The total sugar content of BP and BO was measured using a phenol sulfuric acid method and was calibrated with a D-mannose standard (Dubois et al., 1956). The reducing sugar content was quantified using a 3, 5-dinitrosalicylic acid colorimetry method and calibrated with a D-mannose standard (Miller, 1959). The total proteins were detected by a Bradford method using the standard curve prepared with bovine serum albumin (Bradford, 1976). The uronic acid content was measured by a meta-hydroxybiphenyl method using glucuronic acid as the standard (Blumenkrantz and Asboe-Hansen, 1973).

The molecular weights of BP and BO were measured by a high-performance gel permeation chromatography (HPGPC) method (Chen et al., 2021). The analysis was performed on Waters1525 equipped with TSK gel GMPWXL column (7.8 mm × 300 mm). The column was eluted with ultrapure water at a flow rate of 0.6 mL/min at 30°C. Before analysis, the sample solution was prepared with ultrapure water and filtered through a 0.22 μm filter. The injection volume was 20 μL. Dextrans with different molecular weights (1−71 kDa, Sigma, MO, USA) were used as the standards to obtain the calibration curve.

The monosaccharide composition of BP and BO was determined using a PMP derivatization method (Fu et al., 2018; Chen et al., 2021). Briefly, 50 mg of saccharide sample was hydrolyzed with 3 mL of trifluoroacetic acid (2 M) at 120°C for 2 h. The excess trifluoroacetic acid was removed by evaporation under reduced pressure at 45°C. After that, 600 μL of NaOH (0.3 M) and 600 μL of PMP (0.5 M) were added to the sample solution. The mixture was incubated at 70°C for 30 min, and the mixture was neutralized with 0.3 M HCl and extracted with chloroform three times. The aqueous layer solution was filtered through a 0.22 μm filter before analysis.

The samples were analyzed with Uranus 5u C₁₈ column (4.6 mm × 250 mm, 5 μm) at 30°C. The mobile phase was acetonitrile (A) and 0.02 M of ammonium acetate (B). The gradient elution conditions were: 0−5 min, 16% A; 5−17 min, 16%−19% A, 17−30 min; 19%–22% A; 30–35 min, 22%−16% A. The flow rate was 1 mL/min, and UV absorption was measured at 254 nm. The injection volume was 10 μL. Fucose, D-glucose, rhamnose, xylose, mannose, galactose, arabinose, glucuronic acid, and galacturonic acid were used as the standards.

The glycosidic bond composition of BP was analyzed by methylation reaction and GC-MS analysis (Chen et al., 2021). BP (5 mg) was dissolved in 2 mL of dimethyl sulfoxide and reacted with 50 mg NaH for 1 h. The solution was incubated with 1 mL CH₃I at 25°C in the darkness for 2 h. Methanol was added to the solution to complete the reaction. After that, 5 mL H₂O was added to terminate the reaction. After the extraction with chloroform, the methylated polysaccharide was subjected to acid hydrolysis, NaBH₄ reduction, and acetic anhydride acetylation. The sample was finally extracted with CH₂Cl₂ for GC-MS analysis.

The simulated digestion in saliva was carried out according to the reported method (Huang et al., 2019; Huang et al., 2020). Simulated salivary fluid (SSF) was prepared by dissolving 0.0764 g/L NaCl, 0.0133 g/L CaCl₂, and 0.1491 g/L KCl into distilled water. The pH of the solution was adjusted to 7.0 using 0.1 M of NaHCO₃. BP or BO was dissolved in distilled water with a final concentration of 2.5 mg/mL. Then, 20.0 mL BP solution (BP group) or BO solution (BO group) was mixed with 20.0 mL SSF and 1 mL salivary α-amylase (1500 U/mL), followed by the incubation in a water bath (37°C). An equal volume of distilled water (without BP or BO dissolved in it) was used in the Control group. During the digestion, a 2.0 mL reaction mixture was withdrawn at 0, 5, 15, and 30 min, respectively. The collected mixture was boiled in a water bath for 5 min to deactivate salivary amylase. After the simulated salivary digestion, the reducing sugar content and molecular weights of BP and BO in the digestion solution were determined. The experiment was repeated three times.

