Lycium Barbarum Polysaccharides Regulate the Gut Microbiota To Modulate Metabolites in High Fat Diet-Induced Obese Rats

Previous studies had indicated that the gut microbiota was a main internal factor leading to obesity through energy storage and metabolic disorders. Lycium Barbarum Polysaccharides (LBP) have been discovered with a more protective effect on intestinal ora. But it is unclear whether LBP could regulate the gut microbiota to modulate metabolites, nally relieving obesity. A related study of high-throughput 16S rRNA sequencing and serum metabolomics proling in LBP intervention on high fat diet-induced obese rats then explored the benecial effects of LBP and the underlying mechanism. LBP affected lipid parameters such as total cholesterol, Triglyceride, and High-density lipoprotein. The gut microbiota result detected 16 types of the phylum of bacteria in total, while four of them (Bacteroidetes, Firmicutes, Proteobacteria, Deferribacteres) were signicantly different. LBP upregulated the level of Firmicutes of obese rats. LBP might associate with the gut microbiota that participates in the membrane transport and metabolism of amino acid, carbohydrate, energy, and lipid. The serum metabolomics proling of high-fat diet-induced obesity rats found over 30 differential metabolites between model and intervention groups. Primary metabolites include cortisol, glycohyocholic acid, homo-L-arginine, ursodeoxycholic acid, isoursodeoxycholic acid, glycoursocholic acid, 4-ethylphenylsulfate, deoxycholic acid, 7-hydroxy-3-oxocholanoic acid isomers, gly-phe, pipecolic acid, proline betaine, and pyrocatechol sulfate. Pathway analysis in serum found four disorder pathways: glycerophospholipid metabolism, glycine-serine-threonine metabolism, biosynthesis of unsaturated fatty acids, and linoleic acid metabolism. The studies revealed that LBP treatment increased the diversity of fecal microorganisms and reduced metabolic disorders in obese rats. LBP ameliorated metabolic disorders and the on the serum metabolome the potential predicted function of biosynthesis of 12-, 14- and 16-membered macrolides, aminobenzoate degradation, MAPK signaling pathway-yeast, lysosome, glycosphingolipid biosynthesis-globo series, adipocytokine signaling pathway, isoavonoid biosynthesis, pentose and glucuronate protein bile secretion, glycosaminoglycan biosynthesis-chondroitin sulfate, cytochrome P450, glycosylphosphatidylinositol (GPI)-anchor biosynthesis


Introduction
As an epidemic disease, obesity is generally caused by fat accumulation and is characterized by increasing an unproportionate weight and height [1,2]. Obesity has shown a rising trend globally, and 13.1% of adults are obese [3]. Obesity is usually comorbid with cardiovascular, metabolic, and psychosocial diseases. These comorbidities are leading to increased disease burden, morbidity, and mortality [4,5].
Previous studies indicated that the intestinal microbiota had been linked to obesity, and a meta-analysis suggests the promotion of probiotics in weight loss in adults [6]. Some data showed that the microbiota might connect with host metabolism through its production of metabolites; for example, the abundance of B.thetaiotaomicron was decreased in obese rats and the weight-loss intervention [7,8]. Metabolites derived from intestinal ora, such as 2-hydroxyisobutyrate and xanthine, were linked to obesity [9]. The typical obese metabolites were lost after bariatric surgery-induced weight loss [9,10]. Therefore, researches on gut microbiota and metabolism could play an important role in therapeutic and preventive interventions [10][11][12][13].
Lycium barbarum polysaccharides (LBP) are biologically active components extracted from Goji berries and have been extensively explored for their health effects [14]. Evidence shows that LBP can lower blood lipid levels in a hyperlipidemic animal model [15]. Another study elucidated that LBP could improve the amount of intestinal microbiota [16,17]. However, whether LBP regulates the disorder of bacteria then adjusted metabolites of high-fat-diet (HFD), its mechanism is still unclear, and its anti-obesogenic effect was not reported [18].
This study used high-throughput 16S rRNA sequencing and un-target metabolomics in LBP intervention on high fat diet-induced obese rats to examine the gut bacteria and metabolites. It aimed to clarify the changing of gut microbiota and metabolites, explore the bene cial effects of LBP and the underlying mechanism of the gut microbiota modulating metabolites. Consequently, we provided important information for the scienti c basis of obesity prevention. Gut microbiota and metabolite composition of rat serums were investigated by 16S rRNA sequencing and Ultra Performance Liquid Chromatography-High Resolution Mass Spectrometry (UPLC-HRMS), respectively. It aimed to identify the feces of microbiota and revealed the relationship of serum metabonomic between groups control, model, and different LBP doses intervention on obese rats.

