In vitro methods

In vitro digestibility methods began as simple one step incubations with pepsin or other proteases such as trypsin, papain, pronase or rennin⁽Reference Boisen²⁾. These early single enzyme methods appeared to be satisfactory for comparing the effects of various treatments on a single foodstuff but gave lower digestible protein values than those obtained in vivo (faecal nitrogen digestibility). Other methods have simulated gastric and intestinal digestion using a 2-stage in vitro digestion involving digestion in a pepsin-HCl mixture followed by neutralisation, and then digestion in pancreatin⁽Reference Akeson and Stahmann^8, Reference Büchmann⁹⁾, trypsin⁽Reference Saunders, Connor and Booth¹⁰⁾, or intestinal fluid from pigs⁽Reference Furuya, Sakamoto and Takahashi¹¹⁾. In general, there has been good agreement between these in vitro results and in vivo rat true faecal nitrogen digestibility.

Multi-enzyme processes that measure the pH drop after a 10 minute digestion have been developed⁽Reference Hsu, Vavak and Satterlee^12, Reference Satterlee, Marshall and Tennyson¹³⁾. Good correlations (r>0·8) were found with in vivo digestibility (rat true faecal) particularly when the protein sources were analysed by plant or animal origin⁽Reference Pedersen and Eggum^14, Reference Pedersen and Eggum¹⁵⁾, and this method was found to be highly reproducible across six laboratories⁽Reference McDonough, Sarwar and Steinke¹⁶⁾. This approach assumes that the rate of change in pH is correlated with protein digestibility and that there is a direct relationship between the observed pH drop and the extent of protein hydrolysis⁽Reference Barbeau and Kinsella^17, Reference Mozersky and Panettieri¹⁸⁾. The components of some food materials, however, interfere with the pH drop due to their buffering capacity⁽Reference Pedersen and Eggum^14, Reference Moughan, Schrama and Skilton^19, Reference Urbano, Lopez-Jurado and Frejnagel²⁰⁾.

A modification of this approach is the pH stat method where the pH is kept constant by automatic titration with 0·1 M NaOH and the total amount of alkali used at the end of the incubation is recorded. This modification improved the prediction of in vivo protein digestibility, was reproducible, and gave high correlations (r>0·90) with in vivo (rat faecal) digestibility for highly digestible animal protein sources⁽Reference Pedersen and Eggum¹⁵⁾.

The immobilised digestive enzyme assay (IDEA) system⁽Reference Porter, Swaisgood and Catignani²¹⁾ uses a bioreactor containing enzymes immobilised on glass beads, eliminating contamination of the digest with digestive enzymes and preventing autolysis. However, this assay takes two and a half days to complete so is not as rapid as other in vitro techniques. Predicted digestibility values from the IDEA assay have been found to correlate (r = 0·8, 0·83) with rat faecal protein digestibility for a range of foods⁽Reference Thresher, Swaisgood and Catignani^22, Reference Chang, Catignani and Swaisgood²³⁾.

Gauthier et al. ⁽Reference Gauthier, Vachon and Jones²⁴⁾ developed an in vitro digestion under constant dialysis (molecular weight cut off 1000 Daltons) with specialised apparatus (dialysis cell method) to address the concern that enzyme activity is suppressed by the products of digestion. Continuous dialysis, however, needs extra and complex equipment, and some researchers do not regard it as necessary⁽Reference Boisen and Eggum^3, Reference Büchmann⁹⁾. The proteins undigested following a pepsin-pancreatin digestion were determined using polyacrylamide gel electrophoresis (PAGE), allowing the estimation of the molecular weight of the polypeptides and proteins remaining after hydrolysis⁽Reference Kim and Barbeau²⁵⁾. This procedure was time consuming, taking 5 days to complete, and is likely too labour intensive to be used in a rapid routine assay.

