Impact of Glycosidic Bond Configuration on Short Chain Fatty Acid Production from Model Fermentable Carbohydrates by the Human Gut Microbiota

Short chain fatty acids (SCFA) are the major products of carbohydrate fermentation by gut bacteria. Different carbohydrates are associated with characteristic SCFA profiles although the mechanisms are unclear. The individual SCFA profile may determine any resultant health benefits. Understanding determinants of individual SCFA production would enable substrate choice to be tailored for colonic SCFA manipulation. To test the hypothesis that the orientation and position of the glycosidic bond is a determinant of SCFA production profile, a miniaturized in vitro human colonic batch fermentation model was used to study a range of isomeric glucose disaccharides. Diglucose α(1-1) fermentation led to significantly higher butyrate production (p < 0.01) and a lower proportion of acetate (p < 0.01) compared with other α bonded diglucoses. Diglucose β(1-4) also led to significantly higher butyrate production (p < 0.05) and significantly increased the proportions of propionate and butyrate compared with diglucose α(1-4) (p < 0.05). There was no significant effect of glycosidic bond configuration on absolute propionate production. Despite some differences in the SCFA production of different glucose disaccharides, there was no clear relationship between SCFA production and bond configuration, suggesting that other factors may be responsible for promoting selective SCFA production by the gut microbiota from different carbohydrates.


Introduction
There is increasing evidence that bioactive molecules produced during colonic bacterial fermentation play a central role in the beneficial effects of non-digestible carbohydrates (NDC) [1]. Dietary NDC are the major fermentable component of dietary fiber and have been associated with positive health outcomes related to appetite regulation, body composition and metabolic health [2,3]. Short chain fatty acids (SCFA) are the main products of saccharolytic bacterial fermentation of NDC, with acetate, propionate, and butyrate being the principal SCFA produced [4]. Although there is considerable variability between individuals, SCFA are produced in the approximate ratio 60:20:20 for acetate, propionate and butyrate respectively [5]. With increased interest in the effects of SCFA on human metabolic health [2], there is a growing need to understand how NDCs and the gut microbiota interact to yield different patterns of SCFA production and if certain profiles are related to improved health outcomes.
Acetate is a central metabolite in intermediary metabolism and is involved in lipid and cholesterol synthesis [6,7] and central appetite regulation [8]. In addition to inhibiting de novo cholesterol and butyrate [32] and in contrast both, Ruminococcus obeum related species and Roseburia inulinivorans are involved in the production of propionate [33]. The relationship between monosaccharide composition, intra-molecular glycosidic bond configuration and chain length appears to be complex. However, attempts to model the fermentation process in silico have shown some promise. Using a differential equation model, many of the changes in the microbiota and key metabolites, such as SCFA, have been simulated in continuous flow fermenters inoculated with human fecal microbiota [34].
In the present study, the relationship between glycosidic bond orientation and position and SCFA production was investigated using glucose-glucose disaccharides as model NDC substrates. This allowed for experimental control of confounding factors such as solubility and monosaccharide compositions on SCFA production.

Substrates
Glucose-glucose disaccharides (diglucoses) were used as model substrates to explore all bond options on SCFA production with the exception of diglucose β(1-1) β (β, β-trahalose) which was not commercially feasible. Disaccharide substrates were obtained from Carbosynth (Berkshire, UK). Bond linkages investigated were; diglucose A blank (no-substrate) control was also used to adjust for background SCFA production. Due to the cost of some substrates, miniaturized batch fermentations of those used previously were established [35,36]. In the present study, the method was proportionally scaled down by a factor of 20 (to 2.5 mL) and validated against our standard larger (50 mL) fermentation system.

Batch Fermentations
Fermentation vials contained: 50 mg of each substrate, 0.1 mL of reducing solution (per 50 mL 312.5 mg cysteine hydrochloride, 2 mL 1 M NaOH, 312.5 mg Na 2 S·9H 2 O, dH 2 O to 50 mL) and 2.1 mL of pre-boiled, and cooled under oxygen free nitrogen (OFN) fermentation medium (per liter 2.25 g tryptone, 450 mL dH 2 O, 112.5 µL micromineral solution (13.2 g CaCl 2 ·2H 2 O, 10 g MnCl 2 ·4H 2 O, 1 g CoCl 2 ·6H 2 O, 8 g FeCl 3 ·6H 2 O to 100 mL using dH 2 O), 225 mL bicarbonate solution (2 g NH 4 HCO 3 , 17.5 g NaHCO 3 to 500 mL with dH 2 O), 225 mL macromineral solution (providing buffering) (2.85 g Na 2 HPO 4 , 3.1 g KH 2 PO 4 , 0.3 g MgSO 4 ·7H 2 O to 500 mL with dH 2 O) and 1.125 mL 0.1% resazurin (an indicator of anaerosis), and adjusted for a starting fermentation pH of 7 was sealed with airtight seals and degassed under oxygen free nitrogen (OFN) for one minute. Per vessel, 250 µL of 32% fecal slurry prepared by adding pre-boiled and OFN cooled sodium phosphate Sorensen's buffer to a homogenized stool and blended (with a standard household blender) and strained through nylon (to remove large food residues) before being injected into the vial. The vials underwent a further degassing with OFN for 1 min before being incubated in a shaking water bath at 37 • C. After 0, 8 and 24 h of incubation, 400 µL of fermentation fluid were taken for pH measurement (Mettler Toledo pH meter), and 150 µL of 1 M NaOH added to stabilize the SCFA before extraction.

