Dried roots of Platycodon grandiflorum (Jacq.) A. DC. were purchased from Hefei Mandi New Pharmaceutical Co., Ltd., and identified and stored at the College of Pharmacy, Anhui University of Chinese Medicine. Platycodon grandiflorum was used as a raw material to extract polysaccharides via hot water extraction and 85% ethanol/water precipitation. The extract was deproteinised using the Sevag method, which was repeated several times. The polysaccharide-rich fraction was dissolved in distilled water (1:100), filtered with a 0.45 μm filter, and purified using cellulose DEAE-52 and Sephadex G-100 columns (60 × 1.6 cm). Finally, the purified neutral Platycodon grandiflorum (PGNP) was obtained for further identification ( Figure 1A ). The phenol-sulfuric acid method (24, 25) was used to determine the carbohydrate content of the PGNP. With ascorbic acid as the control, the scavenging ability of PGNP on DPPH radical scavenging was determined.

The FT-IR spectra were recorded using an FT-IR spectrometer (Thermo Electron, USA) in the range of 4000–400 cm⁻¹ (26). Thereafter, ion chromatography (IC) was performed to detect the monosaccharide composition of polysaccharides. Briefly, the dried PGNP (20 mg) was hydrolysed with 6 M HCL at 105°C for 8 h. The filtrate was eluted with distilled water and 15 mM NaOHC,15 mM NaOH, and 100 mM NaOAC at a flow rate of 0.3 mL/min. The monosaccharide standards were purchased from Sigma-Aldrich Chemical Co., Inc. The homogeneity and molecular weight of PGNP were detected using HPLC on an Agilent 1260 instrument (Agilent, USA) (27). The PGNP was methylated using previously reported methods (26). PGNP (60 mg) was dissolved in 1.0 mL D₂O. The ¹H NMR, ¹³C NMR, and two-dimensional NMR (including COSY, HMBC, and HSQC) spectra were recorded on a Bruker 600 MHz NMR spectrometer.

This animal trial was conducted under the supervision of the Ethics Committee of the Anhui University of Chinese Medicine, Hefei, China. Eight-week-old male C57BL/6 mice were supplied by Anhui Medical University (Hefei, China) (Reg. No. SCXK (Wan) 2019-0004). The timeline of animal administration is outlined in Figure 2A . After a 1-week acclimation period, mice were randomly divided into three groups: one group was fed a chow diet (D12450B, 10% energy derived from fat, Research Diets, Inc.), while the remaining two groups were fed an HFD (D12492 60% of energy from fat, Research Diets, Inc.) with or without PGNP (300 mg/kg/day) via oral gavage. Each cage contained three animals and was housed under pathogen-free conditions with temperature control (25−28°C) and light-dark cycles. Body weight and fodder intake were recorded weekly throughout the experimental period. At 14 weeks, the main organs of animals and their serum were collected after killing with carbon dioxide. The epididymal white adipose tissue (EWAT) was collected and weighed, and the caecal contents of five mice in each group were collected in sterile cryovials, frozen in liquid nitrogen, and stored at -80°C for the analysis of gut microbiota.

Liver and colon tissues were fixed in 4% paraformaldehyde and embedded in paraffin, as previously described (19). All the biochemical parameters in serum were measured according to the manufacturer’s instructions. The serum collected from all mice was analysed for TC, total cholesterol, TG, triglyceride, LDL-C, low-density lipoprotein cholesterol, AST, aspartate aminotransferase, ALT, and alanine aminotransferase levels using commercial kits (Nanjing Jiancheng Institute of Bioengineering, Nanjing, China). Serum interleukin-1β (IL-1β) and tumour necrosis factor-α (TNF-α) levels were measured using commercial ELISA kits (Dogesce, China).

The impaired glucose tolerance range was detected after HFD feeding and PGNP intervention for 12 weeks. Mice in each group fasted for 12 h before testing. Subsequently, 2 g/kg D-glucose was administered to mice via oral gavage (28). The tail blood glucose levels of mice were monitored at 0, 30, 60, and 120 min using glucose kits. The blood glucose-time correlation curves were plotted, and the area defined by the glucose concentration curve at the top and the x-axis (time) at the bottom was determined using SPSS 11.5 software (SPSS Inc., Chicago, IL, USA).

