An in-vitro dynamic intestinal model, food system was developed, that can describe the effect of intestinal motility on mass transfer. The importance to the contribution of mixing to mass transfer, during bread digestion and glucose absorption can help to point out the likely molecular mechanisms underpinning glycaemia-lowering contributions of dietary fibres observed in-vivo. The model introduce mixing through segmentation and peristaltic wave-like motions, similar to in-vivo phenomena, to simulate in-vivo mixing mechanism. The effect of segmentation, peristalsis and viscous digesta properties on the kinetics of glucose absorption and bread digestion were studied. It was found that segmentation had a greater effect on transportation of sugars than peristalsis, enhancing uptake by 37%. Viscosity of the lumenal phase was also found to have a significant effect, decreasing uptake by 33%. However, there is a threshold, where where mass transfers becomes independent/ limiting of both viscosity and mixing, described by the Sherwood number as a function of viscosity. Different bread formulations containing non-digestible soluble carbohydrates (guar fibre) were investigated and shown to slow starch digestion by 52%. Breads containing arabinoxlyan, fibre fractions (AX), from two different cultivars in the bread recipe at ~15% (AX1) and ~17% (AX2) slowed digestion and showed different digestibility profiles. AX1 showed greater digestibility than AX2, absorbing a total of 8.9 mM and 5.9 mM respectively at the end of 2.5 hours. AX1 displayed an increase in overall mass transport coefficient 1.08 x10-07 m/s, by one order of magnitude faster than AX2. These reduction suggests the benefits of soluble fibres on blood plasma glucose levels. However, the overall extent and rate of starch breakdown to blood glucose in-vitro may not be due primarily on the viscosity of the digesta alone.
Introduction
When we expand our knowledge on food digestion, we improve our understanding on the effects of food on health. The challenge to maintain good health continues to climb globally, with diet-related diseases such as, diabetes (Jenkins et al., 2002) and obesity (WHO, 2000) being two of the main threats. Absorption of any nutrient in reduced or amplified quantities can be harmful to human health. Therefore, in order to optimize foods for health benefits, knowledge underlying digestion on the molecular scale is needed to understand how foods are degraded to mobilise nutrients for absorption during the process that enables digestion in the bowels (Norton, 2007; Jenkins et al., 1981; and Englyst et al., 1996).
In-vitro digestion studies are highly desired because of their uses in food and nutrition (Frei et al., 2003; Hur et al., 2011; Guerra et al., 2012) and highly advantageous in saving time and cost, and have demonstrated some levels of reproducibility when compared to clinical studies (in-vivo) in (Kong and Singh 2008a, 2008b; Kong and Singh 2010). However, without physical models that can closely simulate and represent digestion “near real”, in-vitro to in-vivo correlations will leave gaps that will reduce the true understanding of the digestion processes. Therefore, a lot of effort has been made to improve the understanding and functioning of the human gastrointestinal tract and to engineer novel in-vitro models with greater biomechanical and biochemical relevance (Boulby et al. 1999; Faas et al. 2002; Kunz et al. 2005; Goetze et al. 2007, 2009; Kwiatek et al. 2006; Marciani et al. 2001a, 2007, 2012; Marciani 2011; Schwizer et al. 2002, 2006; Steingoetter et al. 2005; Treier et al. 2006; Mackie et al. 2013). These biomechanical and biochemical relevant models follows a dynamic simulation approach, when performing digestion (Guerra et al., 2012 and Alegria et al., 2015). They simulate the physical processes of digestion (Alegria et al., 2015) such as, shearing, mixing, hydration and peristalsis (Fernandez-Garcia et al, 2009 and Wickham et al., 2009). This effort, to closely represent digestion to model the gastrointestinal complexities that will allow better interpretation of in-vitro results, makes dynamic digestion models a favourite choice in simulating aspects of digestion (Hur et al., 2011; Guerra et al., 2012).
