Dissecting Efficacy and Metabolic Characteristic Mechanism of Taxifolin on Renal Fibrosis by Multivariate Approach and Ultra-Performance Liquid Chromatography Coupled With Mass Spectrometry-Based Metabolomics Strategy

Taxifolin (TFN) is an important natural compound with antifibrotic activity; however, its pharmacological mechanism is not clear. In this study, our aim is to gain insight into the effects of TFN and its potential mechanisms in unilateral ureteral obstruction (UUO) animal model using metabolomics approach to identify the metabolic biomarkers and perturbed pathways. Serum metabolomics analysis by UPLC-Q-TOF/MS was carried out to discover the changes in the metabolic profile. It showed that TFN has a significant protective effect on UUO-induced renal fibrosis and a total of 32 potential biomarkers were identified and related to RF progression. Of note, 27 biomarkers were regulated by TFN treatment, which participate in eight metabolic pathways, including phenylalanine, tyrosine and tryptophan biosynthesis, and phenylalanine metabolism. It also showed that metabolomics was a promising strategy to better dissect metabolic characteristics and pharmacological mechanisms of natural compounds by multivariate approach and ultra-performance liquid chromatography coupled with mass spectrometry.


INTRODUCTION
Renal fibrosis (RF) is a common pathological characteristic of chronic kidney disease (CKD), which leads to final-stage renal disease (Hu et al., 2018;Yin et al., 2018;Kakitapalli et al., 2020;Rayego-Mateos and Valdivielso, 2020;Zeeh, 2020). It is generally recognized that RF is caused by activating various pathogenic factors such as inflammation reaction, injury, and drug stimulation (Zeisberg et al., 2002;Qin et al., 2011;Zhao et al., 2019). Then, the fibrogenic factors such as cytokines, growth factors, and chemotactic adhesion factors are released (Zeisberg and Kalluri, 2004;Liu, 2011;Vasko, 2016;Zhou et al., 2020). Up till now, renal biopsies and conventional biochemical detection are commonly applied to appraise the degree of RF. However, they are invasive, of high cost, and unstable and even have severe side-effects, which make accurate and repeated monitoring difficult in patients in the early stage (Jenkins et al., 2011;Chen et al., 2019). The common clinical treatment for RF, that is, to dilate the renal artery to increase the systemic blood perfusion, improve microcirculation disorder to enhance metabolism, alleviate the disorder of internal environment caused by hypoxia, and reduce toxic symptoms, is not ideal for patients (Boor et al., 2010;Zhang et al., 2018;Rayego-Mateos and Valdivielso, 2020).
Natural herbs characterized by multiple components, targets, and pathways have been widely utilized to cure various diseases for thousands of years (Li et al., 2020). Antifibrotic natural medicines with unique advantages have gained more and more attention for the treatment of RF (Shen et al., 2019;Sachan et al., 2020). Taxifolin (TFN), also called 3,5,7,3,4-pentahydroxy flavanone or dihydroquercetin, is a well-known natural flavonoid ingredient abundant in the Pinaceae tree family (Supplementary Figure S1) (Slimestad et al., 2007;Schauss et al., 2015;Topal et al., 2016). It possesses a wide range of biochemical and pharmacological properties in the management of oxidative stress, inflammation, tumors, microbial infections, and cardiovascular and liver disorders (Oi et al., 2012;Sun et al., 2014;Guo et al., 2015;Manigandan et al., 2015;Zhao et al., 2015;Gocer et al., 2016;Inoue et al., 2019;Zheng et al., 2019). Previous studies have proved that TFN exerts significant antioxidant effects that weaken cerebral ischemia-reperfusion injury by restraining oxidative enzymes and reactive oxygen species (ROS) production (Vladimirov et al., 2009;Voulgari et al., 2010). It enhances capillary microcirculation and antiplatelet aggregation and decreases the dose-dependent production of lipid-free radicals. In the transverse aortic constrictioninduced animal model, TFN eliminates the phosphorylation of Smad2 and Smad2/3 nuclear translocation and restrains the superfluous production of ROS, ERK1/2, and JNK1/2 to weaken left ventricular fibrosis and collagen synthesis (Guo et al., 2015). It also suppresses the cholesterol esterification, triacylglycerol and phospholipid synthesis, apolipoprotein B secretion, and microsomal triglyceride synthesis in liver cells (Casaschi et al., 2004). Recent studies revealed that TFN protects against RF by upregulating the intracellular Nrf2 level and promoting nuclear translocation of Nrf2, regulating redox metabolites, and preventing TGF-β1induced fibroblast activation and collagen synthesis (Wang et al., 2020). In addition, it cut back the concentrations of blood uric acid and creatinine and recovered the levels of caveolin-1/NF-κB signaling-related mRNA and proteins in diabetic nephropathy (Zhao et al., 2018). Unfortunately, the pharmacodynamics effect and potential molecule of TFN on RF have not been fully known owing to the limited scientific research data.
