MODÉLISATION DU TRANSPORT, DE LA

DÉGRADATION ET DE L’ABSORPTION DES

ALIMENTS DANS L’INTESTIN GRÊLE
Masoomeh Taghipoor

To cite this version:

Masoomeh Taghipoor. MODÉLISATION DU TRANSPORT, DE LA DÉGRADATION ET DE
L’ABSORPTION DES ALIMENTS DANS L’INTESTIN GRÊLE. Systèmes dynamiques [[Link]].
Université François Rabelais - Tours, 2012. Français. �NNT : �. �tel-00762216v2�

HAL Id: tel-00762216

[Link]
Submitted on 6 Dec 2012

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 UNIVERSITÉ
FRANÇOIS RABELAIS
DE TOURS
Écoles Doctorales SSBCV et MIPTIS
Unité de Recherche Avicole (INRA-UR 0083)
Laboratoire de Mathématiques et Physique Théorique (CNRS-UMR 7350)

THÈSE présentée par :

Masoomeh TAGHIPOOR
soutenue le : 24 septembre 2012

pour obtenir le grade de : Docteur de l’Université François Rabelais - Tours

Discipline/ Spécialité : Mathématiques/ Biomathématique

MODÉLISATION DU TRANSPORT, DE LA
DÉGRADATION ET DE L’ABSORPTION DES ALIMENTS
DANS L’INTESTIN GRÊLE

THÈSE DIRIGÉE PAR :

L ESCOAT Philippe Ingénieur de Recherche HDR, INRA - Nouzilly
BARLES Guy Professeur, Université François Rabelais - Tours

R APPORTEURS :
L AROCHE Béatrice Directeur de Recherche, INRA - Jouy en Josas
F IORAMONTI Jean Directeur de Recherche, INRA - Toulouse

JURY :
BARLES Guy Professeur, Université François Rabelais - Tours
B ROSSARD Ludovic Ingénieur d’études, INRA - Rennes
D UITTOZ Anne Professeur, Université François Rabelais - Tours
F IORAMONTI Jean Directeur de Recherche, INRA - Toulouse
F OUILLET Hélène Chargé de Recherche, INRA - Paris
L AROCHE Béatrice Directeur de Recherche, INRA - Jouy en Josas
L ESCOAT Philippe Ingénieur de Recherche HDR, INRA - Nouzilly
S UPPO Christelle Professeur, Université François Rabelais - Tours
Remerciements

L’accomplissement d’une thèse de doctorat représente une étape importante dans la

vie professionnelle et personnelle d’un doctorant. La particularité de cette thèse, d’avoir
été réalisée dans un milieu pluridisciplinaire entre l’INRA et le LMPT, donne à ce travail
de recherche une dimension humaine très enrichissante.

Je tiens à exprimer ma profonde gratitude aux deux personnes qui ont dirigé cette
thèse de façon extraordinaire, Guy BARLES et Philippe LESCOAT. Vous avez toujours
été disponibles pour répondre à mes questions même les plus naïves. J’ai pu apprécier
votre soutien, votre assistance, votre optimisme et votre enthousiasme constants. Merci
de m’avoir confié ce projet de recherche passionnant et de m’avoir envoyé à droite et à
gauche pour en apprendre plus. Merci de m’avoir fait découvrir le monde magnifique du
croisement de deux domaines scientifiques : les mathématiques et la biologie. Bref, merci
de m’avoir appris mon métier.

J’exprime également toute ma gratitude à Jean FIORAMONTI pour avoir accepté

d’être rapporteur de ma thèse et de participer à mon jury, de même qu’à Béatrice LA-
ROCHE que je dois aussi remercier pour les discussions scientifiques qu’on a pu avoir
à des différentes occasions. Je les remercie pour le temps qu’ils ont consacré à l’éva-
luation de mes travaux. Je remercie aussi Anne DUITTOZ, Christelle SUPPO, Ludovic
BROSSARD et Hélène FOUILLET pour l’intérêt qu’ils portent à mon travail doctoral en
acceptant de participer à mon jury de thèse.

J’adresse mes plus vifs remerciement à l’équipe MODINGRE. Je pense avoir pro-
fité de conditions assez exceptionnelles pour une thèse. Des échanges régulièrs avec les
biologistes et les mathématiciens de l’équipe m’ont permis de mieux explorer le monde
merveilleux de la modélisation de la digestion. Je les remercie aussi d’avoir fait partie
de mon comité de thèse. Je souhaite donc formuler de chaleureux remerciement à Jaap
VAN MILGEN, Hélène FOUILLET, Lucile MONTAGNE, Alain TRUBUIL et Jean Marc
CHOBERT de même qu’à Jean-René LICOIS et Christine GEORGELIN que je dois aussi
remercier pour leur présence dans nos réunions internes et pour leurs nombreux commen-
taires qui ont contribué à l’amélioraton de ce travail.

