The extract enriched of alpha-amylase inhibitors were purified from white kidney bean as described previously by Chi (21). The white kidney beans were first ground into a fine powder by sieving through an 80-mesh sieve. Two hundred grams of white kidney bean powder were added to distilled water with a ratio of 1:6 material and solvent, and agitated at 4°C to be fully dispersed. After an 8 h extraction, the supernatant was centrifuged at 4,000 g for 15 min at 25°C to obtain crude white kidney beans water-soluble extract. Then, the crude water-soluble extract was micro filtered with a 0.2 μm filter. According to our previous knowledge, most of the molecular weight of α-amylase inhibitors in common beans are in the range of 10–60 kDa. The extract was firstly filtered by using membrane of 50 kDa to remove large molecules. Then the filtrate supernatant was filtered again with a membrane of 10 kDa molecular mass, to remove small molecules. Finally, the retention supernatant with 10–50 kDa molecules was freeze-dried, pulverized, and used as the PVE in this study. The yield of the PVE was 4.56 g/100 g dried bean.

The protein, fat, and carbohydrates determination procedure was adopted from Association of Official Analytical Chemists (AOAC) (1999) (22). The alpha-amylase inhibitory activity of the extract was carried out according to a modified DNS (3,5-dinitrosalicylic acid) method (23). In brief, the PVE was dissolved in a sodium phosphate buffer (pH 6.9) containing 40 U/mL of porcine pancreatic α-amylase (Sigma-Aldrich, St. Louis, USA). The reaction solution was incubated in a water bath at 37°C for 30 min. Next, 400 μL of 2% starch solution as a substrate was added to 200 μL of each test solution and incubated at 37°C for 5 min. Then, 1,000 μL of DNS was added to the mixture and boiled for 10 min. After that, 150 μl of the mixed solution was transferred to a transparent-bottomed 96-well microplate and the absorbance at 540 nm was measured using microplate reader (Bio-TEK Instruments, Winooski, VT, USA). The activity of the α-amylase inhibitor (U/g) was defined as inhibition of α-amylase which cuts the starch chain into maltose (μmol) within 1 min.

The animal experiments in this research were approved by the Ethics Committee of Beijing Institute of Genomics, Chinese Academy of Sciences with approval number of 2017H023. Sixty male C57BL/6J mice (12-month-old) were purchased from Sibeifu Laboratories (Beijing, China) and maintained on 12 h light/dark with access to food and water ad libitum. After a 1-week acclimatization period, mice were randomly divided into two groups; one group was provided with standard chow diet (n = 12) and the other group was fed with high fat diet (n = 48, 45% kcal from fat, D12451) for 12 weeks.

After 12 weeks, 48 of high fat fed mice were randomized again into four different experimental groups (12 animals per group) including HFD, PVE, Y, and YPVE groups. The HFD group received high fat diet only. According to dose conversion between animals and human (24), the PVE group received high fat diet with 180 mg/kg of PVE once per day, the Y group received high fat diet with plain yogurt (30 ml/kg), and the YPVE group was fed with a mixture of PVE with yogurt (180 mg/kg PVE+30 ml/kg yogurt). The nutrition fact of yogurt is fat content 3.1 g/100 g, total carbohydrates 8.5 g/100 g, protein 2.8 g/100 g, sodium 75 mg/100 g, and calcium 90 mg/100 g. The control group was fed with the standard chow diet. The supplementations of PVE, Y, and YPVE groups were administered via oral gavage.

Animals were treated for 8 weeks. Body weight and food intake were measured weekly throughout the study. At the end of experiment, mice were sacrificed to collect the blood. The blood was kept at room temperature for 30 min for coagulation. Serum samples were isolated by centrifugation at 1,000 g for 15 min at 4°C, then were snap-frozen and stored at −80°C for further analysis.

OGTT was performed after 8 weeks of treatment to test the function of pancreatic islet β cells and the glucose regulation ability of host. Mice with overnight fasting were orally administered two grams of glucose per kilogram body weight. Blood samples were collected from the tail vein at 0, 15, 30, 60, and 120 min after oral glucose loading. Blood glucose levels were measured at the indicated time points with a glucometer (Roche, ACCU-CHEK Active). The OGTT curve were made from glucose level vs. time. Then, Area under curve (AUC) of the OGTT curve were calculated according to the trapezoid method (25) to reflect the host body's metabolism of oral glucose.

