**3.3** *In vivo* **methods**

*In vivo* methods, albeit the most complex ones, are at the same time the most informative ones. The two possible strategies within this group are based on the evaluation of either the remnants of the substrate within or at the end of digestive tract, or of the assimilation of the digestion products by the organism.

The methods based on protein nitrogen utilization by an organism are subject of several great reviews, and we refer a curious reader to them for more details (Bender 1958; Darragh & Hodgkinson 2000; von der Decken 1983). Below, a short description of method families will be given.

The first group of methods (protein efficiency ratio, net protein retention, rat-repletion method) use direct determination of the weight of animals, fed with a test protein. The

Methods of Protein Digestive Stability Assay – State of the Art 221

studies have shown that anastomosed animals have an altered physiology compared with intact animals. It should be noted that human ileostomates may also have an altered physiology compared with intact humans, which could also call into question the validity of using human ileostomates for the collecttion of ileal digesta (Darragh & Hodgkinson 2000). The insertion of a re-entrant cannula involves transecting the terminal ileum and sealing the two ends. A cannula is inserted into each end of the sealed ileum and the two cannulae are joined. This allows a quantitative collection of digesta. The surgery to insert a re-entrant cannula is complex, however, and involves total transection of the ileum. Blockage of the cannula is a common complication. Simple T-piece cannulation and postvalve T-cecum cannulation have the distinct advantage of maintaining the ileocecal valve intact and avoiding ileal transection; in this case there is also no surgical interference with the small intestine. In addition to that, most of the digesta should pass through such cannula during sampling, as the ileocecal value protrudes directly into it (Darragh & Hodgkinson 2000).

The *in vivo* methods of protein bioaccessibility determination, such as retained nitrogen analysis, cannula incision and swallowed probe sampling, undoubtedly provide the most relevant data. However, their major drawback is experimental complexity and practical

Most of the aforementioned models, excluding the animal ones, lack two important characteristics of digestive tract — the mechanical forces, exerted by stomach and intestinal walls, and the fact that food arrives into stomach and upper intestine not as a solution or an emulsion, but rather as a bolus with different accessibility of its internal and external contents for the enzymatic machinery. The digestive tract simulators try to overcome this situation. Usually they represent complex devices, consisting of several mixing chambers with programmatically controlled pH, mixing speed, temperature and valves between them, which provide a definite transition time from one compartment to another. The compartments that imitate intestine contain immobilized enzymes, either trypsin and chymotrypsin, or more complex mixtures including proteases of intestinal wall. The system is filled with simulated gastric and intestinal fluid plus phospholipids and/or bile, and the test substances are delivered into the system either in form of solution or within agarose beads (a way to mimic food bolus). The result of protein digestion in such system could be analyzed at any stage; the commonly employed analytical methods are SDS-PAGE (mass

The most representative example of such device is the Dynamic Gastric Model, described by Vardakou et al. *(*Vardakou et al. 2011*)*. It is composed of three parts, the main body (fundus), the antrum and the valve assembly. In the main body of the model the inhomogeneous mixing of the stomach is reproduced by gentle contractions induced by computer-controlled changes in the applied pressure of water in the thermostated water bath surrounding the main body. Gastric acid and enzymes are added from a dispenser that is floating on the top of the main body contents. The dispenser is designed in such a way as to deliver the enzyme and acid evenly from the sides of the main body, replicating the human gastric secretions originating from the walls of the stomach. The rate of addition of both enzymatic and acid secretions is also computer-controlled. The food material is

impossibility of any reasonable screening.

range 2-15 or 14-60 kDa) and reverse phase HPLC.

**3.4 Digestive tract simulators** 

variations of this method lie in the employment of non-protein diet fed rats as a control group or starvation of animals before the introduction of a test meal (Bender 1958; Henry 1965). A variation of these methods suggests the determination not of the animal weight, but rather of amount of a specific tissue (liver, muscle, *etc.*), laid down while feeding on a test protein (Bender 1958; Mokady et al. 1969). These include protein retention efficiency and liver protein utilization. It should be noted, that the original growth determination methods are very animal-sparing, while the more advanced assays usually implicate animal death in order for some parameter to be determined.

