-Pyridoxic acid was detected in urine. The expression was downregulated (log 2 FC < 0), and participated in the Vitamin B6 metabolism pathway. Vitamin B6 deficiency is common in diabetic patients [26], and studies have shown that the ability of vitamin B6 of diabetic patients to decompose into 4-Pyridoxic acid was heightened [27]. The up-regulation of 4-Pyridoxic acid was related to renal insufficiency or inflammation, and an excess of 4pyridoxic acid was rapidly excreted in urine [28]. Diabetic nephropathy was the most common cause of end-stage kidney disease as well as renal failure [26]. The up-regulation of 4-pyridoxic acid means that ginsenoside Re could improve diabetic kidney disease.
The differential metabolite dehydroepiandrosterone sulfate (DHEA-S) was detected in urine, the expression was up-regulated (log 2 FC > 0), and participated in steroid hormone biosynthesis and bile secretion. The DHEA-S was the sulfated form of dehydroepiandrosterone (DHEA), and most DHEA was found in the form of DHEA-S, which is about 300 times higher than that of free DHEA. They were the most abundant adrenal steroids in human circulation [29]. DHEA-S could enhance insulin secretion of the pancreas and improve insulin sensitivity of muscle and liver. Moreover, it had a beneficial effect on diabetes and obesity [30]. In addition, a low concentration of DHEA-S in serum could lead to insulin resistance, while high insulin concentrations could down-regulate the concentration of DHEA-S by inhibiting promotion and producing its clearance [31]. There was a significant correlation between the onset of T2DM and a decrease in DHEA-S level, and a higher DHEA-S level in serum had a significant protective effect on the onset of T2DM in older men [32]. It has been reported that DHEA-S can up-regulate the expression of ACC1, and promote insulin secretion, which had a useful therapeutic effect on diabetes mellitus, especially for patients with insulin secretion deficiency [29]. Furthermore, the degree of diabetic nephropathy in men was inversely related to the level of DHEA-S [33]. The up-regulation of DHEA-S means that ginsenoside Re could improve its concentration and promote insulin secretion, so as to exert a part in the improvement and treatment of T2DM.
2-Arachidonylglycerol (2-AG) was detected in urine, up-regulation (log 2 FC > 0), and participated in retrograde endocannabinoid signaling. Excessive energy intake may lead to obesity, insulin resistance, and various metabolic syndrome complications, and then lead to diabetes. 2-AG could activate an endocannabinoid receptor CB 2 , the activation of CB 2 could improve obesity and its related metabolic disorders [34]. It has been reported that 2-AG was a key signal molecule to improve insulin resistance and inflammation. 2-AG can activate AMPK by stimulating cannabinoid type-1 receptor (CB 1 ) and Ca 2+ /calmodulindependent protein kinase β (CaMKK β), inhibit insulin resistance induced by TNF α, improve glucose uptake, and improve diabetes mellitus [35]. The up-regulation of 2-AG meant that ginsenoside Re could activate the AMPK signal pathway, against insulin resistance and improve diabetes. The metabolite Riboflavin, Allocholic acid from serum, Estrone glucuronide, Tetrahydrocortisol, and 17-β-estradiol-3-glucuronide from urine were also enriched in the KEGG pathway, as shown in Table 3.
In the recent decade, due to the expansion and application of metabolomics, the latest understanding of T2DM pathophysiology has been continuously expanded. In addition to branched-chain amino acids, phospholipid metabolism, and glutamine glutamate cycle, which were previously unknown diabetes pathways and biomarkers, vitamin B6 metabolism pathway were also reported to be associated with diabetes [21], which was also consistent with the screening results in this study. The appearance of metabolomics helped realize the discovery of these pathways and biomarkers. Moreover, it has been reported that the anti-diabetic mechanism of ginseng may be related to its regulation of insulin secretion, improvement of glucose uptake, and oxidative stress [6], and the mechanism of ginsenoside Re's anti-diabetic effect was basically consistent with previous reports.
According to the results of the hypoglycemic experiment, ginsenoside Re could reduce the fasting blood glucose of T2DM rats. After 30 days of intervention, the hypoglycemic rate could reach 21.97%, but the hypoglycemic effect was lower than that of ginseng berry total ginsenoside extract (27.35%) [20]. The pharmacological effects of traditional medicinal herbs such as ginseng had the characteristics of multi-component and multi-target, that is, the pharmacological effects of herbal medicines were usually the result of the joint action of multiple active ingredients on different biological target molecules. Although the monomer drugs were better than traditional herbal medicines in quality control, and ginsenoside Re was also the highest monomer saponin contented in ginseng berry, from the research results, it cannot replace the total ginsenosides of Panax ginseng berry in lowering fasting blood glucose and treating T2DM.