LBP
LBP was obtained from Shanghai Yuanye Biotechnology Co., Ltd (No. B20460, China). LBP was dissolved in pure water in a solution of 10 mg/ml, 30mg/ml, and 90 mg/ml. The content in the LBP product was analyzed by the phenol-sulfuric acid method [19].

Animals breeding and maintenance diets
Male Sprague Dawley rats of clean grade (8-weeks old; body weight 200.0 ± 2.0g) were supplied with certi cate no. SCXK (Ning) 2020-0001 from the Laboratory Animal Center of Ningxia Medical University (Ningxia, China). All rats were fed in the laboratory animal center of Ningxia Medical University and kept a 12-h-dark/12-h-light cycle relative humidity (60%-65%) at 25°C. We supplied sterile water and food in this whole experiment. All procedures were getting permission from the Ethics Committee of Ningxia Medical University.
To create the HFD induced obesity model for 12 weeks, fty-seven rats with body weights were separated into two groups (control n = 11, modeling n = 46) randomly based on their diet (HFD, High Fat Diets, 60% energy from fat; MD12033, medicience. Ltd, Jiangsu, China) and control (10% energy from fat; 1010009, Xietong, Jiangsu, China). Obesity was de ned as 10% more weight than the standard diet group at 12 weeks [20]. The experimental group received LBP for another 12 weeks at different doses, low (10mg/kg bw), medium (30mg/kg bw), and high (90mg/kg bw), separately. In comparison, the control and the model group were given the same volume of physiological saline via intragastric administration.
Bodyweight was determined weekly during the whole experiment.

Sample Collection
The rats were weighed, measured body length under anesthesia, and sacri ced. Blood was collected from the aorta ventralis after 12 weeks of LBP exposure. Then blood samples were gathered after the centrifugation of 4000 × g for 10 minutes at 4°C, stored at -80°C for the next experiment. Fresh feces were picked into individual 5ml Eppendorf tubes from the intestine after the execution and snap-frozen with liquid nitrogen for decreasing its pollution by environment and stored in the ultra-low temperature refrigerator.

Sample preparation for 16S rRNA
The instruction extracted genomic DNA from the feces using a MagPure Soil DNA LQ Kit (Magen, D6356-02). Amplicon quality was monitored by gel electrophoresis, puri ed with AMPure XP beads (Agencourt), and performed another round of PCR after ampli cation. We were using the Qubit dsDNA assay kit to quantify the nal amplicon. Illumina novaseq6000 (Illumina, San Diego, USA) was used to survey the concentration pooled puri ed amplicon. The 16S rRNA V3-V4 region was magni ed by PCR. The primer sequences followed by: 343F (5'-TACGGRAGGCAGCAG -3') and 798R (5'-AGGGTATCTAATCCT-3') primers. The rst-round PCR used an initial step at 94°C for 5 min; then took 94°C for 30 s, 56°C for 30 s, 72°C for 20 s in 26 cycles, and 72°C for 5 min for the nal extension.

