Dan Liu

General discussion
Figure 6.4. Methods for performing standardized mastication.
Standardized in vivo mastication can be achieved by following an instructed chewing protocol that define bite size, chewing frequency and number of chews. These parameters regarding chewing behavior need to be determined based on natural in vivo mastication. Standardized in vivo mastication can be performed either by a group of participants (Chapter 3) or by one person (Chapter 5). The boli collected form instructed in vivo mastication by a group of participants are pooled. Bolus samples for subsequent in vitro digestion are taken from the pooled boli. This method reduces inter-individual differences, but the properties (e.g., particle size distribution) of bolus particles sampled from pooled boli may differ from those of complete boli expectorated after in vivo mastication. Standardized in vivo mastication performed by a single person allows for studying the digestion of complete expectorated boli. However, the bolus properties would depend on the oral physiology of the person who chews the samples.

The alternative of standardized in vivo mastication is standardized in vitro oral phase simulation following a fixed protocol. An easy way to simulate mastication is mincing foods while mixing with simulated saliva fluid, as described in the INFOGEST 2.0 protocol (Brodkorb et al., 2019). Some studies applying this method to produce artificial boli that had similar bolus particle size as real boli collected from in vivo mastication on the same samples (Guo et al., 2015, 2016). Most in vitro digestion studies simply mince their samples without using real bolus properties as guidance (Bayrak et al., 2021; Hiolle et al., 2020; Homer et al., 2021; Xie et al., 2022; Zhang et al., 2023b). Consequently, the boli produced through this method might exhibit significantly different properties from real boli formed by in vivo mastication. For instance, simulating mastication by blending roasted peanut produced boli with higher moisture content and larger mean particle size compared to real boli formed through in vivo mastication (Xu et al., 2024). Simulating mastication using advanced mastication simulators that closely mimic oral physiology could help producing boli with more realistic bolus properties (Xu et al., 2024). Multiple mastication simulators such as the artificial masticatory advanced machine (AM2) (Peyron et al., 2019), the bite master II (Meullenet & Gandhapuneni, 2006), the chewing robot (Wang et al., 2015) and the in vitro bio-inspired oral mastication simulator (iBOMS-Ⅲ) (Xu et al., 2024) have been developed over the past decades. These models are developed to study the kinetics of food oral processing including oral breakdown and the release of flavor compounds (Guo et al., 2024). AM2 has been applied in several in vitro digestion studies (Blanquet-Diot et al., 2021; Peyron et al., 2021; Ribes et al., 2024). These studies applied the number of chews during in vivo mastication to set the operation parameters of AM2 to obtain artificial boli. Standardizing the design of these simulators and developing relevant operational protocols for different food categories according to in vivo oral processing could provide a reproducible and objective method for collecting boli for in vitro digestion studies.

To sum up, multiple methods are available for performing standardized mastication. The strengths and limitations along with other aspects need to be considered when selecting methods. For instance, instructed in vivo mastication for elderly individuals or patients can be challenging, whereas simulating mastication using advanced mastication simulators may be unfeasible due to a lack of facilities. Regardless of the method used, information on natural in vivo mastication, such as bite size, number of chews, chewing frequency, and bolus properties, is required to provide guidance for standardizing mastication.

6.3.2 In vitro gastrointestinal digestion studies
The applied in vitro digestion model remarkably influences the in vitro digestion outcomes as shown in Chapter 5 and a few other studies (Homer et al., 2021; Mella et al., 2021). Therefore, in vitro digestion studies need to carefully consider the choice of in vitro models which vary from simple static models to near-real dynamic models. Additionally, other aspects need to be considered alongside digestion models, including study samples (ranging from simplified model foods to full meals), whether the study focuses on digestion kinetics or endpoints, and the targeted populations (general healthy adults or specific groups such as infants or elderly) (Duijsens et al., 2022). For studies aiming to understand the interactions between food properties and macronutrient digestibility, the author suggests to start by investigating the digestion kinetics using simplified model foods and simple static models. These simple static models allow to manipulate variables easily and the interpretation of outcomes are usually straightforward. One of the advantages of in vitro digestion studies compared to in vivo studies is the great flexibility to control digestive conditions. Starting directly with complex methodologies could introduce unnecessary difficulties in interpreting results. After identifying relevant properties and gaining mechanistic insights through simplified methodologies, the complexity can be gradually increased to investigate more detailed interactions. Finally, advanced dynamic digestion models that closely mimic digestive physiology can be applied to explore the digestion of complex food or meals. The added complexity would help to bridge the gap between in vitro studies and in vivo reality, while the outcomes influenced by multiple variations can be properly explained by the underlaying mechanisms learned from simple models.