Currently, metabolomics based on high throughput and multivariate statistical analysis is generally applied in early diagnosis to discover the pathways associated with disease processes and drug treatment, disease classification, and prognosis (Zhang et al., 2013a;Zhang et al., 2013b;Zhang et al., 2013c;Wang et al., 2013;Zhang et al., 2014a;Zhang et al., 2014b;Wang et al., 2016;Li et al., 2017;Xie et al., 2017;Zhao et al., 2017;Sun et al., 2018a;Sun et al., 2018b;Li et al., 2018;Xie et al., 2019). Due to limited sensitivity and high data complexity, clear identification is usually limited to less than 100 metabolites Zia et al., 2019). Mass spectrometry (MS) combined with triple quadrupole instruments show exceptional sensitivity and specificity for the measurement of approximately 1,000 metabolite peaks. However, it is necessary to know the precursor ion and the product ion of each metabolite in advance (Sun et al., 2012;Zhang et al., 2016a;Zhang et al., 2017;Fang et al., 2019;Tokuoka et al., 2019). To date, most kidney metabolomics studies have applied NMR-or MSbased methods. For example, a GC/MS-based metabolomics found that the urine alkane-alpha,omega-diamine and alpha,omega-dicarboxylic acid were abnormally upregulated in UUO-induced RF rats primarily involved in amino acid and sugar metabolism (Fang et al., 2016). There is a powerful connection between renal tubule interstitial fibrosis and glycerophospholipid metabolism and L-carnitine metabolism in the development of chronic renal failure (CRF) using UPLC-QTOF/HDMS-based plasma lipidomic and metabolomic approaches, and rhubarb extracts ameliorate glycerophospholipid, fatty acid, and amino acid metabolisms in adenine-induced chronic tubule interstitial nephropathy animal (Zhang et al., 2016b). In the present study, we emphatically explored the anti-RF efficacy and the underlying mechanisms of TFN by metabolomics strategy based on UPLC-Q-TOF/MS to identify the perturbed metabolic biomarkers and pathways changes.  Figure S2. Interleukin-1β (IL-1β), tumor necrosis factor-α (TNF-α), superoxide dismutase (SOD), and malondialdehyde (MDA) enzyme-linked immunosorbent assay (ELISA) kits were bought from Jiancheng Bioengineering Institute (Nanjing, China). Serum creatinine (SCr) and blood urea nitrogen (BUN) ELISA kits were purchased from Longton Co., Ltd. (Shanghai, China). The antibody of transforming growth factor-β1 (TGF-β1) and small mothers against decapentaplegic (Smad-2) were purchased from Abcam (Cambridge, MA, United States) and used for immunohistochemical staining. The antibody of α-smooth muscle actin (α-SMA) and connective tissue growth factor (CTGF) were obtained from Beijing Laiyao Biological Technology Co., Ltd. (Beijing, China), and Beijing Century Aoke Biological Technology Co., Ltd. (Beijing, China), respectively. Primary antibodies against collagen type I, nuclear factor-kappaB (NF-KB), and fibronectin (FN) were produced by Cell Signaling Technology (Danvers, United States).

Ethics
Fifty male SD rats in specific pathogen-free-(SPF-) grade (8 weeks old, weighing 180-200 g) were obtained by the Laboratory Animal Center of Guilin Medical University. After one week of adaptive feeding for animals under temperature of 24 ± 1°C, humidity of 55 ± 10%, and 12 h light/dark cycle with free access to standard chow and water, all the rats were randomized and divided into five groups (n 10/group): sham operation group (control group), UUO rats group (model group), UUO rats treated with losartan group (UUO + LS), UUO rats treated with a high dose of TFN (UUO + TFN high), and UUO rats treated with a low dose of TFN (UUO + TFN low). All experimental procedures and animal care measures were executed in the light of the Guide for the Care and Use of Guilin Medical University.