Another group of methods, developed with the aim to overcome the not-so-obvious relationship between protein in diet and growth rate, introduces direct determination of nitrogen either in food before and after passage through digestive tract, or in animal carcass; a variant of latter analysis implicates the division of animal group in several parts with subsequent determination of carcass nitrogen at different time points (Bender 1958). The methods in this group are Thomas-Mitchell method, N-balance index method, carcassnitrogen method and growth and nitrogen balance method. In some variations of nitrogen retention methods protein isotope labeling is required.

The further development of nitrogen analysis concept led researchers to the development of some biochemical methods for protein nutritive value determination. These methods are based on several parameters, which strongly correlate with tissue growth: polyribosomal profiles or ribosome activity and content; the activity of enzymes involved in urea metabolism or transamination reactions; 3-methylhistidine (an excreted-only myosin component) production and creatinine excretion (Bender 1958; von der Decken 1983). Ribosome activity is quite informative, as it is sensitive enough to detect small differences in protein quality. The latter two parameters, on the other hand, are measured in urine and thus are suitable for humans.

The current recommendation, when calculating a protein digestibility, is to determine the digestibility of a dietary protein across the entire digestive tract, using the rat as a model animal for humans. This fecal digestibility value is subsequently corrected for endogenous contributions of protein using a metabolic nitrogen value determined by feeding rats a protein-free diet, a task for which the protein digestibility-corrected amino acid score (PDCAAS) was introduced. To calculate a PDCAAS, the availability of the amino acids in a dietary protein is assessed based on the digestibility of total nitrogen (N) in that dietary protein. Digestibility is defined as the difference between the amount of N ingested and excreted, expressed as a proportion of N ingested. Although accepted as the recommended procedure, the use of fecal digestibility coefficients to evaluate amino acid availability is thought to be inherently inaccurate due to the metabolism of both dietary and endogenous proteins by the hindgut microbial population (Darragh & Hodgkinson 2000; Kies 1981).

In fact, the aforementioned techniques are not precisely suitable for protein digestibility evaluation. The problem is that they measure protein assimilation and do not provide any data on how exactly a particular protein was or was not digested and transported. This hindrance is circumvented in the second group of *in vivo* approaches, which include swallowed probe utilization, anastomosis and cannula incision (Darragh & Hodgkinson 2000; Faber et al. 2010).

Anastomosis involves transecting the ileum anterior to the ileocecal junction and attaching this to the descending colon. This allows a quantitative collection of ileal digesta, but many

variations of this method lie in the employment of non-protein diet fed rats as a control group or starvation of animals before the introduction of a test meal (Bender 1958; Henry 1965). A variation of these methods suggests the determination not of the animal weight, but rather of amount of a specific tissue (liver, muscle, *etc.*), laid down while feeding on a test protein (Bender 1958; Mokady et al. 1969). These include protein retention efficiency and liver protein utilization. It should be noted, that the original growth determination methods are very animal-sparing, while the more advanced assays usually implicate animal death in

Another group of methods, developed with the aim to overcome the not-so-obvious relationship between protein in diet and growth rate, introduces direct determination of nitrogen either in food before and after passage through digestive tract, or in animal carcass; a variant of latter analysis implicates the division of animal group in several parts with subsequent determination of carcass nitrogen at different time points (Bender 1958). The methods in this group are Thomas-Mitchell method, N-balance index method, carcassnitrogen method and growth and nitrogen balance method. In some variations of nitrogen

