Effects of Guava (Psidium guajava L.) Leaf Extract on the Metabolomics of Serum and Feces in Weaned Piglets Challenged by Escherichia coli

The effects of dietary supplementation with guava leaf extracts (GE) on intestinal barrier function and serum and fecal metabolome in weaned piglets challenged by enterotoxigenic Escherichia coli (ETEC) were investigated. In total, 50 weaned piglets (Duroc × Yorkshire × Landrace) from 25 pens (two piglets per pen) were randomly divided into five groups: BC (blank control), NC (negative control), S50 (supplemented with 50 mg kg−1 diet GE), S100 (100 mg kg−1 diet GE), and S200 (200 mg kg−1 diet GE), respectively. On day 4, all groups (except BC) were orally challenged with enterotoxigenic ETEC at a dose of 1.0 × 109 colony-forming units (CFUs). After treatment for 28 days, intestinal barrier function and parallel serum and fecal metabolomics analysis were carried out. Results suggested that dietary supplementation with GE (50–200 mg kg−1) increased protein expression of intestinal tight junction proteins (ZO-1, occludin, claudin-1) (p < 0.05) and Na+/H+ exchanger 3 (NHE3) (p < 0.05). Moreover, dietary supplementation with GE (50–200 mg kg−1) increased the level of tetrahydrofolic acid (THF) and reversed the higher level of nicotinamide-adenine dinucleotide phosphate (NADP) induced by ETEC in serum compared with the NC group (p < 0.05), and enhanced the antioxidant capacity of piglets. In addition, dietary addition with GE (100 mg kg−1) reversed the lower level of L-pipecolic acid induced by ETEC in feces compared with the NC group (p < 0.05) and decreased the oxidative stress of piglets. Collectively, dietary supplementation with GE exhibited a positive effect on improving intestinal barrier function. It can reprogram energy metabolism through similar or dissimilar metabolic pathways and finally enhance the antioxidant ability of piglets challenged by ETEC.


INTRODUCTION
Weaned piglets infected with enterotoxigenic Escherichia coli (ETEC) may cause post-weaning diarrhea, which leads to growth retardation and damage to the innate and adaptive immune systems of piglets. These risk factors increase the morbidity and mortality of piglets and result in large economic losses in the swine industry worldwide (1,2). Regarding the mechanisms of ETEC infectious diarrhea, it has been demonstrated that ETEC can produce colonization factors (CFs) and enterotoxins that adhere to the intestinal mucosa of piglets, and this action inhibits intestinal immune function, perturbs hydro-electrolytic secretions in the intestine, and results in the occurrence of diarrhea (3).
As a consequence of this, veterinary antibiotics have been commonly used to treat intestinal infections for improving animal growth and health in several decades. However, concerns about antimicrobial resistance, residue accumulation in animal products, and environmental pollution have led to a limited application of antibiotics as growth promoters (4,5). Due to these factors, searching for alternatives to antibiotic growth promoters, such as pro-and prebiotics, organic acids, enzymes, and plant extracts, have attracted more and more attention (5)(6)(7). Among the candidate alternatives to antibiotics, plant extracts appear to be one of the most widely accepted (8,9).
Guava (Psidium guajava L.) is a tropical fruit and medicinal plant, which is mainly distributed in the tropical and subtropical areas. Guava leaf extract (GE), known as an herbal medicine for the treatment of respiratory and gastrointestinal diseases (10), is reported to contain phenolics, triterpenoids, and other compounds that have antibacterial, antioxidant, and antiinflammatory activities (11)(12)(13). Pruning usually is used to stimulate growth and influence fruiting in guava (14); thus, residual guava leaves from pruned processing are promising sources of natural feed additives, which may be utilized as a potential alternative for in-feed antibiotics.

Preparation of Guava Leaf Extract
Fresh guava leaves were collected from the guava plantations in Qionghai City of Hainan Province, China, during the pruning period in July 2018, and fresh leaves were dried (60 • C, 24 h) and powdered. The powder (50 kg) was exhaustively extracted with 95% ethanol three times at room temperature and then filtered. The solvent was evaporated under reduced pressure using a rotary vacuum evaporator to afford GE (3.47 kg), which was stored at 4 • C for an animal experiment.

