DEVELOPMENT OF A STATIC IN VITRO DIGESTION MODEL FOR DETERMINATION OF GLYCEMIC

While in vivo methods have been used to determine the glycemic response of food, they are time consuming, costly, and not suitable for large-scale applications. As an alternative, in vitro digestion models offer fast, reproducible results to study food digestion kinetics that are less expensive than conducting human trials. While there are several in vitro glycemic index (GI) methods used to determine the GI of food, most do not employ methods of in vivo testing. Therefore, we used a static in vitro digestive system, the Dedicated Ryerson University In-vitro Digester (DRUID), that simulates both gastric and intestinal conditions to determine the glycemic response of commonly consumed carbohydrate-containing foods. Samples were collected at regular intervals over a 2h residence time after digestion in the intestinal phase of the DRUID. The DRUID-determined GI values were compared to published in vivo GI values. A Bland-Altman plot showed that there was agreement between the GI values determined from the DRUID compared with published in vivo GI values. In conclusion, the in vitro DRUID can reliably and reproducibly determine the GI across a spectrum of carbohydrate-containing foods, and has the potential to predict the digestion kinetics of novel food products in vivo that may promote human health.

Traditionally, carbohydrate-containing foods have been categorized according to their structural classification 3 . As a result, it was assumed that when "simple" carbohydrates were consumed, it would cause a rapid glucose response in the body thus being unsuitable for diabetics while "complex" carbohydrates were thought to have caused slower and smaller responses to blood glucose thus were believed to be a healthier option for patients with glucose intolerance 3 . However, a study by Conn and Newburgh (1939) showed that different carbohydrate foods with a similar macronutrient content elicited different glycemic responses 4 . They found that, even though the foods had similar macronutrient composition, there were differences in the rate at which they were metabolized into glucose. Jenkins et al. (1981) later concluded that the glucose response of carbohydrate-containing foods was not only a result of the primary structure of the carbohydrate but also is influenced by the physical form of the carbohydrate (like particle size and degree of hydration) and dietary fibre 5 . All of these factors helped to influence the physiological properties (ability to raise blood glucose levels) of the foods consumed and thus were important for understanding the nutritional and health effects of carbohydrates 6 .
In order to incorporate all the factors influencing the physiological properties of carbohydrates, a system was needed to classify the carbohydrate containing food based on its glycemic response 7 . This led to the development of the glycemic index (GI), which is a quantitative measure of how much a carbohydrate-containing food raises blood glucose levels relative to a standard food 8 . It is based on the degree of glucose release (physiological properties) into the bloodstream after consumption of a carbohydrate-containing food 9 .
The GI of many foods has been determined through traditional in vivo methods. However, in vivo GI methods are time consuming, costly, and not suitable for large-scale applications 10 . As an alternative, in vitro digestion models offer fast, reproducible results to study food digestion kinetics that are less expensive than conducting human trials.
While there are several in vitro GI methods that are used to determine the GI of food, most do not employ methods of in vivo testing and do not simulate the physiological conditions that occur in vivo. The purpose of this research was to develop a static in vitro digestion model called the "Dedicated Ryerson University In-vitro Digester" (DRUID) that simulates both gastric and intestinal conditions to determine the glycemic index of commonly consumed carbohydratecontaining foods.

Overall objectives
This study investigated how food is digested in an in vitro digestion model that mimics a mammalian in vivo digestive tract.
The objectives were to: ▪ Develop a static in vitro gastrointestinal model that simulates an in vivo human digestive system.
▪ Perform a series of tests to validate the DRUID for GI determination.

Specific objectives
1. Validate the model for determining the in vitro glycemic index of commonly consumed carbohydrate-containing foods (ranging from low to high GI).
2. Examine the effect of different amounts of fat on the GI of a carbohydrate containing food.
3. Calculate the available carbohydrates of foods by determining the glucose, fructose, sucrose and starch content of foods without a manufacturer's nutritional label.

Thesis layout
The thesis is presented in 9 chapters as follows:

Chapter 1: Introduction
Chapter 1 presents an introduction to the study. It includes a summary of carbohydrates, the concept and development of GI and the development of a static in vitro digestion model for determining the GI of carbohydrate-containing food.
The problem statement and motivation behind the development of an in vitro digestion model are also stated along with the objectives of the project.

Chapter 2: Literature review
In chapter 2, the background and nutritional significance of carbohydrates, GI and in vitro and in vivo digestion methods are discussed. Current in vitro digestion models are introduced.
This chapter also includes steps to designing an in vitro digestion model.

Chapter 3: Methods and Materials
Chapter 3 presents the methodology for determining the GI of carbohydrate containing food after digestion in the DRUID. This chapter introduces the selected test foods that were used for the determination of GI in vitro. This chapter also introduces the development of the DRUID and the different compartments of the DRUID.

Chapter 4: Results and discussion
In chapter 4 the assessment of the DRUID as an in vitro GI determination model is explored. Results from the determination of GI after digestion in the DRUID were compared to in vivo data from previous literature. A Bland-Altman plot was used to show the agreement between the in vitro and in vivo GI values of the selected foods. The in vitro determination of GI using the DRUID was compared with other GI determining methods. This chapter also presents a study on how fat affects the GI of white rice.

Chapter 5: Conclusion
In this concluding chapter, major findings of the project were summarized. This chapter also describes potential applications of the DRUID for determining GI.

Chapter 6: Future studies
This chapter presents the potential of the DRUID for future research in the in vitro determination of GI.
Chapter 2 -Literature review

Introduction
The focus of this chapter is to provide a background and to show the significance of carbohydrates, GI and in vivo and in vitro digestion. This chapter also introduces the major in vitro digestion models that are currently available.