The in vitro gastric digestion was performed as the described method with minor modification (Li et al., 2020). After the salivary digestion, the pH of the mixture was immediately adjusted to 3.0 by HCl (1 M) for the subsequent gastric digestion. In brief, 20 mL salivary digestion was mixed with 4.72 mg pepsin and 20 mL simulated gastric fluid (GSF) electrolyte solution (3.1 g/L of NaCl, 1.1 g/L of KCl, 0.6 g/L of NaHCO₃, 0.15 g/L of CaCl₂, pH 3.0), followed by the incubation in a water bath (37°C). During the digestion, a 2 mL reaction mixture was withdrawn at 0, 1.0, 3.0, and 5.0 h, respectively. The collected mixture was boiled in a water bath for 5 min to deactivate the pepsin enzyme. After the simulated gastric digestion, the reducing sugar content and molecular weights of BP and BO in the digestion solution were determined. The experiment was repeated three times.

The in vitro simulated intestinal fluid digestion was performed as the described method with some modifications (Li et al., 2020; Ma et al., 2021). After the gastric digestion, the pH of the mixture was immediately adjusted to 7.0 using NaHCO₃ solution (1 M) for the subsequent intestinal fluid digestion. In brief, the simulated intestinal fluid (SIF) was prepared by dissolving 5.4 g/L NaCl, 0.65 g/L KCl, and 0.33 g/L CaCl₂ into distilled water, and the pH of the solution was adjusted to 7.0 by NaHCO₃ solution (0.1 M). The intestinal fluid digestion was mixed with SIF-containing pancreatin solution (7%, w/v), bile salt solution (4%, w/v), and 26 mg trypsin at a ratio of 3:10 (v/v). Then, the solution was incubated at 37°C, and 2.0 mL of the mixture at different time points (0, 0.5, 1.0, 2.0, 4.0, and 6.0 h) was withdrawn and boiled in a water bath for 5 min to deactivate pancreatin enzymes. After the intestinal digestion, the reaction mixture was centrifuged, and the reducing sugar content in the supernatant was determined. Next, 80% (v/v) ethanol was used to precipitate BP and BO, and the molecular weights of both saccharides were determined. The experiment was repeated three times.

The in vitro fermentation of fecal inoculum was conducted based on an established method with minor modification (Huang et al., 2020; Mao et al., 2021). Fecal samples were pooled from five healthy volunteers (three males and two females) who remained on a regular diet and did not use antibiotics within three months. The fresh fecal samples were mixed with sterilized PBS (including 0.5 g/L of L-cysteine) to obtain a 10% (w/v) fecal suspension.

The fermentation medium was prepared following the report (Ding et al., 2019; Huang et al., 2020). Briefly, 1 L basal medium consisted of 2.0 g peptone, 2.0 g yeast extract, 0.1 g NaCl, 0.01 g K₂HPO₄, 0.04 g KH₂PO₄, 0.01 g MgSO₄·7H₂O, 0.01 g CaCl₂·2H₂O, 2 g NaHCO₃, 2 mL Tween-80, 0.05 g hemin, 10 μL vitamin K₁, 0.5 g L-cysteine, 0.5 g bile salts, and 0.01 g resazurin. The basal medium was sterilized at 115°C for 20 min. Then, 1.0 mL fecal slurry was added to a 9.0 mL basal medium containing BP (2.5 mg/mL), BO (2.5 mg/mL), or not (Control group), respectively. The fermentation experiment was carried out in an anaerobic tank. All treatment groups were incubated with the fecal inoculum at 37°C in a thermostatic shaker. During the fermentation, samples were withdrawn at 0, 6, 12, 24, 36, and 48 h for further analysis. The experiment was repeated three times.

The OD₆₀₀ of fermentation broth was detected with a multi-functional microplate reader at 600 nm. The pH value was measured using a pH meter. The contents of total carbohydrates and reducing sugars in fermentation broth at different time points were determined according to the methods as mentioned above.

Gut microbiota analysis was performed as referenced (Hu et al., 2020). In brief, 2 mL of fermentation broth was collected after 24 hours of fermentation. The bacterial genomic DNA was extracted using an intestinal microbial DNA extraction kit. The DNA library was constructed by two-step PCR amplification, and the V3-V4 region of 16S rDNA was sequenced by Illumina miseq. After optimization, the sequence was analyzed by OTU cluster analysis, species diversity analysis, and species taxonomy analysis, as detailed in Supplementary Methods .