Animal Model
UUO model establishment was performed as described in the literature (Kakitapalli et al., 2020;Zeeh, 2020). Then, they were placed on the test bench in the supine position and shaved locally. Skin disinfection was applied on the middle of the abdomen, and the kidney and left ureter were exposed and separated using blunt dissection after opening the abdomen. Nipping with a hemostat at the upper middle section, the left ureter was ligated twice using 4-0 silk thread at the ends and then cut and eliminated between the two ligatures. The incision was cleaned and closed by a suture layer. The surgical method for rats in the control group was the same as that of the operation group, in which the abdominal cavity was separated, but no tissue was ligated or cut. The UUOoperated rats were injected with penicillin into the muscle three times. UUO model establishment was performed as described in the literature (Kakitapalli et al., 2020;Zeeh, 2020). Rats in every group received an intraperitoneal injection of 5% chloral hydrate (0.35 ml/100 g). Then, they were placed on the test bench in the supine position and shaved locally. Skin disinfection was applied on the middle of the abdomen, kidney, and ureter.

Treatment
Before the surgery, UUO + TFN low and UUO + TFN high rat groups were given 8 and 16 mg·kg −1 , respectively, by oral administration. Losartan was given by oral gavage at 10 mg/kg-1 every day for rats in UUO + LS group. Rats in the control and model group received sterile saline solution in the same volume and way. The administration duration of the drug was set to twenty-eight days. At the same time, other experimental operations and animal care procedures were carried out.

Biochemical Indexes Detection
After TFN treatment for four weeks, rats in all groups were anesthetized with 10% chloral hydrate (4.0 ml/kg) on the 29th day. Blood samples were collected from the aorta abdominalis and transferred into the tube with heparin sodium. Blood samples were subsequently centrifuged at 3,500 rpm for 15 min at 4°C and the upper serum was collected for IL-1β, TNF-α, SOD, MDA, Scr, and BUN analysis. Olympus AU640 automatic biochemical analyzer was applied to disclose Scr and BUN levels. The blood of IL-1β, TNF-α, SOD, and MDA content was detected by ELISA according to the instruction in the kits. Mass system (Waters Corporation) was used for UPLC-Q-TOF/MS analysis. For UPLC-Q-TOF/MS detection, the serum needs to be further processed in addition to the above operations. Methanol in an ice-cold state was added to the serum sample to protein removal in the proportion of 3:1; then the mixture was centrifuged at 10000 rpm for 10 min at 4°C after mingling for 1 min using a mixer mill (Retsch GmbH & Co., Haan, Germany). The obtained liquid supernatant was dried under nitrogen and reconstituted with 200 μL methyl alcohol. Quality control samples (QCs) were prepared using a mixture of 10 μL plasma samples obtained from each sample in order to examine the stability of the instrument and optimize the analytical method. A QC filter was used to select and get rid of any ions with a coefficient of variation >15% during the analysis run.

Histological Examination
Renal tissues were immediately harvested and superficial connective tissue was removed after washing with physiological saline. At least eight randomly histologic sections of the kidney in each group were chosen and stained with hematoxylin-eosin (H&E) for histological assessment of UUO model rats before and after TFN treatment. Renal tissues were fixed in 4% formaldehyde for 24 h, dehydrated with gradient alcohol, processed with xylene, paraffin-embedded, and sectioned at 5 µm thickness. The slices were baked and orderly placed in xylene dewaxing, underwent gradient alcohol dehydration, hematoxylin staining at 60°C for 10 min, 0.5% ammonia for 30 s, eosin staining for 10 min, and immersed in 80, 90, 95, and 100% alcohol for 1 min and transparent xylene for 5 min. Then, all the operated samples were observed under an optical microscope.
buffer containing protease inhibitor and benzamidepolyfluoride for homogenization. The supernatant was obtained as the total protein extract of kidney tissue in light of the instructions of the protein concentration detection kit. The obtained proteins underwent SDS-PAGE electrophoresis separation in the conventional operation method and then transferred to PVDF membrane. After incubating 5% nonfat milk powder at room temperature for 2 h, the protein samples were added to primary antibody working solution, reacted at 4°C for 24 h, and washed 5 times at 10 min/time. Then, a secondary antibody working solution was addedto the sample and incubated for 1 h at room temperature; the membrane was washed 5 times with TBS-T at 10 min/time.

