In addition to soy macronutrients, soy contains relevant concentrations of genistein, daidzein, glycitein, and their respective β‐ and malonyl glycosides 4 . When soy plants encounter environmental stress, such as infection by a microorganism, the plant produces glyceollins (glyceollin I, glyceollin II, and glyceollin III) in defense. These molecules act as antibiotics and antioxidants to defend against the infection 5 . Because glyceollin exposure after an oral administration is low 6 , 7 and because they are potent antioxidants with antibiotic activity, we speculated ingesting soy macro‐ and micronutrients would add novel dietary diversity to facilitate a regimen shift in the GI microbiome. In a diet‐induced obese (DIO) mouse model, addition of this GI microbiome modulator (GIMM) to the simple obesogenic diet resulted in a microbiome shift characterized with impeded dietary fat uptake, reduced bile acid pool, and increased production of short chain fatty acids (SCFA). These changes contributed to improved body composition.

The gastrointestinal (GI) microbiome is an ecosystem composed of microbiota communities in a solution of partially digested and undigested foods with microbiota secreted products 1 , 2 . The ultimate regulator of the GI microbiome is diet, but modulation by drugs, dietary supplements, and the host immune system are documented. It is studied by monitoring biomarkers in fecal samples such as microbiota taxa and their secretome, nutrients, intestinal secreted proteins, and xenobiotic metabolites. Diversity of a healthy GI microbiome 1 is significantly restrained in obesity 3 . Simplification of diets, low in variety but high in energy, contributes to the loss in diversity observed in the obese microbiome. Soy offers a diverse chemical library of bioactive molecules that are lacking in most modern diets.

Results are expressed as mean ± 1 standard error of the mean (SEM). Data were analyzed using a generalized linear model and one‐way analysis of variance (ANOVA) on untransformed data (JMP 11.2.0, SAS Institute, Carry, NC). Paired Student's t ‐tests, adjusted for multiple comparisons, were used to compare data for baseline to final values. Comparisons to ObD were made using Dunnett's method when only terminal data for a variable were available. Significance for all tests was set at P < 0.05.

Colons were cut into two pieces and identified as proximal and distal. Each was embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E). Trained pathologists who were blinded to the treatment groups performed the procedures and analyses (Bolder BioPath, Inc, http://www.bolderbiopath.com /). Because there were no differences between histology of the proximal and distal sections, only data from the proximal sections are presented. A second section was cut and stained with Mayer's mucicarmine to better visualize mucus.

The Mothur ( www.mothur.org ) software was used to build contigs from raw sequencing reads, and to filter for size and chimeras following the MiSeq SOP ( http://www.mothur.org/wiki/MiSeq_SOP ) through chimera removal. Taxonomic assignment was done on the resulting containers with the RDP classifier ( https://github.com/rdpstaff/classifier ) in multi‐sample mode at a level of 0.5.

16s libraries were constructed and taxonomy analyses were performed at Cofactor Genomics ( http://cofactorgenomics.com /) who were blinded to the treatment groups. DNA was extracted using the PowerSoil bacterial DNA isolation kit (MoBio, http://www.mobio.com /). Briefly, genomic DNA (10 ng) was amplified with primers flanking the V4 variable region of the 16s ribosomal gene. Following amplification, DNA was AMPure XP bead purified. Library quality was assessed and a multiplexed sample library was made by qPCR quantification of individual samples that were pooled in an equimolar ratio. Samples were diluted to a 10 nM stock solution and prepared for Illumina sequencing.

Gross energy content of the feces was measured using an isoperibol calorimeter (Parr Instrument model 6200, Parr Instruments, Moline, IL) and mineral oil as a standard. A mineral oil spike (0.2 ml, ∼0.16 g) was added to five dried mouse fecal pellets (∼0.040 g). Caloric content of the sample was calculated in calories per gram of feces after correcting for the calorific value of the oil spike.

About 1.0 ml of 75% ethanol was added to 50 mg of dried, pulverized feces, incubated at 50°C for 2 h, and centrifuged at 1,050 g for 10 min. About 100 μl of supernatant was added to 500 μl phosphate buffered saline solution (PBS), vortexed, and were assayed using the Crystal Chem mouse total bile acids kit (Downers Grove, IL) according to the manufacturer's instructions.

