Critical Reviews in Food Science and Nutrition

ISSN: 1040-8398 (Print) 1549-7852 (Online) Journal homepage: http://www.tandfonline.com/loi/bfsn20

Can dynamic in vitro digestion systems mimic the

physiological reality?

D. Dupont, M. Alric, S. Blanquet-Diot, G. Bornhorst, C. Cueva, A. Deglaire,

S. Denis, M. Ferrua, R. Havenaar, J. Lelieveld, A. R. Mackie, M. Marzorati, O.
Menard, M. Minekus, B. Miralles, I. Recio & P. Van den Abbeele

To cite this article: D. Dupont, M. Alric, S. Blanquet-Diot, G. Bornhorst, C. Cueva, A. Deglaire,

S. Denis, M. Ferrua, R. Havenaar, J. Lelieveld, A. R. Mackie, M. Marzorati, O. Menard, M.
Minekus, B. Miralles, I. Recio & P. Van den Abbeele (2018): Can dynamic in vitro digestion
systems mimic the physiological reality?, Critical Reviews in Food Science and Nutrition, DOI:
10.1080/10408398.2017.1421900

To link to this article: https://doi.org/10.1080/10408398.2017.1421900

Published online: 23 Jan 2018.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=bfsn20
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION
https://doi.org/10.1080/10408398.2017.1421900

Can dynamic in vitro digestion systems mimic the physiological reality?

D. Duponta, M. Alricb, S. Blanquet-Diotb, G. Bornhorstc, C. Cuevad, A. Deglairea, S. Denisb, M. Ferruae, R. Havenaarf,
J. Lelieveldf, A. R. Mackieg, M. Marzoratih, O. Menarda, M. Minekusf, B. Mirallesd, I. Reciod, and P. Van den Abbeelei
a
INRA Agrocampus Ouest, STLO, Rennes, France; bUniversit
e Clermont Auvergne, Clermont-Ferrand, France; cUniversity of California, Davis, USA; dCSIC
Universidad Autonoma de Madrid, CIAL, Madrid, Spain; eFonterra, Palmerston North, New Zealand; fTriskelion, Zeist, The Netherlands; gUniversity of
Leeds, Leeds, United Kingdom; hUniversity of Ghent, Ghent, Belgium; iProDigest BVBA, Gent, Belgium

ABSTRACT KEYWORDS
During the last decade, there has been a growing interest in understanding the fate of food during Dynamic in vitro digestion;
digestion in the gastrointestinal tract in order to strengthen the possible effects of food on human health. food; stomach; small
Ideally, food digestion should be studied in vivo on humans but this is not always ethically and financially intestine; colon
possible. Therefore simple static in vitro digestion models mimicking the gastrointestinal tract have been
proposed as alternatives to in vivo experiments but these models are quite basic and hardly recreate the
complexity of the digestive tract. In contrast, dynamic models that allow pH regulation, flow of the food
and injection in real time of digestive enzymes in the different compartments of the gastrointestinal tract
are more promising to accurately mimic the digestive process. Most of the systems developed so far have
been compared for their performances to in vivo data obtained on animals and/or humans. The objective
of this article is to review the validation towards in vivo data of some of the dynamic digestion systems
currently available in order to determine what aspects of food digestion they are able to mimic. Eight
dynamic digestion systems are presented as well as their validation towards in vivo data. Advantages and
limits of each simulator is discussed. This is the result of a cooperative international effort made by some
of the scientists involved in Infogest, an international network on food digestion

Introduction (several compartments). The different systems available have been

described recently (Guerra et al., 2012) and a general description of
Digestion is a complex process that will provide nutrients to the
the different systems investigated is presented in Table 1. In this
body and release molecules in the gastrointestinal tract that can
review, we particularly would like to focus on their ability to simu-
have a beneficial or a deleterious effect on human health.
late the physiological reality and recreate what happens in the gas-
Therefore, understanding the fate of food in the digestive tract
trointestinal tract of animals or humans. This is the contribution of
is a way to increase our knowledge on the effect of food on
scientists involved in the international Infogest network (www.
health. When entering in the gastrointestinal tract food will be
cost-infogest.eu) that aims at understanding the fate of food in
disintegrated in the different compartments (mouth, stomach,
the gastrointestinal tract.
small and large intestine), macronutrients will be hydrolyzed
and micronutrients will be absorbed.
Investigating food digestion using in vivo models (animals Mono-compartmental systems
or humans) is rather difficult, expensive and sometimes ethi-
The Dynamic Gastric Model (DGM)
cally questionable. For this reason, several in vitro models
have been developed. Most of the numerous protocols Origins of the system
described in the literature are static ones and consist in The Dynamic Gastric Model was developed at the Institute
placing the food in a series of bioreactors where the physico- of Food Research (Norwich, UK) to address the need for a
chemical and enzymatic environment of each digestive model that could simulate both the biochemical and
compartment is recreated. However, digestion is a dynamic mechanical processes occurring during human gastric diges-
process and therefore these models exhibit strong limitations: tion in a physiologically relevant manner. The DGM was
there is no flow of the food between the different compart- initially developed to further food research and to enable
ments and the pH, digestive enzymes and bile concentrations the study of parameters such as nutrient bioaccessibility,
are kept constant. For these reasons, dynamic systems have effect of food structure on dissolution and bioactive deliv-
been designed and protocols are available for simulating food ery, nutrient interactions, and the survival and delivery of
digestion. functional foods. Its close simulation of physiological con-
Dynamic systems are either monocompartmental (simulate one ditions has meant the DGM has also increasingly been
compartment of the gastrointestinal tract) or multicompartmental used by the pharmaceutical industry as an in vitro tool to

CONTACT Dr. D. Dupont, PhD [email protected] INRA Agrocampus Ouest, STLO, 65 rue de Saint-Brieuc, 35042 Rennes, France.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/bfsn.
© 2018 Taylor & Francis Group, LLC
Table 1. The dynamic digestion systems investigated and their main characteristics. 2

Body Control of Mixing in Gastric Control of Digestive Intestinal Intestinal Intestinal

System temp. gastric pH the stomach emptying intestinal pH secretions transit absorption microbiota Photo

DGM C C Contractions by water pressure C N/A Saliva N/A N/A N/A

with piston and barrel Gastric fluid
HCl
D. DUPONT ET AL.

HGS C C Contractions by mechanical C N/A Saliva Gastric fluid N/A N/A N/A
driving device HCL

ARCOL C N/A N/A N/A N/A N/A C C C

colonic retention time controlled passive absorption of
by the inlet flow of nutritive water and microbial
medium, outlet flow of metabolites with a
fermentation medium and dialysis system using
volume inside the bioreactor hollow fibers

DIDGI C C Rotational stirring C C Saliva C ¡ ¡

Gastric fluid
HCl
Intestinal fluid
Bile
NaHCO3

TIM-1 C C C C C C C C C
Meal specific curves Contractions flexible wall by Simulation of pyloric Site specific for Saliva Based on meal specific gastric Jejunum and ileum: TIM-2: Complex high
water pressure. sphincter. Meal duodenum, Gastric fluid emptying, intestinal transit Dialysis for water density (>1011 cfu/
Also for small intestinal specific curves Jejunum and HCl and ileal- emptying curves, soluble compounds. g) microbiota of
compartments ileum Intestinal fluid controlled with peristatic Filtration for lipid soluble human or animal
Pancreatic juice valve-pumps. compounds origin
Bile
NaHCO3
tiny-TIM C C C C C C C C C C
AGC Meal specific curves Contractions flexible walls for Simulation of pyloric Conditions for Saliva Based on meal specific gastric Dialysis for water soluble TIM-2
Incl. infant corpus, proximal and distal sphincter. Meal duodenum of Gastric fluid emptying. compounds.
conditions antrum. specific curves ‘overall’ small HCl Optional: ileal emptying curves Filtration for lipid soluble
Also for small intestinal intestine Intestinal fluid controlled with peristatic compounds
compartment Pancreatic juice valve-pump.
Bile
NaHCO3

SHIME C C § C C C C C C
Magnetic Stirring Saliva Luminal and mucosal
Gastric fluid microbiota
HCl
Intestinal fluid
Bile
NaHCO3

ESIN C C C C C Saliva C C ¡
HCl for the stomach 2 pistons inside cylinders with Differential gastric Gastric fluid Passive absorption of
and NaHCO3 back and forth frequency emptying of HCl water and digestion
for the small fixed at 3 movements/min solids and liquids Intestinal fluid products with dialysis
intestine with 2 separate Bile fibers in the jejunum
pumps NaHCO3 and ileum

SIMGI C C Contractions C C Saliva C C C

Water pressure Gastric fluid
HCl
Intestinal fluid
Bile
NaHCO3
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION
3
4 D. DUPONT ET AL.

study the effect of food matrices on the disintegration and the DGM through the use of a “dead volume,” i.e. a defined
dissolution of drug formulations and the delivery profile of space between barrel and piston whose volume is main-
drugs to the duodenum. This success is in part due to its tained during ejection thereby allowing large, dense particles
ability to realistically process any complex food matrix for to remain in the antrum and undergo repeated processing
direct comparison with the results of in vivo/clinical stud- cycles. At the end of a simulated digestion, any material
ies. The design of the DGM is based on extensive research remaining in this dead volume is ejected to simulate the
into gastric digestion and the physiology of the human phase III contraction (housekeeper wave) which fully emp-
stomach, both biochemical and mechanical (Wickham ties the human stomach at the end of gastric digestion. Fol-
et al., 2012). lowing ejection from the DGM, samples can be subjected to
further digestion using a static duodenal model. To this
Short description of the system end, the pH of the samples is elevated and a physiological
The gastric digestion of food involves secretions from the mix of bile salts with lecithin and cholesterol and pancreatic
gastric mucosa and a change in peristaltic contractions. enzymes, is added to simulate conditions found within the
Within the DGM, acid and enzyme solutions are added duodenum.
through a perforated loop situated at the top of the fundus
and allowing a flow of secretion down the wall of the stom- Validation of the system towards in vivo data animal and/or
ach. The flow rates of secretions are controlled dynamically human
and the rate of acid addition slows in response to the drop The grinding forces of the DGM and a Dissolution Appara-
in pH as detected by the pH electrode positioned in the tus USP-II operated at two rotational speeds (50 and
fundus. The DGM simulates the fundus and the antrum of 100 rpm) were measured using the breakdown of agar gel
the stomach. Within the fundus/main body, the food bolus beads of various fracture strengths in high and low-viscosity
is subjected to rhythmic squeezing brought about by cyclical meals and compared to in vivo data collected on human
pressurization of the 37
C water jacket surrounding it. The volunteers (Vardakou et al., 2011a). For this experiment,
DGM antrum consists of a barrel and a piston, which move the DGM was designed to replicate the real-time changes in
within a water jacket. While the piston draws portions of pH, enzyme addition, shearing, mixing, and retention time
food bolus through an inlet valve from the fundus into the of an adult human stomach. The model can be fed ‘meals’
antrum, it is the upward and downward movement of the ranging from a glass of water to high fat meals (i.e. the
barrel during processing which exerts shear stresses on the FDA high fat American breakfast) and deliver samples from
antral contents. This is due to a flexible annulus mounted its ‘antrum’ in the same processed form and at the same
within the top part of the barrel through which food (and rate as seen in vivo. The data used to program the DGM
formulations) passes during every stroke, thereby simulating were derived from echo-planar imaging studies (Marciani
the rhythmic peristaltic contractions of the human stomach. et al., 2009; Marciani et al., 2001b) and from published
While the speed of movement has been calibrated to pro- references detailing physiological ranges for the rate of pro-
vide physiological shear forces (Vardakou et al., 2011b), the duction of gastric secretions (Geigy, 1981). All beads tested
actual volume of food bolus processed within the antrum at in the DGM broke after a certain amount of gastric proc-
any one time, as well as duration of processing are tailored essing. The results expressed as MBT obtained for the beads
to the specific meal used (volume, composition, calorific at the four strengths administered in low (LV LBG) and
content). At pre-defined intervals, the inlet valve closes and high viscosity Locust Beam Gum (HV LBG) meals are rep-
the outlet valve opens, allowing the processed chyme to be resented in Figure 1b in order to facilitate a direct visual
ejected from the DGM. Gastric sieving is simulated within comparison with the in vivo data, Figure 1a. Increasing the