3
REMERCIEMENTS

Je remercie Jean-René non seulement pour son enthousiasme surtout quand il s’agit de
programmation mais aussi pour avoir été mon parrain pour l’enseignement. Merci d’avoir
toujours laissé ta porte ouverte et d’avoir toujours été prêt à répondre à mes questions. Je
remercie Christine pour toutes les discussions qu’on a pu avoir pendant les réunions, sa
sympathie et son sens de l’humour même après 3h de réunion.

Une thèse repose aussi sur les conditions de travail dans lesquelles elle peut être réali-
sée et les relations amicales et cordiales du doctorant avec son entourage. J’ai pu profiter
des ambiances assez extraordinaires au LMPT et à l’URA. Je remercie ainsi les membres
du Laboratoire de Mathématiques et Physique Théorique et l’Unité de Recherche Avi-
cole de l’INRA : les responsables administratives et techniques du LMPT, Olivier, Ro-
main, Anne Marie, Bernadette, Anouchka et Sandrine et de l’URA, Claudette, Marie-José
et James pour leur disponibilité et leur gentillesse. Merci beaucoup à Ahmad Elsoufi,
Saïd ILIAS, Amaury MOUCHET, Hector GIACOMINI du LMPT et à Michel DUCLOS,
Agnes NARCY, Yves NYS, Anne COLIN, Bértrand MEDA et Isabelle BOUVAREL de
l’URA pour leur soutien et leurs conseils. Mes remerciements vont également à Jean-
Claude PICAUD pour son amitié et pour la relecture d’une partie de mes travaux, ses
commentaires et recommandations judicieuses.

Je tiens à remercier également l’ensemble des stagiaires, doctorants et jeunes docteurs

du LMPT et de l’URA : Amine, Asma, Kévin, Mohamad, Safaa, Rym, Tai, Anh, Hung,
Steve, Ali, Thomas, Isaline, Xavière, Dolores. Je remercie tout particulièrement Romain
et Élisa que je devrais aussi remercier pour la relecture d’une partie de mes travaux.

Une pensée toute spéciale à mes deux (aller je le dis) sœurs au LMPT, Élisa et San-
drine, avec qui j’ai partagé de chaleureux “tea-times”, des moments de joie et aussi de
déception ! Merci d’avoir été toujours présentes pour me soutenir.

Je n’oublirai pas de remercier mes amis en France et ailleurs : Toktam, Zara, Saeid,
Mandana, Leila, Solmaz, Imen, Florence, Chady, Aida, Sarah, Ola, Younes, Yousef, Ha-
maseh. Je remercie également les familles KHOSHNIAT, LE RHUN et PEER HOSSAINI
pour leur soutien, leurs conseils et leur présence.

Merci de tout coeur à ma famille, toujours présente à mes côtés. Merci à mes parents,
Nahid et Fathali, de m’avoir transmis le courage et l’envie pour atteindre mes objectifs.
Merci Madar, de m’avoir appris à me surpasser dans ma vie professionnelle, tu as toujours
été de bon conseil. Merci Pédar, d’avoir été toujours fière de moi, avant même que je ne
prouve quoi que ce soit. Je tiens à remercier particulièrement Mehdi et Hosnieh que je
considère comme mes parents, pour leurs conseils et leur gentillesse.

4
 Merci à ma petite sœur Shima pour sa joie de vivre et son optimisme et pour toutes
nos discussions téléphoniques. Merci à mes frères, Sadegh, Amir et Mohamad d’avoir
toujours été présents pour me soutenir. Je ne t’ai pas oublié ma petite Assale, merci d’être
restée avec nous.

Enfin, un immense merci à celui qui m’a soutenue, supportée, encouragée, motivée,
comprise, ... . Merci Mojtaba de t’être montré aussi enthousiaste que moi quand je parlais
de la digestion. Je te remercie pour ton pouvoir magique qui a transformé chaque instant
de ces trois années de thèse en un magnifique souvenir. Malgré la distance tu t’es toujours
montré présent pour moi. Merci de m’avoir apporté la tendresse et l’attention dont j’ai
eues besoin tout au long de cette thèse.
REMERCIEMENTS

6
Aux épaules sur les quelles j’ai pu mieux voir le monde ...
REMERCIEMENTS

8
Abstract

The purpose of this study is to represent a generic model of digestion in the small
intestine.

In the first part of this work, a model based on ordinary differential equations is used
to represent the digestion : the equations describe the evolution of the position and com-
position of the bolus coming from the stomach. Each bolus is identified as a cylinder.
This model considers simultaneously the different aspects of digestion i.e. transport of the
bolus all along the small intestine, feedstuffs degradation according to the enzymes and
local physical conditions, and nutrients absorption.

In the second part of this study, we use the homogenization method to sustain the
simplified digestion model developed in the first part. We show that this model can be
considered as a macroscopic version of more realistic models which contain the biological
phenomena at lower scales of the small intestine : (i) the high frequency peristaltic pulses
(microcopic time scale) and their effects on the velocity of bolus, (ii) the presence of the
intestinal villi (microscopic scale of space) and their influence on the digestion.