Feeding Trial and Experimental Design
The current feeding trial is from our previous published study (17). Fifty 21 ± 3 day-old crossbred weaned piglets (Duroc × Yorkshire × Landrace, 7.35 ± 0.18 kg) were selected and transported from the piggery to the barn, where they were randomly allotted to five groups of five replicate pens per group (two piglets per pen). The five groups were as follows: (1) blank control group (BC), piglets were fed diet without supplements and ETEC challenge; (2) negative control group (NC), piglets were fed diet without supplements and challenged by ETEC; (3) S50 group (S50), piglets were fed diet supplemented with 50 mg kg −1 GE and challenged by ETEC; (4) S100 group (S100), piglets were fed diet supplemented with 100 mg kg −1 GE and challenged by ETEC; (5) S200 group (S200), piglets were fed diet supplemented with 200 mg kg −1 GE and challenged by ETEC. The diet was formulated to meet the nutrient recommendations of the National Research Council (2012). The ingredient and nutrient composition of basal diet were presented in Supplementary Table 1.
Feed and water were available ad libitum during the 28-day experimental period. All piglets were housed in a weaner facility temperature maintained at 25 ± 0.5 • C, with 12 h of light and dark. On day 4, all piglets (except BC) were orally challenged with about 1.0 × 10 9 colony-forming units (CFUs) of ETEC according to the method developed by Wu et al. (20). ETEC was obtained from the China Veterinary Culture Collection Center (Beijing, China). The occurrence of diarrhea during the whole experiment for each group was observed.

Sample Collection
On day 29, one pig was randomly selected from each pen, and the blood samples were collected from the jugular vein, and serum was prepared by centrifugation at 700 × g for 15 min at 4 • C and stored at −80 • C until metabolomics analysis. After sampling, all piglets were anesthetized by an intraperitoneal injection of 50 mg kg −1 pentobarbital sodium and were killed by exsanguination. Fecal samples were collected directly in 10ml sterile plastic tubes from the rectum of piglets and stored at −80 • C until analysis. The small intestine was removed, and a piece (4-cm length) of the middle jejunum was collected, gently rinsed with 0.1 M phosphate-buffered saline (PBS) at pH 7.2, and then fixed in 10% formaldehyde-phosphate buffer for subsequent immunohistochemical analysis.