Metabolomics Analysis
Detailed parameters of chromatographic separation and mass spectrometry were depicted as follows: the metabolomics profiling analysis was performed on a Waters BEH C18 (2.1 × 100 mm, 1.7 μm) using 0.05% (v/v) formic acid water solution as mobile phase A and 0.05% (v/v) formic acid acetonitrile solution as mobile phase B running with a gradient program of 5-30% B in 0-3.5 min, 30-65% B in 3.5-6 min, 65-85% B in 6-9 min, 85-95% in 9-10 min, 95% B in 9-11 min, and 2 min of balance back to 5% B; the flow rate was set constant at 0.4 ml/ min and the column temperature was maintained at 35°C for all samples; the injection volume was set to 2 μL; under ESI+ and ESI-ion scanning mode, capillary voltages were, respectively, 3,500 and 3000 V; nebulizer pressure was controlled at 32 psi; nozzle voltage was set at 250 V; drying gas temperature was set to 350°C at 20 L/min; sheath gas was set to 400°C at 16 L/min. The acquisition mass range was from m/z 50 to m/z 1,500 in fullscan mode. The precision and reproducibility were analyzed for evaluating the above-mentioned developed UPLC-Q-TOF/MS method. Raw data were processed using the Progenesis QI data analysis software (Version 2.0, Nonlinear Dynamics, Newcastle, United Kingdom), which permits deconvolution, alignment, and data reduction in order to export a list of mass, retention time, m/z, and corresponding intensities for all the detected peaks from each sample. The principal parameters were controlled as follows: retention time range, 1-11 min for metabolomic analysis; mass range, 50-1,500 m/z; mass tolerance, 0.01; minimum intensity, 1%; mass window, 0.05; retention time window, 0.20; noise elimination level, 6.
Multivariate analysis, such as principal component analysis (PCA), orthogonal partial least squares discriminant analysis (OPLS-DA), and variable weight value (VIP) plot, was carried out using SIMCA software (version 14.0, Umetrics AB, Umeå, Sweden). PCA score plot of spectral data has the ability to visualize overall clustering, appraise the main sources of variation, and remove outliers beyond the confidence interval (95%). For the purpose of filtering endogenous metabolites that play a vital role in metabolic profile, the OPLS-DA model performed one hundred permutation validations to assess the fitting of the discriminant analysis.
In the VIP scatter plot that highlight the variable weight value of the relevant reactive ion contribution degree, the ion fragments from the bottom with small VIP value have a smaller contribution to differential metabolism. On the contrary, the top ions in V-shape with larger VIP value provide more contribution to the metabolic profile trajectory difference among groups. Meanwhile, the ions that keep away from the original point in the loading plot obtained from OPLS-DA analysis seem as potential metabolites. Differential metabolites meet the conditions as follows: VIP value more than 1.5 and calculated p value in both Student's t-test and Mann-Whitney test less than 0.05.
The identification of metabolites was firstly performed to retrieve by available biochemical databases, such as HMDB, KEGG, and Chemspider, and subsequently was verified by MS data, MSE fragments, molecular weights, and chemical element compositions. The topological trait of metabolic pathways related to the RF model and anti-RF efficacy of TFN was described by analysis software (MetPA) in MetaboAnalyst 4.0 combined with various advanced path analysis programs to highlight the role of perturbed metabolic biomarkers and pathways in the biological system.
All the acquired data were confirmed by the Kolmogorov-Smirnov test and were found to meet the peculiarity of the normal distribution. The results of conventional efficacy evaluation and metabolomics analysis were carried out in SPSS (version 22.0, SPSS Inc., Chicago, IL, United States), which were expressed as mean ± standard deviation (SD). The differences of mean values among groups were tested by Student's t-test or one-way ANOVA followed by Tukey's post hoc test. p values <0.05 were deemed to be statistically significant, and p values <0.01 were deemed to be more statistically significant.

Histological Changes Analysis
To appraise the changes of tissue fibrosis before and after TFN treatment, high-power field optical microscope was used for each section from different groups. Pathological results showed that the kidney tissue of the control group presented clear glomerular and renal tubular structures, and there was no inflammatory cell infiltration in the renal interstitium, no fibrous tissue hyperplasia, and no mesangial cell or stromal tissue hyperplasia (Supplementary Figure S3). In the model group, glomerular stromal tissue hyperplasia, tubular epithelial cell swelling, vacuolar degeneration, focal necrosis, flaky atrophy, and even epithelial cell necrosis and shedding can be seen in the tubular lumen. Renal interstitial inflammatory cells are diffusely infiltrated, leading to fibrous tissue proliferation and even sheet fibrosis. After 8 and 16 mg·kg −1 of TFN treatment, the UUOrelated histopathologies were moderated, and the pathological results were similar to those of rats under sham operation. There were less inflammatory cell infiltration, neatly arranged renal tubules, and a small interstitial fibrosis area.