Sample preparation for metabolomics
The serum sample was thawed at 4℃, absorbed and transferred 50µl accurately to EP tube, added 200µl methanol-acetonitrile (1/1, V/V) mixing with internal standard, vortexed for 120s, centrifugation at 15,000 g × 15 min, transfer 200 µl supernatant to 1.5 ml EP tube. Low-temperature vacuum drying in centrifugal concentrator (Thermo Scienti c, USA), sealed and cryopreserved before analysis. Before LC-MS analysis, 100 µL 50% methanol was used after dissolving.
The hydrophilic fraction of metabolite extracts were submitted to the non-targeted metabolomics analysis platform by Ultimate TM 3000 ultra-high-performance liquid chromatography coupled to Q Exactive TM quadrupole-Orbitrap high-resolution mass spectrometer system (UPLC/HRMS) (Thermo Scienti c, San Jose, USA) scanned at 50-1000 m/z mass range with ESI in positive and negative modes separately. The metabolite pro le with reverse-phase chromatographic separation mode with positiveand negative-ion detection. Metabolites were apart by an Acquity TM HSS C18 column (Waters Corporation, Milford, USA, 1.8 µm, 2.1 × 100 mm). The gradient followed by: 2% organic phase 0 min ramped to 100% in 10 mins, column washing and equilibrating for another 5 min. The ow rate, injection volume, and column temperature were 0.4 ml/min, 5 µl, and 50℃, respectively.
Quality control (QC) samples assessed the UPLC/HRMS system's reproducibility and reliability.

Data analysis and results interpretation
For bodyweight, length, and all biochemical indexes, the data were expressed as the mean ± SD and used Student's t-test. A signi cant difference was considered p < 0.05.
The data were graphically plotted using GraphPad Prism 7.

16s rRNA assessment
Sequences and alpha diversity (Chao1 and Shannon indices) analysis were used Quantitative Insights into Microbial Ecology (QIIME 1.8.0 version). We limited operational taxonomic units (OTUs) by the results of clustering into sequence tags. Sequences over 97% similarity were de ned to the same OTUs to species annotation and taxonomic analysis on the SILVA database. Beta diversity and Principal coordinate analysis (PCoA) were calculated from Bray Curtis distance matrices to show differences between the gut microbiota composition and structure. The unweighted pair-group method with arithmetic mean (UPGMA) was used to measure the hierarchical clustering of samples. We used PICRUSt software to predict functional 16r RNA pro les and produce KEGG pathways [23].
A non-parametric test (Mann-Whitney U test and Kruskal-Wallis) was analyzed gut microbiota by SPSS software (IBM SPSS 21.0; SPSS Inc., Chicago, IL, USA).

Metabolite pro le analysis
The data from the untargeted metabolomic analysis was obtained by Mass Lynx v.4.1 software (including ESI + and ESI − modes). The metabolomic data from different measurements were normalized using the support vector regression method in the MetNormalizer R package [24].
The MS/MS metabolic data (hydrophilic fraction) were processed with Compound Discoverer software (Thermo Scienti c, San Jose, USA) to help accomplish content extraction.
For identifying or structural annotation of metabolites, retention time and high-resolution MS/MS spectrum similarity were strictly used to determine the structural annotation of metabolites. Used XCalibur Quan Browser information to extract the area under the curve as the quantitative information of metabolites and exported the peak area data of all labeled metabolites to Excel software for trimming and statistics. The chemical identi cation results were nally annotated with classi cation criteria proposed by MSI (Metabolomics Standardization Initiative).
The output tables of Metabolites were de ned for SIMCA 14.1 software (Sartorius AG Umetrics, Goettingen, Germany). We conducted principal component analysis (PCA), orthogonal projections to latent structures discriminant analysis (OPLS-DA), and coordinates analysis was performed with R package [25]. To avoid chemical noise, all models adopted the Pareto transformation. Potential metabolites were selected based on variable importance (VIP) value (VIP > 1.5) combined with the t-test p-value (p < 0.05) in the OPLS-DA model. We used MetaboAnalyst 4.0 (http://www.metaboanalyst. ca) to visualize the metabolites differences between two groups by generating the heatmap. Venn diagram showed differential metabolites in intervention groups and model group[26].