Serum and Fecal Sample Preparation and Analysis by Ultra-High-Performance Liquid Chromatography Coupled With a Hybrid Quadrupole Time-of-Flight Mass Spectrometry
Serum and fecal samples were extracted prior to analysis by ultra-high-performance liquid chromatography coupled with a hybrid quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS) in positive ionization mode. Serum samples (100 µl) were prepared via one-step protein precipitation with 400 µl of methanol (TEDIA, Fairfield, USA). The samples were left at −80 • C for 6 h after a 2-min vortex. After that, the samples were centrifuged at 20,000 × g for 10 min at 4 • C. Then 400 µl of supernatant was transferred into an Eppendorf Tube, and the supernatant was concentrated in a vacuum centrifugal concentrator for 1 h using SPD121P SpeedVac Concentrator (Thermo Fisher, Germany), then reconstituted with 100 µl of acetonitrile (ACN) (MERCK, Darmstadt, Germany) for UHPLC-QTOF-MS analysis as described previously (22).
For extraction of the fecal samples, fecal samples (100 mg) were prepared via protein precipitation with 500 µl of methanol. Then the samples were vigorously vortexed for 5 min and centrifuged at 20,000 × g at 4 • C for 10 min. After the supernatant was collected, the residue again was extracted according to the above extraction procedure and combined with the previous supernatant. At last, 200 µl of supernatant was transferred into the sampling vial for UHPLC-QTOF-MS analysis as described previously (23).
Drying gas temperature and the ion spray voltage were set at 550 • C and 5,000 V, respectively. Atomization gas pressure, auxiliary heating gas pressure, and curtain gas pressure were set at 45, 45, and 35 psi, respectively. The instrument was mass calibrated by automatic calibration infusing the Sciex Positive Calibration Solution (Sciex, Foster City, CA, USA) after every sixsample injections. One quality control sample and one blank vial were run after each cycle of 10 samples.
Automatic peak extraction, peak matching, peak alignment, and normalization preprocessing on the acquired data were performed using MarkerView software (Sciex, USA). The retention time and m/z tolerances were 0.1 min and 10 ppm, respectively; the response threshold was 100 counts and the isotope peak was removed. After Pareto scaling, principal component analysis (PCA), and partial leastsquares discrimination analysis (PLS-DA) models were carried out to visualize the metabolic difference among BC, NC, S50, S100, and S200 groups. The quality of the models was described using R 2 X and R 2 Y. To avoid model overfitting, 999 cross-validations in SIMCA-P 13.0 were performed throughout to determine the optimal number of principal components. R 2 X, R 2 Y, and Q 2 Y values of models were nearly 1.0, indicating that these models retain the ability to explain and predict variations in the X and Y matrix.
Furthermore, the value of fold change (FC) was calculated as the average normalized peak intensity ratio between the two groups. Differences between data sets with FC > 1.10 or and p < 0.05 (Student t-test) were considered statistically significant. The structural identification of differential metabolites was performed by matching the mass spectra with an in-house metabolite library, including accuracy mass, retention time, MS/MS spectra, and online databases Metlin (http://www.metlin.scripps.edu) and HMDB (http://www.hmdb.ca).
The impact of ETEC and GE on metabolic pathways was evaluated based on the MetaboAnalyst platform, a tool for metabolomics data analysis platform, which is available online (https://www.metaboanalyst.ca). The pathway analysis module combines results from powerful pathway enrichment analysis with pathway, topology analysis to help researchers identify the most relevant pathways involved in the conditions being investigated. The analysis report was then presented graphically as well as in a detailed table. Potential biomarkers for GE efficacy were identified based on the metabolic pathway enrichment and statistics analysis.

Statistical Analysis
Statistical analysis of the integral optical density among the groups was evaluated by using the one-way analysis of variance (ANOVA), performed using SPSS 23.0 (IBM-SPSS Inc., Chicago, USA). The results were presented as mean ± standard error of mean (SEM). Orthogonal polynomial contrasts were used to test for linear and quadratic effects of GE by comparing with the NC group. Significant differences and extremely significant differences were evaluated by Tukey multiple comparisons test at p < 0.05 and p < 0.01, respectively.

Immunohistochemistry
As seen in Figure 1 and Table 1, the color signals and the integral optical density of occludin and claudin-1 in the S100 and S200 groups were significantly higher than in the NC group (p < 0.05), and NHE3 in the GE groups (GEs) was significantly higher than that of the NC group (p < 0.05). Supplementation of GE in the diet linearly and quadratic increased the integral optical density of claudin-1 (p < 0.01), occludin (p < 0.01), and NHE3 (p < 0.01) compared with the NC group.

Analysis of Fecal Metabolomics
Using the optimal UHPLC-QTOF-MS condition described above, the representative total ion chromatograms (TICs) for fecal samples are presented in Supplementary Figure 1A. The score plots of PCA overlapped partly both in the direction of PC1 and PC2 based on the fecal samples from the NC vs. BC group ( Figure 2A1), S50 vs. NC group (Figure 2B1), S100 vs. NC group (Figure 2C1), and S200 vs. NC group (Figure 2D1). Supervised PLS-DA analysis suggested that there were significant differences between two groups, indicating the distinct metabolic profiling of the NC vs. BC group (Figure 2A2), S50 vs. NC group (Figure 2B2), S100 vs. NC group (Figure 2C2), and S200 vs. NC group ( Figure 2D2).
Metabolic profiling in the feces was significantly changed based on the results of the NC vs. BC group, S50 vs. NC group, S100 vs. NC group, and S200 vs. NC group, respectively (Supplementary Table 2). L-pipecolic acid is a unique differential metabolite between the NC group and BC group. Different metabolic pathways were enriched from the groups of S50 vs. NC, S100 vs. NC, and S200 vs. NC, respectively (Supplementary Figure 2). The details of the top 4 ranked metabolic pathways and relevant differential metabolites between the BC, NC, and GEs groups are presented in Table 2. It shows that the S50 group significantly boosted the production of 3methoxytyramine (p < 0.05) and decreased the production of epinephrine, normetanephrine, N-acetylserotonin, melatonin, and caffeine (p < 0.05) compared with the NC group. The S100 group significantly upregulated the levels of biliverdin and Lpipecolic acid (p < 0.05), and downregulated the levels of 5aminolevulinic acid and phosphorylcholine (p < 0.05) compared with the NC group. Moreover, the S200 group significantly downregulated the levels of uridine 5 ′ -monophosphate (UMP), deoxycytidine monophosphate (dCMP), deoxyguanosine, and L-phenylalanine (p < 0.05) compared with the NC group.