Biochemical Indexes
The changes of IL-1β, TNF-α, SOD, MDA, Scr, and BUN content after twenty-eight days of TFN administration are shown in  Figures 2A-G, TGF-β1, Smad-2, α-SMA, CTGF, collagen type I, NF-KB, and FN contents in kidney tissue were upregulated when compared with the normal renal tissues. Compared with the model group, TFN could decrease TGF-β1, Smad-2, α-SMA, CTGF, collagen type I, NF-KB, and FN levels in a dose-dependent manner.

Metabolic Profile Changes
In this research, serum samples from five groups were searched in ESI+ and ESI− ion mode under the UPLC-Q-TOF/MS system, in which typical BPI metabolic profiles of control and model group in Figures 3A,4A were obtained by SIMCA V14.0 software. It was not hard to see that metabolic profiling of serum is similar and there were only distinct content differences in mapping. For further seeking endogenous differentiated metabolites to assess UUO-induced RF animal model and TFN efficacy, a nontargeted metabolomics strategy was carried out for the multivariate data analysis. 2D PCA score plots of both ion modes were shown in Figures 3B,4B, in which each spot represented a sample, and the control groups were clearly clustered and separated to the model group, indicating that the RF rat model has been successfully established at serum metabolism level. From Figures 3C,4C, the parameters R2X, R2Y, and Q2 obtained by cross-validation were, respectively, 0.923, 0.8996, and 0.642 in positive mode, and those parameters were, respectively, 0.931, 0.908, and 0.710 in negative mode, indicating the data model possesses good prediction ability and reliability. In addition, the 2D OPLS-DA score plot showed a notable separation. The loading plot (Figures 3E,4E) and VIP plot diagrams of OPLS-DA ( Figures 3D,4D) were generated to know the contribution rate between different groups. The further away they were from the origin, the greater the contribution was to the clustering of the control group and model group.
Frontiers in Pharmacology | www.frontiersin.org January 2021 | Volume 11 | Article 608511 dodecanoic acid, and dityrosine. Sixteen metabolic pathways with impact value more than zero were closely related to pathogenesis and development of model rats in serum metabolism level, including PA, tyrosine and tryptophan biosynthesis, PA metabolism, arachidonic acid metabolism, sphingolipid metabolism, terpenoid backbone biosynthesis, citrate cycle (TCA cycle), alanine, aspartate and glutamate metabolism, arginine and proline metabolism, arginine biosynthesis, pyrimidine metabolism, glutathione metabolism, tryptophan metabolism, glyoxylate and dicarboxylate metabolism, cysteine and methionine metabolism, purine metabolism, and glycerophospholipid metabolism, as shown in Supplementary Figure S4.

TFN Effect on Perturbed Biomarker
From typical BPI metabolic profiles of UUO + TFN and UUO + LS group in Figures 5A-C, it was shown that the peaks in metabolic profiling exist in content differences. PCA was performed on the serum metabolic profile of five groups of rats, which were generally clustered together in an individual group with similarity, and a clear separation between groups was detected, suggesting that these five groups were differential. As shown in Figures 5D,6E, the UUO + TFN group and UUO + LS located between the control and model group present a similar changing trend, where UUO + TFN high group is closer to the control group than UUO + TFN low group, manifesting TFN plays a vital intervention role in reversing the metabolic profile of RF model animals to make them in a healthy state in dosedependent manner. Compared with the model group, 17 metabolite level expressions in the UUO + TFN group were significantly increased, such as isocitric acid, 3-hydroxyanthranilic acid, picolinic acid, citric acid, asparagine, tryptophan, glutamine, SM (d18:1/22:0), kynurenic acid, cyclic GMP, 20-hydroxyeicosatetraenoic acid, deoxyuridine, LysoPC (17:0), LysoPC (15:0), galabiosylceramide, oleic acid, and sphinganine, whereas the levels of 10 metabolites were significantly decreased such as ornithine, uric acid, mevalonic acid-5P, hydroxytyrosol, prostaglandin F2a, PA, arachidonic acid, palmitoleic acid, SM (D18:0/16:1), and LysoPC (16:1 (9Z)). Among them, UUO + TFN low group can affect the Frontiers in Pharmacology | www.frontiersin.org January 2021 | Volume 11 | Article 608511 content of 16 metabolites, whereas UUO + TFN high group can regulate 27 metabolites. In Figure 6A, the brightness difference of color in the heatmap reveals the relative content changes of 27 potential metabolites in five groups to highlight the TFN pharmacological activity during treatment. The relative peak areas of the above-mentioned metabolites were shown in a bar graph from Figure 6B. TFN in low dosage callback serum biomarkers mainly involved arachidonic acid metabolism, sphingolipid metabolism, citrate cycle (TCA cycle), alanine, aspartate and glutamate metabolism, and arginine and proline metabolism with impact value greater than 0.1 in Supplementary Figure S3A, and TFN in high dosage regulated serum biomarkers such as PA, tyrosine and tryptophan biosynthesis, PA metabolism, arachidonic acid metabolism, sphingolipid metabolism, terpenoid backbone biosynthesis, citrate cycle (TCA cycle), alanine, aspartate and glutamate metabolism, and arginine and proline metabolism with impact value greater than 0.1 in Supplementary Figure S3B. It was indicated that TFN plays a role in preventing and treating RF by interfering with the above metabolic pathways. The information of single-nucleotide polymorphisms (SNPs) loci and dysfunctional enzymes were detected by genome-scale network model of human metabolism in Figure 7.