It is crucial to keep in mind that all in vitro models have specific features that can introduce bias. For example, static models usually overemphasize the accessibility of enzymes to substrates, thereby overestimating the impact of food properties that directly affect enzyme accessibility, such as microstructure. In contrast, gastric-emptying-mimicking models could overestimate the impact of food properties that directly affect gastric emptying, such as particle size, due to their lack of metabolic regulation. Interpretating in vitro results with awareness of the features and limitations of the applied models can avoid misinterpretations caused by shortcomings of the simulated digestive conditions.

The strengths and limitations of various in vitro digestion models have been reviewed along with suggestions for improvements (Bohn et al., 2018; Duijsens et al., 2022; Dupont et al., 2019; Li et al., 2020; Mackie et al., 2020). These reviews give constructive advice including standardizing dynamic digestion models and integrating current in vitro gastrointestinal digestion models with in vitro absorption systems that involve intestinal digestion and absorption via brush border membranes (Mackie et al., 2020). However, efforts to improve in vitro digestion systems are often focused on the gastrointestinal tract, with less attention given to the oral phase. Integrating in vivo mastication or realistic in vitro oral phase simulation in in vitro gastrointestinal digestion methods is highly recommended. Developing and standardizing oral-to-intentional in vitro digestion protocol could significantly benefit studies on the role of oral processing in food digestion.

6.4 Designing food structures to modify protein digestibility
Food structure design refers to the concept of controlling the release of dietary lipids and carbohydrates or delivering bioactive components in the gastrointestinal tract (McClements et al., 2008, 2009; Pellegrini et al., 2020). A systematic concept of food structure design focusing on modifying protein digestibility is missing. Here, we propose a preliminary concept of food structure design to optimize protein digestion based on the outcomes of this thesis and previous studies.

6.4.1 Proposed guideline: from molecular to macroscopic scale
The following aspects need to be considered when designing food structures with the aim to maximize gastric protein digestibility (Bornhorst et al., 2016). Examples of structures at difference length scales are given in Figure 6.5.

a) Molecular / nanometer scale (< 1 μm): First, the molecular protein composition, i.e., the amino acid composition, can be improved by mixing various proteins (Gorissen et al., 2018). Proteins differ significantly in their amino acid profiles and combinations of various proteins aid in balancing the essential amino acids, enhancing protein quality (Adhikari et al., 2022; Jiménez-Munoz et al., 2021; Sá et al., 2020b). Secondly, it has been shown that pepsin tends to cleave the peptide bonds involving hydrophobic aromatic amino acids which are often directed toward the interior of the protein structure (Gajdos et al., 1963). Therefore, at molecular scale, the key to improve protein digestibility is to expose these peptide bonds as much as possible to pepsin by unfolding the protein structure. Increasing the fraction of disordered structure such as random coil while decreasing the ordered structure such as β-sheet can aid in protein hydrolysis of legume proteins (Sun et al., 2020; Yang et al., 2016). These modification on the secondary structure can be achieved by processing (e.g., microwave or heating) or changing the protein sources. Modifications on proteins (i.e., denaturation) that can open the tertiary or quaternary structure can also help proteolysis. It should be noted that processes such as severe heating result in formation of protein aggregates which increases the number of disulfide bonds and inhibits proteolysis (Duodu et al., 2002). b) Microscopic scale (1 μm - 1 mm): At the micrometer scale, micro-phase separation can occur due to interactions between components or protein aggregations (Chapter 2; Singh et al., 2014). Generally, for heterogeneous microstructures, porous structures and less compact protein networks are preferred. For instance, soy protein gels with porous protein network with pores of approximately 10 μm diameter showed higher protein hydrolysis than gels with compact, thin, layer-like protein networks (Zhao et al., 2020). In contrast, at larger scales, around 100-200 μm, a homogeneous structure is preferred instead of microphase separation (Chapter 2). This is because at larger scale, a heterogeneous structure often indicates a dense protein network in a protein-rich phase (Singh et al., 2014). Specially, at this length scale, spatial barriers such as cell walls should be avoided (Zahir et al., 2018). These findings are based on model systems as study subjects. Modifying the microstructure of foods is possible by adjusting the formulation and processing conditions. The translation of these findings from model systems to commercially available foods should be explored in future studies. c) Macroscopic scale (> 1 mm):
At the macroscopic scale, the impact of structure on digestion could be associated with multiple factors. For individual food particles, an open structure, such as the large pores in TVPs, could facilitate protein hydrolysis. Additionally, increasing macroscopic surface area enhances static in vitro gastric protein digestion (Chapter 4). Furthermore, solid foods are often broken down into food fragments ranging from millimeters to centimeters during oral mastication. Increasing the fraction of small fragments (i.e., decreasing the mean particle size) tended to enhance protein hydrolysis during in vitro dynamic gastric digestion (Chapter 5), although in vivo studies are needed to validate this finding. Food structures at the millimeter scale can be modified by food processing, such as altering TVP structure by adjusting extrusion conditions. Control of macroscopic surface area or particle size of food boli could be achieved by modifying food texture, for example, brittle foods tend to form more and smaller food particles, thereby increasing total surface area. These associations enable food designers to modify food structure and texture through processing techniques to optimize protein digestibility.