Animal Experiments
30 seven-week-old male Wistar rats (190 ± 10 g) were obtained from the Experimental Animal Center of Jilin University (Jilin, China). The animal experiment was approved by the Institutional Animal Care and Use Committee (IACUC) of the Changchun University of Chinese Medicine; permit number: CPCCUCM IACUC 2019-008. Animals were reproduced in an SPF barrier system with standard environmental parameters: constant temperature of 21-25 • C, relative humidity of 60% ± 20%, and 12 h light-dark (6:00 a.m.-6:00 p.m.) cycle and received food and water freely. After 7 days of acclimatization, 8 rats were randomly divided into healthy control (CON) group and fed a normal diet. The others were fed a high-fat and high-sugar diet (HFD), including 10% lard oil, 20% saccharose, 2.5% cholesterol, 1% sodium cholate, and 66.5% standard rat diet, and 8 weeks later, STZ (dissolved in sodium citrate buffer, 0.1 mol/L, pH 4.3-4.5) was intraperitoneally injected into rats at a dose of 40 mg/kg b.w. and the rats in the CON group were only injected with sodium citrate buffer as parallel control. Seven days later, after fasting for 12 h, the tail-vein blood glucose was measured. The rats with blood glucose ≥7.8 mmol·L −1 , polyuria, and polydipsia were successfully established as T2DM model.
Twenty-two T2DM rats were randomly divided into three groups. Among them, 6 rats in the Re group continued to feed with an HFD diet and were given ginsenoside Re (dissolved in saline) at a dose of 25 mg·kg −1 (b.w.) once a day by gastric irrigation. Eight rats in the MET group were also fed an HFD diet and given metformin (dissolved in saline) at a dose of 100 mg·kg −1 (b.w.) once a day. In the T2DM group, 8 rats were fed an HFD diet and only saline was given by gastric irrigation. The CON group continued to be fed with a normal diet and was also given saline in parallel. The blood glucose level of rats was measured every 7 days (after fasting for 12 h). The weight of rats and the food and water intake in 24 h were measured every 4 days.

Urine Samples
The intervention lasted for 30 days, at the endpoint of treatment, rats in each group were placed in a metabolism cage (1 rat per cage). Twenty-four-hour urine samples were collected in tubes on ice from 4 groups. Then, urine samples were centrifuged at 8000 rpm at 4 • C for 15 min, and the supernatant was divided into 250 µL/portion at 4 • C and stored at −80 • C until analysis. The urine samples were thawed at 4 • C before analysis, after thawing, 250 µL of urine was exactly aspirated, added 750 µL methanol (4 • C), mixed by vortex for 30 s, and then centrifuged at 12,000 rpm at 4 • C for 15 min. The centrifuged supernatant was sucked into a new tube and evaporated under nitrogen. Then, the dried sample was redissolved with 1 mL ultra-pure water (4 • C), centrifuged at 4 • C and 12,000 rpm for 15 min, and the supernatant was filtered by 0.22 µm membrane and analyzed by UHPLC-MS. All operations were performed on ice.

Serum Samples
To collect serum samples, the rats were fasted (12 h, overnight) and anesthetized by intraperitoneal injection of 10% chloral hydrate (3 mL/kg body weight). Blood samples were vacuum-collected from the abdominal aorta. Then, the blood sample was placed at room temperature for 60 min and centrifuged at 3000 g at 4 • C for 15 min. The supernatant was separated from the blood sample and the serum sample was obtained. The serum was packed into 300 µL/portion and immediately frozen in liquid nitrogen, then stored at −80 • C until analysis. One part was used for determining rat serum biochemical parameters according to the commercial kit instructions, and the other was used for UHPLC-MS detection. Before the analysis, the serum was thawed at a temperature of 4 • C. After thawing, 250 µL of serum was accurately aspirated and added to 750 µL of methanol solution (4 • C) in order to precipitate the proteins, which was mixed by vortex (30 s) and centrifuged at 12,000 rpm for 15 min at 4 • C. After the supernatant was transferred into a new tube and dried by a nitrogen-blowing instrument, 500 µL of methanol was again added, vortexed, and centrifuged at 12,000 rpm for 10 min. After the supernatant was collected again and dried, the solid sample was dissolved in 1 mL of ultra-pure water and centrifuged at 12,000 rpm at 4 • C for 15 min. Then, the supernatant was filtered by a 0.22 µm membrane filter, and then analyzed by UHPLC-MS. All the processing was performed on ice.

Quality Control Sample
In order to test the stability and reproducibility of UHPLC-MS, quality control (QC) samples were prepared. Among all the samples, small parts of the same volume were removed, mixed, and treated according to the pretreatment method of the samples to obtain the QC samples of serum and urine samples, respectively. Before the experiment, the QC sample was detected six times to balance the analysis system, then QC samples and blank samples (75% acetonitrile in water) were detected once between every six real samples in the sequence.