DISCUSSION
Metabolomics research has begun to outline the changes in metabolites from blood and urine at different stages of nephropathy in order to provide insights into nephropathy at the molecular level. However, some current challenges limit the interpretation of modern research (Schrimpe-Rutledge et al., 2016). In particular, metabolomics cannot fully cover the cognition of metabolites on different platforms and there is a crisis of reproducibility in the study of clinical biomarkers (Casadei et al., 2018). These limitations may be resolved through continuous developments in MS sensitivity and mass accuracy and strength to add currently unknown m/z Frontiers in Pharmacology | www.frontiersin.org January 2021 | Volume 11 | Article 608511 peak identities and standardize reagents and terminology (Gowda and Djukovic, 2014). The combined dataset will provide enhanced statistical capabilities for integrating metabolomics data with genomics and other functional genomics outputs (Telenti, 2018). In turn, these achievements will provide insight into the genetic determinants of selected metabolites alteration and whether metabolite markers of kidney disease have causality or association relationship with the emphasized pathways. The epidemiological scale should be combined with physiological and experimental research to provide more direct insights into organ specificity and the underlying mechanisms of certain metabolite changes (Wu et al., 2016;Lai et al., 2018).
The results of HE staining in this study showed that RF rats displayed glomerular hyperplasia in the glomeruli, swollen tubule epithelial cells, tubule vacuoles degeneration, necrosis, atrophy, and infiltration of renal interstitial inflammatory cells. In the control group, renal tubular epithelial cells and renal interstitium were generally normal. The above morphological studies are consistent with relevant literature reports (Chen and Li, 2018), indicating that the animal model replication method used in this experiment is feasible and successful. Administration of TFN for 4 weeks can improve the infiltration of inflammatory cells and the swelling of renal tubular epithelial cells and reduce the area of interstitial fibrosis. Tissue staining shows that the renal tubular structure is clear and there are no degeneration and necrotic cells in the lumen; in addition to the above changes, the epithelial cells of the tube wall are regularly arranged without atrophy, the renal interstitial structure is clear, and there is no fibrosis. It also shows that the antirenal fibrosis of TFN is mainly manifested in renal tubular epithelial cells. The study found that TNF-α can directly induce apoptosis of renal tubular epithelial cells and simultaneously upregulate the expression of IL-1β, which is consistent with the results of this experiment; that is, the content of TNF-α and IL-1β in the serum of the model group was increasing (Donnahoo et al., 1999). SOD that can block and resist the damage of oxygen free radicals to renal tubular epithelial cells and may repair damaged renal tubular epithelial cells is the most important substance in the body to scavenge oxygen free radicals. It is an intuitive way to observe cell damage and repair index (Son et al., 2008). Our results showed that TFN could effectively reduce the TNF-α and IL-1β activity and increase the SOD activity of rats caused by UUO for enhancing the antioxidant capacity, repairing damaged renal tubular epithelial cells to alleviate the process of renal interstitial fibrosis. MDA level can indirectly reflect the degree of damage to renal tubular epithelial cells and is a cytotoxic substance produced during the peroxidation reaction (Nomani et al., 2018). Scr and BUN are two important indicators for evaluating renal function, which both are excreted by the kidney. Creatinine is the end product of creatine metabolism in muscle tissue, and urea nitrogen is the end product of protein metabolism (Hosten, 1990). Scr and BUN reflect the ability of the kidney to clear creatinine from the blood and concentrate it in urine. This study found that the Scr and BUN contents of the model group were upregulated. Compared with the model group, MDA, Scr, and BUN levels were downregulated after 8 and 16 mg·kg −1 TFN in gavage way, indicating that TFN can ameliorate UUO-induced tubular epithelial damage. TGF-β 1 is a key renal interstitial fibrosis-promoting factor, and TGF-β 1 is one of the strongest known fibrotic cytokines (López-Hernández and López-Novoa, 2012). In the TGF-β 1 signal transduction pathway, the Smad-2 pathway is recognized as one of the most important downstream pathways and the only known intracellular kinase receptor substrate. TGF-β 1/ Smad-2 signaling