Fecal pH was performed after mixing four lyophilized fecal pellets with water using a wt:vol ratio of 100:1 (mg/ml). Fecal SCFA were quantified using the method by Hernot et al. 10 with some modifications. Concentrations of acetate, propionate, and butyrate were determined in the supernatant using a gas chromatography mass spectrometry (GC‐MS) system consisting of an Agilent 7890A (Agilent Technologies, Palo Alto, CA) gas chromatograph coupled to an Agilent 5975C mass selective detector (Agilent Technologies, Palo Alto, CA) and a fused‐silica capillary column with a free fatty acid phase (DB‐FFAP 125‐3237, Agilent Technologies, Palo Alto, CA). Quantification was accomplished by measuring the peak areas for acetate ( m / z 60), propionate ( m / z 74), and butyrate ( m / z 73) relative to the internal standard 2‐ethylbutyric acid (target ion at m / z 88).

Daily collections of feces were pooled to provide a baseline and a final sample for each mouse. Four fecal pellets (∼10 mg/pellet) were homogenized in 100 μl diluent (10 ml/l Triton X‐100, 6 ml/l Brij® 30 and 0.1 mm/l HCl in saline) and thoroughly mixed by vortexing. After 30 min at room temperature, samples were centrifuged at 1,050 g for 15 min. The supernatants were assayed using a Beckman Coulter AU480 chemistry analyzer. The assays were validated for feces by spiking standards.

Fecal pellets were collected on days 10‐13 (inclusive) of the 2‐week acclimation period to serve as baseline samples and were stored at −70°C until assay. Body composition was measured using EchoMRI‐900 NMR whole body analyzer (Houston, TX) without restraint or anesthesia. Body weight and food intake were recorded weekly. Fresh food replaced any remaining at weekly intervals. Feces were collected again on days 24‐27 (inclusive) and were stored at −70°C until assay. Final body composition was measured on day 30 and the mice then were euthanized by decapitation. Trunk blood was collected and plasma was stored at −20°C for further analysis. Colons were collected, placed into 10% formalin solution, and shipped to Bolder BioPath (Boulder, CO) for tissue preparation and histological evaluation.

Mice weighing 19.89 ± 0.22 g (mean ± SEM) were chosen and were fed an obesogenic diet (D12266B; Research Diets, New Brunswick, NJ) on arrival. After 2 weeks of acclimation, 10 mice were randomly assigned to continue consuming D12266B (ObD) or 10 mice were assigned to ObD, which was modified to contain 15% ASPF (ObD‐ASPF, Supporting Information Table S1). The diets were isocaloric (Supporting Information Table S2) and balanced for macronutrient content (Supporting Information Tables S1 and S3).

Male C57Bl/6NTac mice were ordered from Taconic Biosciences ( http://www.taconic.com /) and were shipped to PreClinOmics (PreClinOmics, Indianapolis, IN) where they were individually housed and maintained on a 12:12 light:dark cycle (lights on at 21:00 h) with controlled room temperature (20‐21°C). House water was available ad libitum . The Institutional Animal Care and Use Committee of PreClinOmics approved the protocol and all procedures. The trained staff was masked to the treatments.

Sugars in ASPF were quantified by HPLC as described by Smiricky‐Tjardes et al. 8 . Other analytical assays were performed by Medallion Labs (Minneapolis, MN; www.medallionlabs.com ) using methods of analysis of the Association of Official Analytical Chemists 9 . Dietary fiber was analyzed by AOAC 991.43, fat by AOAC 996.06, and protein by AOAC 968.06. Carbohydrate was measured by difference (carbohydrates = 100 − moisture − ash − fat − protein; absorbable carbohydrates = 100 − ash − fat − protein − insoluble fiber).

Pod cross‐sections were placed on trays and then positioned into plastic containers (25.4 cm height × 27.9 cm wide × 27.9 cm deep) using spacers between each tray. Saturated potassium chloride was placed below the bottom tray to fix the humidity at 83% after the container was sealed and placed in the dark for 72 h. After incubation at 22.5°C, tissue was transferred into gallon‐sized plastic bags, sealed, and stored at −80°C. Sections were dried in a lyophilizer and milled using a Retsch Cutting Mill SM 100 with a 0.5 mm screen to produce activated soy pod fiber (ASPF). High performance liquid chromatography (HPLC) was performed to measure soy glyceollins as described previously 6 .