Classification of carbohydrates
Carbohydrates (CHO) are classified by their monomer composition, molecular size, degree of polymerization and type of linkages 11 . This classification divides carbohydrates into four groups: monosaccharides, disaccharides, oligosaccharides and polysaccharides (Table 2.1) 12 .
The most common monosaccharides include glucose, fructose and galactose 12 . Disaccharides consist of two sugar monosaccharides linked via a glycosidic bond and common examples include sucrose (glucose + fructose), lactose (galactose + glucose) and maltose (glucose + glucose) 12 .
Oligosaccharides consists of three to nine glycosidic bond-linked residues that are water soluble.
They are generally resistant to digestion in the upper digestive tract but are fermented in the large intestine by the microflora 12 . Polysaccharides have a degree of polymerization that is greater than nine 12 . There are two main type of dietary polysaccharides that are important in human nutrition: starch and non-starch polysaccharides (such as dietary fibre).

Starch
Starch is the dominant storage carbohydrate in plants such as cereals, seeds and legumes.
Starch consists of two main types of macromolecules, amylose and amylopectin 13 . Amylose is a long, unbranched, helical chain containing 500-2,000 glucose residues linked by α-1,4-glycosidic  The helical structure of amylose arising from the twist after six glucose residues and the hydrogen bonding between the glucose chains confers amylose a compact structure (Figure 2.1).
This makes amylose less accessible to enzymatic digestion than the more branched amylopectin 14 . A study by Behall et al. (1988) showed that a meal containing cornstarch with 70% of amylose resulted in a lower glycemic response than a meal containing cornstarch with 70% of amylopectin 16 .
Non-starch polysaccharides are the other important type of polysaccharide in the human diet. According to Trowell et al. (1976), non-starch polysaccharides such as dietary fibre are plant cell wall material that do not get digested in the small intestine due to a lack of appropriate digestive enzymes 17 . They are made up of long chains of glycosidic bond-linked monosaccharides that are not always glucose residues 17 . One example of a non-starch polysaccharide is cellulose, which is the structural polysaccharide of the plant cell wall. It is made of β-1,4-glycosidic bondlinked glucose residues. It is able to form beta sheets via the higher bonds and thus is highly resistant to degradation 18 . Hemicelluloses are branched heteropolymers made up of an array of monomers including glucose, mannose, arabinose and xylose 18 .

Nutritional classification of carbohydrates
Available carbohydrates are the portion of dietary carbohydrates that can be metabolized by the body 19 . These include soluble sugars and starch (which are digested and absorbed by the body) but exclude dietary fibre 20 . In contrast, unavailable carbohydrates are the portion of carbohydrates (including dietary fibre like cellulose, hemicellulose and pectin) that are not digested by the small intestine but provide the body with energy following fermentation by the microbiota in the large intestine 21 .
Carbohydrates can also be classified by a nutritional factor as proposed by Englyst et al.
(1992) 22 . One of these factors is known as resistant starch. Resistant starch (RS) is defined as the fraction of starch and as well as its byproducts that are not absorbed by the small intestine 18 .
Englyst defined the term rapidly digestible starch (RDS) and slowly digestible starch (SDS) 22 . RDS are found in cooked bread and potatoes and these are rapidly digested in the small intestine within 20 min 22 . On the other hand, SDS which are found in seeds and grains are slowly digested in the small intestine (between 20 and 120 min) since these foods hinder access as a physical barrier for enzymatic digestion 22 .

Understanding glycemic index
The concept of glycemic index (GI) was developed by Jenkins (1981) to classify carbohydrate-containing foods according to their physiological properties (which are not evident just by the structural composition of carbohydrates) 5 . The GI is measured by determining the incremental area under the blood glucose response curve after the consumption of a test food containing 50 g of available carbohydrates and expressed as a percentage of the response to an equivalent carbohydrate portion of a reference food (either glucose or white bread) taken by the same individual 11 . The GI is calculated as follows: where iAUC stands for incremental area under the curve. GI foods such as white bread and processed cereals have GI values greater than 70 23 . Low GI foods are slowly digested and have gradual release of glucose into the bloodstream 23 . While high GI foods are quickly digested and lead to a high glycemic response with short duration 23 .
The concentration of glucose in the human body is maintained due to homeostatic equilibrium between insulin and glucagon hormones. The glycemic response is a disruption of this equilibrium due to an increase in the blood glucose concentration after consumption, digestion and absorption of a carbohydrate-containing food. Depending on factors like the type of food consumed can affect how rapidly this change in blood glucose concentration can occur.
Slow carbohydrate digestion and absorption leads to a low glycemic response, which is gradual and is long lasting. On the other hand, fast carbohydrate digestion and absorption leads to a high glycemic response, which is high in amplitude, fast and is temporary in duration. There are two main hormones that maintain the homeostatic equilibrium of blood sugar, insulin and glucagon.
Insulin is secreted by the β -cells of the pancreatic islets of Langerhans and glucagon is secreted by the α-cells of the pancreatic islets of Langerhans. When blood glucose levels are high (hyperglycemia), the secretion of the hormone insulin is triggered and insulin causes the cells of the body (specifically muscle, adipose and liver tissue cells) to take in the glucose from the blood 24 . Glucagon works to oppose the action of insulin and restore normoglycemia 25 . In response to hypoglycemia, glucagon works by causing the breakdown of glycogen (a multibranched glucose polysaccharide) which is stored in the liver to be released as glucose into the blood stream 25 .
The rapid rise in glucose concentration after the consumption of a high GI meal has been documented in the literature. A study done by Ludwig (2002) showed that within 2 hours of consuming a high GI meal, the blood glucose concentration was twice as much as that found after a low GI meal 26 . This high glucose concentration from the high GI meal triggers the increased release of insulin in order to oppose activity from glucagon 26 . Due to the activity of the high levels of insulin, blood glucose levels drop. After 4 hours, almost all the nutrients are absorbed but a high insulin to glucagon ratio can remain 27 . This imbalance in the homeostatic hormones further decreases the blood glucose concentrations. Since blood glucose is the primary source of fuel for the brain, hunger starts to develop 6 . Moreover, the high level of insulin also suppresses lipolysis and thus helps to prevent the use of free fatty acids as fuel 26 . Therefore, a high GI food causes the body to undergo similar conditions to fasting.
It is after 4 -6 hours of consuming a high GI meal that low glucose levels start to stimulate the release of glucagon, epinephrine, and cortisol to restore normaglycemia 27 . From the stimulation of these regulatory hormones, cortisol and glucagon, the body attempts to perform glyconeogenesis (synthesis of glucose from non-carbohydrate sources) and glycogenolysis (breakdown of glycogen to glucose), respectively 26 . Epinephrine helps the body to mobilize fat from adipocytes to restore the blood free fatty acid concentrations. However, due to the high levels of insulin, the body's attempts to restore normoglycemia are prevented. It is at this point that hunger intensifies and leads to over consumption of the following meal 23 .
These drastic changes in blood glucose following a high GI meal do not occur with low GI meals. In a low GI meal, the carbohydrates take a longer time to be digested and absorbed which results in a gradual change in the blood glucose level 23 . The resulting ratio of insulin to glucagon does not rise dramatically, hence gluconeogenesis, glycogenolysis and lipolysis are not inhibited 23 .