The further development of nitrogen analysis concept led researchers to the development of some biochemical methods for protein nutritive value determination. These methods are based on several parameters, which strongly correlate with tissue growth: polyribosomal profiles or ribosome activity and content; the activity of enzymes involved in urea metabolism or transamination reactions; 3-methylhistidine (an excreted-only myosin component) production and creatinine excretion (Bender 1958; von der Decken 1983). Ribosome activity is quite informative, as it is sensitive enough to detect small differences in protein quality. The latter two parameters, on the other hand, are measured in urine and thus are suitable for humans. The current recommendation, when calculating a protein digestibility, is to determine the digestibility of a dietary protein across the entire digestive tract, using the rat as a model animal for humans. This fecal digestibility value is subsequently corrected for endogenous contributions of protein using a metabolic nitrogen value determined by feeding rats a protein-free diet, a task for which the protein digestibility-corrected amino acid score (PDCAAS) was introduced. To calculate a PDCAAS, the availability of the amino acids in a dietary protein is assessed based on the digestibility of total nitrogen (N) in that dietary protein. Digestibility is defined as the difference between the amount of N ingested and excreted, expressed as a proportion of N ingested. Although accepted as the recommended procedure, the use of fecal digestibility coefficients to evaluate amino acid availability is thought to be inherently inaccurate due to the metabolism of both dietary and endogenous proteins by the hindgut microbial population (Darragh & Hodgkinson 2000; Kies 1981).

In fact, the aforementioned techniques are not precisely suitable for protein digestibility evaluation. The problem is that they measure protein assimilation and do not provide any data on how exactly a particular protein was or was not digested and transported. This hindrance is circumvented in the second group of *in vivo* approaches, which include swallowed probe utilization, anastomosis and cannula incision (Darragh & Hodgkinson

Anastomosis involves transecting the ileum anterior to the ileocecal junction and attaching this to the descending colon. This allows a quantitative collection of ileal digesta, but many

order for some parameter to be determined.

2000; Faber et al. 2010).

retention methods protein isotope labeling is required.

studies have shown that anastomosed animals have an altered physiology compared with intact animals. It should be noted that human ileostomates may also have an altered physiology compared with intact humans, which could also call into question the validity of using human ileostomates for the collecttion of ileal digesta (Darragh & Hodgkinson 2000).

The insertion of a re-entrant cannula involves transecting the terminal ileum and sealing the two ends. A cannula is inserted into each end of the sealed ileum and the two cannulae are joined. This allows a quantitative collection of digesta. The surgery to insert a re-entrant cannula is complex, however, and involves total transection of the ileum. Blockage of the cannula is a common complication. Simple T-piece cannulation and postvalve T-cecum cannulation have the distinct advantage of maintaining the ileocecal valve intact and avoiding ileal transection; in this case there is also no surgical interference with the small intestine. In addition to that, most of the digesta should pass through such cannula during sampling, as the ileocecal value protrudes directly into it (Darragh & Hodgkinson 2000).

The *in vivo* methods of protein bioaccessibility determination, such as retained nitrogen analysis, cannula incision and swallowed probe sampling, undoubtedly provide the most relevant data. However, their major drawback is experimental complexity and practical impossibility of any reasonable screening.

#### **3.4 Digestive tract simulators**

Most of the aforementioned models, excluding the animal ones, lack two important characteristics of digestive tract — the mechanical forces, exerted by stomach and intestinal walls, and the fact that food arrives into stomach and upper intestine not as a solution or an emulsion, but rather as a bolus with different accessibility of its internal and external contents for the enzymatic machinery. The digestive tract simulators try to overcome this situation. Usually they represent complex devices, consisting of several mixing chambers with programmatically controlled pH, mixing speed, temperature and valves between them, which provide a definite transition time from one compartment to another. The compartments that imitate intestine contain immobilized enzymes, either trypsin and chymotrypsin, or more complex mixtures including proteases of intestinal wall. The system is filled with simulated gastric and intestinal fluid plus phospholipids and/or bile, and the test substances are delivered into the system either in form of solution or within agarose beads (a way to mimic food bolus). The result of protein digestion in such system could be analyzed at any stage; the commonly employed analytical methods are SDS-PAGE (mass range 2-15 or 14-60 kDa) and reverse phase HPLC.