Soy plants (variety Pioneer 95Y61) were grown at the Louisiana State University Agricultural Center Dean Lee Research and Extension Center, Alexandria, LA. One hundred twenty thousand pods were harvested at reproductive stage 6 or when the green pod contained beans that filled the pods. Pods were collected into plastic bags that were sealed and maintained at 4°C overnight. Pods containing seeds were thinly sliced by placing pod in the food pusher of a food processor (KitchenAid® Model KFP720WH1) that was modified with a 20 ml syringe in the center to hold the pod vertical about 2 mm from the disc‐slicing blade. The thin cross‐sections were transferred as a single layer onto cafeteria trays (22.9 cm × 30.5 cm) containing a paper towel presoaked with 80 ml distilled sterile water.

Plasma proinflammatory levels after 4 weeks of feeding ObD or ObD‐ASPF. Bars are the mean ± SEM. Statistical differences between the two groups are indicated by ( P < 0.05). N = 10 for both groups.

Shifts in (A) fecal lactate and ( B ) fecal glucose content from baseline after 4 weeks (final) of feeding ObD‐ASPF. Bars are the mean ± SEM. Statistical differences between the two groups are indicated by ( P < 0.05). N = 10 for both groups.

Increased fermentation from baseline to 4 weeks (final) of feeding the ObD‐ASPF. Evidence includes (A) decreased fecal pH with (B) increased fecal short chain fatty acid content for acetate, butyrate, and propionate. Bars in panel A are the mean ± SEM. Stacked bars in panel B are the mean change from baseline for ObD and ObD‐ASPF. N = 10 for both groups.

Most genera that were increased by the modified diets are capable of fermenting carbohydrates. Therefore, we assayed for evidence of fermentation. We observed a decrease in fecal pH (Figure 5 A) with a significant increase in fecal content of SCFAs (Figure 5 B) acetate, butyrate, and propionate. Lactococcus utilize glucose to create lactate and abundances of the genus were decreased (Figure 4 ). Consistent with that decrease was a decreased fecal lactate‐ (Figure 6 A) and increased fecal glucose‐content (Figure 6 B).

Change from baseline in genera abundances of gut microbiota. Genera shown were the only statistically significant ( P < 0.05) shifts when compared to change observed in feces from the ObD group. Bars are the mean change. Abundances were the change in mean number of reads from fecal samples at 4 weeks from baseline for all 10 mice in each group.

Statistically significant shifts in abundance at the genus level (Figure 4 ) were observed. Species in five genera, Flavonifractor , Barnesiella, Bacteroides , Oscillibactor and Alistipes , were significantly increased by consumption of ObD‐ASPF. Species that were significantly decreased in abundance belong to five genera. Species in Parabacteroides , Ruminococcus , Hydrogenoanaerobacterium , and Lactococcus were significantly decreased in feces from ObD‐ASPF fed group. Species of mucolytic genera were also significantly decreased. Akkermansia species were reduced in abundance by ObD‐ASPF.

Bile acids are transformed by some intestinal bacteria and are toxic to others, which may regulate GI microbiota communities 11 . Bile acid production and secretion by the liver are stimulated by dietary fat so we characterized the microbiota profile in feces of the mice. We observed phylum‐level shifts in the microbiota composition. The shifts between the two dominant phyla, Bacteroidetes (70% of community) and Firmicutes (25% of community), tended to increase and decrease in abundance, respectively, in feces from the ObD‐ASPF group, but those differences were not statistically different. However, there was a significant ( P < 0.05) decrease in abundance of Verrucomicrobia in feces from ObD‐ASPF fed mice.

Fecal total bile acid content was decreased in mice fed ObD‐ASPF. Fecal bile acid content was measured at baseline and after 4 weeks (final) from mice fed ObD or ObD‐ASPF. Bars are the mean ± SEM. Statistical difference between the two groups is indicated by ( P < 0.05). N = 10 for both groups.

Decreased energy absorption. (A) Fecal output was evaluated weekly, and both (B) fecal caloric density and (C) fecal triglyceride (TG) content were evaluated at baseline and after 4 weeks (final) from mice fed ObD or ObD‐ASPF. Symbols and bars are the mean ± SEM. Statistical differences between the two groups are indicated by ( P < 0.05). N = 10 for both groups.