The bacterial genomic DNA was extracted using an intestinal microbial DNA extraction kit (Solarbio, Beijing, China). In brief, the bacterial copies were measured by qRT-PCR using a SYBR QPCR mixture on a Bio-Rad CFX Connect Real-time system (Bio-Rad, CA, USA). The specific primer sequences were displayed in Supplementary Table 6 . PCR mixture was initially heated at 95°C for 10 min, followed by 34 cycles of 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s. Vectors containing specific bacterial 16S rDNA sequences were constructed as standards. The copies of specific bacteria in the fermentation broth were calculated according to the standard curves.

After 24 hours of fermentation, the broth was centrifuged at 13,000 ×g for 5 min. The supernatant (0.5 mL) was evenly mixed with 1.0 mL pre-cooled methanol. After the low-temperature ultrasound for 10 min, the mixture was incubated overnight at 4°C. Next, the mixture was filtrated through a 0.22 μm membrane for the determination of SCFAs. Acetic acid, propionic acid, butyric acid, and valeric acid were used as SCFA standards. GC-MS was used to detect the SCFAs of all samples according to the reported method (Huang et al., 2020).

The capacity of fermentation broth samples to scavenge 2,2-Diphenyl-1-picrylhydrazyl (DPPH) radicals were measured as referenced (Chen et al., 2021). The fermentation supernatant (200 μL), 0.2 mM DPPH (400 μL), and methanol (400 μL) were mixed and reacted at room temperature in the darkness for 20 min (Chen et al., 2021). The absorbance of samples was measured at 515 nm. With the basic medium as the control, the DPPH free radical scavenging activity was calculated according to the following formula: Scavenging rate (%) = (1 − As/Ac) × 100. Ac, absorbance of the control group; As, absorbance of the sample.

The capacity of fermentation broth samples to scavenge superoxide radicals was measured as referenced (Chen et al., 2021). The fermentation supernatant, 338 μM of nicotinamide adenine dinucleotide (NADH), 30 mM of phenazine methosulfate (PMS), and 72 μM of nitro blue tetrazolium chloride monohydrate (500 μL each) were mixed and reacted at room temperature for 15 min. The absorbance of samples was measured at 560 nm, and the essential medium was used as the control. The superoxide radical scavenging activity was calculated according to the following formula: Scavenging rate (%) = (1 − As/Ac) × 100. Ac, absorbance of the control group; As, absorbance of the sample.

The OD₆₀₀ change reflected the growth of gut microbiota in bacterial media. As shown in Figure 2A , the OD₆₀₀ value of the BP group increased from 0.16 ± 0.00 to 0.72 ± 0.07 during the first 24 hours, and that of the BO group raised from 0.16 ± 0.02 to 0.70 ± 0.04. For the Control group, the OD₆₀₀ value raised from 0.15 ± 0.02 to 0.47 ± 0.03. The growth rates of the BP and BO groups in the logarithmic growth period were higher than that in the Control group. Also, the OD₆₀₀ in the final platform period was significantly higher compared with the Control group (p < 0.05). The result shows that either BP or BO could improve the growth of gut microbiota.

The sugar products of bacteria metabolism can reduce the pH of culture media (Koh et al., 2016). As indicated in Figure 2B , the pH value of the BP group decreased from 7.81 ± 0.20 to 6.24 ± 0.43 after the fermentation for 6 h, and that of the BO group reduce from 7.88 ± 0.19 to 6.20 ± 0.41. The pH values of both groups were statistically lower than those of the Control group at all time points (p < 0.05), which decreased from 7.88 ± 0.18 to 7.24 ± 0.04. The reduction in pH values of the BP and BO groups may contribute to the rapid metabolite production of polysaccharides or oligosaccharides by gut microbiota (Huang et al., 2020).

