The three substrates used were CMC (NOVON, Beijing, China), purity ≥ 99%; BG (YIKANG, Zhangjiakou, Hebei, China), purity ≥ 80%; and GOS (Solarbio, Beijing, China), purity ≥ 57%. Substrates with different DPs have different molecular weights. Therefore, we examined the molecular weights of CMC and GOS, determined their average DP and studied the effects of DP difference on the in vitro fermentation of intestinal bacteria. The molecular weight of CMC and GOS was determined by DAWN HELEOSII laser scattering (Wyatt Technology Corporation, Santa Barbara, CA, United States), using a Shodex OHPak SB806M HQ column (SHOWA DENKO K.K., Tokyo, Japan). The CMC polymerization degree was calculated to be 2194, the BG polymerization degree was calculated to be 17–22, and the GOS polymerization degree was calculated to be 1.29.

This study was carried out in accordance with the recommendations of provisions on Article 11 of the “Ethics Review Methods for Human-Related Biomedical Research (Draft for Soliciting Opinions)” (National Health and Family Planning Commission of China). The protocol was approved by the Human Research Ethics Committee of Institute of Food Science and Technology, Chinese Academy of Agricultural Sciences. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

Human fecal samples were obtained from 10 healthy donors (5 females and 5 males) who were not treated with antibiotics for more than 2 months before the trial. All participants were recruited by the following inclusion criteria: aged between 22 and 36 years old, Chinese citizens, non-smoking, and healthy. Exclusion criteria included clinically significant deviations from normal depending on investigator judgment; history or suspicion of diabetes, liver disease, kidney disease, or having a food allergy; or taking dietary fiber supplements or lipid-lowering drugs. The inoculum was prepared by suspending the fecal sample in 0.1 M phosphate-buffered solution (pH 6.5, 0.2 M NaH₂PO₄ and 0.1 M Na₂HPO₄) supplemented with 1.0% L-ascorbic acid. The concentration of the fecal suspension was 10% (wt/vol).

In vitro fermentation was performed in a penicillin vial fermenter with ten parallel and independent vials. Each vial contained 5 mL of YCFA medium (Schwab et al., 2017) and 1% substrate (wt/vol). A control (CK) was prepared with no substrate. A mixture of N₂ and CO₂ gases (80:20; 0.02∼0.04 MPa) was used to induce anaerobic conditions, and the mixture was sterilized using a 0.2-μm PTFE membrane. Each vial was autoclaved at 121°C for 20 min. Using a 1 mL injection syringe, 0.5 mL fecal suspension was inoculated into the corresponding vial [1% (wt/vol)] and anaerobic fermentation cultivation was conducted at 37°C. After 24 h of cultivation, air pressure was measured by a BMP-Test System pressure gauge (WAL Mess- und Regelsysteme GmbH, Oldenburg, Germany) and the pH value of the supernatant was measured using a compact pH meter (Model B-212, Horiba, Japan). Samples were harvested for the detection of SCFA production and bioinformatics analysis.

Short-chain fatty acids (acetic acid, propionic acid, and butyric acid) and branched SCFAs (isobutyric acid, valeric acid, and isovaleric acid) were analyzed by gas chromatography. Harvested samples (500 μL) were mixed with 100 μL crotonic acid, incubated at −20°C for 12 h, centrifuged (16060 × g for 3 min), and 100 μL was injected into GC-9720 (Zhejiang Fuli Analytical Instrument Co., Ltd) with an HP-FFAP column (30 m × 0.25 mm × 0.25 μm; Agilent Technologies Inc., Santa Clara, CA, United States). The following gas chromatography conditions were used: feed inlet parameters: 250°C, sample injection: 1 μL; carrier gas type: N₂, blow sweep flow rate 3.0 mL/L, shunt ratio 5:1; column flow rate: 2.3 mL/min; column box parameters: starting temperature: 75°C, warming rate: 20°C/min rose to 180°C, maintain 1 min, 40°C/min rose to 220°C, maintain 1 min, complete procedure time 8.05 min; FID detector parameters: temperature 250°C, tail blowing type N₂; tail blowing gas flow: 30 mL/min; hydrogen flow: 40 mL/min, air flow: 400 mL/min. Using external calibration curves, acetic acid, propionic acid, butyric acid, isobutyric acid, valeric acid, and isovaleric acid were quantified in the samples.