In vivo measurement of the glycemic index
In vivo methods are commonly used to determine the GI after consumption of food. In

Anatomy and physiology of the gastrointestinal tract
The gastrointestinal tract (GIT) is a single hollow tube that is 9 -10 meters in length 28 . Its main function is to process and digest the food into a form that can be absorbed into the body via the lymphatic system and the blood stream.
Mechanical breakdown in the mouth by chewing and churning action in the stomach helps to break down the digesta into a smaller size and increase the surface area for increased enzymatic digestion. This is referred to as the bolus. Peristaltic action helps to move the digesta through the GIT and also assists in mixing. Several accessory glands and organs like the salivary glands, liver, gallbladder and pancreas work with the GIT to help with chemical breakdown of the digesta.
Changes in pH in the different compartments of the GIT help maintain the optimal activity of digestive enzymes 29 .

In vivo digestion
In vivo digestion is a multi-step process involving many organs ( Table 2.2). The process of digestion begins in the mouth with the act of chewing food (mastication) which helps to breakdown food and increase surface area for enzymes to act on. Mastication also helps to form food into a bolus (lubricated and salivated portion) by mixing the food with saliva thus helping the food to be swallowed easily. Saliva is produced by serous and mucous acinar cells and secreted mostly by the submandibular salivary glands 30 . Depending on the location saliva is secreted, the magnitude of the salivary α-amylase activity changes but is typically 60 -70 U/ml 31,32 . Saliva is mostly composed of water, while also containing mucus (which helps to lubricate food and protect the mouth), electrolytes, enzymes such as salivary amylase (which starts the process of carbohydrate digestion) and lysozyme (which protects against bacteria) 29 .
Humans typically produce 0.9-1.5 L of saliva daily 29 . Water in the saliva helps to moisten and soften the food and the mucus helps to bind the bolus thus allowing the food to be easily swallowed.
Food is passed into the esophagus where involuntary contractions push the food into the stomach. Upon reaching the stomach (by passing through the lower oesophageal sphincter), salivary amylase becomes inactivated due to the acidic nature of the stomach. This acidic environment brought on by parietal cells that secrete HCl to lower the pH is needed for the activation of pepsinogen (secreted by chief cells in the wall of the stomach) to pepsin, a digestive enzyme that aids in breakdown of protein 28 . Coupled with the proteolysis by pepsin, the lower part of the stomach contracts in a rhythmic manner to chum the food inside and mix it with the gastric acid and pepsin ( Table 2.2).
The human stomach is a J-shaped organ subdivided into four portions: fundus, body, antrum and pylorus ( Figure 2.2). As it is elastic, it can expand to accommodate 1 -1.5 L of food 28 .
The stomach contains different types of cells such as chief cells and parietal cells that secrete the needed components to maintain and regulate digestion. Churning, the mechanical process that occurs in the stomach, helps to mix the digesta with gastric juice (composed of HCl, pepsin, mucus and salts) to help with enzymatic digestion. Peristaltic contractions of the antrum help to break down the digesta and push it into the small intestine 28 .
Two types of contractions occur in the stomach: peristaltic and regular tonic. Peristaltic contractions lead to grinding and mechanical breakdown of food 33 whereas the latter are responsible for moving food from the top to the bottom of the stomach.
The extent of peristaltic contractions is affected by factors such as gender, age, body mass index (BMI) and certain disorders 34 as well as the physical properties of the food such as fat and solid content 34 . Normally, as the peristaltic waves move toward the pylorus of the stomach, the width of the contractions increase causing the pylorus to contract and the sphincter to narrow 33 .
Due to this, only liquids and small particles of the digesta can be driven into the small intestine.
The larger particles experience retropulsion, where they are pushed back into the stomach for further breakdown 33 . Churning and mixing results in the formation of chyme, which is passed into the duodenum (first part of the small intestine) through the pyloric sphincter. After it reaches the duodenum, the chyme is mixed with pancreatic juices that contain bicarbonate (produced by duct cells) that help neutralize it 28 . The majority of food is digested by pancreatic enzymes in the small intestine secreted by acinar cells of the exocrine pancreas 35 . Prior to secretion, the pancreatic juices are mixed with bile secreted by the liver and stored in the gall bladder via bile duct. Bile is a cholesterol derivative made of inorganic ions, bicarbonate, bile salts and phospholipids 28 . Due to its amphiphilic nature, it disperses fats into micelles to increase their surface area which assists in lipolysis by pancreatic lipase 28 . Just as the small intestine is the site where the majority of digestion occurs, it is also the location where most of the absorption of the digested products occurs. The absorption of the resultant products after digestion by the mucosal epithelia is increased due to the mucosal epithelia taking the form of finger like projections (villi) 28 . During digestion, the three macronutrients, proteins, carbohydrates and fats are broken down into their monomers.
Protein and carbohydrates are broken down into amino acids and monosaccharides respectively and these products are taken up and transported into the blood stream and end up in the liver through the hepatic portal vein. Fats on the other hand are broken down into glycerol and fatty acids and transported as chylomicrons (water-soluble lipoproteins) which help fats and cholesterol to be transported into the lymph and finally into the bloodstream 28 . Absorption across the in mucosal epithelia (in the small intestine) includes processes such as passive and active transport, pinocytosis and even carrier-mediated transport 36 . The last stop for food is the large intestine (colon). This is also the last chance for the body to absorb any water or minerals still remaining. The remaining indigestible content in the large intestine such as fibre is passed on to the rectum where it is expelled out of the body.