pathway is the core pathway of RF. Through experiments, compared with the control group, the expressions of TGF-β 1 and Smad-2 in the rat kidney were increased in the model group, indicating that the TGF-β1/Smad-2 signal transduction pathway is involved in UUO-induced renal interstitial fibrosis. TGF-β1 promotes renal interstitial fibrosis to destroy the basement membrane, then enhances the expression of α-SMA and other myofibroblast markers, and promotes the synthesis of collagen types I and III in the extracellular matrix . CTGF as a newly discovered fibrogenic factor can be produced by interstitial cells such as fibroblasts under the action of TGF-β1 and participates in the effect of TGF-β1 on interstitial cells to promote cell proliferation and extracellular matrix synthesis to accelerate the process of tissue fibrosis. In resting cells, NF-κB and IκB form a complex that does not play a role in the cytoplasm. When cells are stimulated by extracellular inflammatory signals, the inhibitor of NF-κB dimer k (IκBk) is inactivated, leading to phosphorylation of IκB, and the nuclear localization site of NF-κB is exposed. The rapid translocation of κB and specific IκB induces target gene transcription, promotes the proliferation of target cells, such as TNF-α and ICAM-1, and triggers renal damage (Oh et al., 2014;Suh et al., 2015). TGF-β1 can also induce NF-κB-mediated renal inflammation. FN, as a glycoprotein with a relative molecular mass of 250,000, interacts with a variety of matrix proteins and regulates a variety of cellular processes (Stevens et al., 2007;Xie et al., 2012). NF-κB activation can regulate the overexpression of adhesion molecules ICAM-1, FN, and other ECM components, leading to continuous inflammation and kidney damage. Different concentrations of TFN can reduce the expression of TGF-β1, Smad-2, α-SMA, CTGF, I collagen, NF-KB, and FN. Fatty acid biosynthetic pathways include palmitic acid, oleic acid, and arachidonic acid. Oleic acid is a monounsaturated w-9 fatty acid, which is the most widely distributed fatty acid with the highest fat Frontiers in Pharmacology | www.frontiersin.org January 2021 | Volume 11 | Article 608511 content in nature. Interstitial cell apoptosis can be induced by the production of ceramide. Studies have shown that palmitic acid can induce apoptosis of renal tubular epithelial cells, upregulate cPLA2, produce free fatty acids and lysophosphatidylcholine and other biologically active components such as arachidonic acid, and participate in the occurrence and development of tissue fibrosis (Mu et al., 2001;Allison, 2015). Arachidonic acid can be hydrolyzed to release and generate various active substances, such as 20hydroxyeicosatetraenoic acid. As the lead compound synthesized by prostaglandin F2a, arachidonic acid metabolism plays an important role in the inflammatory reaction process and is related to kidney disease. It can be used for platelet depolymerization and vasodilation, reducing hypoxia in kidney tissue, improving the body's ability to resist hypoxia, scavenging free radicals, improving hemodynamics, regulating hyperlipidemia, and improving hypercoagulability (Skibba et al., 2017). PCs are also important metabolites in lipid metabolism related to the pathogenesis of RF. In this study, two types of lipid metabolism were discovered, that is, glycerophospholipid and sphingolipid metabolism (Ferro et al., 2018). The decreased LysoPC (17:0) and LysoPC (15:0) that could induce RF by causing disorder of glycerophospholipid metabolism, along with the increased LysoPC (16:1 (9Z)), were observed in the model group. The elevated levels of SM (d18:1/16:0) and decreased sphinganine and galabiosylceramide were observed in the model group. Sphingosine can promote cell growth, adhesion, migration, and death. SM (d18: 1/16: 0) has the activity of hydrolysis to ceramide, which significantly promotes the formation of tissue damage and fibrosis (Dincer et al., 2019;Gai et al., 2019). Citric acid and isocitric acid are intermediates of the tricarboxylic acid cycle (Biasioli et al., 1987). According to these reports, citric acid can significantly shorten the recovery time of urine and stay in the ICU, improve renal function indicators, blood biochemical indicators, and inflammation indicators, maintain the stability of the internal environment, and reduce the risk of bleeding. This study found that the content of citric acid in the model group was decreased, resulting in a decrease in the synthesis of its downstream product isocitrate. After TFN treatment, the content of the two has been called back. Glutamine, as a coding amino acid in protein synthesis is a nonessential amino acids in mammals, can be converted from glucose