Analysis of Serum Metabolomics
The UHPLC-QTOF-MS system can picture metabolic profiling of five groups with TIC (Supplementary Figure 1B). To investigate the global metabolic rewiring in the serum among NC, BC, S50, S100, and S200 groups, all observations were integrated and analyzed using PCA (Figure 3). The score plots of PCA overlapped partly both in the direction of PC1 and PC2 based on the serum samples from the NC vs. BC group ( Figure 3A1) and the S50 vs. NC group (Figure 3B1). The score plots of PCA from the S100 vs. NC group ( Figure 3C1) were separated from each other. To further explore the differences between the two groups, supervised PLS-DA was applied for chemometrics analysis (Figure 3). The score plots of PLS-DA showed that the NC vs. BC group (Figure 3A2), S50 vs. NC group (Figure 3B2), S100 vs. NC group (Figure 3C2), and S200 vs. NC group ( Figure 3D2) could be clearly separated, which reflected the remarkably distinct metabolic status of the serum samples among the BC, NC, S50, S100, and S200 groups. Many metabolites in the serum were significantly altered based on the results of the NC vs. BC group, S50 vs. NC group, S100 vs. NC group, and S200 vs. NC group, respectively (Supplementary Table 3). Finally, different metabolic pathways were identified and further enriched as referred to as the KEGG pathway (Supplementary Figure 3).
As shown in Table 3, the top 4 ranked metabolic pathways between the BC, NC, and GE groups displayed characteristic differences in the serum of piglets, respectively. It is interesting that nicotinamide-adenine dinucleotide phosphate (NADP), as a node molecule, was upregulated in the NC group in comparison with the BC group and downregulated in the GE groups in comparison with the NC group ( Figure 4A). Notably, tetrahydrofolic acid (THF) is a key node molecule affected by GE, and GE supplementation significantly upregulated the level of THF (p < 0.05) compared with the NC group ( Figure 4B). Additionally, considering dosage influence on the GE groups, the important top 4 ranked metabolic pathways and relevant metabolites affected also yielded dissimilar results. Especially, it suggested that GE in the S100 group significantly increased the synthesis of adenosine triphosphate (ATP), L-glutamic acid (Lglu), and L-glutamine (Gln) (p < 0.05) compared with the NC group. On the other hand, the NC group significantly reduced the production of thiamine pyrophosphate (TPP) (p < 0.05) compared with the BC group. However, GE in the S200 group significantly increased the production of TPP and L-aspartic acid (p < 0.05) compared with the NC group.