Previous studies show that gut microbiota can degrade polysaccharides into reducing oligosaccharides and monosaccharides by secreting glycosidases (La Rosa et al., 2019; Singh, 2019). In this study, the fermentation of intestinal flora resulted in a decreased total carbohydrate and changed reducing sugar contents at different time points ( Figures 2C, D ). The concentration of total carbohydrates in the BP group decreased from 2.54 ± 0.30 mg/mL to 0.35 ± 0.08 mg/mL, but the concentration of reducing sugar had no significant change. For the BO group, the concentration of total carbohydrates decreased from 2.56 ± 0.23 mg/mL to 0.37 ± 0.01 mg/mL, and the content of reducing sugar reduced from 1.61 ± 0.22 mg/mL to 0.00 ± 0.02 mg/mL. There was no change in the concentration of either total carbohydrate or reducing sugar for the Control group during the whole fermentation process. Compared to the Control group, the concentrations of total carbohydrates in both the BP and BO groups decreased remarkably within the first 24-hour fermentation (p < 0.05). Though the content of the original reducing sugar in the BO group was much higher than that of the BP group ( Table 1 ), it was consumed rapidly and almost fully utilized within 24 hours ( Figure 2D ).

The TLC spectrum of BP fermentation products is shown in Figure 3A . During the fermentation, the imprint of BP at the original site was gradually weakened, but no new oligosaccharide bands appeared. We presume that the newly produced oligosaccharides may be rapidly utilized by gut microbiota, which was consistent with the change of reducing sugar in the fermentation broth of the BP group ( Figure 2D ). Based on the HPGPC analysis in Figure 3B , the average molecular weight of BP gradually decreased from 221.17 kDa to 3.00 kDa within 24 h. After the fermentation for 48 h, a new oligosaccharide peak was generated at 17 min, and its average molecular weight stabilized at about 3.00 kDa ( Supplementary Table 3 ). The result suggests that gut microbiota preferentially chose low molecular weight carbohydrates as their substrates.

By TLC assay, the oligosaccharide bands of BO were observed with different polymerization degrees after the fermentation ( Figure 3C ). And these oligosaccharide products were quickly utilized by gut microbiota within 12 h as indicated by the weakened TLC detection band ( Figure 3C ). Due to the low polymerization degree of BO degradation products, we failed to obtain them by alcohol precipitation. Thus, the average molecular weight of BO in the culture medium did not change significantly during the fermentation ( Figure 3D and Supplementary Table 3 ).

In our previous studies, BO displayed a regulatory effect on the composition of gut microbiota in obese mice (Hu et al., 2020). In this study, 16S rDNA from human fecal fermentation samples were sequenced to verify the effects of BP and BO on the composition of gut microbiota. After removing the unqualified sequences, a total of 21,649 valid ones were obtained for each sample ( Supplementary Table 4 ). The filtered sequences were clustered into an OTU with 97% similarity. The number of OTUs of each sample mainly represented the richness of sample diversity, and a total of 245 OTUs were obtained ( Supplementary Table 5 ). The α diversity index of samples from different groups was indicated in Figures 4A−D . There were no marked differences in the α diversities of gut microbiota among the three experimental groups. However, they were significantly lower than that of the OFF (Original fecal flora) group (p < 0.05), similar to the results of previous in vitro fermentation studies (Chen et al., 2017; Chen et al., 2019; Zhou et al., 2020). The decreased gut microbiota diversity might be attributed to that the present in vitro culture system can’t entirely meet the growth conditions of all gut microbiota. The β diversity reflected the difference of species composition between different experimental groups. The results of PCA and NMDS ( Figures 4E, F ) suggested that both the BP and BO groups displayed a close cluster, which was away from the Control and OFF groups.

Next, we analyzed the composition and structure of gut microbiota among fermentation groups with different treatments by 16S rDNA sequencing ( Figure 5 ). At the phylum level, the dominant bacterial communities in initial fecal samples of the OFF group were mainly Firmicutes, Bacteroidetes, and Proteobacteria ( Figure 5A ). And the relative abundances of Firmicutes, Bacteroidetes, and Proteobacteria reached 99.28% of the total bacterial community. After the fermentation for 24 h, the gut microbiota composition among the three groups was significantly changed. Compared to the Control group, the abundance of Bacteroides in the BP and BO groups was statistically reduced (38.11% for the Control group, 5.92% for the BP group, and 10.21% for the BO group). In comparison, the content of Firmicutes was significantly elevated (52.38% for the Control group, 66.26% for the BP group, and 61.20% for the BP group). The result indicated that both BP and BO treatment could regulate the composition of gut microbiota during fermentation. At the family level, BO and BP increased the abundances of Veillonellaceae and Streptococcaceae, but reduced the contents of Enterobacteriaceae and Bacteroidaceae. It seemed that BP displayed a stronger regulatory effect on gut microbiota communities than BO.