The most representative example of such device is the Dynamic Gastric Model, described by Vardakou et al. *(*Vardakou et al. 2011*)*. It is composed of three parts, the main body (fundus), the antrum and the valve assembly. In the main body of the model the inhomogeneous mixing of the stomach is reproduced by gentle contractions induced by computer-controlled changes in the applied pressure of water in the thermostated water bath surrounding the main body. Gastric acid and enzymes are added from a dispenser that is floating on the top of the main body contents. The dispenser is designed in such a way as to deliver the enzyme and acid evenly from the sides of the main body, replicating the human gastric secretions originating from the walls of the stomach. The rate of addition of both enzymatic and acid secretions is also computer-controlled. The food material is

Methods of Protein Digestive Stability Assay – State of the Art 223

Our group has recently suggested an assay, that tries to replicate the conditions of parietal digestion in small intestine, at the same time maintaining capability for high-throughput sample analysis (Akimov et al. 2010). The active component of this system is a fragment of rat stomach or intestinal wall, immersed in simulated gastric or intestinal fluid. The products are analyzed by HPLC. The utilization of tissue fragments allows for reduction of animal consumption, while still maintaining the whole set of intestinal peptidases and providing some phosphatidyl choline in the medium. The assay design is suitable for the

To simultaneously determine protein digestibility in small intestine and transport through epithelial cells, a hybrid model is often employed (Dhuique-Mayer et al. 2007; Jin et al. 2006; Versantvoort et al. 2005). The intestinal epithelium is represented by the Caco-2 cell culture, which is grown to the monolayer state on a membrane that separates two chambers. A test protein solution with digestive enzyme or after *in vitro* digestion is applied in one chamber and, if transport is possible, the products are collected from the second chamber. Two optimizations of this experimental design have been proposed: first, a mucin layer could be applied on the surface of the cell culture, thus mimicking the real mucus layer, which exists in intestine. Mucin protects cells from damage by luminal digestive enzymes and limits the passage of large (over 1500 Da) molecules to the cell surface (Jin et al. 2006). Second, Caco-2 cells could be co-cultured with HT29-MTX cells, which imitate mucus-producing cells of small intestine; this approach could be considered a more natural variant of previous modification (Yao et al. 2010). The core complication of such methods is the requirement for the facilities for operations with animal cell cultures. On the other hand, the model could be used for medium-sized screening, has a decent degree of intestinal wall imitation and thus

The final and most promising approach is the molecular modeling. In a work by Foltz et al*. (*Foltz et al. 2009) a special database was constructed to study the relationship between peptide structure and activity, permeability, and digestive stability. For this purpose, a total of 228 dipeptides were synthesized and their intestinal stability was evaluated by *in vitro* digestion. Then, a quantitative structure-activity relationship (QSAR) modeling was performed using partial least squares regression based on 400 molecular descriptors. The correlation coefficient for the best fit model was 0.76, and proteolytic stability for 12 new peptides was successfully predicted. As it is seen from the description above, molecular modeling has one major drawback: it requires a huge amount of preliminary data to build an efficient model. After that, it offers great speed and low experimental costs, but the aforementioned initial stage makes this approach suitable only for large screening

The common approaches to the determination of the protein and peptide digestibility in

 Allergenicity studies are usually the simplest ones and rely on direct hydrolysis of test substrate by pepsin and a combination of trypsin with chymotrypsin in simulated

**4. Current view of protein stability evaluation in different fields** 

Biological value is most commonly determined using animal assays.

processing of 40 samples per day.

the obtained results are quite relevant.

different areas of research varies significantly:

gastric and intestinal fluids.

experiments.

allowed to move from the main body into the antrum, and *vice versa*. The artificial antrum simulates the strong shear forces of the human antrum to reproduce the breakdown of the food particles and the preferential sieving observed *in vivo*. The mechanical processing of the food within the antrum is achieved by the sliding of a piston within a barrel, which forces the material through an elastic annulus where selective sieving takes place. Once ready, the processed bolus is ejected through the valve assembly and can be collected for further analysis.