Figure 1. Mean Breaking Time (MBT) of agar gel beads for four bead breakdown forces for both low-viscosity (LV—grey bars) and high-viscosity (HV—black bars) meals.
n D 9 for the LV and n D 8 for the HV for each bead strength. Panel b in the DGM; n D 5 for each bead strength. p < 0.05 vs. each of the two lower beads strengths for
the LV meal.
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 5

Figure 2. Correlation of the in vitro MBT (DGM) and in vivo MBT for the beads at the different breaking forces. Panel a LV meal. Panel b HV meal. Straight lines are the
regression lines. Bars represent SE

viscosity of the meal reduced the survival time of the harder Human Gastric Simulator (HGS)
beads. The interrelationship between the in vitro data
Origins of the system
obtained with the DGM and those observed in vivo (Mar-
The human gastric simulator (HGS) system was developed
ciani et al., 2001a) is clearly visible from the graphs
at the University of California, Davis to enable measure-
depicted in Figure 2. Even though the R2 in Eqs. 5 and 6
ment of gastric food breakdown in a system with physio-
are not very high, there is a clear correlation between the in
logically-relevant physical and chemical conditions to the
vitro (DGM) and the in vivo data. One-way ANOVA analy-
stomach. This mono-compartmental system focuses on
sis showed that no statistical difference exists between any
gastric digestion. However, oral and/or small intestinal
of the data collected from the DGM and those found in
stages may be incorporated either before or after testing
human. This indicates that the forces produced during the
using the HGS, respectively. Two generations of this model
DGM processing are within the range of forces exerted by
have been developed (Guo et al., 2014; Kong and Singh,
the human gastric compartment in vivo. Furthermore, the
2010; Phinney, 2013), both incorporating a flexible gastric
DGM showed to be able to discriminate between the two
vessel, continuous peristaltic contractions provided by roll-
meals, similarly to the finding of (Marciani et al., 2001a).
ers, controllable secretions (enzymes, pH), and gastric
The different behavior observed for the beads in the LV
emptying.
and HV meals is of special interest particularly when con-
sidering the effect that it may have on dosage forms for
which the drug release is greatly susceptible to the shear Short description of the system
forces applied to its surface, as in the case of erodible The HGS model (Table 1) consists of a flexible outer vessel
matrixes. to simulate the stomach. Although the vessel may be filled
with several liters of material, the typical amount of mate-
Advantages and limitations of the system rial (e.g. the “meal” and secretions) is » 0.9–1.0 L, which is
The DGM accurately processes real food items and meals as the volume that can be ingested without resulting in
eaten and simulates gastric mixing, transit and breakdown increases in gastric luminal pressure (Ferrua and Singh,
forces within the normal physiological range. The system 2010). The rollers that simulate peristaltic contractions are
adjusts acid and enzyme additions (quantity and rate) controlled by a variable-speed motor, which can be used to
according to the calculated emptying rate and physical change the contraction frequency. To simulate normal adult
processing based on the food matrix, allowing fed and gastric digestion, a frequency of » 3 contractions/minute is
fasted state comparisons and studies of the impact of differ- used, according to previous in vivo studies in humans
ent meals/food items on dosage form behavior. It provides (Hocke et al., 2009; Marciani et al., 2001c). The gastric
samples of digested materials at any sampling time, within secretions are added through tubes entering the top of the
the total digestion period. Thus it provides an accurate sim- vessel. The secretion rate and specific secretion composition
ulation of gastric dissolution. However, the DGM only (pH, enzymes, salts, mucin) can be varied, depending on
models the behavior of the gastric compartment, necessitat- the goal of the study. Samples are emptied through a small
ing a method of simulating the oral phase (e.g. chew and tube (0.01 m inner diameter) in the distal portion of the
spit) and the intestinal phase if the fate of nutrients and vessel. A mesh with 1 mm openings is may be used to con-
bioactives is to be investigated fully. Perhaps one further trol the gastric emptying, such that only smaller particles
limitation is that it must be fed an adult meal and there is (< 1 mm diameter) can exit the stomach. The entire unit is
no way to alter emptying based on gastric phase behavior. kept inside of a temperature-controlled chamber maintained
The emptying is based on the average properties of the at 37
C using a small heater and fan (Guo et al., 2014;
meal. Kong and Singh, 2010; Phinney, 2013).
6 D. DUPONT ET AL.

Validation of the system towards in vivo data animal and/or

human
Data from the HGS model systems have been compared
with previously published in vivo data, and a systematic val-
idation of the model is currently ongoing. Specific compari-
sons can be made from the second generation HGS model
to in vivo animal studies using the growing pig. For both in
vitro and in vivo studies, meals of white rice (medium
grain, Calrose variety) were cooked following a standardized
procedure (Bornhorst et al., 2013a; Bornhorst et al., 2013b).
For the in vivo study, the growing pig (20.9 § 0.2 kg) was
used as a model for digestion in adult humans. Digestion
was monitored for up to 8 h. For the in vitro study, the
Figure 4. Comparison of intragastric pH between an in vivo study with growing
same meals of white rice were mixed with simulated saliva pigs (Bornhorst et al., 2014) and from the HGS in vitro system (Phinney, 2013).
and fed into the second generation HGS model. Digestion Samples were taken at two intragastric locations: Pylorus, representing a location
was monitored for up to 3 h. Specific experimental details near the pyloric sphincter (or HGS emptying tube), and Fundus, representing a
location at the top of the gastric fundus (or top of HGS gastric vessel). Values are
are given elsewhere (Bornhorst et al., 2013a; Bornhorst averages (n D 3, in vitro; n D 6, in vivo) with error bars representing the standard
et al., 2013b; Bornhorst et al., 2014; Phinney, 2013). deviation.
Figure 3 shows the correlation between the gastric emp-
tying rate of dry matter from the in vitro and in vivo fundus was 4.9 § 1.3 in vivo compared to 6.2 § 0.4 in
experiments. The solid line represents a 1:1 correlation (e.g. vitro. These differences may be the result of varying gastric
gastric emptying rate in vitro is the same as the gastric secretion rates in vivo that are difficult to predict and con-
emptying rate in vivo). It can be seen that the dry matter trol in vitro.
gastric emptying rate was similar between the in vitro and Additionally, texture changes were monitored to quantify
in vivo studies, especially at later digestion times. For exam- the white rice breakdown during gastric digestion in the
ple, after 3 h gastric digestion, the in vivo and in vitro sys- proximal and distal stomach regions in both the in vivo
tems both had 64% dry matter remaining. The intragastric and in vitro models. Differences between the proximal and
pH distribution between the HGS and the in vivo study distal stomach regions can be observed from both in vivo
also showed similarities at certain locations. pH measure- and in vitro models, where rice grains from the proximal
ments were taken at ten intragastric locations (Bornhorst region have greater hardness compared to the distal region.
et al., 2014), and values from the location closest to the This trend is observed at all digestion time points (20, 60,
pylorus (or HGS emptying tube) as well as values from the 120, and 180 min). At the shorter time points, the in vivo
top of the fundus (or top of HGS gastric vessel) were com- and in vitro systems have similar hardness values within
pared over the 3 h gastric digestion period (Figure 4). The each region. For example, after 60 min digestion in the dis-
pH values varied significantly between location (e.g. pylorus tal region, the in vivo hardness was 26.2 § 1.9 N compared
vs. fundus), but the values were similar between the in vitro to 28.2 § 1.4 N in the HGS in vitro system. However, at
and in vivo systems. For example, after 60 min digestion, longer digestion times, the hardness was lower in the in
the pH in the fundus location was 6.9 § 0.1 in vivo and 7.0 vivo model compared to the in vitro system in both the
§ 0.1 in vitro. Although most values compared here were proximal and distal stomach regions. For example, after
similar between the in vitro and in vivo system, some differ- 180 min digestion in the proximal region, the hardness
ences were observed. After 180 min digestion, the pH in the from the in vivo model was 23.7 § 5.7 compared to 34.1 §
5.6 in the HGS in vitro system. These promising results
indicate that the HGS in vitro model has the capability of
producing a similar gastric emptying rate, similar pH values
at certain gastric locations, and similar trends in food
breakdown in a white rice meal. However, a more complete
validation, including additional measurements, longer
digestion times, and varying meal types is necessary for the
HGS in vitro model to be utilized in a wide variety of
applications.

Advantages and limitations of the system

The advantages of the HGS are that it can be used to study both
the physical and chemical breakdown of food and other materi-
als in the stomach with physiologically relevant parameters.
The gastric secretion rate, pH, and gastric emptying can be con-
Figure 3. Comparison of gastric emptying of dry matter from an in vivo study with trolled and varied as needed to represent different physiological
growing pigs (Bornhorst et al., 2013) and from the HGS in vitro system (Phinney,
2013). Values are averages (n D 3, in vitro; n D 6, in vivo) with error bars represent- conditions (e.g. elderly, people taking certain medications that
ing the standard deviation. The solid line represents a 1:1 correlation. alter gastric secretions, etc.). The system can be used with larger
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 7