Site of action Enzymes Action
Mouth Salivary αamylase

Digestion and absorption of carbohydrates
The majority of the dietary carbohydrates are in the form of starch and the remainder is in the form of oligosaccharides. These polysaccharides must be broken down to their monomers prior to absorption by the body. Starch is made up of amylose (a linear α-1,4glycosidic linkage polysaccharide) and amylopectin (α-1,4-glycosidic linkage polysaccharide with α-1,6-glycosidic linkage branches) 28 .
The digestion of starch begins in the mouth with the help of salivary α-amylase secreted by the serous acinar cells. During the process of mastication, the salivary α-amylase is mixed with the bolus and breaks down the α-1,4-glycosidic linkage of carbohydrates. As the stomach is acidic, salivary α-amylase becomes deactivated and plays a small role in digestion. After the acidic chyme passes into the small intestine, pancreatic amylase further digests the carbohydrates present 28 . The result of the α-1,4-glycosidic linkage cleavage in starch (for both amylose and amylopectin) produces maltose and maltotriose (trisaccharide of glucose monomers). However, in amylopectin, α-limit dextrins (glucose polymers with α-1,6-glycosidic linkage branch points) are also produced. This breakdown of starch happens within ten minutes of the acidic chyme entering the duodenum 37 . These products are further digested by a number of brush border enzymes found on the apical membrane of the small intestine. Thus, maltose is broken down into glucose by maltase. The terminal glucose residues on maltotriose and α-limit dextrins are cleaved by glucoamylase 28 . Lactose cleaves lactose into glucose and galactose. Finally, sucrase cleaves sucrose to glucose and fructose. The α-1,6-glycosidic linkages of amylopectin are cleaved by isomaltase. Moreover, isomaltase is responsible for cleaving α-dextrin to maltose.
After the digestion of dietary carbohydrates into monosaccharides (glucose, fructose and galactose), they are absorbed. During this process, the body converts fructose and galactose to glucose 38 . Monosaccharide absorption is highly regulated; there are two ways in which they can enter the enterocytes: transcellular or paracellular transport via tight junctions.
Glucose absorption occurs predominantly in the proximal small intestine via sodium glucose cotransporters at the luminal membrane and GLUT2 transporters at the basolateral membrane 39 . Glucose then diffuses into the intestinal villus capillary beds 40 where it travels to the portal vein, then into the liver after which it is circulated to the entire body via systemic circulation. This is when the glycemic response is generated since glucose is a major source of energy for all tissues.

In vitro digestion
In vivo methods that use human subjects to determine the GI of food from their glycemic response provide the most accurate results, but they are expensive, highly variable, labourintensive, time-consuming and can make large studies impractical 41 . Therefore, there is a strong case for the development and application of in vitro models that closely mirror the conditions and processes that occur in vivo. Such models have to be sufficiently refined to allow the process of digestion to be followed in some detail and have to be validated against in vivo data 42 . These models are non-invasive, economical and allow for the analysis of large sample sizes 42 . Ideally, an in vitro digestion model should offer the advantages of rapid representative sampling at any time point, and permit testing of whole foods.

Advantages of in vitro digestion
In vitro digestion systems have proven to be powerful tools for understanding and monitoring the complex transformation processes that take place during in vivo digestion 43 . In vitro digestion models provide a useful alternative to animal and human models by rapidly screening food ingredients. There are many benefits to initial screening experiments using these models. They provide significant physicochemical insights into the digestive processes since samples are easier to collect and analyze. They are also more efficient to conduct than animal or human studies in terms of cost and time, allowing for higher turnover of potential delivery systems under study.
A typical in vitro digestion model should consider 4 main stages: (i) mouth (oral phase), (ii) stomach (gastric phase), (iii) duodenum (small intestinal phase) and iv) large intestine (large intestinal phase). These four phases can be considered separately or in combination depending on the purpose of the study. There are two models of in vitro digestion -static and dynamic.

General background on static and dynamic digestion models
Static methods are simple to use and usually include 2 or 3 sections of the GIT (oral, gastric, and intestinal). The products of digestion are not removed during the digestion process.
This approach may only simulate a limited number of parameters relevant to digestion as physical processes such as shearing and peristalsis may not be mimicked 44 . These models are good for limited digestion but are less applicable for total digestion. These models are used for digestion studies of simple foods or isolated nutrients. Many of these models involve homogenization of the food, then acidification with HCl with the addition of gastric enzymes followed by a delay to simulate gastric residence time, and finally neutralization with sodium hydroxide with the addition of pancreatic enzymes and bile salts all the while stirring at 37 o C 44 . The progress of the reaction is measured by the rate of loss or appearance of a component 44  By contrast, dynamic digestion models simulate the continuous changes of the physicochemical conditions that occur during digestion, including changes in pH, peristaltic forces, shear, mixing, hydration and secretion release 44 . Some dynamic digestion models can also simulate nutrient absorption (like the TNO TIM-1). Digestion System (IViDiS) use a beaker alignment, which is similar to static models.