The content of LBP
The content of LBP was measured to be 85.74%. Drew a standard graph by the concentration of glucose as x-axis and absorbance on y-axis, the regression equation was y = 7.8596x-0.0012, R 2 = 0.9988.

Biochemical indexes
Fifty rats were t for the subsequent experiment treatment after developing the model successfully. The bodyweight of the intervention rats changed smoothly contrasted with the control rats. Treatment with LBP for 12 weeks decreased the trend of obesity, especially in the medium group. The weight of rats changed signi cantly. Lee index changed obviously compared intervention groups to control and the model group. Lee index in high LBP group lower than the model and low LBP group. It also signi cantly decreased with LBP dose (Figure 1. a-c).
The results of the biochemical analysis were shown in Table 1. Medium and high LBP intervention groups had descended in TC, TG, ALT, AST, CREA, UA, and LDL levels, increasing HDL simultaneously, compared with the control and model group. However, the effect was not noticeable (p > 0.05). The medium group had a different trend for biochemical indexes (Table 1). Data were expressed in mean (SD). Estimation of p-value by using ANOVA test*

Gut microbiota analysis
There was evidence that the gut microbiota may be conducive to relieving or treating obesity [27]. We explored was the role of LBP related to obese gut microbiota for obesity; the study used 16S rRNA (V3-V4 region). It analyzed the fecal ora of obese rats after 12 weeks of treatment. The species diversity of the samples was evaluated at the OTU level. A total of 4679 OTUs for 40 samples were generated (Figure 2. a).
The Firmicutes and Bacteroidetes ratio in the gut were closely related to obesity [28]. More than sixteen phylum of bacteria were detected, and four (Bacteroidetes, Firmicutes, Proteobacteria, Deferribacteres) were quite different between the obesity and control groups. Speci cally, Bacteroidetes was the predominant phylum in all studied groups, over 60% of species abundance, but it was not signi cant in all groups (Figure 2. b-c). And Firmicutes, the following intestinal ora, was signi cantly lower in the obesity group than the control group. The relative abundances of Proteobacteria and Deferribacteres were higher in the obesity group than in the control group. However, there were no signi cant differences between the obesity group and the control. The Firmicutes/Bacteroidetes ratio was decreased on obese rats, which changed signi cantly on low and high LBP groups versus control (Figure 2. d).
We measured the α diversity using four indicators: Observed species, Chao 1, Shannon index, and Simpson index. The higher values of the rst three indicators, the lower the Simpson value represented, the higher the richness. The plots showed Observed species, Chao 1, Shannon index statistical signi cance respectively (p = 0.000, 0.001, 0.000), but 0.059 for Simpson index. Then, we found the Shannon index of obese rats was lower than that of the control and high LBP groups, while low and medium LBP groups did not have e cacy almost (Figure 2. e-g).
Beta diversity was calculated to examine and plot the distributions of the 40 samples of obesity with control groups, PC1 (22.56%) and PC2 (10.05%). In comparison, the red points (control) were distributed in a different quadrant of the gure, and other points of obesity circulated in the bottom quadrants of the graph far away from the red points. It is suggested that there were differences in clustering patterns between them, which points of the model group and the different obesity groups overlapped. So, further explored the obesity group by dividing them using OPLS-DA for two groups. Finally, the samples of LBP intervention groups showed overall differences (Figure 2. h).
The correlations between bacterial abundance and 16S rRNA predictive function showed that more than 40 predicted functions were modi ed, especially amino acid metabolism, carbohydrate metabolism, energy metabolism, lipid metabolism, and membrane transport in ve groups (Figure 2. i). Thus, LBP altered ve potential gut microbiota metabolic functions mainly involving an amino acid, carbohydrate, energy, lipid, and membrane metabolism.