meal volumes (up to 1 L), which may be important if greater Validation of the system towards in vivo data animal and/or
amounts of sample are needed for analysis (e.g. physical prop- human
erty analysis). Samples can be taken from the system to repre- ARCOL has been validated towards in vivo data in human, pig
sent gastric emptying (e.g. emptied through the sampling tube or calves regarding the composition of the colonic microbiota
as they would be emptied into the small intestine in vivo) or (main bacterial populations followed by qPCR or plating), its
detailed intragastric property analysis can be done after certain metabolic activity (production of major end products of fer-
digestion time points. The limitations of this system are that mentation, such as short chain fatty acids) and/or the composi-
the mixing and physical property changes of sample meals tion of the nutritive medium used to feed the fermentor
needs to undergo more complete validation with in vivo data. (Gerard-Champod et al., 2010; Thevenot et al., 2015). The rele-
An additional limitation may be the sample size. The typical vance of the ARCOL model for probiotic studies was also
sample (“meal”) size used in the HGS varies from 150 – 300 g. shown as the survival of probiotic yeasts and their influence on
For certain high-value compounds or ingredients, this large SCFA production obtained in vitro corroborate the available
sample size may be prohibitive or extremely costly. In addition, data in human adult volunteers (Blanquet-Diot et al., 2012;
the HGS does not account for the oral or small intestinal phases Cordonnier et al., 2015; Thevenot et al., 2015).
of digestion, although it may be coupled with other static or
dynamic digestion model systems. Advantages and limitations of the system
Among available colonic in vitro models, ARCOL is one of the
The artificial colon: ARCOL few wireless systems that allows the maintenance of anaerobic
conditions by the unique activity of intestinal microbiota and
Origins of the system which is equipped with dialysis fibers in order to mimic passive
ARCOL (Artificial colon) is a one-stage fermentation model absorption of microbial products. The effect of single or
that reproduces the colonic environment of humans or animals. repeated administration of compounds of interest (food com-
This model has been developed by the University of Clermont ponents, probiotic, prebiotic, drug, xenobiotics…) on intestinal
Auvergne (Clermont-Ferrand, France). It’s the first one that microbiota composition and activity can be evaluated in the
has allowed the maintaining of anaerobiosis inside the fermen- ARCOL model. In its current configuration, ARCOL reprodu-
tor by the sole metabolic activity of the microbiota and not by ces the conditions that can be found in average in the human
flushing with N2 or CO2, as usually done in other colonic in or animal colon but does not simulate the different biotic and
vitro models. Up to date, ARCOL has been used to reproduce abiotic conditions (e.g. pH, retention time, availability of sub-
the colon of humans (Blanquet-Diot et al., 2012; Cordonnier strates, microbiota) associated with the three parts of human or
et al., 2015; Thevenot et al., 2015; Thevenot et al., 2013), pre- pig colon. ARCOL currently undergoes optimization by inclu-
ruminant calves (Gerard-Champod et al., 2010) pigs and sion of a mucosal contact surface, comparable to M-SHIME
piglets. technology (see below), to distinguish the mucosal and luminal
microenvironments in the colon.
Short description of the system
ARCOL integrates the main parameters of in vivo fermentation
in the large intestine, such as pH, temperature, anaerobiosis, Multicompartmental systems
supply of simulated ileal effluents, colonic residence time, pres- DIDGIÒ
ence of a complex, high-density, metabolically-active micro-
biota and passive absorption of water and microbial Origins of the system
metabolites. The DIDGIÒ system was built up at INRA in order to monitor
ARCOL is a bioreactor equipped with various ports and the disintegration and the kinetics of hydrolysis of the food
probes that is used in semi-continuous conditions. The fermen- occurring during a simulated digestion. It focuses on the upper
tor is inoculated with fresh feces from healthy volunteers or parts of the digestive tract, i.e. the stomach and the small intes-
animals, after suspension into phosphate buffer and filtration tine. To be physiologically realistic, the computer-controlled
through a double layer of gauze. A culture medium, reproduc- system reproduces the gastric and intestinal transit times, the
ing the composition of ileal effluents and containing various kinetics of gastric and intestinal pH, the sequential addition of
carbohydrate, protein, lipid, mineral and vitamin sources, is digestive secretions and the stirring of the stomach and small
sequentially introduced into the bioreactor, while fermentation intestine contents.
medium is sequentially withdrawn from the bioreactor. During
fermentation, the fermentation medium and the atmospheric Short description of the system
phase are continuously stirred. The pH and temperature are The DIDGIÒ system consists of two consecutive compartments
kept at a constant value by adding NaOH and heating with a simulating the stomach and the small intestine. Each compart-
water double-jacket. After initial sparging with O2-free N2 gas, ment is surrounded by a glass jacket filled with water pumped
the fermentative process allows the maintenance of anaerobic using a temperature-controlled water bath. The system is
conditions in the bioreactor. This favors the maintenance of equipped with temperature, pH and redox sensors and variable
hydrogenotrophic microorganisms inside the bioreactor. A speed pumps to control the flow of meal, HCl, NaHCO3, bile,
dialysis system using hollow fiber membranes (cut-off 30 kDa) enzymes and the emptying of each compartment. Flow rates
maintains the appropriate electrolyte and metabolite concen- are regulated by specific computer-controlled peristaltic pumps.
trations and the operating volume. Anaerobic conditions can be simulated by purging air with
8 D. DUPONT ET AL.

nitrogen. A Teflon membrane with 2 mm holes is placed before

the transfer pump between the gastric and the intestinal com-
partment to mimic the sieving effect of the pylorus in human,
as described previously (Kong and Singh, 2008). The computer
program was designed to accept parameters and data obtained
from in vivo studies in animals or human volunteers, such as
the quantity and duration of a meal, the pH curve for the stom-
ach, the secretion rates into the different compartments and
the gastric and small intestine emptying rates. The system is
controlled by software named StoRMÒ for Stomach regulation
and monitoring (Guillemin et al., 2010). To control the transit
time of the chyme in each compartment, a power exponential
equation for gastric and intestinal delivery is used

f D 2-(t) t / 1/2 b

where f represents the fraction of the chyme remaining in the

stomach, t is the time of delivery, t1/2 is the half time of delivery
and b is the coefficient describing the shape of the curve, as
described previously (Elashoff et al., 1982).

Validation of the system towards in vivo animal data

Example – Digestion of infant formula
The DIDGIÒ system is a very recent one. Although several
matrices (dairy, meat, fruits and vegetables, emulsions) have
been submitted to digestion using the DIDGIÒ system, only
data obtained on the digestion of infant formulas (M
enard
et al. 2014), cheese (Adouard et al., 2016), bovine skim milk Figure 5. Comparative residual concentration of total casein (a) and b-lactoglobu-
(Sanchez-Rivera et al., 2015) and human milk (de Oliveira lin (b) determined by ELISA after in vitro (white) and in vivo (black) digestion.
et al., 2016a; de Oliveira et al., 2016b; Deglaire et al., 2016) have
been published so far. In order to demonstrate that this system immunoreactive b-lactoglobulin in the small intestine was sig-
was physiologically-relevant, a comparison of the in vitro and nificantly higher in vitro than in vivo. The correlation coeffi-
in vivo digestion of an infant formula was performed. The in cient, between in vitro and in vivo ELISA determination for
vivo trial was conducted on 18 piglets that were fed the infant caseins and b-lactoglobulin was 0.987 (p < 0.001), proving a
formula for which the concentration in lipids and proteins was good agreement between in vitro and in vivo proteolysis during
increased compared to a standard one, but the ratio lipids/pro- digestion.
teins was kept constant. In parallel, in vitro gastro-intestinal
digestion was performed on this enriched infant formula using Advantages and limitations of the system
the newly developed system and the extent of milk proteolysis The main advantage of this system is that, since it is basic, it is
was monitored and compared to the one obtained in vivo. All quite robust and can handle real foods and full meals up to
the details regarding the experimental conditions used for this 200 g. The compartments are transparent allowing to see in
validation have been described previously (Menard et al., 2014). real time the evolution of the food structure during digestion.
Volumes of the stomach content observed in vitro with the Biochemical processes (enzymes and digestive fluids) and phys-
dynamic digestion system were compared to the ones observed ical (gastric and intestinal emptying) are well mimicked. In
in vivo in piglets. No significant differences were observed 30, contrast, the mixing in the compartments consists only in basic
90 and 210 min after ingestion confirming that the parameters stirring using a propeller and the peristalsis occurring in the
chosen for mimicking the gastric transit of infant formula in stomach leading to the mixing of the food is not realistic in this
vitro were physiologically relevant. Evolution of caseins and system. So far, another limit, is the absence of nutrients absorp-
b-lactoglobulin throughout in vitro and in vivo digestion, as tion in the small intestine due to the lack of dialysis membranes
determined by ELISA, was compared. Results showed that the in the intestinal compartment. This limitation is currently
kinetics of hydrolysis of both proteins during in vitro and in being overcome by the development of a new version of the
vivo digestion were similar. The proportion of immunoreactive DIDGIÒ .
caseins appeared not to be significantly different between both
experiments for samples collected in the stomach as well as in
the small intestine after 30, 90 and 210 min of digestion TIM
(Figure 5a). Similarly, the percentage of immunoreactive b-lac-
Origins of the TIM systems
toglobulin showed no significant differences for samples col-
lected in vivo and in vitro in the stomach after 30, 90 and In 1992 the authors initiated the development of in vitro gastro-
210 min (Figure 5b). However, the percentage of intestinal (GI) models at TNO. Realizing the limitations of
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 9

static models, from the start we focussed on dynamic systems.

Extensive literature data about anatomy and physiology of the
GI tract were ‘translated’ to the TIM technology. The gastric
and small-intestinal model (TIM-1) was described in details in
1995 (Minekus et al., 1995), and patented in the EU and USA.
After that, the large-intestinal model (TIM-2) was developed
(Minekus et al., 1999). Even today it is a continuous process of
optimization, such as simulation of infant GI conditions (Have-
naar et al., 2013a; Roussel et al., 2016) or elderly (Denis et al.,
2016) and development of the advanced gastric model ‘TIMagc’
(Bellmann et al., 2016). Over the years a broad variety of valida-
tion studies in nutrition research (section below) and pharma-
Figure 6. Cumulative gastric and ileal delivery of a meal expressed as a percentage
ceutical research (not part of this review) have been published. of total meal intake: in vivo (human n D 7) gastric () and ileal (O) delivery of
Although the focus is on humans, the GI conditions and yoghurt and gastric (■) and ileal (&) delivery of blue dextran in the TIM-1 system
colon microbiota of pigs (Avantaggiato et al., 2007; Martinez simulating the gastrointestinal transit of yoghurt.
et al., 2013) and dogs (Smeets-Peeters et al., 1999) can also be
simulated in TIM. been published relating studies performed on all kinds of foods
and micro/macronutrients. In the next paragraph, only the
papers showing a comparison between in vivo and TIM data
Short description of the TIM systems
will be presented.
The TIM-1 system comprises compartments for the stomach, duo- In a pioneer work, the gastric and ileal deliveries of the TIM
denum, jejunum and ileum, connected by peristaltic valves and model were shown to simulate accurately the pre-set curves for
linked with semi-permeable membrane units. In these compart- slow and fast deliveries of chime calculated from in vivo data
ments the successive dynamic conditions in the upper GI tract are obtained from studies with human volunteers (Figure 6).
simulated. Also a tiny-TIM system is available, comprising one
compartment for the small intestine (Verwei et al., 2016). TIMagc
Macronutrients
simulates the specific conditions in the corpus and antrum part of
The digestion and fermentation of carbohydrates and dietary
the stomach, including peristaltic motility and pressure forces
fibres in TIM-1 and TIM-2 (Venema et al., 2003; Venema et al.,
(Bellmann et al., 2016). The TIM-2 system simulates the dynamic
2005), respectively, showed reliable results for the human situa-
conditions in the colon with a high density of metabolic active
tion. The human glycaemic response curve after carbohydrate
microbiota of human origin (Aguirre et al., 2015).
intake can be predicted by combining TIM digestion studies
The settings in the computer software accurately and reproduc-
with in silico modelling of the insulin response (Figure 7) (Bell-
ibly control the TIM system, e.g. for temperature, peristaltic mix-
mann et al., 2010).
ing, transit times, pH curves, and secretion of GI fluids (e.g.
The digestion of proteins and bioaccessibility of amino acids
salivary and gastric juice, bile, pancreatic juice prepared according
in TIM was compared with in vivo data (Schaafsma, 2005),
to SOPs). The settings and composition of secretion fluids can be
showing a high predictive quality. Therefore, the TIM system is
adapted related to the type of drink and food, age, health status,
a suitable in vitro tool to determine the true ileal protein digest-
and drug use. It can vary from rapid gastric emptying with low
ibility and amino acid bioaccessibility (Havenaar et al., 2016),
secretion after intake of water, up to slow gastric emptying with
e.g. to determine protein quality according to DIAAS.
high initial gastric pH and high secretion of digestive fluids after
intake of a high fat meal. The average dynamic GI conditions as
well as the biological day-to-day and inter-individual variation can
be simulated based on available physiological data. For example,
the GI conditions of neonates, infants and toddlers in tiny-TIM
and the consequences it has for oral drugs and digestion vs. adult
conditions have been documented (Havenaar et al., 2013b).
Related to the research question, the TIM-2 systems can be
inoculated with pooled or individual faecal samples (Aguirre
et al., 2015; Aguirre et al., 2014b), from healthy volunteers, e.g.
on different diets (Tabernero et al., 2011), from obese persons
(Aguirre et al., 2014a), or patients with GI disorders (Rose
et al., 2010). Phylogenetic analysis showed that the microbial
density and composition in TIM-2 was rather similar to the
human faecal microbiota (Kovatcheva-Datchary et al., 2009).