At the genus level, the change trends of gut microbiota after BP and BO treatment were similar to the above results ( Figure 5C and Supplementary Figure 7 ). Overall, BP had a more significant impact on intestinal flora than BO. For instance, BP presented a more substantial effect on increasing the abundances of Veillonella and Streptococcus, and decreasing the contents of Escherichia and Bacteroides in comparison with those by BO treatment. Further, we confirmed the sequencing results of gut microbiota by qRT-PCR ( Figure 5D and Supplementary Figure 7 ). In parallel, BP and BO greatly promoted Veillonella and Streptococcus, and suppressed Escherichia.

Linear discriminant analysis effect size (LEfSe) analysis is an algorithm for high-dimensional biomarker discovery and interpretation, which can find statistically different biomarkers between different groups. To identify the specific bacterial taxa among three groups with different treatments, we compared their gut microbiota compositions using the LEfSe method ( Figure 6 ). The cladogram in Figure 6A revealed those taxa enriched in each group (‘p_’, phylum; ‘o_’, order; ‘c_’, class; ‘f_’, family). The results of LEfSe analysis are shown in Figure 6A . In the control group, BP group, and BO group, there were five, seven, and four dominant taxa with statistical significance identified in the Control, BP and BO groups, respectively. Then, linear discriminant analysis (LDA) coupled with LEfSe screened the significantly different taxa in individual experimental groups. As indicated in Figure 6B , ten, ten, and seven taxa with LDA scores higher than 2.0 were separate from the Control, BP, and BO groups. Among these bacteria, some belong to the probiotic ones. For example, Streptococcaceae, Lactobacillales, and Bacilli were the dominant characteristic bacteria in the BP group, while Christensenellaceae, Alcaligenaceae, Burkholderiales, and Betaproteobacteria dominated the BO group.

As shown in Figure 7 and Table 2 , the concentrations of SCFAs in the culture medium were changed during in vitro fermentation of gut microbiota with BP or BO treatment. Within 24 hours, the acetic acid content of the BP group increased from 61.45 ± 9.90 μM to 287.76 ± 56.52 μM and from 11.41 ± 0.24 μM to 161.39 ± 2.41 μM for butyric acid ( Table 1 and Figures 7A, C ). The concentrations of both SCFAs were significantly higher than those of the Control group (p < 0.05 or 0.01). For the BO group, the concentration of propionic acid in the culture medium increased from 7.09 ± 0.29 μM to a maximum of 95.37 ± 0.29 μM at 48 h ( Table 1 and Figure 7B ) (p < 0.01, vs. Control group). Compared to the Control and BO groups, the gut microbiota of the BP group produced a more considerable amount of total SCFAs during fermentation (p < 0.01). Based on the above, although BO was easier to be utilized by gut microbiota, BP had more advantages in promoting SCFA production during in vitro fermentation.

Finally, we detected the scavenging effect of fermentation broth on free radicals in the BP and BO groups. After the fermentation for 12 h, the scavenging capacity against DPPH free radical increased to 68.27 ± 0.27% for the BO group, 40.13 ± 1.95% for the BP group, and 39.06 ± 0.04% for the Control group ( Figure 8A ). During the whole fermentation period, the BO group displayed higher scavenging effect on DPPH (p < 0.01, vs. Control and BP groups). Only after the fermentation for 24 h, the BP group shows a better suppressive activity for DPPH than that of the Control group (p < 0.01). On the contrary, the BP group had a much higher scavenging effect on O₂- than that of either the BO or Control group (p < 0.01), both of which failed to scavenge the production of O₂- ( Figure 8B ). The maximal activity of BP was reached at 48 h (22.16 ± 0.16%). The results implied that glucomannans with different polymerization degrees had specific selectivity for the scavenging of free radicals.