Another widely used system is the Immobilized Digestive Enzyme Assay (IDEA), developed by Schasteen et al*. (*Schasteen et al. 2002). This system is somewhat more simple, and concentrates not on the grinding forces, but rather on the mixing of the «ingested» components and the food transition times through various digestive tract compartments. The overall digestion procedure consists of a stepwise acid solubilization, pepsin digestion, neutralization, trypsin, chymotrypsin and intestinal peptidase digestion followed by analysis of hydrolysis products. As follows from the model name, it relies on the glass beads-immobilized enzymes, and this allows for substantial costs reduction. The downside of this particular model is very long analysis time — up to 2.5 days.

The digestive tract simulators possess several advantages. They are more efficient at predicting of the fate of tablets within the gastrointestinal tract and more accurate at the simulation of the interaction of investigated protein or peptide with other meal and chyme components. The use of immobilized enzymes allows the reduction of experiment costs. However, the pay-off of these systems lies within their design: the complete analysis procedure usually takes a day or even two, and the complexity of the system dramatically reduces the throughput, as usually only one protein could be analyzed by one device at once. Finally, such machines are usually too complex to build by a research team, and thus require a substantial capital investment at the initial stage of a research project.

### **3.5 Other approaches**

Beyond aforementioned widely used approaches to study of protein digestibility, several less common or hybrid ones also exist.

First of all, when it comes to the evaluation of protein nutritive value, the simplest way to determine it is to calculate relative quantity of essential amino acids in the test protein relative to some well-known food protein. Hansen et al. *(*Hansen 1975) describes two such indices: Essential Amino Acid Index (EAAI) and Arnoulds index. The reference protein in both cases is egg protein; the latter index differs by including of the sum of the nonessential amino acids (NEAA) as part of equation. An unquestionable advantage of the amino acid scores is very low requirement of experimental work for production of preliminary estimates about a protein. However, this method does not take into account any properties of a protein, thus making impossible to draw any conclusions on its real bioaccessibility.

Then, von der Decken et al. point out to the fact, that the protozoan *Tetrahymena pyriformis W* has a proteolytic enzyme to digest proteins and similar to human needs for essential amino acids, and thus could be a model for protein digestion (von der Decken 1983). Time consumption of such model is, however, rather high (48 to 66 hours), and the details of animal digestion process are missed as well.

allowed to move from the main body into the antrum, and *vice versa*. The artificial antrum simulates the strong shear forces of the human antrum to reproduce the breakdown of the food particles and the preferential sieving observed *in vivo*. The mechanical processing of the food within the antrum is achieved by the sliding of a piston within a barrel, which forces the material through an elastic annulus where selective sieving takes place. Once ready, the processed bolus is ejected through the valve assembly and can be collected for

Another widely used system is the Immobilized Digestive Enzyme Assay (IDEA), developed by Schasteen et al*. (*Schasteen et al. 2002). This system is somewhat more simple, and concentrates not on the grinding forces, but rather on the mixing of the «ingested» components and the food transition times through various digestive tract compartments. The overall digestion procedure consists of a stepwise acid solubilization, pepsin digestion, neutralization, trypsin, chymotrypsin and intestinal peptidase digestion followed by analysis of hydrolysis products. As follows from the model name, it relies on the glass beads-immobilized enzymes, and this allows for substantial costs reduction. The downside

The digestive tract simulators possess several advantages. They are more efficient at predicting of the fate of tablets within the gastrointestinal tract and more accurate at the simulation of the interaction of investigated protein or peptide with other meal and chyme components. The use of immobilized enzymes allows the reduction of experiment costs. However, the pay-off of these systems lies within their design: the complete analysis procedure usually takes a day or even two, and the complexity of the system dramatically reduces the throughput, as usually only one protein could be analyzed by one device at once. Finally, such machines are usually too complex to build by a research team, and thus