Figure 7. Prediction of glycaemic response in humans based on the digestion of

Validation and application in food and nutrition research carbohydrates and bioaccessibility of glucose and fructose in tiny-TIM in combina-
tion with in silico modelling of the insulin response: correlation (r D 0.94) between
TIM is a pioneer in vitro digestion system and has been widely predicted blood glucose Cmax and measured blood Cmax in humans for 22 differ-
used during the last 2 decades. More than 100 papers have ent carbohydrate products.
10 D. DUPONT ET AL.

composition of secretion fluids the experiments are highly

reproducible, (vii) by changing only a single parameter (e.g.
gastric emptying or pH), the effect of this parameter on the fate
of the test compound can be tested to study more extreme con-
ditions in a population or mechanistic aspects.
Limitations of the TIM systems are that (i) there is no feed-
back on energy density of the food on the GI conditions; these
parameters should be set in advance in the TIM-software; (ii)
there is no intestinal mucosa, therefore absorption should be
studied in combination with intestinal cell lines (D
eat et al.,
2009; Haraldsson et al., 2005) or tissues (Westerhout et al.,
2014); (iii) in TIM the availability for absorption (bioaccessibil-
Figure 8. Measurement of the concentration of the sulfasalazine in vivo (stomach, ity) is measured and not the bioavailability including metabo-
small intestine and end of the colon) and related 5-ASA production in vitro. Data
on the concentration of sulfasalazine in the ascending colon in vivo are not avail- lism and excretion; this can be overcome by combining TIM
able and no pro-drug was detected in the fecal samples. Adapted from Molly et al. with in silico modelling (Naylor et al., 2006; Verwei et al., 2006).
1994 (Molly, et al., 1994). In conclusion, the TIM system is a broadly validated, time-
and cost-efficient, reliable in vitro tool to study the digestibility
Micronutrients of foods, the bioaccessibility of nutrients, and the fate and effi-
Human plasma concentrations after long-term intake of folate cacy of functional ingredients under simulated dynamic human
was accurately predicted using TIM in combination with in sil- adult and infant GI conditions.
ico modelling (Verwei et al., 2006).
The stability and bioaccessibility of fat-soluble vitamins such
Simulator or the Human Intestinal Microbial Ecosystem
as lycopenes and tocopherol (D
eat et al., 2009) and fat-soluble
(SHIMEÒ )
phytochemicals (Ribnicky et al., 2014) were studied in TIM
showing food matrix and food preparation effects consistent Origins of the system
with in vivo data. The reactor setup was adapted from the original SHIMEÒ
Different aspects of minerals and metals in TIM showed model developed at Ghent University (Belgium), representing
good correlation with human data such as the bioaccessibility the gastrointestinal tract (GIT) of the adult human, as described
of iron from various food products (Larsson et al., 1997) or the by Molly et al. (Molly et al., 1993). During the years the system
risks for young children of unintended lead intake via polluted has been improved and nowadays, it is a computer-controlled
soil (Van de Wiele et al., 2007). device that can be used to simulate the gastrointestinal micro-
bial ecology and physiology of healthy humans, babies, elder-
Functional foods lies, some specific disease conditions (e.g. IBD, pathogen
Studies with functional foods vary from probiotics and prebiot- infection) and also pigs, dogs and cats (ProDigest, Belgium).
ics to anti-oxidants. The survival of probiotic strains during
transit through TIM-1 was first validated in 1997 by Marteau Short description of the system
et al. (Marteau et al., 1997) and during the years, many different The SHIMEÒ consists of a succession of five reactors simulat-
bacterial or yeast strains were tested (Blanquet-Diot et al., 2012; ing the different parts of the gastrointestinal tract. The first two
Cordonnier et al., 2015). Examples of anti-oxidants studies in reactors are of the fill-and-draw principle to simulate different
TIM are about the bioconversion of phenolic acids (Gao et al., steps in food uptake and digestion, with peristaltic pumps add-
2006) and fermentation of cereal fibre fractions by the colon ing a defined amount of SHIME nutritional medium (3x/day)
microbiota (Anson et al., 2011b). The anti-inflammatory capac- and pepsin to the stomach and pancreatic enzymes with bile
ity measured in TIM samples using a macrophage assay (Anson liquid in the small intestine. A specific software allows the sub-
et al., 2010) was confirmed in an ex-vivo human study (Anson sequent simulation of the physiological conditions occurring in
et al., 2011a). the duodenum, jejunum and ileum. The last three compart-
ments are continuously stirred reactors with constant volume
and pH control. Retention time and pH of the different vessels
Advantages and limitations of the TIM system
are chosen in order to resemble in vivo conditions in the differ-
Advantages of the TIM system are that (i) they simulate accu- ent parts of the gastrointestinal tract. Upon inoculation with
rately the dynamic physiological GI conditions; (ii) they can fecal microbiota, these reactors simulate the ascending, trans-
handle specific food ingredients and drugs as well as complete verse and descending colon. Upon stabilization of the microbial
meals; (iii) they can simulate average GI conditions, biological community in the different regions of the colon, a representa-
variation, and disease conditions for different age groups; (iv) tive microbial community is established in the three colon com-
therefore, they can be used for a broad scope of applications in partments, which differs both in composition and functionality
the food and pharma research and are not limited to a specific in the different colon regions. Inoculum preparation, retention
application; (v) samples can be collected from the compart- time, pH, temperature settings and reactor feed composition
ments during transit of the chyme for analysis, which results in were previously described by Possemiers et al. (Possemiers
detailed information about the fate of test products in the GI et al., 2004). In order to investigate different compounds at the
tract in time; (vi) due to the strict control over all settings and same time, a TWINSHIMEÒ setup was developed by operating
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 11

two systems in parallel at the same time. This makes the model healthy individuals and IBD patients (Vermeiren et al.,
an ideal system for direct comparison of two products or to 2012; Vigsnaes et al., 2013).
perform placebo-controlled studies. More recently a Triple-  Multiple case studies have also demonstrated that specific
SHIME and a QuadSHIME model have been introduced to enzymatic conversions can be accurately simulated. As an
compare 3 or 4 conditions, respectively. example, Possemiers et al. (Possemiers et al., 2006) eluci-
The most recent developments in relation to the SHIME dated the mechanism of the intestinal activation of
technology consist in the automation of the process control phyto-estrogens and showed that a high inter-individual
(i.e. liquid transfer, pH, flushing), data acquisition and the variability exists in the capacity of the intestinal bacteria
development of an additional absorption unit that can be used to perform this activation. Selection of specific metabolic
to simulate the small intestinal absorption processes. This unit phenotype in vivo and use of a fecal sample from that
is connected directly in line with the main operation unit and donor, resulted in the establishment of a SHIME with the
operated with the same software. Using the so-called M- same metabolic phenotype ( D SHIME allows to maintain
SHIMEÒ it is possible to mimic the mucosal microbial coloni- in vivo functionality). Animal (Possemiers et al., 2008)
zation by incorporation of mucin-covered microcosms there- and human trials (Bolca et al., 2007) confirmed these in
fore maintaining in vitro unique features of an individual’s vitro data.
microbiome in terms of its mucosal composition (Van den  Sulfasalazine is a pro-drug historically used for the treat-
Abbeele et al., 2013a). Systems have been developed to simulate ment of inflammatory diseases in the gut. Sulfasalazine is
the specific physiological conditions occurring in babies and partially absorbed in the small intestine (approx. 30%).
elderly, as well as pig, dog and cat. Moreover, by combining the The residual part enters into the colon, where it is reduced
SHIMEÒ with the so-called HMITM module (Marzorati et al., by the metabolic activity of the gut microbiota to to sulfa-
2014), it is possible to simulate online the host-microbiota pyridine and 5-ASA. The pro-drug behaved similarly in
interaction occurring at the level of the gut wall (i.e. biofilm for- vivo and in the SHIME (Molly et al., 1994) (Figure 7).
mation under a shear stress and concomitant presence of enter-  A high similarity between in vitro and in vivo data was
ocytes to evaluate the impact of a treatment in terms of gut wall also found for the metabolism of prebiotics. When intro-
modulation). ducing the same human fecal sample in germfree rats
Last but not least, specific protocols have been developed to (Van den Abbeele et al., 2011) and in the SHIME model
simulate diseased conditions: inflammatory bowel disease, (Van den Abbeele et al., 2013b), similar fermentation pro-
treatment with antibiotics, infection with Clostridium difficile files by specific microbial groups were found to be
(PathoGutTM model). enhanced by specific prebiotics (i.e. arabinoxylans and
inulin). Another study with inulin (Van de Wiele et al.,
Validation of the system towards in vivo data animal and/or 2004) confirmed that the administration of inulin to the
human SHIME model led to a 2-times increase of butyrate and
The SHIME model was initially developed in 1993 and was val- propionate production by the microbiota and induced
idated based on a comparison with in vivo human data regard- specific quantitative (1 log unit) and qualitative changes
ing indicator bacterial groups, short-chain fatty acid in the bifidobacterial community. The effects of inulin
production (SCFA), enzymatic activities, headspace gases and administration in a clinical validation study confirmed
microbiota-associated characteristics (MACs) (Molly et al., the predictive power and scientific quality of the SHIME
1994). Over the years, a large number of experiments (i.e more with highly similar effects on bifidobacteria and butyrate
than 100 papers) has been performed in which SHIME results production.
were compared with in vivo animal and human experiments. In the probiotic field, a typical example of validation of
Below, we summarize some key findings. SHIME results is a study related to cholesterol-lowering activity
 The application of a high-resolution phylogenetic micro- of Lactobacillus reuteri. Using the SHIME model, it was shown
array (i.e. HITChip) pointed out that a wide range of that this probiotic strain exerted a high specific bile salt hydro-
intestinal microbes of in vivo human samples can be lase activity, which alters bile salt circulation in the intestine
maintained in the SHIME model and are colon region- and the body. This altered bile salt metabolism may then lead
specific, similar to in vivo data (Van den Abbeele et al., to a cholesterol-lowering effect. Validation of the effect of the
2010). One critical remark of this study was that the shift probiotic on cholesterol levels in pigs, showed a significant
from an in vivo to an in vitro environment resulted in an decrease of total and LDL cholesterol (De Smet et al., 1998).
increased Bacteroidetes/Firmicutes ratio as also occurs in
other in vitro models (Rajilic-Stojanovic et al., 2010). In Advantages and limitations of the system
this respect, Van den Abbeele et al. (Van den Abbeele The advantages correlated with the use of a SHIME technology
et al., 2012) introduced a simulated intestinal surface in platform for experimental purposes can be listed as follows: i)
the SHIME (M-SHIMEÒ ). As a result, in contrast to con- presence of two to four full GIT in the same system (i.e. TWIN-
ventional models, washout of relevant mucin-adhered SHIME to QuadSHIME) to study the mechanism of action of
microbes was avoided. This resulted in the fact that products and ingredients; ii) possibility to work with volumes
unique inter-individual differences among human sub- close to the in vivo ones; iii) possibility to culture the intestinal
jects are preserved in this in vitro model (Van den microbiota in the different colonic compartments for periods
Abbeele et al., 2013a). Since then, the M-SHIME has also up to several months. This allows studies based on repeated
been applied to e.g. investigate the differences between daily dosing strategy to evaluate the adaptation of the activity
12 D. DUPONT ET AL.