Participants
Stool samples were obtained from 15 healthy Caucasian individuals (8 females, 7 males, median age, 28; 20-52 years), who had not taken antibiotics in the previous 6 months and had no gastrointestinal disease. Ethical permission was granted by the College of Medical, Veterinary and Life Sciences Ethics Committee, University of Glasgow (Application No.: 2011023) with the sample providers giving informed signed consent. Stool samples were prepared and incubated within 2 h of being voided.

SCFA Extraction and Analysis
For SCFA extraction, 200 µL of the fermentation fluid were extracted. Orthophosphoric acid (50 µL) and 50 µL internal standard (2-ethyl butyric acid in 2 M NaOH; 73.78 mM) were added and mixed by vortexing. To this mixture, 1 mL of diethyl-ether was added and vortexed for 1 min. The ether phase was removed, and the ether extraction repeated twice more with the three extracts combined and stored in gas-tight vials at −20 • C until analysis by gas chromatography with flame ionization detection (GC-FID). SCFA concentrations were calculated using the ratio of each SCFA to the internal standard (73. 8  Statistical analysis was conducted using SPSS version 22 (IBM, Chicago, IL, USA). The Shapiro Wilk test was used to test for normality of the data. One-way ANOVA with post hoc Bonferroni analysis were carried out on log transformed data on all occasions. Statistical significance was identified at p < 0.05. Statistical power was measured by post hoc power analysis using Minitab Inc. Version 16 (Minitab Inc., State College, PA, USA). With 10 incubations the study was powered (0.8) to observe a 9.5 mM (0.47 mmol/g carbohydrate/day) difference in propionate production, an additional 5 incubations provided the power (0.8) to observe a 7.5 mM (0.37 mmol/g carbohydrate/day) difference in propionate production. Previous in vitro fermentation studies have shown similar differences in propionate production between different fibers [35] and between simple sugars [21].

Subject SCFA Variability
Significant variability in SCFA production was observed between the individual donor samples. A small subset of donors (n = 3) appeared to have much higher propionate production compared with other donors (Figure 3). Butyrate production was also variable between individuals with a small subset of donors (n = 4) appearing to have much higher butyrate production. There was no consistent pattern across the substrates and participants for high SCFA yield.

Subject SCFA Variability
Significant variability in SCFA production was observed between the individual donor samples. A small subset of donors (n = 3) appeared to have much higher propionate production compared with other donors (Figure 3). Butyrate production was also variable between individuals with a small subset of donors (n = 4) appearing to have much higher butyrate production. There was no consistent pattern across the substrates and participants for high SCFA yield.

Subject SCFA Variability
Significant variability in SCFA production was observed between the individual donor samples. A small subset of donors (n = 3) appeared to have much higher propionate production compared with other donors (Figure 3). Butyrate production was also variable between individuals with a small subset of donors (n = 4) appearing to have much higher butyrate production. There was no consistent pattern across the substrates and participants for high SCFA yield.