The bulk of nutrient absorption phenomenon takes place in the small intestine, where ca. 10% of the total micro-nutrients from bulk digest escapes the absorption process (Guerra et al., 2012: Borgström et al., 1957; Guyton and Hall, 2015). The mechanisms removing up to 9/10ths of the nutrient content from bulk across the epithelium border during digestion in the body are not clearly understood. One major question emerges from this observation, ‘how does mixing and formulation influence these processes?’. This study examines this question. In particular, intestinal flow and mixing processes were simulated to perform digestion using a dynamic duodenal in-vitro model. If the digested contents were to simply propel through the small intestine, then the digestion and absorption processes would have been poorly performed, as admix of enzymes to digesta would be partial, and bulk digesta would rarely come in contact with epithelium cells for absorption to take place (Hall., 2015; Ganong and Ganong, 1995). Mixing is crucial to the digestion process in the small intestine bowel, facilitating reactions and transportation (Dikeman et al. 2006; Lentle and Janssen, 2010). Mixing is facilitated by two main processes, a combination of segmentation or peristalsis or either (Hall., 2015). However, segmentation is known to have a more powerful effect on mixing (Stoll et all, 2000; Ganong and Ganong, 1995 and Barrett et al., 2010) than do peristalsis, and is often overlook in most dynamic in-vitro models (Gouseti et al., 2014).
A lot of in-vitro procedures simulate intestinal digestion by simply mixing food and intestinal fluids using an over-head mixer (Oomen and others 2003), a shaking bath (Muir and O’Dea 1992), or magnetic stirrer (De Boever and others 2001). Clearly, these procedures oversimplify the mixing process, and will not reproduce the fluid mechanics and the mechanical forces that the digesta would face in the small intestine resulting from contractions of the gut wall. Evidence strongly supports that, the environment brought about from fluid-mechanical events during digestion have a crucial function on the material response to making nutrient becoming biologically available (Dikeman et al. 2006; Lentle and Janssen, 2010; Guerra et al., 2012; Gouseti et al., 2014; Tharakan et al., 2010).
Some of these complex dynamic in vitro models able to simulate aspects of intestinal digestion include the TIM, a gastro intestinal model, simulating peristalsis and absorption (Minekus et al., 1995 and Kheadr et al., 2010), the soft tubular bio-inspired reactor model, which simulate peristalsis frequency and amplitude to understand mass transfer and mixing (Deng et al., 2016), and the human duodenum model (HDM), which mimics the geometry of the human duodenum, consisting of ascending and descending regions capable of irregular segmenting patterns to understand transit rate (Wright et al., 2016). These models clearly simulate one or two aspects of intestinal digestion, but not all or almost real. A small intestine model (SIM), developed by Tharakan (2010) was improved by Gouseti (2014) (the Dynamic Duo) which was further improved to the dynamic duodenal model (DDM). It has enhanced features and is able to simulate most aspects of small intestinal digestion in-vitro (Tharakan et al., 2010 and Gouseti et al., 2014). Longland (1991) has stated that an effective in-vitro model should incorporate the following: sequential and physiological relevant use of enzymes; appropriate pH and simulated GI fluids; the elimination of the digested product; suitable biological transit times and mixing at each compartment for each step of digestion. The improved dynamic duodenal model was designed with these functionalities.
This study explores these effects on these contributions on digestibility and absorption while varying the digest’s viscosity. Manipulating the luminal content’s viscosity using soluble dietary fibre such as arabinoxylan and guar may affect digestibility, and has shown to alter physiologic responses (Dikeman et al., 2007; Dorota et al., 2012; Edmund et al., 2014; Pernille et al., 2013). The soft tubular material of the dynamic duodenal model vessel allows the walls and inner lumen to be actively engaging in the mixing, reactions and simulated mass transfer during digestion. It also has at pancreatic duct positioned to release intestinal fluids onto the incoming pyloric outflow. Overall, the model was developed, and the new tool attempts to create a realistic environment to perform intestinal digestion, than can describe the resistance to mass transfer during intestinal motility. To date, the dynamic duodenal model has been used to investigate digestibility changes experienced by different food products during digestion, and also to study the effect of intestinal motility on the outcomes of the digestion processes, such as glycaemic indices.
Materials and method
Preparation and characterisation of biopolymer model fluids
Aqueous guar gum-glucose solutions containing; 0.5, 0.75, 1, 1.5 and 2.5 (%w/v), gum guar supplied by Sigma Aldrich (G4129-500G) and glucose (1% w/v: 55 mM) were prepared. Gum guar was added to the water, with simultaneous stirring, and temperature raised and maintained at 80 0C for 10 minutes until fully mixed and allowed to cool at room temperature, then left overnight for complete hydration, before each use. Shear experiments were carried out on the model fluids using a Bohlin Instruments rheometer (KTB 30, Crawley, UK) equipped with a cone-plate geometry 40/1 with gap size of 1mm. For each test, 5 ml of freshly prepared solution was allowed to equilibrated at room temperature (240C) prior to experiments. The viscosities were then measured as a function of shear rates which logarithmically increased from 0.01 to 103 s-1. This was repeated in triplicates for each concentration of model solution.