Metabolism analysis
The presentative method of UPLC-HRMS detected and collected the metabolites information of serum in positive and negative ions. The metabolites in serum samples were identi ed based on MS/MS data, and 375 metabolites were annotated.
Firstly, the data of serum samples from the ve groups were analyzed using PCA and OPLS-DA. The score plots showed signi cant clustering in serum samples of the control, model, and different intervened groups. The metabolic phenotypes model parameter was R 2 X = 0.611 for unsupervised PCA score plots between the ve groups (Figure 3. a). OPLS-DA model scoring chart maximized the difference between the ve groups of metabolomics data, R 2 Y = 0.541, Q 2 = 0.349 (Figure 3. b). The same trend existed in medium and high LBP versus model modeling, respectively (Appendix-1, 2).
The score scatter plots for the PCA and OPLS-DA models showed inner-differentiation of metabolomics data between model and control group, R 2 Y = 0.991, Q 2 = 0.934. It noted that speci c signi cant biochemical changes occurred following HFD induced obesity in rats.
Compared with the mass spectrometry data and their changing trends in 3 intervention groups versus the model group. Score plot of OPLS-DA modeling indicated inner-group differentiation of metabolomics data between model and low, R 2 Y = 0.648, Q 2 = 0.204. (Figure 3.c). The parameters of two groups for R 2 and Q2 and the permutation results to test the robustness of OPLS-DA modeling (Q 2 = -0.223), and the slope was positive, which illustrated the model validity of the low LBP to model (Figure 3.d).
The multi-group analysis, including medium LBP versus model and high LBP versus model, evaluated three groups' metabolomic changes. Based on these data, the OPLS-DA model displayed a clear separation between medium LBP, high LBP, and model groups (Figure 3. e-f). These results suggested that the LBP intervention condition of obesity signi cantly affected the serum metabolic pro le in rats.
Finally, the differences in perturbed pathways between obesity and the LBP intervention group were analyzed by comparing the different metabolites based on the impact value greater than 0.1. Pathway analysis found four disorder pathways according to the screened differential metabolites: glycerophospholipid metabolism, glycine, serine and threonine metabolism, biosynthesis of unsaturated fatty acids, and linoleic acid metabolism. These pathways denoted their potential as the targeted pathways of LBP against obesity. (Figure3. g-i).

Correlation between the gut microbiota and the metabolome
To comprehensively analyze the correlation between the gut microbiota composition and the host metabolome by calculating Spearman's correlation coe cient after 12 weeks' treatment with LBP in rats ( Figure 5. a-c). Analyze the correlation with the predicted function based on gut microbiota and the serum metabolome by calculating Spearman's correlation coe cient after 12 weeks' treatment in rats with LBP. After LBP intervention, the potential predicted function of biosynthesis of 12-, 14-and 16-membered macrolides, aminobenzoate degradation, MAPK signaling pathway-yeast, lysosome, glycosphingolipid biosynthesisglobo series, adipocytokine signaling pathway, iso avonoid biosynthesis, pentose and glucuronate interconversions, protein digestion and absorption increased; bile secretion, glycosaminoglycan biosynthesis-chondroitin sulfate, cytochrome P450, glycosylphosphatidylinositol (GPI) -anchor biosynthesis and Renin-angiotensin system decreased. These relations suggested that gut microbiota and host metabolites pro le affect each other.

LBP improved biochemical index in obese rats
At present, biological activities of LBP had been previously observed that used to treat hypertension and atherosclerosis in Chinese traditional medicine that could be improved on animal and cellular models [29][30][31]. In this study, we provided a more integrated understanding of obesity and the role of LBP in regulating intestinal microbial to modulate metabolic processes. This experiment evaluated the effect of LBP-regulated gut microbiota to elucidate its effective mechanism connected to metabolomics.
The obesity model was established successfully by feeding HFD for 12 weeks, as evidenced by ndings consistent with previous reports [32][33][34]. LBP softly reversed the biochemical parameters, including TC, TG, ALT, AST, CREA, UA, HDL, LDL. It suggested that LBP may affect lipid function in our experimental and have a role to in uence the obese rats and could be used for further experiments.