For analyzing the influence of different DPs on the microbiota communities in the in vitro fermentation, genomic DNA from 40 samples was extracted using the E.Z.N.A.^® Soil DNA Kit (Omega Bio-Tek, Norcross, GA, United States) following the manufacturer’s protocols. The V3-V4 hypervariable regions of the bacterial 16S rRNA genes were amplified with bacteria-universal primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) using a thermocycler PCR system (GeneAmp 9700, ABI, United States). The pooled purified amplicons were sent to Shanghai Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China), and sequenced on an Illumina Miseq sequencing platform (Illumina, San Diego, CA, United States). All of the sequences were clustered into Operational Taxonomic Units (OTUs) based on a 97% identity threshold by the SILVA database (Quast et al., 2013). The raw reads used in this study have been deposited into the NCBI Sequence Read Archive (SRA) (Accession number: PRJNA573754).

Data processing and statistical analyses were performed using GraphPad Prism version 7.0.4 for Windows (GraphPad Software, La Jolla, CA, United States¹). Principal coordinates analysis (PCA) based on R language was used to study the similarity or difference in the community composition of the three treatments, including β-diversity analysis. Independent-samples Kruskal-Wallis test was performed to evaluate the significant difference (P < 0.05 or P < 0.01) in species among the three substrates. Canonical correspondence analysis (CCA) was performed to determine the correlation between community distance matrices and environmental factors (acetic acid, propionic acid, and butyric acid). Correlation Heatmap Analysis was performed using the spearman rank correlation coefficient, visually representing the data in a defined shade of color.

The air pressure of the fermentation with CMC, GOS, and BG as substrates was compared with that of CK using a paired t-test. Results in Figure 1 show that the air pressure of the fermentations with the three substrates increased. Significantly higher air pressure was obtained in the BG-treated fermentation than in the CMC- and GOS- treated fermentations (P = 0.0012 and 0.0027, respectively). The air pressure of the BG-treated fermentation was 31.65 MPa, which was nearly twice that of the CMC-treated fermentation (15.68 MPa).

Results in Figure 2 show that the pH value decreased with decreasing fiber DP during fermentation. The pH value of the CMC-treated fermentation was close to that of CK, which was 6.5. However, the pH values of the BG- and GOS-treated fermentations were 4.6 and 4.1, respectively.

Short-chain fatty acids are universally considered to be beneficial for human health. To study the effects of in vitro fermentation on the production of SCFAs in three different DP treatments, the concentrations of acetic acid, propionic acid, butyric acid, isobutyric acid, valeric acid, and isovaleric acid were measured. The data were compared with those of CK using the paired t-test. SCFA generation after 24 h of in vitro fermentation is shown in Figure 3. Production of acetic acid, propionic acid, and butyric acid increased after the addition of the three substrates. As shown in Figure 3A, the acetic acid concentration from the GOS-treated fermentation (3197 μM) was nearly 60% more than that of the CMC- and BG-treated fermentations (1988 and 1818 μM, respectively). Similar to the results for air pressure, the propionic acid concentrations in the BG-treated fermentation were significantly different from those in the CMC- and GOS-treated fermentations (p = 0.0052 and 0.014, respectively) (Figure 3B). The butyric acid concentrations varied among CMC-, BG-, and GOS-treated fermentations (Figure 3C) with a trend of CMC (485 μM) > BG (450 μM) > GOS (292 μM). Total acid concentrations of the six acid types in the CMC-, BG-, and GOS-treated fermentations were 2980, 2795, and 3779 μM, respectively (Figure 3D). The production of isobutyric, valeric, and isovaleric acids was less than 150 μM. All of the metabolomics measurements for each sample are provided in Supplementary Table S1.

Each of the three fiber types (CMC, BG, and GOS) and CK were fermented using 10% fecal dilution from ten healthy donors. The microbiota communities were analyzed with 16S rRNA gene amplicons. A total of 3,596,733 sequences were generated from 40 samples, and after quality and chimera checking and removal of low-quality reads, a total of 1,800,373 valid sequences were obtained (Supplementary Table S2). All of the cleaned sequences were clustered into operational taxonomic units (OTUs) that shared ≥ 97% sequence identity (Edgar, 2013). The minimum read number was adopted to subsample sequences from all other samples for comparing the treatments and control samples at the same sequencing level (Supplementary Table S3). After subsampling to equal sequencing depth, 27,749 reads per sample were used for further analyses. The rarefaction curve (Supplementary Figure S1) tended to be flat and the amount of sequencing data was large enough to reflect most of the microbial diversity information in samples. The Shannon diversity index of all intestinal bacteria was calculated (Figure 4). By comparing the diversity of each substrate with that of the control, we obtained significant differences in the estimators of community Shannon diversity between each substrate (2.05, P = 0.0003 for CMC-treated fermentation; 1.85, P = 0.0002 for BG-treated fermentation; 1.72, P = 0.00003 for GOS-treated fermentation respectively) and CK (2.67), indicating that a lower microbiota diversity was induced with the addition of the three substrates.