Current in vitro digestion models
There have been several different digestion models that have been developed for studying the complex processes of digestion.

Dynamic Gastric Model (DGM)
The dynamic gastric model (DGM), developed at the Institute of Food Research (Norwich, UK), is made of two compartments: the stomach and small intestine 45 . The first compartment, the stomach (mainly the fundus), mimics the dynamic conditions that occur in the human stomach like the diffusion profiles of gastric juice as well as gastric emptying. This is based on data from echo-planar magnetic resonance Imaging (EPI) and from human trials 46 . The stomach compartment of the DGM also simulates the antrum and causes the digesta to undergo high shear and mechanical breakdown. The second stage mimics the human small intestine where digesta empties from the stomach and is subjected to simulated intestinal secretions. The DGM was developed to examine the impact of digestion on the bioaccessibility and delivery profiles of nutrients to the duodenum 46 . The DGM also assesses the effects of food structure on nutrient delivery, nutrient interaction and survivability of pharmaceutical drugs 46 .

Human Gastric Simulator (HGS)
The Human Gastric Simulator developed by University of California, Davis, is based on the vertical alignment approach. It has a round cylindrical gastric compartment that is squeezed by rollers on a latex chamber 33 . The latex body has a diameter of 102 mm and is able to hold up to 5.7 L of content. The rollers help to simulate the contractile forces of the antrum. Peristaltic action in the HGS is simulated by having 12 rollers, 4 conveyor belts, and a pulley system 33 . The HGS simulates the actual human stomach contraction cycles by creating three contractions per minute on the latex vessel.

The TNO Gastro-Intestinal Model (TIM)
The TNO Gastro-Intestinal Model (TIM) is a multicompartmental model that was developed by TNO in Zeist (the Netherlands) in the 1990s to simulate the lumen of the GI tract.
It is a dynamic model that is controlled by a computer to adjust the physiological conditions (such as temperature, peristalsis, secretion of enzymes, flow rates, pH values) in the GIT 47 . This model also removes digested compounds and water and has been used to study an array of foods and pharmaceutical products.  48 . This model has also been used to study the bioaccessibility of food 48 .

Oral stage
The process of mastication (chewing) in the mouth is influenced by many factors such as the composition and the size of the food, the condition of the teeth, the number of chewing cycles, the force of the bite, and the volume of food 49 . These factors contribute to the size and area of the particle size in the bolus. These differences were observed in a study by Peyron et al.
(2004) who compared the boluses of raw vegetables and nuts 50 . They found that the raw vegetables and nuts gave similar boluses made with similar particle sizes to each other 50 . To simulate mastication, the particle size of solid food should be standardized by using a mincer. One of the major enzymes in the oral phase is salivary amylase. This enzyme which aids in the digestion of carbohydrates has an optimal pH of 6.8 52 . Since digestion of starch happens in the oral stage, this was another argument for the 2 minutes digestion time in the oral stage 43 .

Gastric stage
A study by Tyssandier et al. (2003) found that it takes approximately 3 -4 hours for solid western food (such as hamburgers) to leave the stomach and enter the duodenum. This process is known as gastric emptying 53 . Studies have shown that the gastric emptying of liquid food digestion is very rapid (30 min-1 h) 54 . As there are many factors that influence the rate of gastric emptying, e.g., the type of nutrients, the changes in pH, etc., it is difficult to mimic the complex digestive processes that occur in the stomach. Taking these factors into account, Minekus et al.

Small intestine
When the acid chyme from the stomach reaches the small intestine, it is neutralized by bicarbonate and reaches 6.

Oral phase
The oral phase involves the mechanical and chemical digestion of starch by α-amylase Other models have made use of mincers, sieves and food processors to simulate chewing.
These methods do not correctly simulate the oral phase since grinding or homogenizing does not produce food particle sizes similar to chewing. Since these processes only mimic chewing, Brighenti et al. (1993) incubated the food with human salivary α-amylase to introduce salivary enzymatic digestion 8 .

Gastric phase
In vivo digestion in the gastric phase involves the digestion of the protein fraction of food by enzymatic hydrolysis by pepsin. Also, the low acidic condition of the stomach allows for further denaturation of food. Gastric emptying of the acidic chyme into the duodenum is affected by factors such as viscosity and the quantity of food.
To mimic these conditions, in vitro methods involving starch digestion started to include the gastric phase with pepsin proteolysis as a way to closely simulate the physiological conditions and also to disrupt protein-starch interactions that may have occurred 58 . restricted α-amylolysis in the small intestine 60 . A study by Englyst (1992), which looked at the determination of nutritionally important starch, did not include the gastric phase. This was rectified in a following study where Englyst included a pepsin step for 30 minutes at pH of 2.0 22 .

Intestinal phase
Once the acidic chyme reaches the duodenum from the pyloric antrum, it is mixed with the pancreatic enzymes including α-amylase. The enzymes at the brush border of the small intestine also help in the complete hydrolysis of starch into their end product monosaccharides.
In vitro digestion models for the small intestine subject the acidic chyme to pancreatic enzyme secretion that is reflective of what is seen in vivo. Also, a buffer is added to increase the pH of the solution. However, depending on the type of study being performed, samples can be taken during the digestion process, or at the end or not even taken at all for studies involving the removal of digestible carbohydrate for resistant starch isolation.