The serum levels of insulin were determined by enzymatic methods using commercial kits (R&D Systems, Minneapolis, MN, USA). Insulin resistance (HOMA-IR) was measured by combining values of their fasting insulin and fasting glucose using a glucometer (Roche, ACCU-CHEK Active) with the following formula:

Before and after the establishment of the mice model and additive feeding testing, total triglycerides (TG), total cholesterol (TC), low density lipoprotein (LDL) cholesterol, high density lipoprotein (HDL) cholesterol of the hosts serum were recorded using an automatic biochemistry analyzer (ACCUTE TBA-40FR, Japan).

The serum levels of insulin and circulating inflammatory cytokines, such as TNF-α and IL-1β, were determined by enzymatic methods using Mouse Quatikine ELISA kits (R&D Systems, Minneapolis, MN, USA).

The frozen intestinal contents samples underwent fecal DNA extraction using QIAamp DNA stool mini kit (QIAGEN, cat#51504) according to the manufacturer's instructions. Concentration of the genomic DNAs was determined by spectrophotometry (Nanodrop ND1000; Thermo Scientific, USA) and its integrity was visually assessed by 1% agarose gel electrophoresis.

DNA libraries were constructed according to the manufacture's instruction of the NGS Fast DNA Library Prep Set for Illumina (Illumina, cat# CW2585M) and prepared with fragment length of ~300 bp and paired-end sequencing of the V4 hypervariable region of the bacterial 16S rRNA gene. Amplicon sequencing was performed on an Illumina MiSeq platform using MiSeq Reagent Kit (Illumina, cat#RH1021001). Paired-end reads were merged, demultiplexed, and analyzed by Quantitative Insights in to Microbial Ecology (QIIME) software (v2.2019.10) using a published bioinformatics pipeline (26). Sequences were assigned to OTUs with an open reference approach against the SILVA 16S rRNA reference database (release 128) clustered at 97% similarity. QIIME was used for the alpha diversity metrics to determine taxa richness (observed species) and faith's phylogenetic diversity (PD whole tree). The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (27) in BIG Data Center (28), Beijing Institute of Genomics, Chinese Academy of Sciences, under accession numbers CRA001468 and publicly accessible at http://bigd.big.ac.cn/gsa.

Taxa with zero counts were normalized to a single count across all samples. Changes in taxon relative abundance were determined by calculating the log2 values of fold change in relative abundance against the anaerobically processed matching controls across all taxa. Taxon relative abundance average fold change was computed based on the inverse logarithm of the sum of log2 fold change divided by the number of donors in which the taxa was detected. Only bacterial taxa that were present (sequence count ≥ 2) in at least one comparison group and were present in at least 97% of samples were included in the analysis. Quantitative variables were expressed as mean value and standard deviation. Variables were compared using alpha and beta diversity analysis. Microbial metabolic function was predicted using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) software. Then, dominant species were analyzed by calculating interactions among all of the microflora using Spearman's rank correlation coefficient. The threshold for significance was two-tailed, and p-value ≤ 0.05 was considered statistically significant. All of the statistical analyses were performed using R (version 3.5.2) software.

The correlation between the abundance of each bacterial genus and each blood glucose indicator was analyzed to obtain the Pearson correlation coefficient. The gut microbiota were ranked by the sum of their orders under each indicator according to the coefficients. The top 10 and bottom 10 bacterial genera were deemed glucose-related microbiota.

Data were expressed as means ± SEM of sample replicates. Unless otherwise specified, significance was determined using one-way ANOVA followed by Duncan's post-hoc tests. A p-value < 0.05 was considered to indicate statistical significance.

Firstly, the chemical components of the PVE were analyzed. We found out that the main constituent was protein of up to 75.4 ± 1.2 gram per 100 gram of the bean (g/100 g). Besides, carbohydrates were also detected with content of 14.5 ± 0.6 g/100 g. And the detection of fat showed that it was 2.8 ± 0.2 g/100 g. We also detected that the α-amylase inhibitor from our testing white kidney bean original in China was a structurally uniform glycoprotein with high content of protein, which is consistent with the previous studies in other white kidney beans (21). In our study, the α-amylase inhibitory activity against porcine pancreatic amylase of the extracts was 1,360 ± 30.2 U/g. The activity was slightly lower than that of previous report which was 1,840–5,777 U/g (29).