To evaluate the relationship between the potential benefit of gut bacteria and DP, the bacterial communities from the CMC-, BG-, and GOS-treated and CK in vitro fermentations were analyzed at the phylum and genus levels. At the phylum level (Figure 5A), the trend of Actinobacteria abundance was proportional to fiber DP, whereas that of Proteobacteria was inversely proportional to fiber DP. The dominant bacterial communities in CK were Firmicutes, Proteobacteria, Bacteroides, and Actinobacteria, whose relative abundance accounted for 97.4% of the total bacterial communities in the BG- and GOS-treated fermentations, and 89.2% in the CMC-treated fermentation because of the higher relative abundance of Fusobacteria. The relative abundance of Bacteroides dramatically increased with the addition of the three substrates. Proteobacteria abundance increased with CMC treatment. A decrease in the relative abundance of Firmicutes was observed in the three treatments compared to that in CK.

At the genus level (Figure 5B), the relative abundance of Escherichia-Shigella, Fusobacterium, Dorea, and Klebsiella was proportional to fiber DP, whereas the relative abundance of Bifidobacterium, Streptococcus, and Lactobacillus was inversely proportional to fiber DP. The alpha-diversity values of microbiota in CK were more abundant than those in the three fiber treatments. The dominant bacterial community in CK was Blautia, which decreased after treatment with the three fibers. The relative abundance of Bacteroides greatly increased after treatment with the three fibers. CMC treatment increased the relative abundance of Escherichia-Shigella compared to that in CK. Fusobacterium accounted for approximately 10% of the community abundance after CMC-treated fermentation but accounted for only 2.7 and 0% after BG and GOS treatments, respectively. Additionally, with the addition of BG and GOS, the relative abundance of Prevotella_9 and Megamonas was enhanced compared to that under the CMC treatment and CK. The relative abundance of Bifidobacterium in the GOS-treated fermentation was obviously greater than that in the BG- and CMC-treated fermentations.

Through Principal Coordinate Analysis (PCoA) using the weighted UniFrac distance matrix at the OTU level, spatial separation and clustering of the fermentation bacteria from the CMC- and CK-treated fermentations were clear. The microbial community was similar in BG- and GOS-treated fermentations as part of their microbial bacteria overlapped. As shown in Figure 6, a unique microbial bacterial distribution from top to bottom with increasing values of fiber DP was observed, which suggests that fiber DP affects the gut microbiota.

The genera with significant differences by the Kruskal–Wallis H test in each treatment are shown in Figure 7. The bacteria with high relative abundance bacteria were Escherichia-Shigella (P = 0.0007), Bifidobacterium (P = 0.0322), Bacteroides (P = 0.0002), Blautia (P = 0.00003), Fusicatenibacter (P = 0.0001), Collinsella (P = 0.0388), and Dorea (P = 0.00003), which varied significantly between the three fiber treatments and CK (Figure 7A). As shown in Figures 7B–D, CMC, BG, and GOS treatment can stimulate the growth of Bacteroides, whereas growth of Blautia (P = 0.0002 for CMC-treated fermentation; P = 0.0002 for BG-treated fermentation; P = 0.0002 for GOS-treated fermentation, respectively), Eubacterium_hallii_group (P = 0.0002 for CMC-treated fermentation; P = 0.0002 for BG-treated fermentation; P = 0.0002 for GOS-treated fermentation, respectively), and Fusicatenibacter (P = 0.0002 for CMC-treated fermentation; P = 0.0376 for BG-treated fermentation; P = 0.0002 for GOS-treated fermentation, respectively) was significantly impaired. There were also unique microbiota changes in each treatment. After CMC treatment, the abundance of Escherichia-Shigella (P = 0.0010) and Lachnoclostridium (P = 0.0017) was significantly increased, whereas Bifidobacterium (P = 0.0233), Streptococcus (P = 0.0258), Fecalibacterium (P = 0.0073), and Ruminococcus_torques_group (P = 0.0173) were significantly decreased (Figure 7B). In the BG treatment group, Dorea significantly decreased (P = 0.0041) (Figure 7C), while in the GOS treatment group, Fecalibacterium significantly decreased (P = 0.0312) (Figure 7D). Some microbiota showed a higher relative abundance, such as Fusobacterium in CMC, Prevotella_9 in BG, and Lactobacillus and Megamonas in GOS treatment, which differed distinctly from those in CK. However, this difference was not significant because the individual variations in the microbial communities from each fiber treatment were greater than those between groups (Supplementary Figure S2).