LBP improved gut microbiome diversity and metabolite disorders in obese rats
LBP intervention for obese rats altered the diversity and comprised the ora because of the α and beta diversity analysis [16]. The fecal microbiota analysis showed that LBP ameliorated dysbacteriosis by obesity. The α-diversity was signi cantly reduced in the model group versus the control group in this study. However, the α-diversity was restored like control in rats with LBP treatment. Bacteroidetes was the predominant phylum over 60% of all species abundance in all studied groups. Proteobacteria and Deferribacteres were decreased after LBP treatment, although still higher in each obesity group. Proteobacteria helped to reduce producing lipopolysaccharide [35]. Speci cally, Firmicutes in the obese group were signi cantly lower than those in the control group, and it increased after intervention, which may suggest the LBP effect upregulated its level. Proteobacteria and Firmicutes played a crucial role in regulating obesity for obese rats [36,37]. LBP might be related to latent metabolic functions, such as the amino acid, carbohydrate, energy, lipid metabolism, and membrane transport of gut microbiota.
For the levels of metabolites, the intervention group displayed signi cant changes, including fatty acid, bile acid, and some of the amino acids, and it suggested dysfunctional glycerophospholipid metabolism, glycine, serine, and threonine metabolism, biosynthesis of unsaturated fatty acids, linoleic acid metabolism. Glycerophospholipid metabolism was related to obesity, and another metabolic syndrome was reported in a study about cohort [38]. We found various cephalin associated with obesity, like PE The bile acid level changed in different groups, including 4-ethylphenyl sulfate, deoxycholic acid glycine conjugate, glycohyocholic acid. Early researchers had a similar report [39,40]. And some microbial metabolites, for example, 4-ethylphenyl sulfate and 7-hydroxy-3-oxocholanoic acid isomers, also increased in intervention groups compared with the model.

Conclusion
Our study of LBP treatment testing showed a different dose of LBP affected lipid biochemical differently, and LBP restored microbial diversity, especially for the high LBP group, more than 20 genera positively or negatively associated with different metabolites by 16S rRNA. And it also found over 30 metabolites clustered in four metabolic pathways by an untargeted metabolomics analysis platform. It was associated with glycerol phospholipid metabolism, amino acid metabolism, unsaturated fatty acid biosynthesis, and linoleic acid metabolism and was the same trend as the predictive function of intestinal ora. Different doses of LBP could affect the intestinal micro ora of rats to improve the host metabolome. The current study supplied substantial evidence and a better understanding of molecular mechanisms and helped develop treatments for LBP.  Effect of LBP on the compositions of gut microbiota. a) ower plot about the number of OTUs; b) comparison of relative abundance at phylum; c) heatmap of species among the ve groups; d) the Firmicutes/Bacteroidetes ratio; -diversity analysis: e) observed species violin-plot, f) Shannon, and g) chao1 violin-plot; h) PCoA plot; i) Metabolic pathways from KEGG module predictions using 16S rRNA data with PICRUSt (n = 8 per group). Pairwise comparisons using Wilcoxon rank sum test; *p < 0.05, **p < 0.01, ***p < 0.001 for versus model group; ns means the difference was not statistically signi cant. All the differences were analyzed using Kruskal-Wallis test; *p < 0.05; **p < 0.01 for versus model group.

Figure 3
The results of metabolomics data analysis: a) unsupervised PCA score plots of metabolic for ve groups; b) OPLS-DA score plots of metabolic for ve groups; c) OPLS-DA score plots of metabolic between low LBP and model group; d) test robustness of OPLS-DA modeling compared low LBP to model; e) OPLS-DA score plots between med LBP and model group; f) OPLS-DA score plots between high LBP and model group; Metabolic pathway analysis result of differential metabolites (p < 0.05 of t-test after FDR adjusting) g) low LBP and model group; h) med LBP and model group; i) high LBP and model group.

Figure 4
Venn diagram for differential metabolites. a) all different metabolites in compared groups; b) increased metabolites in compared groups.