and composition of the microbiota to a specific treatment; iv) degraded. Two openings, each connected to a peristaltic pump
the M-SHIME allows to accurately mimic the mucosal micro- allow the differential gastric emptying of “liquids” and “solids”,
bial colonization. Due to its close proximity to host epithelial respectively. These two pumps are programmed to follow spe-
cells, the mucosal microbiome is thought to have an intrinsi- cific profiles as observed in human: the “liquids” emptying fol-
cally higher potency to modulate gut health, and by extension, lows an exponential “Elashoff” curve (Elashoff et al., 1982)
human health; v) the modular setup, which characterizes the without a lag phase period, while “solids” emptying fulfills a lin-
SHIME, makes possible to explore the inter-individual variabil- ear law after a 30 min lag phase (Siegel et al., 1988).
ity in microbiome behavior upon specific treatments; vi) the
system can be easily adapted to simulate other monogastric ani- Validation of the system towards in vivo data animal and/or
mals (i.e. dog, cat, pig, etc…) or diseased conditions; vii) finally, human
an important read-out from SHIME experiments consists of the The model has been validated for pharmaceutical applications
evaluation of host-microbe interactions. Colon suspension can against in vivo data in human (Guerra et al., 2016). Two model
be brought in direct contact with host epithelial cells. This drugs were studied: an immediate release form of paracetamol
allows assessing to what extent changes in microbiome compo- and a sustained release form of theophylline. Both in vitro and
sition, microbial metabolites, signaling molecules or antigens in vivo, the drugs were ingested with a glass of water. In ESIN,
have differential effects at the level of the host in terms of gut the amount of absorbed paracetamol and theophylline was
barrier permeability and parameters related to inflammation. measured in the dialysis samples while in human, saliva (para-
As any other in vitro simulator, the SHIME suffers of the cetamol) or blood samples (theophylline) were collected (Souli-
absence of a physiological environment. Moreover, water and man et al., 2007; Souliman et al., 2006). Paracetamol and
metabolites absorption are not routinely simulated in the theophylline tablets showed similar absorption profiles in ESIN
colonic compartment. Movement and mixing along the GI and in healthy subjects (Figure 9). For theophylline, a level A in
tract is performed by pumping and stirring and not by vitro in vivo correlation (IVIVC) was established with a slope
peristalsis. of 1.097 and a correlation coefficient (r2) of 0.989, showing the
predictive value of the in vitro system. These results demon-
Engineered Stomach and small INtestinal – ESIN strate the high level of efficacy of ESIN in mimicking the behav-
ior of soluble drugs in the human gastrointestinal tract.
Origins of the system
The Engineered Stomach and small INtestinal -ESIN- system is
a new multi-compartmental dynamic in vitro model of the Advantages and limitations of the system
human stomach and small intestine (Guerra et al., 2012). This ESIN is the only multi-compartmental model including gastric
model has been developed by the University of Auvergne (Cler- and small intestinal compartments which has been designed to
mont-Ferrand, France) to overcome some limitations identified digest real-size food particles. This is particularly important in
in the current in vitro multi-compartmental gastrointestinal food applications, because mixing which is commonly done in
models, even in the most complete like TIM and SHIME. other in vitro systems, widely impacts nutrient release and bio-
Indeed, such models do not allow a close imitation of real food accessibility. ESIN can be potentially used in a wide range of
bolus entering the stomach, as they proceed with mixed food applications including nutritional, pharmaceutical, toxicologi-
rather than with food particles of a realistic size. They also do cal or microbiological applications. Nevertheless, as ESIN is a
not reproduce the differential gastric emptying of liquids and new model, it has been validated up to now only for pharma-
solids as observed during digestion in human. Then, ESIN ceutical applications during liquid digestion. Additional valida-
presents an original architecture, especially for the gastric com- tion experiments are necessary to validate the model during
partment that has been patented (Alric and Denis, 2009). digestion of solid foods and for nutritional or microbiological

Short description of the system 120

ESIN is composed of six successive compartments: a meal res-
ervoir allowing a progressive introduction of food particles 100
with a realistic size into the gastric compartment, a salivary
Amount of paracetamol

80
ampoule dedicated to a progressive mixing of food with saliva,
absorbed (%)

the stomach and the three parts of the small intestine, the duo- 60
denum, jejunum and ileum. This model reproduces the main
parameters of human digestion: body temperature, temporal 40
and longitudinal changes in pH, salivary, gastric, pancreatic
20
and biliary secretions, transit times, chyme mixing and passive
absorption of digestion products. 0
The most striking innovation of ESIN is the architecture of 0 30 60 90 120 150 180 210 240
its gastric compartment that enables to reproduce the biphasic Time (min)
nature of gastric emptying observed in vivo. An indentation
inside the gastric chamber allows the passage of small size par- Figure 9. Paracetamol absorption in ESIN and in healthy human volunteers.
Results are expressed as mean cumulative percentages § standard deviations (n
ticles (< 2 mm) and liquids in a second chamber. Large size D 3 in vitro and n D 8 in vivo.  In vitro percentages statistically different from in
particles (> 2 mm) stay in the main chamber to be further vivo ones (P < 0.05)).
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 13

applications. In its current state, the model doesn’t include resi- analysis. Comparison of the resistant protein sequences with
dent microbiota, but the small intestinal compartments of the those reported in duodenal effluents from mini-pigs fed milk
model has been designed to allow inoculation with human (Barbe et al., 2014), that correspond to the end of the gastric
intestinal microbiota (or a consortium of bacteria from human digestion (Barbe, et al., 2014), showed a remarkably close pat-
intestinal microbiota) and their maintaining under anaerobic tern. From the identified sequences in the dynamic model, 73%
conditions by flushing with nitrogen. were common with those reported in the porcine in vivo study.
The flexible-modulating characteristics of the system and
the computer-control of physiological parameters open pos-
SIMulator of the Gastro-Intestinal tract: simgiÒ sibilities for variation of conditions that would allow the
Origin of the system simulation in the simgiÒ as model of microbial dysbiosis
associated to pathological conditions or due to unbalanced
The simgiÒ (SIMulator of the GastroIntestinal tract) has been diets. Using this model, short fatty acids (SCFA) and
developed at the Institute of Food Science Research CIAL ammonium formation under high energy diet (during
(CSIC-UAM, Madrid, Spain). It is a computer-controlled gas- microbiota stabilization period) followed by a low energy
trointestinal in vitro model designed to simulate the physiologi- diet (during dietary intervention) have been compared. Shift
cal processes taking place during digestion in the stomach and from high to a low energy diet resulted in a two-fold
small intestine, as well as to reproduce the colonic microbiota decrease in the average content of total SCFA of the three
responsible for metabolic bioconversions in the large intestine. colon compartments. Besides, a two-fold increase in the
ammonium content in the distal colon compartments (TC
Short description of the system and DC) and a remarkable six fold increase in the proximal
colon compartment (AC) were accounted when changing
The simgiÒ comprises five interconnected compartments that from high to low energy diet (Barroso et al., 2015a). The
simulate the stomach, small intestine and three stages of the SCFA and ammonium results were contrasted with in vivo
large intestine that can operate jointly or independently. The data from obese subjects where a significant decrease of
gastric compartment consists of two cylindrical transparent SCFA and increase of proteolytic products were observed
and rigid methacrylate plastic modules covering a reservoir of when the individuals consumed high protein diets reduced
flexible silicone walls where the gastric content is mixed by in total carbohydrates (Russell et al., 2011).
peristaltic movements. The peristalsis is achieved by changing The system allows the development of a stable and colon
the pressure of water that flows in the jacket between the plastic region specific microbial ecosystem that has been shown
modules and the reservoir. The stomach compartment has dif- representative of the in vivo situation in terms of microbial
ferent ports for input of experimental food components, gastric composition and activity (Barroso et al., 2015b). The evalu-
juice, and acid. ation of the polyphenol metabolic activity of the colonic
The small intestine consists in a double jacket glass reactor microbiota of two volunteers using the simgiÒ has demon-
vessel stirred that receives the gastric content and mixes it with strated that moderate red wine consumption produces a sig-
pancreatic juice and bile. The stages of the large intestine are nificant increase in 3,5-dihydroxybenzoic acid, 3-O-
simulated in three double jacket stirred glass reactors. The pH in methylgallic acid, vanillic acid, protocatechuic acid and
the colonic units named ascending (AC), transverse (TC) and syringic acid (Cueva et al., 2015). This rise was consistent
descending (DC) is controlled by addition of NaOH and HCl. with previous data obtained in human feces in an interven-
When the digested content of the small intestine is transferred tion study using the same wine (Munoz-Gonzalez et al.,
to the proximal colon compartment, the transit of colonic con- 2013). However, it has to be noted that the microbiota met-
tent between the AC, TC and DC compartments is simulta- abolic activity observed was individual-dependent.
neously initiated at the same flow rate. The intestinal and
colonic vessels contain ports for the transit of intestinal content,
sampling, continuous flushing of nitrogen allowing a permanent Advantages and limitations of the system
anaerobic atmosphere and control of pH and temperature. The advantage of the model is associated to its flexible
Flow rates, compartment volumes, pH, temperature and modulating characteristics that permit to work separately in
pressure are computer controlled through a programmable the different compartments or to perform a joint, sequential
logic panel (Unitronics Vision 120TM) and the system stores use, with controlled transit both under acute and chronic
the on-line monitored values such as volumes pumped, temper- administrations, using volumes close to the human ones.
ature, and pH during the whole experiment. Distinctive technical features of the stomach are the peri-
staltic mixing movements and controlled emptying. Appro-
priate design, including a transparent window in the central
Validation of the system towards in vivo data
teflon module, makes possible to take samples from the
Milk whey proteins have been used as model proteins to follow stomach content at any time and to observe the food
the gastric digestion outcome (Miralles et al., 2017). Progress of appearance during digestion.
protein degradation was followed by SDS-PAGE and band inte- Although the simgiÒ model has not yet incorporated
gration. Intact protein decline agreed with data reported in devices simulating the gut microbiota-host interactions,
human subjects after whey proteins ingestion (Sullivan et al., assays for evaluating this type of crucial interaction is cur-
2014). This study incorporated a detailed peptide profile rently approached by co-culturing colon-region specific
14 D. DUPONT ET AL.