YSI Biochemical Analyzer
The use of traditional methods such as HPLC (high performance liquid chromatography) and GC (gas chromatography) requires a fair level of expertise to operate, high maintenance and a fairly long time for analysis. The YSI biochemical analyzer provides for real-time, accurate analysis of key food components for various analytes such as many monosaccharides 61 .
This system uses an enzymatic membrane specific to the analytes of interest and measures changes in current across a membrane. An enzyme specific for the substrate of interest (for example, glucose) is immobilized between two membrane layers 61 . The substrate is oxidized as it enters the enzyme layer, producing hydrogen peroxide, which passes through cellulose acetate to a platinum electrode, where the hydrogen peroxide is oxidized 61 . The resulting current is proportional to the concentration of the substrate, e.g., glucose. YSI membranes contain three layers -a porous poly-carbonate which limits the diffusion of the substrate into the second enzyme layer, preventing the reaction from becoming enzyme-limited 61 . The third layer, cellulose acetate, permits only small molecules such as hydrogen peroxide to reach the electrode, eliminating many electrochemically-active compounds that could interfere with the measurement 61 .

Materials
A spectrum of test foods within their expiration date along with a standard food (white bread) was used for in vitro digestion and for the determination of GI. The in vitro digestion of these test products was performed by using 50 g of available carbohydrates of that product. Each experiment was performed in triplicate. The in vitro digestion process was divided into three main phases: oral, gastric and intestinal phase.

Step 1: Preparation for in vitro digestion
Prior to in vitro digestion in the DRUID, a spectrum of carbohydrate-rich foods was selected. These test foods were selected on the basis of their physical and structural properties as well as to reflect a range of GI values from low to high, namely skim milk, red kidney beans,

Preparation for in vitro digestion with an addition of fat
An additional experiment was performed to examine the dose response of adding butter during cooking on the GI of white rice. In this study, three different amounts of butter (20 g, 40 g and 80 g) were added to white rice (50 g of available carbohydrate portion). The rice was cooked in a rice cooker for 20 minutes before being subjected to digestion in the DRUID.

Determination of glucose, fructose, sucrose concentration and starch content
For foods without a nutritional label, the 50 g of available carbohydrate was calculated by performing glucose, fructose, and sucrose concentration determination assays. Sucrose, fructose and glucose concentrations were determined by following the procedure given by the Sucrose/D-Fructose/D-Glucose Assay Kit (Megazyme, Bray, Ireland). Using the kit, six different solutions were prepared (labelled from solution 1 -6). A homogenizer was used to homogenize the solid sample. A 10 g sample from the homogenate was treated to potassium hexacyanoferrate, zinc sulfate and sodium hydroxide to clarify the sample. A filter was used to filter and obtain a clear, non-pigmented solution from the homogenate. A blank, standard, and the sample were prepared by adding the prescribed amount of the six prepared solutions and water as instructed by the assay's protocol. A spectrophotometer was used to read the absorbance of the sample at 340 nm. A dilution calculation along with Beer's Law was used to determine the concentration of glucose, fructose, and sucrose from the absorbance of the sample.
The starch content was also determined by following the procedure given by Megazyme.
Using the kit, six different solutions were prepared (labelled from solution 1 -6). The food sample was passed through a 0.5 mm screen and 100 mg was added to a test tube. To this 0.  In the first phase (oral phase), solid carbohydrate-containing foods were mechanically broken down using a Ninja blender (a mechanical blender). Particles smaller than 4 mm were collected to make up 50 g of carbohydrates of the food. These food particles were subjected to simulated saliva which was prepared with NaCl (150 mM), KCl (2 mM), NaHCO3 (25 mM) and 160 U/ml salivary alpha amylase for two minutes.

Step 2: In vitro digestion in the DRUID
After the oral phase, the food bolus was driven into the gastric phase of the DRUID (a bioreactor) using a peristaltic pump. Eight ml of 1 M HCl were added into the gastric bioreactor.
The bolus was then subjected to 10 ml of 10% pepsin/0.05 M solution (Pepsin EC 3.4.23. from porcine stomach mucosa, Sigma-Aldrich P 7000; 800-2 500 U/mL) along with 300 ml of Milli-Q water. The food was digested in the gastric phase at 37˚C 30 minutes with constant mixing at 150 rpm (via impeller rotation). The pH was adjusted to a range of pH 1.5 -1.75 using 1 M HCl. One All trials were performed in triplicates and all secretions were prepared fresh right before the start of each trial.