Near-infrared spectroscopy (NIRS) is a very rapid technique with low maintenance costs that shows great promise in evaluating nutritive value for human foods. It has been used routinely for determining the chemical composition, organic matter digestibility, energy digestibility of grains⁽Reference Hervera, Baucells and Gonzalez^28, Reference Pujol, Pérez-Vendrell and Torrallardona²⁹⁾ and oil seeds⁽Reference Losada, Garcia-Rebollar and Alvarez²⁶⁾ for monogastric animals. There is a requirement, however, for a large number of reference samples to calibrate the instrument and these reference data must come from in vivo digestibility assays. There have been only a few published studies comparing NIRS and in vivo nitrogen or protein digestibility values. Prediction of wheat ileal crude protein digestibility for broilers⁽Reference Owens, McCann and McCracken²⁷⁾ was found to be highly variable (r ² = 0·23–0·76). Faecal digestible protein of commercial dry extruded foods for dogs showed good predictions using the pH drop (r ² = 0·78) method, and in vitro two-step digestion (r ² = 0·81), but not for NIRS (r ² = 0·53)⁽Reference Hervera, Baucells and Gonzalez²⁸⁾. Pig ileal nitrogen digestibility of barley (r ² = 0·97, 0·72)⁽Reference Pujol, Pérez-Vendrell and Torrallardona^29, Reference McCann, McCracken and Agnew³⁰⁾ but not wheat (r ² = 0·22)⁽Reference Garnsworthy, Wiseman and Fegeros³¹⁾ has been accurately predicted with NIRS.

A complex biochemical model of human adult and infant gastro-duodenal digestion has been developed⁽Reference Dupont, Mandalari and Molle^32, Reference Dupont, Mandalari and Molle³³⁾ to investigate the allergenic potential of protein digestion products. Electrophoresis, immunochemical and mass spectrophotometry techniques were used to characterise the digestion products. This study found that heat treatment of milk increased the resistance of casein to the simulated digestion processes. An inter-laboratory study⁽Reference Mandalari, Adel-Patient and Barkholt³⁴⁾, comparing the digestion of β-casein and β-lactoglobulin using a simulated human gastro-duodenal in vitro assay, found it to be reproducible.

A computer-controlled, dynamic, multi-compartmental model (TIM) that seemingly closely simulates the conditions of the in vivo gastrointestinal tract of humans and monogastric animals has been developed at the TNO Nutrition & Food Research Institute, The Netherlands⁽Reference Minekus, Marteau and Havenaar³⁵⁾. It comprises compartments with flexible walls heated by water jackets with computer controlled pH adjustment, and rotary and peristaltic pumps to provide mechanical movement of the chyme. It simulates gastric pH change, peristaltic movement, gastric emptying rates, intestinal transit times, gastric, biliary and pancreatic secretion and their activities, small intestinal absorption (TIM-1), and the absorption from the large bowel (TIM-2) of volatile fatty acids and water⁽Reference Minekus, Marteau and Havenaar^35, Reference Havenaar and Minekus³⁶⁾. This model has been applied in one published protein digestion study to date⁽Reference Venema and Dohnalek³⁷⁾ in which the model was modified to simulate the infant gut and to measure the degree of protein hydrolysis of whey protein hydrolysate infant formulas. The hydrolysis of the proteins was found to be only one factor in determining the digestibility of these food products. TIM has been used to investigate nutraceutical delivery vehicles⁽Reference Chen, Hebrard and Beyssac³⁸⁾, gut transit time and iron absorption⁽Reference Salovaara, Alminger and Eklund-Jonsson³⁹⁾, and the availability of heterocyclic aromatic amines⁽Reference Krul, Luiten-Schuite and Baan⁴⁰⁾, and has been used in pharmaceutical studies⁽Reference Blanquet, Zeijdner and Beyssac⁴¹⁾. One of the strengths of these models is their flexibility and ability to model different gastrointestinal conditions such as those found in infants ⁽Reference Dupont, Mandalari and Molle^32, Reference Dupont, Mandalari and Molle^33, Reference Venema and Dohnalek³⁷⁾, or other animal species⁽Reference Venema, Havenaar and Minekus^42, Reference Smeets-Peeters, Minekus and Havenaar⁴³⁾.