microbiota suspensions from the AC, TC and/or DC vessels could allow to run digestion experiments in parallel, allow-
with epithelial or immune cells. In this sense, the micro- ing to screen several compounds with high throughput.
biota stabilized in the simgiÒ has demonstrated to induce
the phenotypical maturation of human monocyte-derived
dendritic cells (Barroso et al., 2015a). In addition, the work- References
ing parameters can be adjusted to physiological conditions
Adouard, N., L. Magne, T. Cattenoz, H. Guillemin, B. Foligne, D. Picque,
or to dysbiosis states caused by pathological situations such and P. Bonnarme. 2016. Survival of cheese-ripening microorganisms in
as diabetes, obesity and other metabolic disorders. a dynamic simulator of the gastrointestinal tract. Food Microbiology
However, a limitation is the lack of devices to evaluate the 53:30–40.
simulation of intestinal absorption to prevent inhibition of the Aguirre, M., A. Eck, M. E. Koenen, P. H. M. Savelkoul, A. E. Budding, and
colon microbiota by end products of digestion. On the other K. Venema. 2015. Evaluation of an optimal preparation of human stan-
dardized fecal inocula for in vitro fermentation studies. Journal of
hand, the food degradation read out still needs to be validated Microbiological Methods 117:78–84.
towards in vivo data. Aguirre, M., D. Jonkers, F. J. Troost, G. Roeselers, and K. Venema. 2014a.
In Vitro Characterization of the Impact of Different Substrates on
Metabolite Production, Energy Extraction and Composition of Gut
Conclusion and perspectives Microbiota from Lean and Obese Subjects. Plos One 9.
Aguirre, M., J. Ramiro-Garcia, M. E. Koenen, and K. Venema. 2014b. To
The present paper reviews some of the main in vitro pool or not to pool? Impact of the use of individual and pooled fecal
dynamic digestion systems currently available. It has how- samples for in vitro fermentation studies. Journal of Microbiological
ever to be emphasized that all the systems presented in this Methods 107:1–7.
review are not at the same stage of development. Indeed, Alric, M., and S. Denis. 2009. Dispositif de simulation d’un estomac d’un
systems like the TIM and the SHIME models have been mammifere monogastrique ou d’un ^etre humain. Patent
n
W02009087314,
developed more than 20 years ago and have been regularly Anson, N. M., A. M. Aura, E. Selinheimo, I. Mattila, K. Poutanen, R. van
improved during all these years. Other systems such as the den Berg, R. Havenaar, A. Bast, and G. Haenen. 2011a. Bioprocessing
ESIN, simgiÒ or DIDGIÒ have been developed more of Wheat Bran in Whole Wheat Bread Increases the Bioavailability of
recently. Nevertheless, from these examples, it is clear that Phenolic Acids in Men and Exerts Antiinflammatory Effects ex Vivo.
dynamic in vitro digestion systems, when programmed with Journal of Nutrition 141:137–143.
Anson, N. M., R. Havenaar, A. Bast, and G. Haenen. 2010. Antioxidant and
physiologically-relevant parameters, can mimic the com- anti-inflammatory capacity of bioaccessible compounds from wheat
plexity of the digestive process. However, one can wonder fractions after gastrointestinal digestion. Journal of Cereal Science
whether when a system is validated for the digestion of a 51:110–114.
certain food it is relevant for other types of foods and it Anson, N. M., R. Havenaar, W. Vaes, L. Coulier, K. Venema, E. Selin-
might be useful to validate those systems for, at least, fami- heimo, A. Bast, and G. Haenen. 2011b. Effect of bioprocessing of wheat
bran in wholemeal wheat breads on the colonic SCFA production in
lies of foods with similar rheological properties (liquids, sol- vitro and postprandial plasma concentrations in men. Food Chemistry
ids, gels, foams…). Food structure is not always taken into 128:404–409.
account in this system and food needs sometime to be sub- Avantaggiato, G., R. Havenaar, and A. Visconti. 2007. Assessment of the
mitted to drastic physical dispersion (ultra-turrax, blender multi-mycotoxin-binding efficacy of a carbon/aluminosilicate-based
etc) before being submitted to digestion in order to avoid product in an in vitro gastrointestinal model. Journal of Agricultural
and Food Chemistry 55:4810–4819.
blockage of the system tubes. Connection with a mastication Barbe, F., S. Le Feunteun, D. Remond, O. Menard, J. Jardin, G. Henry, B.
simulator could be an added value for the digestion of solid Laroche, and D. Dupont. 2014. Tracking the in vivo release of bioactive
foods. Other improvements could be envisaged to make peptides in the gut during digestion:Mass spectrometry peptidomic
these systems even more relevant. Absorption is over sim- characterization of effluents collected in the gut of dairy matrix fed
plified but coupling of the dynamic digestion systems with mini-pigs. Food Research International 63:147–156.
Barroso, E., C. Cueva, C. Pelaez, M. C. Martinez-Cuesta, and T. Requena.
cellular models (Caco-2, HT-29, IPEC-J2 or co-culture of 2015a. The computer-controlled multicompartmental dynamic model
Caco-2 and HT-29MTX) could allow to better simulate the of the gastrointestinal system (SIMGI). 319–327.
epithelial transport (D
eat et al., 2009). The absence of Barroso, E., C. Cueva, C. Pelaez, M. C. Martinez-Cuesta, and T. Requena.
microbiota in the distal parts of the small intestine can 2015b. Development of human colonic microbiota in the computer-
appear as a limit. In the future, dynamic digestion systems controlled dynamic SIMulator of the GastroIntestinal tract SIMGI.
LWT-Food Science and Technology 61:283–289.
will probably become compulsory for understanding the Bellmann, S., J. Lelieveld, T. Gorissen, M. Minekus, and R. Havenaar. 2016.
mechanisms of food digestion, especially because of the Development of an advanced in vitro model of the stomach and its
increased ethical and economic constraints of in vivo trials. evaluation versus human gastric physiology. Food Research Interna-
They will also become key players in the field of drug deliv- tional. 88:191–198.
ery which will also require microsystems able to investigate Bellmann, S., M. Minekus, E. Zeijdner, M. Verwei, P. Sanders, W. Basten,
and R. Havenaar. 2010. TIM-Carbo: a rapid, cost-efficient and reliable
the release of expensive pore molecules in small volumes. in vitro method for glycemic response after carbohydrate ingestion.
Some microfluidic devices have already been developed to Wageningen: Wageningen Acad Publ.
perform protein digestion before identification by mass Blanquet-Diot, S., S. Denis, S. Chalancon, F. Chaira, J.-M. Cardot, and M.
spectrometry (Jansson et al., 2012; Kecskemeti and Gaspar, Alric. 2012. Use of Artificial Digestive Systems to Investigate the Bio-
2017) and the devices developed could be interesting start- pharmaceutical Factors Influencing the Survival of Probiotic Yeast
During Gastrointestinal Transit in Humans. Pharmaceutical Research
ing points for the development of new “microdigestors”. 29:1444–1453.
Interestingly, microfluidic has been used to study the diges- Bolca, S., S. Possemiers, V. Maervoet, I. Huybrechts, A. Heyerick, S. Ver-
tion of one lipid droplet (Marze et al., 2014). Microsystems varcke, H. Depypere, D. De Keukeleire, M. Bracke, S. De Henauw, W.
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 15