Discussion
Determining which primary factor(s) drive production of individual SCFA could enable selective manipulation of SCFA production in the colon. Such targeted activation through dietary means would enable selective impact on metabolism and health. For example, Chambers et al. demonstrated that selective increases in propionate production led to improvements in appetite regulation, hepatic and intra-abdominal visceral fat accretion and body weight gain in overweight adults [2]. Although limited amounts of dietary glucose disaccharides are likely to reach the colon due to digestion and/or absorption in vivo, they represent useful model compounds for in vitro studies because they control for other potential confounding factors which can influence SCFA production such as monomer composition and solubility. In the present study, glycosidic bond position had little impact on SCFA production except for diglucose (1-1) which had marked effects on acetate and butyrate production with α(1-1) producing increased butyrate compared to the other α anomers (p < 0.05, and for α(1-2); p < 0.06), and the lowest acetate (p < 0.05 for α(1-1) and α (1-6)). The anomeric orientation on diglucose (1-4) also had modest effects on the production of all three SCFA. Interestingly, there were no significant effects of bond configuration on propionate production. Many of the changes in butyrate production were associated with a reduction in acetate. This is in line with previous isotope based observations that a significant proportion of butyrate production is derived from acetate through inter-conversion [37,38].
As a result of the high cost of the glucose disaccharides used, fermentation experiments were miniaturized and the number of fermentations possible was restricted. Miniaturization is likely have led to increased variability due to propagation of errors when dealing with smaller samples. This may have increased the variability of the data but the study was powered to observe a difference of 9.5 mM (equivalent to 0.47 mmol/g carbohydrate/day) in propionate production from NDC in vitro which also demonstrated an effect on energy intake in humans [39]. The yield of SCFA was lower than expected from stoichiometric equations which may suggest incomplete fermentation of the substrates with the experimental timeframe. This could be supported indirectly by the observed rapid production of SCFA and drop in pH which would become increasingly inhibitory to ongoing fermentation. A subpopulation of individuals did appear to have greater capacity for propionate and butyrate production compared with other sample donors. Previous studies have indicated that SCFA production varies with pH, with lower pH favoring butyrate production and higher pH favoring propionate [32]. In our study, pH was measured but no variations were observed between participants that would explain the higher propionate and butyrate production seen in this small subset of participants. The reasons for this are unclear but variations in bacterial groups involved in propionate and butyrate production have been observed previously [33]. Recent work has also demonstrated that the microbiota responds quickly to dietary changes [40] and the changes in activity observed within the subgroups in this study may reflect variations in dietary intake (which was not controlled) in the days prior to sample collection.
The physicochemical properties of NDC are thought to play a role in determining SCFA production since some NDC appear to selectively increase production of individual SCFA. Changing the physicochemical properties of an NDC can also affect SCFA production. For example, increased solubility as a result of pyrodextrinization of starch, which introduces β bonding into the mainly α bonded starch, led to increased production of propionate in vitro [24]. Other β-glucans have also been associated with increased propionate production including, laminarin (mainly β(1-3) and β(1-6) bonding) and psyllium (mainly β(1-4) bonding) [30]. Resistant starch consisting of α(1-4) and α(1-6) bonding and oligofructose with β(1-2) and α(1-2) bonding are associated with increased butyrate production [36,41,42]. Stewart et al. [43] elegantly demonstrated in the fermentation of inulin type fructans (ITF) that both rate and proportions of SCFAs produced are influenced by chain length. Increased molar ratios of acetate and a decrease in molar ratios of butyrate were observed as ITF chain length increased. No effect was observed for molar proportion of propionate with ITF chain length. Thus whether the selected disaccharides serve as a model for fermentation of larger and more complex oligo and polysaccharides, where both endo-and exo-glycoside hyrolases may play complementary roles in saccharolytic degradation of polymers [44], is not clear. However, the model substrates studied, uniquely allowed the effect of bond position and configuration to be studied free of the confounding effects such as additional bonding configurations invariably present in many oligoand polysaccharides, solubility and chain length-all potential contributors to changes in the rate and molar proportion of SCFA production.
From the data in the present study, it was clear that the position and configuration of glycosidic bonding in diglucose had no major impact on SCFA production profiles. There was some modest impact on SCFA as a result of anomeric orientation for (1-1) and (1-4) diglucose linkages, with both having~80% difference in butyrate production between anomers. These data are supported, in part, by previous small-scale studies. Sanz et al. demonstrated, in a single fecal donor utilizing all glucose disaccharides, no trends were associated with diglucose linkages, anomeric formation and the resulting bacterial population. Although the fermentation of (1-1) did not lead to differences in bacterial populations compared to the control, (1-4) fermentation led to a significant increase in bifidobacteria compared with the control [29]. They did however observe some changes in SCFA production. For example, diglucose isomers containing β bonds (including β(1-1)β trehalose which we were unable to include within our investigation) significantly increased propionate and butyrate production compared to the α bonded diglucose isomers. However diglucose β(1-6) bonding did not follow this pattern [29]. Diglucoses α(1-1) also resulted in significantly reduced propionate production, but significantly increased butyrate production compared to other α bonded diglucoses. The reasons for this selective increase in butyrate production with α(1-1) glycosidic bond configuration are not clear and warrant further investigation. In another study, diglucose α(1-4) and diglucose α(1-6) linkages led to non-significant increases in propionate and butyrate production [21]. However, comparisons between these studies and the present study are difficult to interpret because of their short fermentation duration and the low number of subjects.

Conclusions
In conclusion, our data show that the orientation of glycosidic linkage, between glucose monomers at least, is not the primary factor determining the SCFA production profile. This suggests that the drivers of SCFA production may be related to other facets of carbohydrate structure or microbiology. Further investigations into how other aspects of carbohydrate structure impact upon SCFA production by the microbiota are required.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript: Free fatty acid receptor 2/3 GC-FID Gas chromatography flame ionization detector GLP-1 glucagon-like peptide-1 NDC non-digestible carbohydrate OFN oxygen free nitrogen PYY peptide YY