Some of the reduced energy gain was accounted for in feces. Fecal output was doubled (Figure 2 A) and the energy content (Figure 2 B) was greater ( P < 0.05) in feces from mice consuming ObD‐ASPF when compared to that from mice fed ObD. The differences in calories excreted in feces during the course of study accounted for about 85‐90% of the fat mass not gained by the ObD‐ASPF group. Some energy in feces was from triglycerides (TG) (Figure 2 C), which were increased almost 50‐fold. Fecal pellets were solid for all groups with no detectable anal oil leakage. Reduced absorption of TG may be a consequence of reduced bile acid pool (Figure 3 ). Decreased secretion of bile was consistent with decreased fecal cholesterol (Supporting Information Figure S1).

Phenotypes of mice consuming ObD or ObD‐ASPF. ( A ) Food intake was increased by week 4 in mice fed ObD‐ASPF. (B) Body weight gain tended to be less in the ObD‐ASPF group, a consequence of (C) less fat gain and (D) a slightly increased lean mass gain. Symbols and bars are the mean ± SEM. Statistical differences between the two groups are indicated by ( P < 0.05). N = 10 for both groups.

Supplementing ObD with ASPF was not appetite aversive. Daily food intake increased during the first 3 weeks for the ObD diet group but food intake continued to increase with mice that were fed ObD‐ASPF (Figure 1 A). Despite constant consumption of isocaloric diets, mice assigned to ObD‐ASPF tended to gain weight at a slower rate than mice fed ObD (Figure 1 B). Mice consuming ObD gained 5.9 ± 0.7 g, which comprised of an increased ( P < 0.05) fat mass (3.2 ± 0.8 g, Figure 1 C). Mice fed ObD‐ASPF only gained 4.4 ± 0.5 g that included an insignificant 1.8 ± 0.7 g fat mass accretion ( P > 0.05). There were no differences in lean mass change (Figure 1 D).

ASPF contained about 50% carbohydrate but total dietary fiber content contributed the majority (Table 1 ). About 30% was protein and 15% fat. The activation process was designed to stimulate glyceollin biosynthesis (Table 1 ) but we also observed that this process eliminated free sucrose, and greatly decreased both free glucose and free fructose content (Supporting Information Table S4).

Discussion

Diet is a principal element that determines the character of the GI microbiome, which may predispose an individual to metabolic disorders 1, 2. Indeed, such diseases have increased incidence, coinciding proportionally with a shift toward habitual uniform diets dominated with fat and carbohydrate. Loss of dietary diversity shifts a less diverse GI microbiome as observed in obesity 1. Dietary habits are difficult to change so we developed a GIMM from soybean pods containing glyceollins for a dietary intervention. Here we demonstrate that supplementing an obesogenic diet with the GIMM led to decreased dietary fat uptake by reducing the bile acid pool using a DIO mouse model. In turn, this improved body composition and conferred potential protection from inflammation.

Fruit of the soy plant is a pod containing soybeans. Most soy products are produced from mature soybeans. To add dietary diversity, we speculated that supplementing an ObD with contents of the entire immature pod, which is activated to produce the glyceollins, could provide a novel dietary intervention for metabolic diseases. Addition of ASPF to the ObD did not alter food intake, suggesting that it is not taste aversive. In fact, there was a trend for food intake to be increased. However, rather than an accompanying increase in weight gain, we observed reduced body weights in the ObD‐ASPF fed group. Weight gain of mice consuming ObD was primarily fat mass, which was significantly attenuated by feeding ObD‐ASPF.

Most of the attenuated fat gain can be accounted for by unabsorbed energy excreted in feces. Fecal output from ObD‐ASPF fed mice was about twice and fecal caloric content was about 1.4‐times that for mice fed ObD. Decreased calories absorbed over the course of 30 days explain 80‐90% of the reduced gain in fat mass. Fecal TG content accounts for some of the calories excreted. Loss of TG in the feces was not associated with oil leakage or oily stools as is observed with pancreatic lipase inhibitors 12. Excretion of TG in feces can be a consequence of dietary flavonoids 13, but also may be a result of a GI microbiome shift. Germ‐free mice are inefficient at harvesting energy from dietary fat and excrete 40% more TG in feces than conventionally raised mice 14.