Verstraete, and T. Van de Wiele. 2007. Microbial and dietary factors Development and validation of a new dynamic computer-controlled
associated with the 8-prenylnaringenin producer phenotype: a dietary model of the human stomach and small intestine. Biotechnology and
intervention trial with fifty healthy post-menopausal Caucasian Bioengineering 113:1325–1335.
women. British Journal of Nutrition 98:950–959. Guerra, A., L. Etienne-Mesmin, V. Livrelli, S. Denis, S. Blanquet-Diot, and
Bornhorst, G. M., L. Q. Chang, S. M. Rutherfurd, P. J. Moughan, and R. P. M. Alric. 2012. Relevance and challenges in modeling human gastric
Singh. 2013a. Gastric emptying rate and chyme characteristics for and small intestinal digestion. Trends in biotechnology 30:591–600.
cooked brown and white rice meals in vivo. Journal of the Science of Guillemin, H., B. Perret, D. Picque, O. Menard, and T. Cattenoz. 2010.
Food and Agriculture 93:2900–2908. Logiciel StoRM – Stomach and duodenum Regulation and Monitoring.
Bornhorst, G. M., M. J. Ferrua, S. M. Rutherfurd, D. R. Heldman, and R. P. IDDN.FR.001.230009.000.R.P.2010.000.31235: 290.
Singh. 2013b. Rheological Properties and Textural Attributes of Guo, Q., A. Ye, M. Lad, D. Dalgleish, and H. Singh. 2014. Effect of gel
Cooked Brown and White Rice During Gastric Digestion in Vivo. Food structure on the gastric digestion of whey protein emulsion gels. Soft
Biophysics 8:137–150. matter 10:1214–1223.
Bornhorst, G. M., S. M. Rutherfurd, M. J. Roman, B. J. Burri, P. J. Haraldsson, A. K., L. Rimsten, M. Alminger, R. Andersson, P. Aman, and
Moughan, and R. P. Singh. 2014. Gastric pH Distribution and Mixing A. S. Sandberg. 2005. Digestion of barley malt porridges in a gastroin-
of Soft and Rigid Food Particles in the Stomach using a Dual-Marker testinal model: Iron dialysability, iron uptake by Caco-2 cells and deg-
Technique. Food Biophysics 9:292–300. radation of beta-glucan. Journal of Cereal Science 42:243–254.
Cordonnier, C., J. Thevenot, L. Etienne-Mesmin, S. Denis, M. Alric, V. Liv- Havenaar, R., B. Anneveld, L. M. Hanff, S. N. de Wildt, B. A. E. de Koning,
relli, and S. Blanquet-Diot. 2015. Dynamic in vitro models of the M. G. Mooij, J. P. A. Lelieveld, and M. Minekus. 2013a. In vitro gastro-
human gastrointestinal tract as relevant tools to assess the survival of intestinal model. TIM) with predictive power, even for infants and
probiotic strains and their interactions with gut microbiota. Microor- children? International Journal of Pharmaceutics 457:327–332.
ganisms 3:725–745. Havenaar, R., A. de Jong, M. E. Koenen, J. van Bilsen, A. M. Janssen, E.
Cueva, C., A. Jimenez-Giron, I. Munoz-Gonzalez, A. Esteban-Fernandez, I. Labij, and H. J. M. Westerbeek. 2013b. Digestibility of Transglutami-
Gil-Sanchez, M. Duenas, P. J. Martin-Alvarez, M. A. Pozo-Bayon, B. Bar- nase Cross-Linked Caseinate versus Native Caseinate in an In Vitro
tolome, and M. V. Moreno-Arribas. 2015. Application of a new Dynamic Multicompartmental Model Simulating Young Child and Adult Gas-
Gastrointestinal Simulator (SIMGI) to study the impact of red wine in trointestinal Conditions. Journal of Agricultural and Food Chemistry
colonic metabolism. Food Research International 72:149–159. 61:7636–7644.
de Oliveira, S. C., C. Bourlieu, O. M
enard, A. Bellanger, G. Henry, F. Rous- Havenaar, R., A. Maathuis, A. De Jong, D. Mancinelli, A. Berger, and S.
seau, E. Dirson, F. Carriere, D. Dupont, and A. Deglaire. 2016a. Impact Bellmann. 2016. Herring roe protein has a high digestible indispensable
of pasteurization of human milk on preterm newborn in vitro diges- amino acid score (DIAAS) using a dynamic in vitro gastrointestinal
tion: Gastrointestinal disintegration, lipolysis and proteolysis. Food model. Nutrition Research 36:798–807.
Chemistry 211:171–179. Hocke, M., U. Sch€ one, H. Richert, P. G€ornert, J. Keller, P. Layer, and A.
de Oliveira, S. C., A. Deglaire, O. Menard, A. Bellanger, F. Rousseau, G. Stallmach. 2009. Every slow-wave impulse is associated with motor
Henry, E. Dirson, F. Carriere, D. Dupont, and C. Bourlieu. 2016b. activity of the human stomach. American Journal of Physiology – Gas-
Holder pasteurization impacts the proteolysis, lipolysis and disintegra- trointestinal and Liver Physiology 296:G709–G716.
tion of human milk under in vitro dynamic term newborn digestion. Jansson, E. T., C. L. Trkulja, J. Olofsson, M. Millingen, J. Wikstrom, A.
Food Research International 88:263–275. Jesorka, A. Karlsson, R. Karlsson, M. Davidson, and O. Orwar. 2012.
De Smet, I., P. De Boever, and W. Verstraete. 1998. Cholesterol lowering in Microfluidic Flow Cell for Sequential Digestion of Immobilized Proteo-
pigs through enhanced bacterial bile salt hydrolase activity. British liposomes. Analytical Chemistry 84:5582–5588.
Journal of Nutrition 79:185–194. Kecskemeti, A., and A. Gaspar. 2017. Preparation and characterization of a
D
eat, E., S. Blanquet-Diot, J.-F. Jarrige, S. Denis, E. Beyssac, and M. Alric. packed bead immobilized trypsin reactor integrated into a PDMS
2009. Combining the Dynamic TNO-Gastrointestinal Tract System microfluidic chip for rapid protein digestion. Talanta. 166:275–283.
with a Caco-2 Cell Culture Model: Application to the Assessment of Kong, F., and R. P. Singh. 2008. Disintegration of solid foods in human
Lycopene and a-Tocopherol Bioavailability from a Whole Food. Jour- stomach. Journal of Food Science 73:R67–80.
nal of Agricultural and Food Chemistry 57:11314–11320. Kong, F., and R. P. Singh. 2010. A Human Gastric Simulator. HGS) to
Deglaire, A., S. C. De Oliveira, J. Jardin, V. Briard-Bion, M. Emily, O. Study Food Digestion in Human Stomach. Journal of Food Science 75:
Menard, C. Bourlieu, and D. Dupont. 2016. Impact of human milk pas- E627–E635.
teurization on the kinetics of peptide release during in vitro dynamic Kovatcheva-Datchary, P., M. Egert, A. Maathuis, M. Rajilic-Stojanovic, A.
term newborn digestion. Electrophoresis 37:1839–1850. A. de Graaf, H. Smidt, W. M. de Vos, and K. Venema. 2009. Linking
Denis, S., T. Sayd, A. Georges, C. Chambon, S. Chalancon, V. Sante-Lhou- phylogenetic identities of bacteria to starch fermentation in an in vitro
tellier, and S. Blanquet-Diot. 2016. Digestion of cooked meat proteins model of the large intestine by RNA-based stable isotope probing. Envi-
is slightly affected by age as assessed using the dynamic gastrointestinal ronmental Microbiology 11:914–926.
TIM model and mass spectrometry. Food & Function. 7:2682–2691. Larsson, M., M. Minekus, and R. Havenaar. 1997. Estimation of the bio-
Elashoff, J. D., T. J. Reedy, and J. H. Meyer. 1982. Analysis of Gastric-Emp- availability of iron and phosphorus in cereals using a dynamic in vitro
tying Data. Gastroenterology 83:1306–1312. gastrointestinal model. Journal of the Science of Food and Agriculture
Ferrua, M. J., and R. P. Singh. 2010. Modeling the Fluid Dynamics in a 74:99–106.
Human Stomach to Gain Insight of Food Digestion. Journal of Food Marciani, L., R. Faulks, M. S. J. Wickham, D. Bush, B. Pick, J. Wright, E. F.
Science. 75:R151–R162. Cox, A. Fillery-Travis, P. A. Gowland, and R. C. Spiller. 2009. Effect of
Gao, K., A. L. Xu, C. Krul, K. Venema, Y. Liu, Y. T. Niu, J. X. Lu, L. Ben- intragastric acid stability of fat emulsions on gastric emptying, plasma
soussan, N. P. Seeram, D. Heber, and S. M. Henning. 2006. Of the lipid profile and postprandial satiety. British Journal of Nutrition
major phenolic acids formed during human microbial fermentation of 101:919–928.
tea, citrus, and soy flavonoid supplements, only 3,4-dihydroxyphenyl- Marciani, L., P. A. Gowland, A. Fillery-Travis, P. Manoj, J. Wright, A.
acetic acid has antiproliferative activity. Journal of Nutrition 136:52–57. Smith, P. Young, R. Moore, and R. C. Spiller. 2001a. Assessment of
Geigy. 1981. Geigy Scientific Tables. Units of measurement, body fluids, antral grinding of a model solid meal with echo-planar imaging. Ameri-
composition of the body, nutrition. Basel, Switzerland: CIBA-GEIGY. can Journal of Physiology-Gastrointestinal and Liver Physiology 280:
Gerard-Champod, M., S. Blanquet-Diot, J. M. Cardot, D. Bravo, and M. G844–G849.
Alric. 2010. Development and Validation of a Continuous In Vitro Sys- Marciani, L., P. A. Gowland, R. C. Spiller, P. Manoj, R. J. Moore, P. Young,
tem Reproducing Some Biotic and Abiotic Factors of the Veal Calf and A. J. Fillery-Travis. 2001b. Effect of meal viscosity and nutrients on
Intestine. Applied and Environmental Microbiology 76:5592–5600. satiety, intragastric dilution, and emptying assessed by MRI. American
Guerra, A., S. Denis, O. le Goff, V. Sicardi, O. Francois, A. F. Yao, G. Gar- Journal of Physiology-Gastrointestinal and Liver Physiology 280:G1227–
rait, A. P. Manzi, E. Beyssac, M. Alric, and S. Blanquet-Diot. 2016. G1233.
16 D. DUPONT ET AL.