Preparation of Simulated Digestive fluids
For performing digestion using various breads, all chemicals used, were purchased from Sigma Aldrich unless otherwise stated. The simulated intestinal fluids were prepared according to Mandalari et al., (2009) and Moreno, Mackie and Mills, (2005), except where otherwise stated.
Simulated Intestinal Fluids
Hepatic Mix Solution (HMS): was prepared using the following: 4 mM Cholesterol; 12.5 mM Sodium taurocholate (hydrate); 12.5 mM Sodium glycodeoxycholate; 146 mM NaCl; 2.6 mM CaCl2.2H2O; 4.8 mM KCl dissolved in distilled water
Pancreatic mix solution (PMS): was prepared using the following: 125 mM NaCl; 0.6 mM CaCl2.2H2O; 0.3 mM MgCl2.6H2O; 4.1 µM ZnSO4.7H2O solution, dissolved in distilled water.
Krebs Ringer Buffer –Modified (MKRB): was prepared as follows; of 0.7 mM Na2HP04.12H2O; 4.56 mM KCl; 0.49 mM MgCl2.6H2O; 1.5 mM NaH2P04.2H2O; 80 mM NaHCO3 solution; 54.5 mM NaCl dissolved in distilled water (Mandalari et al., 2009 and Bordoloi et al.,).
Preparation of foods for digestion
Breads
White breads: commercially available (KINGSMILL, soft white thick from Allied Bakeries, Vanwall Road, Maidenhead, SL6 4UF; ABF Grain Products Ltd.; Produced in the UK) sourced from local shops, were used for these experiments. The nutrient profiles of white breads were as follows 100 grams: carbohydrate, 44.1g and protein 7.1g. Standard control white bread were also provided by Campden BRI, United Kingdom, with the same nutrient profile.
Guar-containing white breads: were prepared in the form of dough’s using the 125g automatic KENWOOD rapid bake. Gum guar was incorporated into the recipe (245g water, 13.8g vegetable oil, 350g unbleached white bread flour, 6.5g skimmed milk powder, 5.5g salt, 7.8g sugar, and 3.5g easy blend dried yeast) at a 10% w/w, replacement level for wheat flour, similar to the method done by Brennan et al., (1996), added to the dough maker and left to be baked for 3 hrs. The final carbohydrate and protein contents were 44.5g and 6g respectively per 100 g of bread. Prior to digestion, breads were prepared to reflect a final nutrient profile of 44.1g and 7.1g carbohydrate and protein respectively per 100 grams i.e., the amount of mass was changed to keep carbohydrate levels the same (~32 grams of carbohydrate regardless of the bread type).
Arabinoxylan breads: were prepared with different levels of AX from different cultivars. Grains (Hereward AX1, and Yumai AX2) were supplied by Rothamsted Research (Hertfordshire, UK) and milled by Campden BRI (Gloucestershire, UK) to 81.4% extraction to provide white flour. These were used in the preparation of breads (100 g flour, 2 g yeast, 1.5 g salt, 1 g Bako fat emulsion, 0.01 g ascorbic acid) with fungal a-amylase (Bakezyme P180. DSM, Delft, Netherlands) added to 80 Farrand Units (based on Hagberg Falling Number) and water added to the Brabender arinograph (600 line) water absorption value using the Chorleywood bread process. After baking breads from Yumai AX2, contained 16.65 mg. g-1 total arabinoxylan (dw) and a dry matter of 57%. Herward AX1, contained 14.56 mg. g-1 total arabinoxylan (dry weight) and dry matter of 60.6 %.
Performing the digestion of breads
Simulated oral digestion of breads
Arabinoxylan breads
Batch in-vitro oral digestion was carried out according to Frances (2015), and described briefly. Simulated salivary fluid containing, 0.15 M NaCl, 6 μg/mL lysozyme, human salivary amylase (HSA) 29.7 U/g bread carbohydrate, 3 mM urea, at pH 6.9 was heated to 37°C and then added to bread to mimic chewing. In brief, bread was thawed, crust removed and cut into approximately 4 cm3 pieces and 35 g added to 12.25 mL of (SSF) containing 436 U/g of human salivary amylase before mincing (Eddingtons mincer pro, product 86002, Berkshire, UK) for 30s. Then, 24.5 mL of deionized water was added to the minced bread and mixed by hand for an additional 1 min simulating chew. Samples were taken for starch and protein analysis. The remaining “chew” was then aliquoted (bread –6.56g) on ice prior to gastric digestion. Separate oral in-vitro digests were performed in triplicate.