During the intervention, the daily food intake of HFD, Y, PVE, and YPVE groups were significantly lower than that of the control group (p < 0.05), and there was slight difference among the intervention groups (Figure 1A). The total intake of control, HFD, Y, PVE, and YPVE groups in 8 weeks were 292.32 ± 15.63 g, 222.16 ± 16.65 g, 210.06 ± 17.48 g, 202.17 ± 17.14 g, and 192.75 ± 18.97 g, respectively. As the caloric of the high-fat food was higher than that of the standard control food, the food intake of the control group was higher than that of the model and intervention groups. There was no significant difference among Y, PVE, and YPVE groups, which indicated that different supplementations had no effect on feed intake. Therefore, we believe that the additive of the extract of beans will not influence the appetite of subjects.

Furthermore, at the end of the model making period of the first 12 weeks (data not shown), the average weight of high-fat diet mice significantly heavier (36.22 ± 2.74 g) than chow-fed mice (30.51 ± 1.63 g). Under intervention, the body weight of each group was generally stable, suggesting that PVE and yogurt are generally safe as a food additive and would not affect the basic physical condition of subjects (Figure 1B).

After 8 weeks on the corresponding diets, glucose tolerance and insulin sensitivity were tested. To investigate the effects of different diets on the prevention of postprandial hyperglycemia, OGTTs were performed at Week 8 (Figure 2A). Firstly, we recorded the fasting blood glucose at the beginning of the OGTTs (0 min) and found out that the blood glucose of mice in Y, PVE, and YPVE groups were significantly lower than HFD group (p < 0.05) (Figure 2A). After glucose loading, the blood glucose of HFD group always maintained a high level at 15, 30, 60, and 120 min. However, after administration of Y, PVE, and YPVE, the glucose levels of intervention groups decreased significantly at different time intervals, comparing with HFD group. The areas under the glucose tolerance curve (AUC) were showed 21.07 ± 0.78, 23.12 ± 1.46, 21.24 ± 0.81 for PVE, Y, YPVE, respectively, which were much lower than that of 27.9 ± 1.7 in the HFD group (p < 0.05) (Figure 2B). These indicators revealed that PVE, Y, and YPVE suppressed postprandial hyperglycemia in mice fed with high-fat diets.

After 8 weeks of high-fat diet, the serum insulin was tested. The HFD group was 26.62 ± 1.16 μlU/ml which is higher than control group (23.18 ± 1.90) (p < 0.05). The blood glucose was also high (9.5 ± 0.8 mmol/L) in HFD group. The serum insulin levels in the intervention groups (Y, PVE, and YPVE) were reduced by 53.1, 68.7, and 41.5, respectively comparing with HFD group (Figure 2C). The physiological indicators revealed that hyperinsulinemia was occurred with severe insulin resistance in HFD mice (Figure 2D). While in the intervention groups, the insulin resistances were suppressed and recovered to a similar level of control group in the YPVE group (Figure 2D). These results suggested that the intervention of white kidney bean extract and yogurt decreased the requirement of insulin to maintain comparable glucose concentration, significantly improved glucose tolerance, insulin sensitivity, and insulin resistance.

TG, TC, LDL-C, and HDL-C levels were measured in the serum of mice at the end of feeding period (8 weeks). High-fat diet feeding induced an increasing in the serum TG, TC, LDL-C, and HDL-C. The serum levels of TG, TC, and LDL-C in HFD mice were increased by 95.1% (p < 0.01), 153.8% (p < 0.01), and 68.8% (p < 0.01), respectively (Figure 3). While treatment with different supplementations resulted in a significant decrease in serum concentrations of TG, TC, LDL-C. Comparing with HFD group, PVE administration increased the serum of HDL-C level significantly (2.46 ± 0.26 mmol/L vs. 1.94 ± 0.17 mmol/L, p < 0.05). All of these results could suggest that white kidney bean extract diet ameliorates the lipid profiles in serum of hyperglycemic mice.