The EWAT and colon were placed in a fresh 10% paraformaldehyde solution, wrapped in paraffin blocks, and cut into thin sections (4 µm) for occult IHC. The sections were then dewaxed with xylene followed by graded ethanol, washed three times with PBS, and treated with 10% foetal bovine serum at 37°C for 30 min. The EWAT was then incubated with the anti-F4/80 (Wuhan Servicebio Technology Co., Ltd, Wuhan, China) primary antibody, while the colon was incubated with the ZO-1(Wuhan Servicebio Technology Co., Ltd, Wuhan, China) primary antibody. After three washes with PBS, the cells were incubated with HRP-labelled secondary antibody for 30 min. Finally, all slides were counterstained with haematoxylin.

Bacterial DNA was separated from the caecal contents using a DNeasy PowerSoil kit (Qiagen, Hilden, Germany). The V3–V4 region of 16S rRNA was amplified using the universal primers, 343F and 798R. 16S rRNA gene amplicon sequencing of faecal samples was performed by OE Biotech Co. Ltd. on a MiSeq platform (Illumina, San Diego, CA, USA). The sequence was processed according to a previous method and stored in FASTQ format (29). The raw data were processed using the QIIME2 software package (version 2020.11). Clustering sequences within a 97% similarity threshold for OTUs using Vsearch software. All representative reads were annotated and blasted against Silva database (Version 138) using q2-feature-classifier with the default parameters. QIIME2 software was used for alpha and beta diversity analysis. Alpha diversity was estimated using the Simpson diversity index. The unweighted Unifrac distance matrix performed by R package was used for unweighted Unifrac Principal coordinates analysis (PCoA) to estimate the beta diversity. Then the R package was used to analyze the significant differences between different groups using ANOVA and T test.

The intestinal contents were analysed, as previously described (30). The analytical instrument used in this study was an LC/MS system composed of an ACQUITY UPLC I-Class PLUS ultra-high-performance liquid phase tandem QE plus high-resolution mass spectrometer. The LC-MS analysis data of each sample were collected in negative ion mode and positive ion mode. Spearman correlation analysis was performed using R version 3.6.2 to evaluate the correlations between bacterial taxa and metabolomics and biochemical index.

In vivo experimental data are presented as mean ± SD of body weight gain, while other parameters are presented as mean ± SEM. Differences between groups were analysed using one-way ANOVA or repeated measures two-way ANOVA and Student’s t-test. The outcomes for alpha diversity (Simpson index) are expressed as median ± interquartile range (IQR). The differences between groups were based on Anosim analysis. A significant difference between the data was indicated by p< 0.05. Statistical analysis was performed using SPSS 11.5 software (SPSS Inc., Chicago, IL, USA). Raw LC-MS data were processed using Progenesis QI V2.3 software (Nonlinear, Kinetics, Newcastle, UK) for baseline filtering, peak identification, integration, retention time correction, alignment, and peak normalisation.

After a series of processes ( Figure 1A ), a neutral polysaccharide was successfully purified from P. grandiflorum (PGNP). Based on the calibration curve with standard dextran, PGNP was a purified fraction with an average molecular weight of approximately 13.6 kDa. PGNP was characterised by a single symmetrical peak in HPLC ( Figure 1B ). The sugar content of PGNP measured using the phenol-sulfuric acid method was 92%. The PGNP had obvious DPPH free radical scavenging abilities within a certain concentration range in a concentration-dependent manner but were still lower than the positive control (vitamin C) ( Figure S1F ). As shown in Figure 1C , PGNP was mainly composed of glucose and fructose, with molar percentages of 10.9% and 84.1%, respectively. As the foremost monosaccharide component of PGNP was fructose, the bone chain may consist of fructose residues. PGNP contained almost no protein, and no obvious absorption peak was found at 280 nm in the UV spectrum ( Figure 1D ). Figure 1E shows the infrared spectrum of PGNP. The broad characteristic strong peak near 3384 cm^-1 was attributed to the OH stretching vibration of the polysaccharide (31), while the peak at 2937 cm^-1 was attributed to the C-H stretching vibration (32). PGNP might have a β-configuration glycosidic bond due to the bands at 935 and 871 cm^-1 (33, 34).