Data analysis
All

Introduction
Due to the lack of a standardized method for the in vitro determination of GI, sample preparation is different in each in vitro digestion method 22,24,33 . The sample may be broken down by mincing or by using subjects to chew the samples. In vivo methods for determination of GI employ volunteers to chew food samples, however some criticize chewing as it varies from person to person as the composition of the saliva is different from person to person 86 Table 4.1 were calculated after using glucose as the standard food 63  For the determination of GI in vitro, white bread was used as the standard food (Table   4.1). The reason for this was because when white bread was digested in the DRUID, there was an accumulation of glucose concentration over the digestion period since the DRUID does not simulate nutrient absorption and this yielded a high iAUC. This was also what was found for the test foods (Figures 4.1 and 4.2). However, when glucose was digested, there was only a very small change in the glucose concentration over the digestion period since glucose solution already exists in its monomeric form of glucose units. As a result, when glucose was used as the standard for the test foods, the GI was over 100 since the iAUC for glucose was relatively very small when compared to the iAUC of the test foods. As such, white bread, having the highest iAUC of all the test foods was used as the standard for in vitro determination of GI.
In Table 4.1, unlike the in vitro GI values, the in vivo GI values have relatively high standard deviations or foods were not tested in duplicate. This shows that the digestion in the DRUID can be more reproducible as the determined in vitro GI does not vary significantly from trial to trial.
From looking at Table 4   It is therefore important to determine the available carbohydrates of these foods prior to determining the in vitro or in vivo GI. For a majority of the test foods, nutritional labels were supplied and from this the 50 g of available carbohydrate was calculated. For raw foods like banana which did not have a nutritional label, assay kits were used to determine the concentration of glucose, fructose and sucrose as well as the amount of starch prior to in vitro digestion in the DRUID. It was found that in 10 g of a ripe banana, there was 1.50 ± 1.1g of sugar and 2.13 ± 1.3g of starch. Even without the manufacturer's nutrition label for GI determination, by determining the total carbohydrates present in the ripe banana, the in vitro GI determined after digestion in the DRUID was able to be compared to literature values and thus reducing the percent differences from inter-banana variation and unknown amount of available carbohydrates.
Once the conditions for GI determination were known or the available carbohydrates have been determined via assay kits, the DRUID was able to reliably determine the GI of a spectrum of food (as seen by Table 4.1). This shows that, even with the many factors that effect the in vitro GI determination, namely the type of starch, content of fibre, food processing, and etc., when given the knowledge of composition and preparation of the food the DRUID was able to accurately determine the GI with excellent agreement to published in vivo GI values 63 .  There are several in vitro methods for the prediction of GI. One of these methods involved incubating the test food with a set of digestive enzymes and afterward performing colorimetric assays to measure the glucose release of test food after 20 min of digestion 74 . The correlation between the in vitro GI and the in vivo GI values was reported by Englyst et al. (1996)

Effects of the addition of fat on glycemic index of food
The inclusion of high fat, high protein, or high fibre food in a carbohydrate-rich food may alter the GI. In the human body, proteins can reduce the GI by stimulating insulin and slowing gastric emptying 76 . Fibre also helps to reduce the GI of food by reducing the available carbohydrates for absorption by fibre-glucose binding as well as by slowing the rate of gastric emptying 77 . The consumption of fat with carbohydrates has also been shown to reduce GI and blood glucose response 78 . There have been several studies that have shown an impact of fat on GI. Collier and O'Dea (1983) showed that the addition of butter to a potato meal produced a 50% reduction in glucose response compared to potatoes alone 79 .
The mechanism by which fat reduces the postprandial glycemic response is by delaying the rate of gastric emptying as well as by reducing enzymatic accessibility to carbohydrates 37,48 .
Because the DRUID does not simulate gastric emptying and a fixed incubation period was used in the gastric phase, the effect of fat on reducing enzymatic accessibility was determined. 51 .

Chapter 5 -Conclusions
The in vitro GI values strongly correlated with published in vivo values and showed a positive relationship between the in vitro and in vivo determined GI (r = 0.9761, p < 0.0001), and the Bland-Altman plot showed there was a strong agreement between the DRUID determined and published in vivo GI values. Moreover, the DRUID was sensitive enough to detect changes in GI as a result of preparation. Therefore, the DRUID can reliably, cost effectively and reproducibly determine the GI across a spectrum of carbohydrate-containing foods, and has the potential to predict the digestion kinetics of novel food products in vivo that may promote human health.

Introduction
Milk has been recognized as one of the major sources of protein for people of all ages 81 .
Milk protein contains all nine essential amino acids needed by humans 81 . Due to the resemblance to human milk, cow's (bovine) milk is the most abundant source of milk for dairy industries 82 .
Bovine milk consists of around 3% protein of which 20% is whey proteins and 80% is caseins 64 . Caseins which account for the majority of milk protein are represented by four distinct proteins: αs1, αs2, β, κ 64 . Whey proteins are a complex set of proteins that in bovine milk are mainly comprised of beta-lactoglobulin (ß-LG) and alpha-lactalbumin (α-LA) and small traces of blood-borne proteins such as immunoglobulins and bovine serum albumin. These two categories of proteins are broadly defined by their physical properties. Unlike caseins, whey proteins do not contain phosphorous but contain a large fraction of sulfur-containing amino acid residues (such as cysteine and methionine) 64 . These amino acids form disulfide bonds within the protein thus causing the linear amino acid structure to form a compact and well-defined globular structure that accounts for their solubility. Due to the presence of proline residues and the lack of disulfide bridges, caseins exhibit a loose, flexible structure. Caseins are also hydrophobic and are found as large colloidal particles of 50-600 nm in diameter, known as casein micelles 64 . These are supramolecular structures that hold caseins together by hydrophobic interactions and calcium phosphate 83 . Caseins are isolated from milk by acid or by rennet precipitation. The acid, or isoelectric, precipitation is performed at pH 4.6, where caseins precipitate (at their isoelectric point of pH 4.6) and whey proteins remain soluble 84 . These structural and chemical differences between the two major classes of protein in bovine milk affects the behavior of these proteins in the processing of food as well as their behaviour in the gastrointestinal tract, specifically their degree of hydrolysis by the digestive enzymes.
Over the last few decades, there has been an increase in the consumption of processed dairy products. The well-being of the dairy industry is dependent on the processing of milk into other milk based products. The processability and quality of many dairy products such as cheese mainly depends on the properties of milk protein such as caseins 64 . The structure of casein is crucial in the processing of milk into gelatinous substances such as yogurt and cheese 64 .
Moreover, the heat stability of caseins allows for many milk products to be subjected to high heat treatments 84 . Due to their industrial importance and technological properties, there has been a growing interest in studying caseins from bovine milk 84 . Although much research has been undertaken in studying the casein micelles, there structure is still not well understood.
In evaluating the structure of caseins, it is important to study the physical and chemical changes that occur during digestion in the human gastrointestinal tract 85 . The gold standard for investigating the human digestive process is the use of in vivo approaches. This normally involves a feeding study and acquiring serial samples from the stomach and the upper small intestine.
However, these approaches are impractical for large-scale studies as they are expensive to preform, time consuming and ethically and technically difficult 41 . To overcome the expenses and physiological difficulties, a simple in vitro method was used in this study to simulate the gastric and intestinal conditions that occur in vivo.
This study uses a confocal microscope to visually examine the in vitro digestion of caseins by developing a simple model which allows for real time digestion of caseins in situ. This will provide a better understanding about the physiological changes that caseins undergo during digestion in the stomach and small intestine and thus help yield information about the structure of caseins during human digestion as well as provide information for designing more nutritious milk based products.