In this study, we also investigated the effects of different treatments on chronic inflammation in C57BL/6J mice (Figure 4). The data showed that HFD increased inflammatory cytokines (IL-1β and TNF-α) in C57BL/6J mice vs. that measured in control mice. However, TNF-α content were 34.33 ± 3.8 pg/ml, 40.63 ± 4.11 pg/ml, 34.54 ± 3 pg/ml for PVE, Y, YPVE groups, respectively, which decreased comparing with HFD group. Meanwhile, those values in IL-1β were also smaller in intervention mice than they were in hyperglycemic group mice (Figure 4). The administration of PVE, Y, and YPVE significantly decreased the serum level of IL-1β and the concentration of TNF-α, which illustrated that white kidney bean could improve the inflammatory response of subjects.

To explore whether the intervention of PVE affected the characteristics of intestinal microbiota of experimental animals, we collected fecal samples from all of the subjects and extracted DNAs to profile the structural composition of the microbiota. In general, a total of 1068118 V416S rRNA sequences reads from the 30 samples with an average of 35,603 sequences reads for each sample (the minimum of one sample was 27,250 and the maximum was 61,850) were used for this project. The average length of sequence reads was 300 bp. Based on the results of the bacteria species counts, the richness was no significant change when the sample size was more than 20 which indicated that the sample size was enough for the sequencing analysis.

Based on the composition and structure of intestinal flora from control group and HFD group, we noticed that the high-fat diet can lead to changes in intestinal bacteria. At the phylum level, Bacteroidetes were decreased rapidly in HFD group comparing with control group after 12 weeks feeding (Figure 5A). It was reported as a type of weighted and healthy related bacteria, while has been checked as a positive correlation bacteria with vegetarianism and lean persons (30). Furthermore, Firmicutes and Proteobacteria were both noticeably increased in HFD group over controls (Figure 5A). Generally, Firmicutes is another type of bacteria which is positively related to obesity of their host (31). Besides, Proteobacteria is the largest of bacteria, containing many pathogenic bacteria, such as E. coli, Salmonella, Vibrio cholera, etc. (32). Therefore, the rise of these two bacteria phyla will increase the risk of illness for their host. At the genus level, the control and HFD samples were divided clearly by the significantly changed bacteria (Figure 5B). As shown in Figure 5B, most of the increasing bacteria in HFD group were harmful, including pathogenicity bacteria Streptococcus and Proteus etc. Meanwhile, most of the probiotics were reduced in HFD samples, such as Prevotella, Lactobacillus, and Bacillus. All the above revealed that diet can change the characteristics of intestinal flora, and the high-fat diet leads to an unhealthy characteristic.

Rely on diversity analysis, we observed that the diversities of intestinal microbial species in each group were generally not significantly different (Figure 5C). At the phylum level, the abundance of Firmicutes in YPVE group remained higher level than HFD group. Meanwhile, the Bacteroidetes and Proteobacteria in YPVE group was declined comparing to the rest of groups (Figure 5D). These results reminded us that, as a food additive of compound formula, the YPVE was leading to a healthy direction on the regulation of intestinal flora. In addition, the classification hierarchy tree (Figure 5E) showed that in YPVE group, the most significantly changed bacteria appeared in Clostridia and Bacteroidetes, such as Coprococcus, Allobaculum, Oscillospira, and AF12.

Furthermore, we found that YPVE group showed more stable composing and more positive changes in the intestinal flora (Figure 5F). For instance, S24-7 appears to negatively correlate to type 2 diabetes (33), and it increased in YPVE group much higher than other groups. The abundance of Ruminococcaceae was raised in YPVE group, but was the lowest in HFD group. It could cause lipid metabolism of HFD mice to decrease (34). Also, reported as opportunistic pathogens (31), Desulfovibrionaceae was less in YPVE group than those in other intervention groups. Summing up, we believed that YPVE intervention led to microbiota composition developing toward a healthy pattern.

Moreover, we utilized PCA analysis to discover the similarity of main bacteria between the samples. As Figure 5G showed, the samples in YPVE group were approaching into the same quadrant, which is much closer than the rest of groups. It indicated that YPVE reduced the individual difference in a short time and made the intestinal flora more consistent between individuals. Therefore, YPVE tends to have more forceful regulation ability than single formula additive.