Canonical correspondence analysis (CCA) revealed a connection between the community composition of microbiota at the genus level and different environmental factors for treatments with fibers of different DPs. CMC, BG, and GOS greatly affected the bacterial community composition. The results of CCA analysis (Figure 8) showed that CMC and genus Escherichia-Shigella were closely associated with the butyric acid yield. BG and GOS treatments remarkably affected the production of propionic acid and had an osculatory correlation with Megamonas, Prevotella_9, and Mitsuokella.

The Spearman correlation heatmap was used to analyze the impacts of different environmental factors on the communities of intestinal bacteria after treatment with fibers of different DPs. The results (Figure 9) showed that the intestinal bacterial community structure was simultaneously affected by Body Mass Index (BMI) (Supplementary Table S4), air pressure, acetic acid, propionic acid, and butyric acid. The effect of different microbiota on BMI, air pressure, acetic acid, propionic acid, and butyric acid differed. The relative abundance of [Ruminococcus] gansvus group (P = 0) and Fusobacterium (P = 0) was significantly positively correlated with BMI. The genera Ruminococcaceae UCG-014 (P = 0), Bifidobacterium (P = 0.003) and Ruminococcus 1 (P = 0.001) were significantly negatively correlated with BMI. The genera [Eubacterium] hallii group (P = 0, 0, 0.001, and 0), Anaerostipes (P = 0, 0, 0, and 0), Blautia (P = 0, 0, 0.005, and 0), Fusicatenibacter (P = 0, 0, 0.012, and 0), and Unclassified f Peptostreptococcaceae (P = 0.002, 0, 0.005, and 0) were negatively correlated with SCFA (acetic acid, propionic acid, and butyric acid) concentration and air pressure in turn. Additionally, the butyric acid concentration was significantly negatively correlated with Bacteroides (P = 0), Lachnospiraceae UCG-004 (P = 0), Parabacteroides (P = 0.001), Phascolarctobacterium (P = 0.007), and Escherichia Shigella (P = 0.008).

Diet has a significant influence on gut bacterial communities and their metabolites, of which the major SCFAs, acetic acid, propionic acid, and butyric acid (De Filippo et al., 2010; Tremaroli and Backhed, 2012; Carmody et al., 2015), contribute greatly to gut health. Prior studies have reported that the host genetics-driven rise in the intestinal yield of butyric acid was related to enhanced insulin response after an oral glucose tolerance test (Sanna et al., 2019). In addition, large DP fractions were shown to be potential substrates to induce restoration of other butyrate-producing bacteria, and butyric acid was only produced after fermentation of the large size-fractions (Ladirat et al., 2014). Combined with the research data, our results indicated that fibers with higher DPs can promote the production of butyric acid. Moreover, the concentrations of propionic acid and acetic acid were highest with BG- and GOS-treated fermentations, respectively, which concurred with results from previous studies (Davis et al., 2011; Carlson et al., 2017).

Changes in physicochemical properties reflect microbiota activity. Microbiota structure and abundance were altered by the three fiber treatments. In healthy adults, the most prevalent phyla are Firmicutes, Bacteroides, Actinobacteria, and Proteobacteria, as was observed in the control (Abdessamad et al., 2013). After treatment with three polysaccharides with different DPs, the trend in abundance variation of Actinobacteria, which has been used for antibiotic research for many years (Landwehr et al., 2016), was proportional to DP. The variation trend of Proteobacteria, which enhances susceptibility to intestinal damage (Mirpuri et al., 2014), was inversely proportional to DP. The abundance of Prevotella, which is particularly common in long-chain carbohydrates or plant-based foods dietary patterns (Wu et al., 2011), in the BG- and GOS-treated fermentations was significantly higher than those in the CMC-treated fermentation and CK. In summary, there is a correlation between fiber DP and probiotic effect.