To investigate the modulation of PGNP on the disordered profile of glycolipid metabolism, obese mice were administered PGNP (300 mg/kg) for 14 weeks. As expected, mice fed the HFD had significantly higher body weights than those fed the low-fat diet ( Figure 2B , P<0.001). Notably, PGNP treatment attenuated weight gain in the HFD group (31.734 ± 1.64 vs 25.75 ± 1.52g, P<0.001). During the intervention period, no significant differences in the mean daily food intake were found ( Figure S2A ), which suggested that the effect of PGNP on body weight and fat parameters was not based on decreased fodder utilisation. The average daily food intake (kcal) levels in the HFD group were significantly higher than the CHOW group (P<0.001, Figure S2B ).

As depicted in Figure 2C , EWAT weight markedly increased after 14 weeks in the HFD-fed group, while PGNP supplementation reversed this change (P<0.001 vs HFD group). Histological analysis confirmed that the HFD group had larger adipocytes than the CHOW group. Further, the HFD + PGNP group showed a significant decrease in the mean size of adipocytes ( Figure S2D , P<0.001), which was similar to that of the CHOW group ( Figure 3A ).

The serum lipid product content was analysed in HFD-fed mice. Serum TG, TC, and LDL-C levels in the CHOW group were lower than those in the HFD group (P<0.05, P<0.001, respectively). Furthermore, PGNP treatment significantly reduced the levels of TC and TG in the serum of HFD-fed mice ( Figures 2E, F , P< 0.001, P< 0.05). We assessed the impact of PGNP on glucose metabolism after 12 weeks of HFD feeding. During the OGTT, fasting glucose levels in the HFD group were approximately 15.17 ± 2.9 mmol/L. After PGNP treatment, obese mice developed good glucose tolerance with a smoother line ( Figure 2D ). Similarly, the glucose concentration profile of the PGNP-treated group was reduced compared to that of the HFD group ( Figure S2C , P<0.01). These results suggest that PGNP was remarkably beneficial to body weight gain, impaired glucose tolerance, and disordered lipid profile induced by HFD feeding.

Arresting hepatic steatosis and hypertrophy of hepatocytes were observed in the HFD group through haematoxylin and eosin and Oil Red O staining. Hepatic cells with critical ballooning degeneration and higher hepatic lipid droplets were observed in the HFD group ( Figures 2I, J ). After 14 weeks, PGNP suppressed the abnormal accumulation of lipid droplets in the hepatic tissue ( Figures 2I, J ). Furthermore, PGNP significantly reduced ALT and AST levels, which were increased by the HFD ( Figures 3B, C ).

Alterations in glycolipid metabolism are often accompanied by chronic low-grade inflammation, which is a malignant process. Thus, the accumulation of triglycerides in adipose tissue accelerates the production of the pro-inflammatory factors, IL-1β and TNF-α. Conversely, a strong inflammatory response contributes to adiposity. In this study, PGNP supplementation markedly reduced the serum expression of IL-1β and TNF-α compared to HFD alone ( Figures 3D, E ). Meanwhile, immunohistochemical staining revealed the F4/80 of macrophages in adipose tissue (brown cells; Figure 3A ). Evidently, the HFD group had more brown areas than the PGNP group ( Figure S2F ). In summary, PGNP had positive ameliorating effects on HFD-related metabolic characteristics, which were revealed by the prevention of liver function and adipose inflammation.

Colonic histological examination was performed using haematoxylin and eosin staining. The HFD group displayed inflammation-related histopathological changes. Inflammation in the colon usually manifests as abundant red blood cells, lymphocytes, and loosely arranged intestinal glands. Significant differences in intestinal morphology were observed between the CHOW and HFD groups, whereas no significant differences were observed between the PGNP and CHOW groups ( Figure 3F ). Zo-1 expression (brown) in colon tissue was determined by IHC. The emergence of Zo-1 in the HFD group was markedly lower than that in the PGNP group ( Figures 3G , S2E , P<0.001).