A similar system to TIM is the Institute of Food Research, United Kingdom, dynamic gastric model (DGM) comprising two parts that mimic the fundus and antrum of the stomach, and this is integrated with a simulated intestine that mixes the digesta and adds bicarbonate, bile and digestive enzymes⁽Reference Wickham, Faulks and Mills⁴⁴⁾ and has also been used in pharmaceutical studies. Model stomach⁽Reference Kong and Singh^45–Reference Ferrua and Singh⁴⁷⁾ and small intestine⁽Reference Tharakan, Norton and Fryer⁴⁸⁾ systems have also been developed to better understand the mechanical forces and mass transfer kinetics of food digestion and absorption. To date these systems have not been used for the routine evaluation of protein or amino acid digestibility of foods.

These complex computer controlled systems are expensive to set up and maintain and therefore may not be a useful tool for the routine evaluation of food. Although being far more sophisticated than simple digestion assays, they do not provide an entirely accurate reproduction of the physiological system whereby the digesta interact with the gut cells leading to active transport of nutrients; nor do they simulate the neural and hormonal feedback mechanisms affecting digestion and absorption⁽Reference Yoo and Chen⁴⁹⁾. Their validity for predicting the in vivo digestion of food proteins has not yet been fully tested. The choice of a standard method will be a balance between accuracy and ease of use, and the endpoint may simply be to rank foods in order of nutrient digestibility. Multiple regression equations combining in vitro digestibility coefficients and chemical constituent measures may have application where greater accuracy is required.

Limitations of the in vitro approach

In general, there are some key requirements for the development of in vitro digestibility assays: matching in vivo enzymes in presence, sequence, enzyme:substrate ratios; standardising enzyme activities and specificities; controlling co-enzymes and co-factors, pH and temperature; separating digested from undigested material while considering the inhibition of end products on digestion; and allowing for the effects of sample size, particle size and particle size distribution⁽Reference Moughan⁶⁾. Digestion and absorption involve complex physiological processes that are virtually impossible to reproduce in vitro ⁽Reference Boisen^2, Reference Savoie⁵⁰⁾. The effects of anti-nutritional factors, dietary dry-matter and fibre, endogenous protein secretions, activity of gut enzymes, and gut bacteria are not able to be mimicked in vitro ⁽Reference Moughan⁶⁾. The sensitivity of an assay is a function of both the time of the reaction and the enzyme-substrate ratio⁽Reference Parsons^51, Reference Drake, Fuller and Chesson⁵²⁾. In vitro assays may need to include lipases, carbohydrases, and elastases to better mimic in vivo digestion and the release of proteins from the food matrix. There is an assumption that all soluble material is digestible whereas small peptides may not be absorbed in vivo particularly in heat-treated proteins.

An in vitro method may be precise and reliable in the laboratory but the resulting data must be correlated with in vivo data to provide a physiologically relevant measure of digestible protein. While it is not necessarily realistic to expect absolute agreement between in vitro and in vivo observations, in vitro assays should not be ruled out as useful tools. They can be used to rank food proteins according to their protein digestibility, to further our understanding of protein structure under digestion conditions⁽Reference Monogioudi, Faccio and Lille⁵³⁾, and with further refinement could be useful for the prediction of in vivo nutritive value. The development of multiple regression equations including in vitro digestibility measures combined with chemical constituent measures for individual foods has been recommended⁽Reference Moughan⁶⁾. In addition, in vitro digestion techniques have an important role to play in extending our understanding of the release of amino acids and peptides during digestion and the influence protein structure plays on dietary protein quality⁽Reference Savoie, Agudelo and Gauthier⁵⁴⁾.