UHPLC-MS Analyses
Serum and urine metabolomic profiling was analyzed by the UHPLC Q-Exactive Orbitrap MS system, and chromatographic separation was detected on an UltiMate 3000 system (Sunnyvale, CA, USA) with C 18 column (50 × 2.1 mm, 1.9 µm, Thermo Fisher Scientific) maintained at 35 • C. Mobile phase A was 0.1% formic acid in water and mobile phase B was 100% acetonitrile with a flow rate of 0. Mass spectrometric detection was implemented on a Q-Exactive Orbitrap MS (Thermo, CA, USA) equipped with electrospray ionization (ESI) source in both positive and negative ion modes. The scanning range was m/z 66.7~1000. The ionization source parameters were set as follows: sheath gas flow was 6.125Mpa, aux gas flow was 2.625 Mpa, and sweep gas flow was 0.175 Mpa. The capillary temperature was 320 • C, the auxiliary gas heater temperature was 350 • C, and the capillary voltage was 3.5 kv.
Scanning mode was Full MS with dd-MS 2 , Top N = 8. The resolution was 70,000 in Full MS mode, and 17,500 in data-dependent MS 2 mode, respectively. The normalized collision energy (NCE) was 10%, 20%, and 40%. The software for recording and analyzing data were Dionex TM , Chromeleon TM 6.8, and Xcalibur software Version 2.2.42 (Thermo Fisher Scientific, San Jose, CA, USA).

Multivariate Data Processing and Analysis
Serum and urine raw mass data from the UHPLC-MS was imported into the Progenesis QI v2.3 software (Nonlinear Dynamics, Newcastle, UK), for peak alignment, deconvolution, correction retention time, normalization, and integration. The software could carry out normalization processing of the data based on QC samples. During data collection, insert and collect the same QC sample every six samples, which could be used to simulate signal changes during data acquisition, and then, normalized the data through QI v2.3 software. More than 50% of the metabolites with missing measured values were not used for a follow-up analysis in the extracted ion features. Both the positive and negative ion data were merged into a data matrix and then saved as an excel file.
The unsupervised PCA (principal components analysis) and the supervised OPLS-DA (orthogonal partial least squares discriminant analysis) were performed on R programming language v3.6.2 (Ropls package was used for multivariate statistics). The robustness of the OPLS-DA model was evaluated by 7-fold cross-validation and 200 times response permutation testing (RPT). Univariate statistical analysis was used for basic significance statistics by Stats package (ggplot2 package was used for plotting). The differential metabolite was selected on the basis of two criteria: (a) the VIP value from the OPLS-DA model should be satisfied with VIP > 1 and (b) the differences from the Univariate statistical analysis (Student's t-test) should be satisfied with p < 0.05. Data were shown as the mean ± standard deviations (SD).

Differential Metabolites Identification and Pathway Analysis
The identification of potential metabolites was based on the product ion mass spectra and MS/MS fragment ion and suited with the structural message in the databases, such as HMDB (http://www.hmdb.ca/) (accessed on 7 December 2019), METLIN (https://metlin. scripps.edu/) (accessed on 7 December 2019), and Lipidmaps (http://www.lipidmaps. org/) (accessed on 7 December 2019). The metabolic pathway analysis of the differential metabolites was performed by the online database KEGG (http://www.kegg.jp/) (accessed on 7 December 2019).

Conclusions
UHPLC-MS was used to study the anti-diabetic effect of ginsenoside Re on T2DM rats induced by a high-fat diet and streptozotocin. The biochemical and pharmacological indexes showed that ginsenoside Re could significantly reduce the level of fasting blood glucose and serum lipid, improving oxidative stress and renal function in T2DM rats, which proved that ginsenoside Re could improve T2DM. There were 10 (in serum) and 12 (in urine) differential metabolites that were screened in the ginsenoside Re vs T2DM group. Among the differential metabolites, two compounds (in serum) and six compounds (in urine) were identified and analyzed which could be enriched in the KEGG metabolic pathway. The differential metabolites were mainly involved in arachidonic acid metabolism, Vitamin B6 metabolism, steroid hormone biosynthesis, and the bile secretion metabolic pathway. Moreover, ginsenoside Re could promote insulin secretion, stimulate cannabinoid type 1 receptor (CB 1 ) and CaMKK β to activate the AMPK signaling pathway, inhibit insulin resistance induced by TNF-α, and improve blood glucose uptake and diabetic nephropathy, so as to exert the goal of anti-diabetic.

Informed Consent Statement: Not applicable.
Data Availability Statement: All the data used in this study are available within this article. Further inquiries can be directed to the authors.