Intestinal Mucosal Barrier
In our previous study, the diarrhea incidences of piglets in the BC, NC, S50, S100, and S200 were 1.79, 21.43, 14.29, 8.93, and 7.14%, respectively. It suggested that dietary addition of GE could reduce diarrhea incidence significantly in weaned piglets challenged by ETEC (17). In general, diarrhea disease caused by ETEC infections is a major risk factor for impaired intestinal structure and barrier function of piglets. It has been reported that claudins and occludins are considered in the tight junction protein components, which primarily regulated the permeability of uncharged and charged molecules. Furthermore, ZO-1 is the adaptor protein that modulates the actin cytoskeleton (24,25), and NHE3 is a primary mediator of the absorptive route for Na + entering the intestinal epithelium from the lumen (26). Thus, all of them play important roles in mediating the functional integrity of the junction in epithelial and endothelial cells of Frontiers in Veterinary Science | www.frontiersin.org  ×200). The staining was visualized using DAB (brown), and slides were counterstained with hematoxylin (n = 5). BC, blank control group, piglets were fed diet without supplements and ETEC challenge; NC, negative control group, piglets were fed diet without supplements and challenged by ETEC; S50, piglets were fed diet supplemented with 50 mg kg −1 of GE and challenged by ETEC; S100, piglets were fed diet supplemented with 100 mg kg −1 of GE and challenged by ETEC; S200, piglets were fed diet supplemented with 200 mg kg −1 of GE and challenged by ETEC. the intestines (24)(25)(26). Our results showed that ETEC decreased the expression of epithelial tight junctions, such as claudin-1, occludin, ZO-1, and NHE3, thereby, in turn, increasing cellular permeability and disturbed the intestinal mucosal barrier. Subsequently, luminal antigens rather than bacteria may enter the lamina propria, resulting in inflammation (27). However, weaned piglets fed a diet supplementation with GE (50-200 mg kg −1 in the diet) are characterized by increased expression of claudin-1, occludin, ZO-1, and NHE3, which are crucial for the formation of a semipermeable mucosal barrier and the recovery BC group, S50 vs. NC group, S100 vs. NC group, and S200 vs. NC group, respectively), and (A2,B2,C2,D2) represent the score plots of the PLS-DA models (NC vs. BC group, S50 vs. NC group, S100 vs. NC group, and S200 vs. NC group, respectively). of the barrier function of intestinal tight junctions compared with the NC group. Furthermore, previous studies also suggested that GE was rich in phenolics (28), and phenolics have a positive effect on gut health (29). Specifically, quercetin and myricetin, known as the main phenolic constituents in GE (17,30), have been demonstrated to enhance intestinal barrier function (31). Consequently, it is assumed that the abundant phenolics in GE exerted anti-inflammatory (32) and anti-diarrhea activity (15), and improved the intestinal barrier function and gut mucosal integrity of piglets.