Through the analysis of species composition, we found that the three groups of diet intervention had a certain impact on the intestinal flora of mice, weakened the impact of high-fat diet on the intestinal flora, and made the intestinal flora develop toward a healthy direction. There was an abundance of Bacteroides and Firmicutes in four sample groups at phylum. Y and PVE supplement the abundance of Bacteroides. However, YPVE reduce the abundance and proportion of Bacteroides. In the right panel, YPVE increased the abundance of Firmicutes, contrary to Y and PVE groups (Figure 6A). At the genus level, PVE group increased the abundance of Prevotella in a small amount compared with HFD group, but the other two groups were not able to supplement Prevotella due to the high-fat diet. YPVE group reduced Ruminococcus. The three intervention groups had no significant effect on the genus and family of Desulfovibrionaceae (Figure 6B). The intervention YPVE reduced the increase of Dorea and Streptococcus compared with the HFD group. PVE reduced the increase of Streptococcus by HFD. As known, Streptococcus contained many pathogenic bacteria. All three intervention groups increased the abundance of Allobaculum (Figure 6C). There are few studies on the relationship between Allobaculum and diabetes, suggesting that Allobaculum may grow more easily under the health condition. The addition of yogurt in Y group greatly increased the abundance of Bifidobacteria, while the increase of Bifidobacteria in PVE and YPVE was less than that in Y (Figure 6D). It has been proved that there is an increase in the number of Akkermansia bacteria in Y group. The addition of YPVE reduced the abundance of Proteobacteria and Odoribacter, which contain multiple pathogenic bacteria, indicating that YPVE is more conducive to inhibiting opportunistic pathogens and maintaining the overall balance of intestinal flora.

Through KEGG functional prediction and abundance calculation, we found out that several essential functions of gut microbiota varied in different groups, especially between YPVE and HFD group. The most obvious changes occurred in metabolic functions, including increased functions of “Amino Acid Metabolism,” “Energy Metabolism,” “Nucleotide Metabolism,” and declined functions of “Glycan Biosynthesis and Metabolism” (Figure 7A). Furthermore, the functions of “Environmental Adaptation” and “Transcription” were exhibited a rising and falling trend, respectively (Figure 7A).

We further checked the abundances for all of the sub-functions in each of the “Metabolism” functions. As shown in Figure 7B, the most increasing sub-function was “alanine, aspartate and glutamate metabolism” in “Amino Acid Metabolism” after comparing YPVE with HFD group. In clinical studies, higher alanine metabolism improves the metabolism of triglyceride, which may beneficially affect the cardiometabolic health status of men with type 2 diabetes (35). In addition, “sulfur metabolism” and “carbon fixation pathways in prokaryotes” as parts of “Energy Metabolism” were improved in YPVE group (Figure 7C). These two functions positively associated with the increasing of anaerobic bacteria which known to be probiotics (36). In addition, the functional abundance of a sub-function “ABC-type sugar transport systems” increased significantly in PVE and Y groups compared with HFD (p < 0.05), which proved that the utilization of polysaccharides in food was increased eventually (Figure 7D). In general, YPVE could promote the functions of intestinal flora to be instrumental in the host health.

Based on all the indicators, the mice could cluster into different groups which were consistent with their dietary treatments (Figure 8A). Since each group of mice have their specific gut microflora, the microflora could be associated with physical indicators. Through the correlation analysis between glucose related physiological indicators and the abundance of intestinal flora, we were able to gain the microbiota which was associated with blood glucose regulation. Comparing the intestinal microflora between intervention groups and HFD group, five of the glucose related microflora had simultaneously significant changes, containing three bacteria that positively related to the indicators (Corynebacterium, Granulicatella, and Streptococcus) and two bacteria that negatively related to the indicators (Paraprevotella and Allobaculum) (Figure 8B). These results told us that the food additives in this study could effectively reduce the metabolism of glucose in hyperinsulinemia mice by altering the composition of intestinal flora, especially by induction the change of Corynebacterium, Granulicatella, Streptococcus, Paraprevotella, and Allobaculum.