Dietary fat stimulates bile acid secretion into the duodenum, a process that is negatively regulated by farnesoid X receptors (FXRs) in the liver and GI tract 15, 16. Primary bile acids are reabsorbed in the ileum or can be converted to secondary bile acids by GI microbiota. Most primary and secondary bile acids function as FXR agonists. However, recently, muricholic bile acids of mice were shown to function as FXR antagonists, which are increased by the presence of the GI microbiota 17. Thus, GI microbiota could alter the balance of bile acid agonists and antagonists at the FXR. In this study, we observed total bile acid content in feces of ObD‐ASPF fed mice to be significantly reduced as is also observed in both germ‐free mice and mice treated with antibiotics 17. It is possible that a component of soy protein inhibited the bile salt export pump 18. In addition, soy isoflavones are poorly absorbed, secreted into bile 19, and have agonist activity at the FXR 20. Thus, one or more soy isoflavones may function as FXR agonists.

Conjugated bile acids are toxic to bacteria and to the intestinal mucosa. However, some microbiota have an adapted ability to produce bile salt hydrolase, which benefit microbiota by establishing some resistance to the primary bile acids and also benefit the host from bile acid toxicity 21. Since we observed a decrease in fecal total bile acids, we studied the fecal microbiota profile. We observed 95% of taxa to reside in the two dominant phyla, Bacteroidetes and Firmicutes, that were similar proportions as those reported elsewhere 22. We observed statistically significant shifts in abundances of 10 genera (5 were increased and 5 were decreased).

The five genera that were increased in abundance may be sensitive to bile acids and, thus, reducing this negative influence could offer a selective advantage to these species. In addition, presenting the GI microbiome with an increased TG load could provide a selective pressure favoring some taxa. Abundances of Flavonifractor, Bacteroides, and Barnesiella were increased in feces of ObD‐ASPF fed mice. Barnesiella do not express bile salt hydrolase 21 and Bacteroides are decreased when exposed to bile acids 23. Adding dietary fat increases the bile acid pool that may be responsible for decreased abundances of Alistipes 24 and Oscillibacter 24, 25. The increased abundances observed in these genera may be a result of decreased bile acids 11, but soy isoflavones 26 may also produce a positive influence.

Species that were decreased by ASPF were in genera Parabacteroides, Ruminococcus, Hydrogenanaerobacterium, Lactococcus, and Akkermansia. Some evidence support that species of these genera bloom when exposed to a caloric dense diet. In contrast, all digestion of fat usually takes place by the time the digesta reaches the terminal ileum. TG are highly reduced and it is possible that elevated TG levels in the large intestine presents a selective pressure reducing presence of some species. Parabacteroides are increased by consumption of sugar and saccharine 27. Hydrogenanaerobacterium 22, 28 and Lactococcus 22, 24 are enriched by feeding dietary fat. Interestingly, increased abundances in Ruminococcus and Parabacteroides are associated with autism spectrum disorder 29, 30 and increases in both Ruminococcus and Lactococcus are associated with chronic bowel inflammation 31-33. Decreased abundance of Lactococcus species that utilize glucose to create lactate is consistent with our observation of a decreased fecal lactate and increased fecal glucose when ASPF was added to ObD.

Species of Akkermansia were decreased after feeding the modified diet. Very recently, Chassaing et al. reported that dietary emulsifiers stimulate a bloom in Akkermansia that are associated with low‐grade colon inflammation 34. Our observations of decreased Akkermansia are in accord because bile acids are emulsifiers that directly reduce mucus consistency 35. Akkermansia species are mucolytic. This prompted us to investigate the histology of the colon. We observed no decrease in mucus production or evidence of colon inflammation.

ObD‐ASPF stimulated a bloom of Barnesiella. Since these species are fermenters, we measured short chain fatty acids and pH of the feces. We observed significant decreases in pH with increases in concentrations of acetate, propionate, and butyrate. These data indicate that the fiber in ASPF was fermentable and are evidence of a prebiotic effect. Other prebiotics are efficacious in attenuating fat gain in DIO models 25, 36. In addition, there is an abundance of evidence supporting anti‐inflammatory effects of prebiotics so we measured proinflammatory markers in blood at the termination of the study.

We observed that plasma concentration of the anti‐inflammatory IL‐10, a counter‐regulatory cytokine was statistically increased and CXCL1, which acts as a chemokine to recruit leukocytes, was decreased only by the ASPF‐supplemented ObD. ASPF contains glyceollins 37, 38 and other flavonoids 39 that are known to be anti‐inflammatory and to have low bioavailability.