Evaluation and validation with in vivo assays

Many of the studies evaluating in vitro protein digestibility assays have compared the in vitro data with in vivo faecal digestibility. The influence of the microflora in the large intestine however, does not make this comparison useful. The ileal method for determining crude protein digestibility is more accurate⁽Reference Moughan and Smith⁵⁵⁾ than the faecal approach, and it is more relevant to in vitro measures given that these methods do not simulate microbial digestion in the large intestine⁽Reference Moughan⁵⁶⁾. It has been recommended previously that protein sources that do not contain fibre or anti-nutritional factors should be evaluated against true ileal protein digestibility⁽Reference Moughan, Schrama and Skilton¹⁹⁾, while plant proteins including grains should be evaluated against real ileal protein digestibility⁽Reference Moughan⁶⁾. Further details of these in vivo assays are given elsewhere in this supplement. There are only a few studies that have used these recommended comparisons and it is therefore difficult to properly evaluate the accuracy of current in vitro assays.

The determination of true or real ileal amino acid and nitrogen digestibility requires access to the terminal ileum, which is usually obtained via surgery, as well as a specific feeding or labelling technique to measure endogenous losses. The determination of in vivo protein and amino acid digestibility in humans is generally not possible so researchers use animal models. The pig⁽Reference Rowan, Moughan and Wilson^57, Reference Deglaire, Bos and Tome⁵⁸⁾ and the rat⁽Reference Bodwell, Satterlee and Hackler^59, Reference Rich, Satterlee and Smith⁶⁰⁾ have been found to be suitable animal models for humans. There is very little published information on ileal amino acid digestibility determined for humans directly⁽Reference Fuller and Tome⁶¹⁾, making the evaluation of in vitro digestibility data even more difficult. One of the few studies to compare human digestibility (albeit faecal digestibility) and in vitro digestibility values on the same foods is that of Bodwell et al. ⁽Reference Bodwell, Satterlee and Hackler⁵⁹⁾. They found there was no significant correlation between in vitro, human and rat digestibilities when all the proteins were considered together. However, when the data were analysed with the food proteins identified separately as of animal or plant origin the correlations were high (r>0·90).

There have been mixed results in studies comparing in vitro and in vivo ileal digestibilities in pigs on the same protein sources ranging from no significant correlation⁽Reference Thomaneck, Hennig and Souffrant⁶²⁾, low correlation⁽Reference Cone and van der Poel⁶³⁾, to good correlations⁽Reference Pujol, Torrallardona and Boisen^64, Reference Jezierny, Mosenthin and Sauer⁶⁵⁾. A comprehensive study of a wide range of feedstuffs⁽Reference Jaguelin, Fevrier and Seve⁶⁶⁾ used standardised true ileal nitrogen digestibility determined in the growing pig and a two-step pepsin/pancreatin in vitro digestion. High correlations were obtained within feedstuffs when the prediction equations also included terms for specific chemical components of the foods. As would be expected, true ileal digestibility gave better correlations than when apparent ileal digestibility was used. The authors concluded that due to differences in endogenous protein loss induced by the different foods a single prediction equation based on the in vitro assay would not be appropriate.

Methodology needs to be standardised across research organisations⁽Reference Mosenthin, Jansman and Eklund^67, Reference Stein, Fuller and Moughan⁶⁸⁾ and reference proteins should be included in every assay for comparison across studies. In vivo digestibility is influenced by the experimental conditions, and the way digestibility is calculated such as the presence of endogenous protein, and methods of endogenous loss correction⁽Reference Boisen and Moughan^69, Reference Boisen and Moughan⁷⁰⁾. It is now widely accepted that protein and amino acid digestibility must be determined using accurate measurements of dietary intake and terminal ileal digesta flow corrected for basal endogenous protein loss to give ‘true’ digestibility values. However, some foods also contain components (dietary fibre, anti-nutritional factors) that induce higher endogenous protein losses than the basal losses measured with purified highly digestible protein-free diets. There have been only a few studies⁽Reference Hur, Lim and Decker^7, Reference Pujol and Torrallardona⁷¹⁾ that have attempted standardisation of methodology and more need to published⁽Reference Hur, Lim and Decker⁷⁾, and correlations determined both within and across food proteins.