Fecal Metabolomics
In mammals, L-pipecolic acid has long been recognized as a metabolite of lysine degradation (33). In this pathway, peroxisomal sarcosine oxidase (PSO) can catalyze L-pipecolic acid and oxygen to yield (S)-2,3,4,5-tetrahydropiperidine-2carboxylate and hydrogen peroxide (H 2 O 2 ). As seen in Figure 5 and Table 2, it indicated that L-pipecolic acid was significantly lower in the NC compared with the BC group and suggested that ETEC might have activated the reactions mentioned above, and led to the consumption of L-pipecolic acid and the production FIGURE 3 | Score plots of PCA models in serum metabolomics. (A1,B1,C1,D1) represent the score plots of the PCA models (NC vs. BC group, S50 vs. NC group, S100 vs. NC group, and S200 vs. NC group, respectively), and (A2,B2,C2,D2) represent the score plots of the PLS-DA models (NC vs. BC group, S50 vs. NC group, S100 vs. NC group, and S200 vs. NC group, respectively).  In addition, catecholamines are generally associated with stress events that result in high levels of Gram-negative pathogens, such as Escherichia coli (36). The metabolic results between the S50 vs. NC group suggested that dietary addition with 50 mg kg −1 GE can decrease the production of stress hormones, such as catecholamines (epinephrine and normetanephrine), and increase the production of 3methoxytyramine (an inactive metabolite of dopamine) through tyrosine metabolism and catecholamine biosynthesis pathways and finally inhibit the growth of Escherichia coli against oxidative stress in the gut. In addition, it has been reported that caffeine can increase intracellular calcium levels through direct effects on metabolic phosphorylase-like enzyme (PHOS) regulation and calcium mobilization from the sarcoplasmic reticulum (37). Compared with the NC group, the caffeine level in the feces from the S50 group was downregulated, which may reduce the intracellular calcium content, thereby inhibiting the pathophysiological process that leads to diarrhea (38).
Based on the fecal metabolomics data of the S100 vs. NC group, it showed that 100 mg kg −1 of GE upregulated the level of biliverdin and downregulated the level of 5-aminolevulinic acid in the gut via the porphyrin metabolism pathway. This process started as the condensation of glycine and succinyl-CoA by 5-aminolevulinate synthase (ALAS) and generated 5aminolevulinic acid. Presumably, the resulting 5-aminolevulinic acid has two fates. On one hand, it may be finally converted into biliverdin via a series of metabolic steps, leading to the accumulation of biliverdin in the gut (Figure 5). The biliverdin generated in this process can protect the intestines from oxidants and inflammation (39,40). On the other hand, 5-aminolevulinic acid was also potentially absorbed into the blood, resulting in the high level of 5-aminolevulinic acid in the serum via glycine and serine metabolism (Figure 6). Here, 5-aminolevulinic acid reduced intracellular carbon monoxide and inhibited oxidative stress and inflammation response (41). Moreover, phosphorylcholine was downregulated in the S100 group compared with the NC group, which was associated with the phosphatidylcholine and phospholipid biosynthesis pathways. It means that most of the choline of the S100 group might not be catabolized in the gut but was absorbed into the blood, and then it fluxed into betaine metabolism and was probably utilized for betaine biosynthesis (Figure 6). Notably, L-pipecolic acid in the feces of the S100 group was significantly higher than that of the NC group. Conversely, Lpipecolic acid in the feces of the NC group was significantly lower than that of the BC group. Our findings suggested that 100 mg kg −1 of GE might inhibit the activity of PSO and reverse the lower levels of L-pipecolic acid caused by ETEC, which in turn prevented the production of H 2 O 2 and decreased oxidative stress level.