Samples and reagents
Milli-Q water (

Development of an in situ digestion chamber
An in situ digestion chamber was designed by considering the fluid dynamics of skim milk. A milling machine, a drill machine, and an automatic saw were used in the construction of the digestion chamber. A 1 mm thick acrylic sheet was cut with the dimensions of 75 mm x 25 mm (for optimal positioning on the confocal microscope stage). A reaction chamber with a diameter of 1.47 cm was drilled ( Figure A-1.). A diagonal slit was drilled from the middle of the reaction chamber to the edge of the cover slip for the insertion of simulated secretions. A small slit with a length of 1.5 cm was filed to alleviate the displacement of air by the milk-dye sample.

In vitro gastric digestion
A sample of 200 μl of the milk-dye mixture was added to the in-situ digestion chamber.
The digestion chamber was placed on the microscope stage for the time lapse imaging process.
The milk-dye mixture was acidified to a pH range of 1.

Gastric digestion -casein hydrolysis
In humans, in vitro digestion of milk proteins, begins in the stomach. During the process of digestion, casein micelles first come upon gastric acid which has a very low pH range of 1.0 -2.0. Although casein is relatively stable at high temperatures, it is sensitive to pH and will precipitate at its isoelectric point, pH 4.6 64 . The precipitation of casein in the presence of HCl can be observed in Figure A-2. Over the period of a minute, the pH dropped from pH of 6.77 (in native skim milk) to pH 1.5 (from the introduction of HCl) which caused the casein micelles to undergo denaturation and precipitate thus resulting in a change in shape. This is seen by Figures A-2B-D.
where the shape of the coagulated caseins (casein domain) gets narrower as it is exposed for a longer duration to the low acidic environment of the stomach.
During gastric digestion in the human stomach, this clotting or precipitation of casein under the acidic environment of the stomach, results in a reduction in gastric emptying rate, which delays the release of amino acids to the small intestine 86 . On the other hand, milk soluble proteins such as whey are rapidly expelled out of the stomach 86 .
Before the precipitated caseins are sent into the small intestine via gastric emptying, a digestive enzyme known as pepsin starts to hydrolyze proteins from the surface of the precipitated caseins 87 . Pepsin is an aspartic protease and has a preference for cleaving peptides with amino acids: phenylalanine, tryptophan, tyrosine and leucine 88 .

Intestinal digestion -casein hydrolysis
The second stage of the in vitro digestion of caseins occurs in the small intestine. With the secretion of pancreatin (which contains proteolytic enzymes like trypsin and chymotrypsin) further digestion of caseins takes place. Trypsin cleaves the backbone of positively charged amino acids (lysine or arginine) and chymotrypsin cleaves peptides with bulky hydrophobic amino acid residues 64 . These proteolytic enzymes help to digest any undigested caseins and peptides.

In vitro GI determination of beverages
For Coca-Cola and Fruité Fruit Punch, it was expected that since there was not too much starch to began with, most of the carbohydrate is in the form of glucose. As such, it was expected to observe a high glucose concentration in the oral phase. This was found to be the case. In the stomach, it was expected for salivary alpha amylase to be deactivated due to the very low pH. As such, the concentration of glucose was expected to be constant. From Figure B-1., it was seen that there was a slight increase in the glucose concentrations for these beverages over the twohour residence time. This may be due to some of the sugars like maltose that were present in these drinks as they were being digested by the enzyme maltase.

Validation of the YSI 2700
To validate the YSI 2700 biochemical analyzer that was extensively used to determine the glucose concentrations of samples after digestion in the DRUID, a commonly used colorimetric assay from Megazyme (GOPOD) kit was used to determine the glucose concentration of the samples after digestion in the DRUID as a way of validating the YSI 2700. Unlike the YSI 2700 biochemical analyzer that only requires to aspirate 25 μm of sample to determine the glucose concentration using enzymatic digestion, the Megazyme D-Glucose (glucose oxidase/peroxidase; GOPOD) Assay Kit requires several dilutions and sample preparations in the dark. As such, this assay can lead to several errors. However, due to its inexpensive price and availability, the Megazyme D-Glucose (glucose oxidase/peroxidase; GOPOD) Assay Kit has been used in past literature 90 .
The validation of the YSI 2700 biochemical analyzer to the Megazyme (GOPOD) kit was compared using a Bland-Altman plot. Figure 9.2 showed that when the glucose concentrations determined from the YSI 2700 biochemical analyzer were compared to the Megazyme (GOPOD) kit for white rice, there was an agreement between the two methods (since most of the values were near the line of no difference). Moreover, Figure 9.3. showed that an agreement was also seen between the two glucose determination methods when the glucose values of an apple were compared after digestion in the DRUID.
Therefore, due to the speed of determining the glucose concentration using the YSI 2700 biochemical analyzer and the lack of error from serial dilution as well as the agreement to the Megazyme (GOPOD) kit for determining glucose concentrations, the YSI 2700 biochemical analyzer was preferred for glucose concentration determination.