Figure 6.5. Examples of food structure at different length scales and their influence on in vitro gastric protein digestion. Images are reproduced from Chapter 2, Chapter 4, Zhao et al. (2020) and MacIerzanka et al. (2012).

6.4.2 Challenges and future directions
Modifying protein digestibility by altering food structure as described in section 6.4.1 is challenging for various reasons. The first is the inevitable associations between sensory perception and food structure. Food structure at different length scales is known to impact sensory properties of foods, so it is important to quantify the sensory properties of the structurally modified foods together with the protein digestibility. Additionally, reverse engineering of food structure tailored to a specific digestion of macronutrients requires deep understanding of the structure formation mechanisms during food processing. Furthermore, the complex interactions between food structures and digestion as shown in this thesis require compromises. Non-structural properties, such as pH, temperature of foods and the combination of different food items into meals, may show a greater impact on protein digestion than structure of individual food items.

Future studies should explore sensory perception, oral behavior, breakdown of foods differing in structure and macronutrient bioavailability. Fundamental understanding of the formation of food structure at different length scales in the context of complex formulas and processing techniques could aid in the implementation of structural design. Moreover, the interactions between food structure and special digestive conditions of specific populations such as elderly and gastrointestinal disease patients need further investigation. Development of in vitro digestion systems tailored to these specific groups could facilitate the relevant studies.

6.5 Concluding remarks
In this chapter, the interplay between food structure and in vitro digestion models is discussed, followed by suggestions for in vitro digestion studies. Including real or realistic mastication as a standardized step before in vitro gastrointestinal digestion is recommended. For studies investigating the impact of food properties on macronutrients digestion, simple static in vitro models can be a good starting point. Increasing the complexity of in vitro models helps to better understand potential interactions during digestion. Interpreting in vitro digestion results with awareness of the strengths and limitations of applied models is crucial to avoid misinterpretation caused by simulated digestive conditions. Furthermore, designing food structures to optimize protein digestibility requires collaboration across fields including food physics, food engineering, sensory and nutrition science. Although there is a long way to go, tailoring food structure to meet specific nutritional needs could offer a promising solution to address global challenges related to malnutrition and obesity.

Due to increasing global population, limited resources, and environmental challenges, it is crucial to establish a more efficient and sustainable food system. Enhancing the digestibility of proteins, which are essential for growth and metabolism, is a key focus. The shift away from animal-based proteins, due to their resource intensity and greenhouse gas emissions, has led to the rise of plant-based meat alternatives. However, these alternatives often have low protein digestibility due to the unbalanced amino acid composition and the presence of anti-nutritional factors. Nevertheless, new protein sources and technologies present opportunities to improve food texture and structure, enhancing protein digestibility and thereby improving the nutritional value of meat analogues. A thorough understanding of the interactions between food structure and the protein digestion is essential for developing nutritious foods. This thesis aimed to untangle the interplay between food microstructure, mechanical properties, macrostructural breakdown caused by oral processing, and gastric protein digestion using in vitro digestion models.

We started with model foods (whey protein gels) and static in vitro digestion models to explore the interplay between microstructure, mechanical properties, macrostructural breakdown and in vitro gastric protein digestion (Chapter 2). Whey protein isolate/polysaccharide mixed gels were developed to obtain gels with distinct microstructures (homogeneous, coarse stranded, bi-continuous and protein continuous) but similar mechanical properties (Young’s modulus). During static in vitro gastric protein digestion, homogeneous gels displayed the highest digestion rate followed by protein continuous, coarse stranded and bi-continuous gels. Increasing Young’s modulus led to decrease in protein digestion rate for homogeneous gels, while it did not influence the protein digestion for protein continuous gels. Increasing the total surface area by a factor of 2.6 enhanced protein digestion rate to different extents depending on gel microstructure. We concluded that microstructure has independent impact on in vitro gastric protein and this impact interacts with mechanical properties and macrostructural breakdown.