in the body. It is the most abundant amino acid in the human body and participates in more metabolic processes than any other amino acid. Glutamine reduces kidney damage associated with renal ischemia/reperfusion nitrosation and oxidative stress, attenuates the decrease in Cox-2 expression after I/R, and prevents the increase in AT-1 expression. The reduction of glutamine content in model rats is related to decreased immune system function, imbalance of metabolic nitrogen, blocked protein synthesis, and release of inflammatory factors in the pathogenesis of renal fibrosis (Alba-Loureiro et al., 2010;Li et al., 2019). The abnormal levels of citric acid, isocitric acid, and glutamine lead to the disturbed citrate cycle (TCA cycle), alanine, aspartate, glutamate metabolism, and glyoxylate and dicarboxylate metabolism in RF animals. Ornithine can be used for nutritional supplementation, acute and chronic liver diseases such as cirrhosis, fatty liver, hyperammonemia caused by hepatitis, and central nervous system symptoms. The ornithine cycle converts the more toxic ammonia produced by protein metabolism in the body into the less toxic urea, which is excreted from the body (Montaguth et al., 2019). Cyclic GMP, an important inhibitor of RF synthesized by guanylate cyclase stimulated by nitric oxide or natriuretic peptide, has pleiotropic regulatory functions in the kidney (Lieb et al., 2009). TFN could downregulate the serum content of ornithine and uric acid and upregulate the serum content of glutamine, picolinic acid, and cyclic GMP by adjusting arginine and proline metabolism, arginine biosynthesis, and purine metabolism. Compared with the UUO + TFN low group, UUO + TFN high group can further affect five metabolic pathways, including PA, tyrosine and tryptophan biosynthesis, PA metabolism, terpenoid backbone biosynthesis, pyrimidine metabolism, and tryptophan metabolism to achieve RF treatment. PA, an essential amino acid in the human body, is involved in the formation of various protein components but cannot be synthesized in the human body. Under normal circumstances, about 50% of the PA consumed is used to synthesize proteins of various components, and the rest is converted to tyrosine under the action of PA hydroxylase and then converted into dopamine, epinephrine, norepinephrine, and melanin. When PA hydroxylase is lacking, these metabolites reach abnormally high levels and accumulate in tissues, plasma, and cerebrospinal fluid, which are excreted in large quantities from the urine (Alkaitis and Ackerman, 2016). PA level in blood sample was decreased after TFN treatment involved in PA, tyrosine, and tryptophan biosynthesis and PA metabolism. 3-Hydroxyanthranilic acid (3-HAA) is a tryptophan metabolite with anti-inflammatory activity, in which the immunoregulatory molecular mechanism of 3-HAA on macrophages is inhibiting the production of inflammatory mediators and reducing NF-κB activity. The results show that 3-HAA has an immunomodulatory effect, which may be due to the inhibition of PI3K/Akt/mTOR and NF-κB activation, thereby reducing the production of proinflammatory mediators in tryptophan metabolism (Krause et al., 2011). The activation of the MAP kinase pathway and MC proliferation by mevalonic acid depletion and might have protective effects by inhibiting IGF-1-mediated MC proliferation. TFN regulates terpenoid backbone biosynthesis activity in RF rats to lower the level of mevalonic acid-5P (Shibata et al., 2009). Disturbances of cerebral purine and pyrimidine metabolism have existed in young children with chronic renal failure (Gerrits et al., 1991). In our present study, serum levels of glutamine and deoxyuridine were remarkably increased in the RF model group. It could be concluded that TFN could reregulate the expression of the above metabolites and perturbed pathways, suggesting the renal-protective effects on UUO-induced RF rats.

CONCLUSION
In this study, our work indicated that TFN possesses an important protective effect in an UUO-induced rat RF model by relieving the perturbed level of 27 metabolites associated with the vital pathways. The therapeutic effects of TFN were confirmed in UUO-induced animal model, which was linked with the delayed pathological development and reversal of the perturbed metabolic biomarkers and pathways at the molecule level. Metabolomics, a valuable and promising strategy, is conducive to better understand natural product pretreatment mechanism facing disease and provide novel thought to develop a therapeutic agent in RF. The study is conducive to us for understanding the pathogenesis of RF and bringing about an emerging potential natural antifibrosis agent.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.