Guar containing and white breads
75 g of white bread were diced into 4cm3 cubes (roughly the size of bite). 24.5ml simulated salivary fluid warmed at 37 0C, containing enzymes; α-amylase, 0.018g and lysozyme 14.7µl and water 50 ml, was poured onto the bread and placed in Hammed mincer for 1-3 minutes, to mimic oral processing time, and chewing (Bolhuis et al., 201; Zijlstra, de Wijk, et al., 2009) and time to remove the chew to the gastric phase of digestion. The mincer has multiple exit diameters of 3mm each. Once the bread samples and required salivary solutions were added in complete, the timer was started. The chew (bolus) was collected, sampled for total starch analysis or immediately proceed to the gastric digestion stage. The oral digestion of pasta was similarly followed. However, oral digestion of the oat meal did not include the mincer, and the SSF was added to the meal directly and stirred briefly for 5-10 seconds before it was allowed to enter the gastric phase of digestion. It is known that oral digestion of liquid and semi-sold meals predominantly only goes through the swallowing process of oral digestion, which last only for a few seconds (de Graaf and Kok, 2010; Haber et al., 1977; Viskaal-van Dongen et al., 2011).
Simulated gastric digestion: static model
To the bread bolus (149 ml) coming from the oral phase, SGF 66 ml, containing 0.093g enzyme pepsin, dissolved at room temperature then raised to 37oC was poured onto it. The gastric chyme was then placed into a 300 ml, conical flask (stirred tank) and into an orbital incubator (Stuart orbital incubator SI500) at 37oC, set at 170 rpm for 2 hours. Samples are then taken to duodenal digestion stage if they are same day digestion or sampled for starch analysis in which case, 5 ml ice cold ethanol was added to 200 mg, boiled at 840C for 6 minutes.
Batch in-vitro oral digestion
Gastric digestion in the dynamic gastric model (DGM)
Dynamic in-vitro gastric digestions were performed using the dynamic gastric model that encompassed three compartments (Fig.5.2.), based on models described previously (Mandalari et al., 2009 and Vardakou et al 2011). Gastric digestion was performed in once completely carried out according to Frances (2015), and described briefly below. Simulated gastric fluid (SGF) pre-warmed at (37°C) containing, 0.9 mM NaH2PO4, 3 mM CaCl2, 0.1 M HCl, 0.15 M NaCl, 16 mM KCl, pH 2.5 was prepared containing 63 U of pepsin per mg of bread protein to be digested. Digestion began by adding 3 mL of SGF to each “chewed” bread samples and the pH adjusted to 2.5 by addition of 1 M HCl. After, samples were placed in a shaking incubator at 37°C, 170 rpm and incubated for 0.3, 11, 22, 33, 44, 55, 66, 77, and 120 min. Digestion was stopped at each time point by raising the pH to 7.5 with the addition of 0.5 M NaHCO3. Volumes of all additions were noted to account for dilution effects in subsequent analyses. Samples were placed on ice and then stored frozen at –20°C until required for further analysis, and digestion in the duodenal phase.
Simulated duodenal digestion in the dynamic duodenal model (DDM)
Simulated small intestine duodenal digestion When gastric digestion was completed, the gastric digesta (214.5 ml) was released and added to the dynamic duodenal model at the entry of the lumen (Figure 3.1. describes the model). After addition, simulated segmentation and peristaltic motions were initiated, section 3.3.2., describes and discusses the operation of the model. From the secretory port, pancreatic mix solution (57ml), containing room temperature dissolved trypsin 0.126g from porcine pancreas-lyophilized powder, (~780 BAEE units/mg) from Applichem (A4148), α-Chymotrypsin 0.008g from bovine pancreas-lyophilized powder, 350U/mg (11.9 BTEE U/mg) and α-amylase 1.007g were added, along with HMS (20ml) and KRB (80ml). Prior to intestinal digestion the model luminal membrane was conditioned with 50 ml KRB.