In Chapter 3, the whey protein isolate/polysaccharide mixed gels differing in microstructure were chewed by a group of people following a standard chewing behaviour. Both intact gels and expectorated boli were subjected to static in vitro gastric protein digestion to investigate the impact of microstructure on protein digestion after in vivo mastication. The results show that the increase in in vitro gastric protein digestion was not proportional to the degree of macrostructural breakdown during mastication and depended on gel microstructure. Bi-continuous gels exhibited the largest increase which might be attributed to their highest partition coefficient of pepsin at the gel-gastric juice interface. We concluded that the impact of microstructure on in vitro gastric protein digestion is sustained after great macrostructural breakdown induced by in vivo mastication.

Chapter 4 moved from model gels to textured vegetable proteins (TVPs) which are the main ingredients in plant-based meat analogues (PMBAs). This study aimed to investigate the impact of structural properties of TVPs on in vitro gastric protein digestion of TVPs and TVP-based meat analogue patties in a quantitative way. Eight TVPs differing in structural properties, such as surface area, porosity, pore size and wall density, were used to prepare TVP-based meat analogue patties. Both TVPs and TVP-based patties were subjected to static in vitro gastric protein digestion. Additionally, these TVPs were ground into powders to remove the porous structure as a control group. The results show that macroscopic surface area and pore size were positively correlated with protein hydrolysis, while wall density was negatively correlated with protein hydrolysis when porous structure was removed. We concluded that in addition to macroscopic surface area, pore-related, rather than wall-related properties were primary structural properties influencing in vitro gastric protein digestion of TVPs and TVP-based patties.

In Chapter 5, a dynamic gastric-motility-mimicking model (NERDT) was used for studying in vitro gastric protein digestion of PBMAs in addition to a static digestion model (INFOGEST). This study aimed to explore the impact of mechanical and bolus properties on in vitro gastric protein digestion of PBMA patties using static and dynamic models. Two commercial patties (Beyond Meat and THIS) differing in Young’s modulus and bolus particle size were subjected to static and dynamic digestion. THIS patties had higher Young’s modulus and were broken down into more and smaller particles during in vivo mastication compared to Beyond Meat patties. During static digestion, THIS patties showed lower free amino group concentrations in gastric juice compared to Beyond Meat patties, which was likely due to their stiffer texture. In contrast, during dynamic digestion, THIS patties displayed faster gastric emptying and higher free amino group concentrations in the emptied liquid compared to Beyond Meat patties. These results suggest that bolus particle size had a primary impact on dynamic in vitro gastric protein digestion of PMBA patties. To further investigate the impact of bolus properties, three model PBMA patties were prepared from textured yellow pea proteins differing only in particle size. The patties with bolus particles smaller than 0.18 mm2 exhibited faster gastric emptying and higher free amino group concentrations in the emptied liquid at the early stage of dynamic digestion as compared to patties with larger bolus particles (0.59-0.68 mm2). We concluded that bolus particle size, rather than mechanical properties, primarily impact dynamic gastric protein digestion. Specifically, smaller bolus particles facilitate dynamic in vitro gastric protein digestion by accelerating gastric emptying and modulating intragastric pH.

Chapter 6 discussed the main results of chapter 2-5 and provided suggestions for in vitro digestion studies. For in vitro digestion studies, static models can be a reliable starting point and follow-up experiments using advanced dynamic models can help to bridge the gap between in vitro simulation and in vivo reality. In vivo mastication or realistic oral phase simulation should be integrated with in vitro digestion methods. Moreover, a preliminary concept of designing food structure at different length scales to optimize protein digestibility is provided along with challenges and future directions, suggesting the potential of modifying protein digestibility by manipulating food structure.

In conclusion, this thesis demonstrated the interactions between microstructure, mechanical properties, (oral) macrostructural breakdown and in vitro gastric protein digestion of model foods and complex foods using static and dynamic models. The main findings indicate the potential to optimize protein digestibility by modulate food structures at both the micro- and macroscopic levels. The importance of in vivo mastication or realistic mastication simulation prior to in vitro gastrointestinal digestion is highlighted in this thesis. The conclusions drawn in this thesis are based on optimal oral-to-intentional conditions which are typically shown in healthy adults. Further studies are needed to transfer these conclusions to specific populations such as infants and elderly.