Serum Metabolomics
As seen in Figure 6 and Table 3, based on the NC vs. BC group, ETEC challenge decreases glucosamine 6-phosphate (G6P) and oxidized glutathione (GSSG) levels and increases NADP levels in the serum by affecting the glutamate and glutathione metabolism pathways, which may result in the accumulation of H 2 O 2 in the serum. On one hand, the produced H 2 O 2 cannot be reduced to water (H 2 O), which resulted in peroxide interference and cell damage through oxidation of lipids, proteins, and nucleic acids (46). On the other hand, H 2 O 2 is not a radical but is considered a reactive oxygen species, which can induce a cascade of radical reactions and inactivate pyruvate dehydrogenase (PDH) (47,48), leading to accumulation of TPP in the serum and meaning that pyruvic acid cannot be synthesized into acetyl-CoA, while the latter was closely associated with fatty acid biosynthesis. In addition, the lower levels of cytidine monophosphate (CMP) and deoxyuridine triphosphate (dUTP) in the NC group compared with the BC group revealed that ETEC perturbed pyrimidine metabolism, and then, it might inhibit the process of pyrimidinerelated nucleotide biosynthesis.
Interestingly, based on the S50 vs. NC group, caffeine was significantly downregulated in the feces ( Table 2), while it was significantly upregulated in the serum (Supplementary Table 3) and suggested that most caffeine can be absorbed into the blood through the intestinal mucosa. Furthermore, the high level of caffeine in the serum could lead to an increase in lipolysis, and it is usually accompanied by the accumulation of glycerol in the serum (49). Then the high level of glycerol could raise blood osmolality, and it, in turn, probably plays a favorable role in the increase in intestinal water absorption and the decrease in sodium efflux into the intestinal lumen, and finally resulted in the attenuation of secretory diarrhea caused by ETEC (50). It is worth noting that, based on the S50 vs. NC group, indoleamine 2,3-dioxygenase 1 (IDO1) or tryptophan 2,3dioxygenase 2 (TDO2) drives tryptophan into the kynurenine pathways that produce tryptophan catabolites, such as the high level of kynurenic acid in the serum (Supplementary Table 3). In this process, it is usually accompanied by the production of folic acid and L-glu, meaning that the generated folic acid and L-glu can be synthesized to THF through the folate metabolism pathway. Meanwhile, THF also was biosynthesized in the serum via two pathways, including pterine biosynthesis and histidine metabolism.
Based on the S100 vs. NC group, the higher level of L-glu supplies the amino group for the biosynthesis of other amino acids, is a substrate for glutamine and glutathione synthesis, and is the key neurotransmitter in biological systems. It revealed that after glutamine synthetase or glutaminase liver isoform (GLS2) converts L-glu into Gln, and glutamine-fructose-6-phosphate aminotransferase (GFPT1) subsequently converts Gln and fructose 6-phosphate (F6P) into L-glu and G6P, it suggested that 100 mg kg −1 of GE reversed the ETEC-induced downregulation of G6P. The higher level of G6P in the S100 group, in turn, can be converted into N-acetyl-D-glucosamine-6-phosphate (N-AG6P) (via glucosamine 6-phosphate N-acetyltransferase) compared with the NC group. Here, downregulation of uridine diphosphate-N-acetylglucosamine and upregulation of ATP in serum indicated that most N-AG6P generated likely can be converted into N-acetyl-D-glucosamine (N-AG) (a polysacchatide) and ATP, via N-acetyl-D-glucosamine kinase (NAGK). The resulting N-AG has confirmed its antiinflammatory efficacy for inflammatory bowel disease (51). It is worth mentioning that betaine, which might be synthesized from choline, can be degraded via two pathways. The first pathway involves betaine metabolism. Compared with the NC group, the higher levels of S-adenosyl-L-homocysteine (SAH), L-methionine, THF, and ATP in the S100 group indicated that THF cofactors were probably used to carry and activate one-carbon units via the folate-mediated one-carbon transfer pathway, resulting in the remethylation of homocysteine to L-methionine, and the synthesis of purine nucleotides and thymidylate (52). In the second pathway, betaine can be synthesized to dimethylglycine in the methionine cycle, then the generated dimethylglycine can be converted into sarcosine and enter glycine and serine metabolism. Here, the sarcosine is synthesized via two pathways to create 5-aminolevulinic acid and serine, respectively. Of them, the formation pathway of serine was accompanied by the production of THF, whereas the production of L-methionine, purine nucleotides, and 5-aminolevulinic acid may participate in the processes of attenuated inflammatory responses and inhibited oxidative stress (53)(54)(55).
Furthermore, it showed that 200 mg kg −1 of GE reversed the ETEC-induced upregulation of NADP and TPP in the serum via the transfer of acetyl groups into the mitochondria and phytanic acid peroxisomal oxidation pathways and thereby participated in the production of acetyl-CoA, and the latter was related to the synthesis of fatty acids and sterols and the metabolism of many amino acids (56). Meanwhile, similar to the S50 or S100 groups, the levels of THF and L-methionine in the S200 group were also upregulated via betaine metabolism and the pterine biosynthesis pathway compared with the NC group.
It is worth noting that GE dietary addition can upregulate the level of THF and reverse the high level of NADP induced by ETEC compared with the NC group (Figure 4). It suggested that THF is probably a main antioxidative force for GE indirectly (57,58). Meanwhile, GE downregulating the level of NADP also means that the NADP pool is probably maintained in a highly reduced state, which boosted antioxidant ability in response to oxidative damage (59).

CONCLUSIONS
Our study has demonstrated that dietary supplementation with 50-200 mg kg −1 of GE exhibited a positive effect on the recovery of intestinal tight junctions and barrier function of weaned piglets challenged by ETEC. Meanwhile, serum and fecal metabolomics analysis indicated that dietary GE (50, 100, and 200 mg kg −1 ) addition could reprogram energy metabolism through similar or distinct metabolic pathways and finally enhance the antioxidant ability of weaned piglets challenged by ETEC.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by Institutional Animal Care and Use Committee of the Chinese Academy of Tropical Agricultural Sciences (Haikou, China). animal experiments and determination. All authors reviewed and approved the final version of the manuscript.