Marciani, L., P. Young, J. Wright, R. Moore, N. Coleman, P. A. Gowland, vitro model of the human large intestine by phylogenetic microarray
and R. C. Spiller. 2001c. Antral motility measurements by magnetic analysis. Microbiology 156:3270–3281.
resonance imaging. Neurogastroenterology and Motility 13:511–518. Ribnicky, D. M., D. E. Roopchand, A. Oren, M. Grace, A. Poulev, M. A.
Marteau, P., M. Minekus, R. Havenaar, and J. Veld. 1997. Survival of lactic acid Lila, R. Havenaar, and I. Raskin. 2014. Effects of a high fat meal matrix
bacteria in a dynamic model of the stomach and small intestine: Validation and protein complexation on the bioaccessibility of blueberry antho-
and the effects of bile. Journal of Dairy Science 80:1031–1037. cyanins using the TNO gastrointestinal model (TIM-1). Food Chemis-
Martinez, R. C. R., H. R. Cardarelli, W. Borst, S. Albrecht, H. Schols, O. P. try 142:349–357.
Gutierrez, A. J. H. Maathuis, B. Franco, E. C. P. De Martinis, E. G. Zoe- Rose, D. J., K. Venema, A. Keshavarzian, and B. R. Hamaker. 2010. Starch-
tendal, K. Venema, S. M. I. Saad, and H. Smidt. 2013. Effect of galac- entrapped microspheres show a beneficial fermentation profile and
tooligosaccharides and Bifidobacterium animalis Bb-12 on growth of decrease in potentially harmful bacteria during in vitro fermentation in
Lactobacillus amylovorus DSM 16698, microbial community structure, faecal microbiota obtained from patients with inflammatory bowel dis-
and metabolite production in an in vitro colonic model set up with ease. British Journal of Nutrition 103:1514–1524.
human or pig microbiota. Fems Microbiology Ecology 84:110–123. Roussel, C., C. Cordonnier, W. Galia, O. Le Goff, J. Th
evenot, S. Chalan-
Marze, S., H. Algaba, and M. Marquis. 2014. A microfluidic device to study con, M. Alric, D. Thevenot-Sergentet, F. Leriche, T. Van de Wiele, V.
the digestion of trapped lipid droplets. Food & Function 5:1481–1488. Livrelli, and S. Blanquet-Diot. 2016. Increased EHEC survival and viru-
Marzorati, M., B. Vanhoecke, T. De Ryck, M. S. Sadabad, I. Pinheiro, S. Possem- lence gene expression indicate an enhanced pathogenicity upon simu-
iers, P. Van den Abbeele, L. Derycke, M. Bracke, J. Pieters, T. Hennebel, H. J. lated pediatric gastrointestinal conditions. Pediatric Research 80:734.
Harmsen, W. Verstraete, and T. Van de Wiele. 2014. The HMI (TM) mod- Russell, W. R., S. W. Gratz, S. H. Duncan, G. Holtrop, J. Ince, L. Scobbie,
ule: a new tool to study the Host-Microbiota Interaction in the human gas- G. Duncan, A. M. Johnstone, G. E. Lobley, R. J. Wallace, G. G. Duthie,
trointestinal tract in vitro. Bmc Microbiology 14. and H. J. Flint. 2011. High-protein, reduced-carbohydrate weight-loss
Menard, O., T. Cattenoz, H. Guillemin, I. Souchon, A. Deglaire, D. diets promote metabolite profiles likely to be detrimental to colonic
Dupont, and D. Picque. 2014. Validation of a new in vitro dynamic sys- health. American Journal of Clinical Nutrition 93:1062–1072.
tem to simulate infant digestion. Food Chemistry 145:1039–1045. Sanchez-Rivera, L., O. Menard, I. Recio, and D. Dupont. 2015. Peptide
Minekus, M., P. Marteau, R. Havenaar, and J. H. J. Huisintveld. 1995. A mapping during dynamic gastric digestion of heated and unheated
Multicompartmental Dynamic Computer-Controlled Model Simulat- skimmed milk powder. Food Research International 77:132–139.
ing the Stomach and Small-Intestine. Atla-Alternatives to Laboratory Schaafsma, G. 2005. The protein digestibility-corrected amino acid score
Animals 23:197–209. (PDCAAS) – A concept for describing protein quality in foods and food
Minekus, M., M. Smeets-Peeters, A. Bernalier, S. Marol-Bonnin, R. Have- ingredients: A critical review. Journal of AOAC International. 88:988–994.
naar, P. Marteau, M. Alric, G. Fonty, and J. Veld. 1999. A computer- Siegel, J. A., J. L. Urbain, L. P. Adler, N. D. Charkes, A. H. Maurer, B. Krev-
controlled system to simulate conditions of the large intestine with sky, L. C. Knight, R. S. Fisher, and L. S. Malmud. 1988. BIPHASIC
peristaltic mixing, water absorption and absorption of fermentation NATURE OF GASTRIC-EMPTYING. Gut. 29:85–89.
products. Applied Microbiology and Biotechnology 53:108–114. Smeets-Peeters, M. J. E., M. Minekus, R. Havenaar, G. Schaafsma, and M. W. A.
Miralles, B., R. del Barrio, C. Cueva, I. Recio, and L. Amigo. 2017. Dynamic Verstegen. 1999. Description of a dynamic in vitro model of the dog gastro-
gastric digestion of a commercial whey protein concentratey. Journal of intestinal tract and an evaluation of various transit times for protein and cal-
the Science of Food and Agriculture n/a-n/a. cium. Atla-Alternatives to Laboratory Animals 27:935–949.
Molly, K., M. Vandewoestyne, I. Desmet, and W. Verstraete. 1994. Valida- Souliman, S., E. Beyssac, J. M. Cardot, S. Denis, and M. Alric. 2007. Investi-
tion of the simulator of the human intestinal microbial ecosystem gation of the biopharmaceutical behavior of theophylline hydrophilic
(SHIME) reactor using microorganism-associated activities. Microbial matrix tablets using USP methods and an artificial digestive system.
Ecology in Health and Disease 7:191–200. Drug Development and Industrial Pharmacy 33:475–483.
Molly, K., M. V. Woestyne, and W. Verstraete. 1993. Development of a 5-step Souliman, S., S. Blanquet, E. Beyssac, and J. M. Cardot. 2006. A level A in
multichamber reactor as a simulation of the human intestinal microbial eco- vitro/in vivo correlation in fasted and fed states using different meth-
system. Applied Microbiology and Biotechnology 39:254–258. ods: Applied to solid immediate release oral dosage form. European
Munoz-Gonzalez, I., A. Jimenez-Giron, P. J. Martin-Alvarez, B. Bartolome, Journal of Pharmaceutical Sciences 27:72–79.
and M. V. Moreno-Arribas. 2013. Profiling of Microbial-Derived Phe- Sullivan, L. M., J. J. Kehoe, L. Barry, M. J. M. Buckley, F. Shanahan, K. H.
nolic Metabolites in Human Feces after Moderate Red Wine Intake. Mok, and A. Brodkorb. 2014. Gastric digestion of alpha-lactalbumin in
Journal of Agricultural and Food Chemistry 61:9470–9479. adult human subjects using capsule endoscopy and nasogastric tube
Naylor, T. A., P. C. Connolly, L. G. Martini, D. P. Elder, M. Minekus, R. sampling. British Journal of Nutrition 112:638–646.
Havenaar, and E. Zeijdner. 2006. Use of a gastro-intestinal model and Tabernero, M., K. Venema, A. J. H. Maathuis, and F. D. Saura-Calixto.
GastroPLUS for the prediction of in vivo performance. Journal of 2011. Metabolite Production during in Vitro Colonic Fermentation of
Applied Therapeutic Research 6:15–19. Dietary Fiber: Analysis and Comparison of Two European Diets. Jour-
Phinney, D. M. 2013. Design, Construction, and Evaluation of a Reactor nal of Agricultural and Food Chemistry 59:8968–8975.
Designed to Mimic Human Gastric Digestion. Davis: University of Thevenot, J., C. Cordonnier, A. Rougeron, O. Le Goff, H. T. T. Nguyen, S.
California. Denis, M. Alric, V. Livrelli, and S. Blanquet-Diot. 2015. Enterohemor-
Possemiers, S., S. Bolca, C. Grootaert, A. Heyerick, K. Decroos, W. Dhooge, rhagic Escherichia coli infection has donor-dependent effect on human
D. De Keukeleire, S. Rabot, W. Verstraete, and T. Van de Wiele. 2006. gut microbiota and may be antagonized by probiotic yeast during inter-
The prenylflavonoid isoxanthohumol from hops. Humulus lupulus L.) action with Peyer’s patches. Applied Microbiology and Biotechnology
is activated into the potent phytoestrogen 8-prenylnaringenin in vitro 99:9097–9110.
and in the human intestine. Journal of Nutrition 136:1862–1867. Thevenot, J., L. Etienne-Mesmin, S. Denis, S. Chalancon, M. Alric, V. Liv-
Possemiers, S., S. Rabot, J. C. Espin, A. Bruneau, C. Philippe, A. Gonzalez- relli, and S. Blanquet-Diot. 2013. Enterohemorrhagic Escherichia coli
Sarrias, A. Heyerick, F. A. Tomas-Barberan, D. De Keukeleire, and W. O157:H7 Survival in an In Vitro Model of the Human Large Intestine
Verstraete. 2008. Eubacterium limosum activates isoxanthohumol from and Interactions with Probiotic Yeasts and Resident Microbiota.
hops. Humulus lupulus L.) into the potent phytoestrogen 8-prenylnar- Applied and Environmental Microbiology 79:1058–1064.
ingenin in vitro and in the rat intestine. Journal of Nutrition 138:1310– Van de Wiele, T., N. Boon, S. Possemiers, H. Jacobs, and W. Verstraete. 2004.
1316. Prebiotic effects of chicory inulin in the simulator of the human intestinal
Possemiers, S., K. Verthe, S. Uyttendaele, and W. Verstraete. 2004. PCR- microbial ecosystem. FEMS Microbiology Ecology 51:143–153.
DGGE-based quantification of stability of the microbial community in Van de Wiele, T. R., A. G. Oomen, J. Wragg, M. Cave, M. Minekus, A.
a simulator of the human intestinal microbial ecosystem. FEMS Micro- Hack, C. Cornelis, C. J. M. Rompelberg, L. L. De Zwart, B. Klinck, J.
biology Ecology 49:495–507. Van Wijnen, W. Verstraete, and A. Sips. 2007. Comparison of five in
Rajilic-Stojanovic, M., A. Maathuis, H. G. Heilig, K. Venema, W. M. de vitro digestion models to in vivo experimental results: Lead bioaccessi-
Vos, and H. Smidt. 2010. Evaluating the microbial diversity of an in bility in the human gastrointestinal tract. Journal of Environmental
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 17

Science and Health Part a-Toxic/Hazardous Substances & Environmen- Venema, K., M. H. M. C. van Nuenen, E. G. Van den Heuvel, W.
tal Engineering 42:1203–1211. Pool, and J. M. B. M. van der Vossen. 2003. The effect of lactulose
Van den Abbeele, P., C. Belzer, M. Goossens, M. Kleerebezem, W. De Vos, on the composition of the intestinal microbiota and short-chain
O. Thas, R. De Weirdt, F. Kerckhof, and T. Van de Wiele. 2013a. Buty- fatty acid production in human volunteers and a computer-
rate-producing Clostridium cluster XIVa species specifically colonize controlled model of the proximal large intestine. Microbial Ecology
mucins in an in vitro gut model. Isme Journal 7:949–961. in Health and Disease 15:94–105.
Van den Abbeele, P., P. Gerard, S. Rabot, A. Bruneau, S. El Aidy, M. Der- Venema, K., S. H. F. Vermunt, and E. J. Brink. 2005. D-Tagatose increases
rien, M. Kleerebezem, E. G. Zoetendal, H. Smidt, W. Verstraete, T. Van butyrate production by the colonic microbiota in healthy men and
de Wiele, and S. Possemiers. 2011. Arabinoxylans and inulin differen- women. Microbial Ecology in Health and Disease 17:47–57.
tially modulate the mucosal and luminal gut microbiota and mucin- Vermeiren, J., P. Van den Abbeele, D. Laukens, L. K. Vigsnaes, M. De Vos,
degradation in humanized rats. Environ Microbiol. 13:2667–2680. N. Boon, and T. Van de Wiele. 2012. Decreased colonization of fecal
Van den Abbeele, P., C. Grootaert, M. Marzorati, S. Possemiers, W. Ver- Clostridium coccoides/Eubacterium rectale species from ulcerative
straete, P. Gerard, S. Rabot, A. Bruneau, S. El Aidy, M. Derrien, E. Zoe- colitis patients in an in vitro dynamic gut model with mucin environ-
tendal, M. Kleerebezem, H. Smidt, and T. Van de Wiele. 2010. ment. FEMS Microbiology Ecology 79:685–696.
Microbial community development in a dynamic gut model is repro- Verwei, M., A. P. Freidig, R. Havenaar, and J. P. Groten. 2006. Predicted
ducible, colon region specific, and selective for Bacteroidetes and Clos- serum folate concentrations based on in vitro studies and kinetic
tridium cluster IX. Appl Environ Microbiol. 76:5237–5246. modeling are consistent with measured folate concentrations in
Van den Abbeele, P., S. Roos, V. Eeckhaut, D. A. MacKenzie, M. Derde, W. humans. Journal of Nutrition 136:3074–3078.
Verstraete, M. Marzorati, S. Possemiers, B. Vanhoecke, F. Van Immer- Verwei, M., M. Minekus, E. Zeijdner, R. Schilderink, and R. Havenaar.
seel, and T. Van de Wiele. 2012. Incorporating a mucosal environment 2016. Evaluation of two dynamic in vitro models simulating fasted and
in a dynamic gut model results in a more representative colonization fed state conditions in the upper gastrointestinal tract. TIM-1 and tiny-
by lactobacilli. Microbial Biotechnology 5:106–115. TIM) for investigating the bioaccessibility of pharmaceutical com-
Van den Abbeele, P., K. Venema, T. Van de Wiele, W. Verstraete, and S. pounds from oral dosage forms. International Journal of Pharmaceutics
Possemiers. 2013b. Different human gut models reveal the distinct fer- 498:178–186.
mentation patterns of Arabinoxylan versus inulin. J Agric Food Chem. Vigsnaes, L. K., P. van den Abbeele, K. Sulek, H. L. Frandsen, C. Steen-
61:9819–9827. holdt, J. Brynskov, J. Vermeiren, T. van de Wiele, and T. R. Licht. 2013.
Vardakou, M., A. Mercuri, S. A. Barker, D. Q. Craig, R. M. Faulks, and M. Microbiotas from UC patients display altered metabolism and reduced
S. Wickham. 2011a. Achieving Antral Grinding Forces in Biorelevant ability of LAB to colonize mucus. Sci Rep. 3:1110.
In Vitro Models: Comparing the USP Dissolution Apparatus II and the Westerhout, J., E. V. de Steeg, D. Grossouw, E. E. Zeijdner, C. A. M. Krul, M.
Dynamic Gastric Model with Human In Vivo Data. Aaps Pharmscitech Verwei, and H. M. Wortelboer. 2014. A new approach to predict human
12:620–626. intestinal absorption using porcine intestinal tissue and biorelevant matri-
Vardakou, M., A. Mercuri, S. A. Barker, D. Q. Craig, R. M. Faulks, and M. S. ces. European Journal of Pharmaceutical Sciences 63:167–177.
Wickham. 2011b. Achieving antral grinding forces in biorelevant in vitro Wickham, M. J. S., R. M. Faulks, J. Mann, and G. Mandalari. 2012. The
models: comparing the USP dissolution apparatus II and the dynamic Design, Operation, and Application of a Dynamic Gastric Model. Dis-
gastric model with human in vivo data. AAPS PharmSciTech. 12:620–626. solution Technologies 19:15–22.