To evaluate the regulatory role of PGNP on gut bacteria, the IlluminaMiQ platform was used to analyse the bacterial 16S rRNA in the variable region V3-V4 of the caecal samples based on pyrosequencing. After the exclusion of unqualified sequences, high-throughput barcode sequencing yielded 1,083,465 high-quality sequences from 15 samples. There were 773, 669, and 894 mice in the Chow, HFD, and HFD + PGNP groups, respectively ( Figure 4A ). However, the Chow and HFD groups only had 422 unique OTUs and 364 unique OTUs, respectively. The 555 unique OTUs in the HFD + PGNP group were investigated ( Figure 4A ). The Simpson index of the PGNP group was significantly higher than that of the HFD group, indicating that PGNP can effectively increase the diversity of the microflora ( Figure 4B ). Clustering of the study participants was performed using PCoA based on weighted UniFrac distances. As shown in Figure 4C , PGNP markedly changed the intestinal flora composition.

To further determine the distinct bacterial communities among the groups, we analysed the similarity of bacterial taxa at the phylum level. Accordingly, the top 15 most abundant phyla were assessed. The HFD group had a higher relative abundance of Firmicutes, Desulfobacterota, and Campilobacterota phyla. However, the relative abundance of the Bacteroidetes phylum decreased considerably in the HFD group. Surprisingly, PGNP reversed the above changes in relative abundance ( Figure 4D ). The increased Firmicutes/Bacteroidetes ( Figure 4E ) abundance ratio positively correlates with the obesity phenotype (37). Notably, PGNP treatment led to a reduction in the Firmicutes/Bacteroidetes abundance ratio.

We proceeded to investigate the composition of the gut microbiota at the genus level to further evaluate its structure. The abundance of Bacteroides ( Figure 5B , p < 0.05) and Blautia ( Figure 5C ) were improved after PGNP treatment, which was decreased by the consumption of an HFD; conversely, for Rikenella, the PGNP group exhibited lower levels than the HFD group ( Figure 5D , p < 0.05). The level of Helicobacter was significantly elevated in the HFD group (p < 0.05 vs CHOW group, Figure 4E ). Intriguingly, after PGNP treatment, the levels of the above bacteria in the CHOW group were similar to those in the HFD group ( Figure 5C ).

Altered gut microbiota composition has a documented relationship with adiposity and metabolic syndromes. Intestinal microorganisms metabolise the intestinal contents and produce different metabolites. To explore the relationship between intestinal microbes and metabolites, the intestinal contents were analysed using LC-MS. Partial least-squares discriminant analysis (PLS-DA) clearly revealed predictive and descriptive modelling. Further, the PLS-DA results revealed that the metabolite profiles in the three groups were markedly modified ( Figure 6A ).

The differentially expressed metabolites were identified by the VIP value of OPLS-DA (VIP>1) combined with the Student’s T test ( Figure 6A , P<0.05). A notable difference in metabolites was found between the CHOW and HFD groups. A total of 824 metabolites were altered when the CHOW and HFD groups were compared. Of these metabolites, 333 were upregulated and 491 were downregulated ( Figure S2G ). Furthermore, the crucial altered metabolic pathways of metabolic syndrome, linoleic acid metabolism, PPAR signalling pathway, steroid hormone biosynthesis, and dehydroepiandrosterone sulphate (DHEAS), were available for metabolic enrichment and pathway analysis based on a database search ( Figures 6B, C ).

A comparison of the HFD+PGNP and HFD groups revealed 513 differential metabolites, with 227 upregulated and 286 downregulated metabolites ( Figure S2H ). Figure 6E shows a schematic of the affected metabolic pathways, for example, linoleic acid metabolism, d-arginine and d-ornithine metabolism, and primary bile acid biosynthesis. PGNP treatment significantly altered the HFD profile after 14 weeks, significantly increasing the levels of both saturated fatty acids and bile acids ( Figure 6D ).

To clarify the exact role of PGNP in improving metabolic disorders, the causal links between the host gut microbiota and metabolites were assessed using Spearman correlation analysis. Based on our results, the gut microbiota community structures were closely related to the gut metabolites and pharmacodynamic parameter profiles. Herein 16 microbiota taxa were identified at the genus level. Further, pharmacodynamic parameters in vivo were positively or negatively associated with differential metabolites ( Figures 6F, G ). Among them, HDL-C was positively associated, whereas seven other pharmacodynamic parameters were negatively associated with gamma-linolenic acid.