Beyond aforementioned widely used approaches to study of protein digestibility, several

First of all, when it comes to the evaluation of protein nutritive value, the simplest way to determine it is to calculate relative quantity of essential amino acids in the test protein relative to some well-known food protein. Hansen et al. *(*Hansen 1975) describes two such indices: Essential Amino Acid Index (EAAI) and Arnoulds index. The reference protein in both cases is egg protein; the latter index differs by including of the sum of the nonessential amino acids (NEAA) as part of equation. An unquestionable advantage of the amino acid scores is very low requirement of experimental work for production of preliminary estimates about a protein. However, this method does not take into account any properties of a protein, thus making impossible to draw any conclusions on its real

Then, von der Decken et al. point out to the fact, that the protozoan *Tetrahymena pyriformis W* has a proteolytic enzyme to digest proteins and similar to human needs for essential amino acids, and thus could be a model for protein digestion (von der Decken 1983). Time consumption of such model is, however, rather high (48 to 66 hours), and the details of

of this particular model is very long analysis time — up to 2.5 days.

require a substantial capital investment at the initial stage of a research project.

further analysis.

**3.5 Other approaches** 

bioaccessibility.

less common or hybrid ones also exist.

animal digestion process are missed as well.

Our group has recently suggested an assay, that tries to replicate the conditions of parietal digestion in small intestine, at the same time maintaining capability for high-throughput sample analysis (Akimov et al. 2010). The active component of this system is a fragment of rat stomach or intestinal wall, immersed in simulated gastric or intestinal fluid. The products are analyzed by HPLC. The utilization of tissue fragments allows for reduction of animal consumption, while still maintaining the whole set of intestinal peptidases and providing some phosphatidyl choline in the medium. The assay design is suitable for the processing of 40 samples per day.

To simultaneously determine protein digestibility in small intestine and transport through epithelial cells, a hybrid model is often employed (Dhuique-Mayer et al. 2007; Jin et al. 2006; Versantvoort et al. 2005). The intestinal epithelium is represented by the Caco-2 cell culture, which is grown to the monolayer state on a membrane that separates two chambers. A test protein solution with digestive enzyme or after *in vitro* digestion is applied in one chamber and, if transport is possible, the products are collected from the second chamber. Two optimizations of this experimental design have been proposed: first, a mucin layer could be applied on the surface of the cell culture, thus mimicking the real mucus layer, which exists in intestine. Mucin protects cells from damage by luminal digestive enzymes and limits the passage of large (over 1500 Da) molecules to the cell surface (Jin et al. 2006). Second, Caco-2 cells could be co-cultured with HT29-MTX cells, which imitate mucus-producing cells of small intestine; this approach could be considered a more natural variant of previous modification (Yao et al. 2010). The core complication of such methods is the requirement for the facilities for operations with animal cell cultures. On the other hand, the model could be used for medium-sized screening, has a decent degree of intestinal wall imitation and thus the obtained results are quite relevant.

The final and most promising approach is the molecular modeling. In a work by Foltz et al*. (*Foltz et al. 2009) a special database was constructed to study the relationship between peptide structure and activity, permeability, and digestive stability. For this purpose, a total of 228 dipeptides were synthesized and their intestinal stability was evaluated by *in vitro* digestion. Then, a quantitative structure-activity relationship (QSAR) modeling was performed using partial least squares regression based on 400 molecular descriptors. The correlation coefficient for the best fit model was 0.76, and proteolytic stability for 12 new peptides was successfully predicted. As it is seen from the description above, molecular modeling has one major drawback: it requires a huge amount of preliminary data to build an efficient model. After that, it offers great speed and low experimental costs, but the aforementioned initial stage makes this approach suitable only for large screening experiments.