The abundance of Escherichia-Shigella and Fusobacterium greatly increased during in vitro fermentation using the CMC substrate. Interestingly, Escherichia-Shigella induces inflammatory reactions and ulceration as well as bloody or mucoidal diarrhea (Ud-Din and Wahid, 2014). In addition, the relative abundance of another special genus, Fusobacterium, was significantly higher during CMC-treated fermentation compared with the other fiber treatments with lower DP. Fusobacterium is mainly associated with cancer cells in metastatic lesions (Bullman et al., 2017). Moreover, CMC, which has a higher DP, was used as an inflammation model in previous research (Wen et al., 2014). In summary, a key factor underlying the pro-inflammatory effect of CMC might be the aforementioned bacteria. Previous studies indicated that BG, which has an intermediate DP value among the three substrates, can enhance the abundance of Bacteroides-Prevotella and the production of propionic and butyric acids during in vitro fermentation (Louis and Flint, 2017), and can decrease the relative abundance of Dorea, which is regarded as a pathogenic bacterial genus. Supplementation with GOS, which has the smallest DP value among the three substrates, is considered to exert a “bifidogenic effect” (Davis et al., 2011; Sangwan et al., 2011). Similarly, the relative abundance of Bifidobacterium significantly increased in the GOS-treated fermentation. Moreover, a decrease in the abundance of Fecalibacterium has anti-inflammatory functions in Crohn disease patients (Sokol et al., 2008).

Physicochemical properties are closely linked to microbial community structure and abundance. The yield of butyric acid in the CMC-treated fermentation was the highest among the three substrates. Coincidentally, the abundance of Escherichia-Shigella was also the highest in the CMC-treated fermentation. Butyric acid is beneficial to enterohemorrhagic E. coli by inducing the production of flagella that regulates its motility function (Herold et al., 2009; Tobe et al., 2011). It also promotes virulence factor gene expression in pathogenic E. coli and thus, colonization of the colon where levels of butyric acid are the highest (Nakanishi et al., 2009). CMC might induce inflammation as it promotes the production of large amounts of butyric acid by intestinal bacteria, and a large amount of butyric acid can boost the expression of Escherichia-Shigella virulence factors, thereby triggering inflammation. Reasonably, Prevotella spp., which are overrepresented in the presence of autoinflammatory diseases (Lukens et al., 2014), showed a low relative abundance in the CMC- treated fermentation.

In BG- and GOS-treated fermentations, the abundance of beneficial genera Bifidobacterium, Prevotella_9, Bacteroides, Megamonas, Lactobacillus, and Megasphaera (Kwon et al., 2018) improved compared to that in CK. Increased abundance of Bifidobacterium, an acetic acid producer, is negatively related to diabetes, hepatocellular carcinoma, and non-alcoholic steatohepatitis (Cani et al., 2007; Ladirat et al., 2014). Similarly, in the GOS-treated fermentation, both the acetic acid concentration and abundance of Bifidobacterium were the highest among the three fermentation groups (Figures 3A, 5B). Prevotella, a commensal bacterium, improves the ability of intestinal bacteria to degrade complex polysaccharides and fibers (Petia et al., 2015). Bacteroides, a butyric acid producer (Louis et al., 2004; Shin et al., 2017), Lactobacillus, predominantly a lactic acid producer, and Megasphaera, a lactic acid and SCFA producer (Luo et al., 2017), induced an acidic environment, thereby decreasing the abundance of Escherichia-Shigella (Meng et al., 2017). The pH was close to neutral in the CMC-treated fermentation, which provided Escherichia-Shigella with a suitable growth environment.

Overall, BG and GOS have potential benefits for modulating the community structure of beneficial intestinal bacteria. This study provides new insight for studying the effects of dietary fibers with different DPs on intestinal bacteria. The future development of human intestinal bacteria as food additives can also be quickly assessed using in vitro fermentation techniques. Specific elements of the enteropathogenic microbiota, such as Escherichia-Shigella, dramatically increased during CMC fermentation whereas the majority of the increased microbiota in BG- and GOS-treated fermentations were beneficial to humans. Therefore, the dosage of CMC should be strictly controlled in food. Although interpersonal deviation exists, the data indicate a positive correlation between beneficial bacteria and BG and GOS, indicating that lower fiber DP values induce greater probiotic effects. To follow up, long-term in vitro fermentation (Poeker et al., 2018) and metagenome sequencing analysis (Wang et al., 2020) are expected to provide a more in-depth characterization of how fiber DP affects the human gut and host health. In vivo experiments for additional information regarding the species of bacteria and the mechanisms underlying their effects are warranted.