Finally, in the third part of this study, we investigate the digestion of a non-homogeneous
feedstuffs matrix by integrating the dietary fibre in the bolus. The two main physiochemi-
cal characteristics of dietary fibre which interact with the function of the small intestine,
i.e. viscosity and water holding capacity are modelled. This leads us to consider some
features of the digestion which have not been taken into account previously, particularly
the interrelationship between the evolution of dry matter and water in the bolus. Although
this model is generic and contains a large number of parameters, to our knowledge, it is
among the first qualitative dynamical modelling of fibre influence on intestinal digestion.

Keywords : Modelling, ODE, Digestion, Small intestine, Peristaltic, Homogenization,

Viscosity solutions, Dietary fibres.

9
ABSTRACT

10
Résumé

L’objectif de ce travail est de présenter un modèle générique de la digestion dans

l’intestin grêle.

Dans la première partie de ce travail, un modèle mécaniste basé sur les équations
différentielles ordinaires est utilisé pour présenter la digestion : les équations décrivent
l’évolution de la position et de la composition du bolus provenant de l’estomac. Chaque
bolus est représenté par un cylindre. Ce modèle prend en compte simultanément les dif-
férents aspects de la digestion à savoir le transport du bolus dans la lumière intestinale, la
dégradation des aliments par des enzymes, et l’absorption des nutriments.

Dans la deuxième partie de cette étude, nous utilisons les méthodes d’homogénéisa-
tion mathématique pour justifier le modèle de la digestion développé dans la première par-
tie. Nous montrons que ce modèle peut être considéré comme une version macroscopique
des modèles plus réalistes, qui contiennent des phénomènes biologiques à des échelles
inférieures de l’intestin grêle : (i) les ondes péristaltiques à haute fréquence (échelle du
temps microscopique) et leurs effets sur la vitesse du bolus, (ii) la présence des villosités
intestinales (échelle microscopique de l’espace) et leur influence sur la digestion.

Enfin, dans la troisième partie de ce travail, nous étudions l’influence du changement

de la structure de bolus sur la digestion en intégrant les fibres alimentaires dans sa compo-
sition. Les deux principales caractéristiques des fibres alimentaires qui interagissent avec
la fonction de l’intestin grêle, à savoir, la viscosité et la capacité de rétention d’eau ont été
modélisées. Ce modèle nous a permis de considérer, en particulier, la relation entre l’évo-
lution de la matière sèche et de l’eau au sein du bolus. Bien que ce modèle est générique
et contient un grand nombre de paramètres, à notre connaissance, il est parmi les premiers
modèles dynamiques qualitatives de l’influence des fibres sur la digestion intestinale.

Mots clés : Modélisation, EDO, Digestion, Intestin grêle, Péristaltique, Homogénéisa-

tion, Solutions de viscosité, Fibres alimentaires.

11
RÉSUMÉ

12
Table des matières

Liste des tableaux 17

Table de figures 20

Nomenclature 21

Introduction générale 23

1 Intestin grêle et digestion des aliments . . . . . . . . . . . . . . . . . . . 25

2 Méthodes mathématiques . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.1 Équations de transport-dégradation-absorption . . . . . . . . . . 29

2.2 Système d’équations différentielles ordinaires couplées . . . . . . 32

I Mathematical Modeling of Transport and Degradation of Feedstuffs in the

Small Intestine 37

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2 General Hypothesis and Synthetic Presentation of the Different Models . 40

3 Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4 Digestion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.1 Model 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.2 Model 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

13
TABLE DES MATIÈRES

4.3 Model 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.4 Model 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.1 Digestion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5.2 Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5.3 Model evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . 51

5.4 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . 52

6 Conclusion and Perspectives . . . . . . . . . . . . . . . . . . . . . . . . 53

7 Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

II Mathematical Homogenization in the Modelling of Digestion in the Small In-

testine 57

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

2 Transport Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

2.1 Position of the problem . . . . . . . . . . . . . . . . . . . . . . . 61

2.2 The asymptotic behavior . . . . . . . . . . . . . . . . . . . . . . 63

3 On the Effects of Intestinal Villi . . . . . . . . . . . . . . . . . . . . . . 67

3.1 Position of problem . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.2 Formal asymptotic . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.3 The rigorous result and proof . . . . . . . . . . . . . . . . . . . . 77

III Digestion Modelling in the Small Intestine : Impact of Dietary Fibre 83

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

2 Biological Background on Water and Dietary Fibre . . . . . . . . . . . . 85

3 Key Model Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4 Physiological Aspects and Bolus Composition . . . . . . . . . . . . . . . 89

14
TABLE DES MATIÈRES

5 Model Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

5.1 Transport of bolus . . . . . . . . . . . . . . . . . . . . . . . . . 93

5.2 Volumic transformation . . . . . . . . . . . . . . . . . . . . . . 93

5.3 Surfacic transformation . . . . . . . . . . . . . . . . . . . . . . . 94

5.4 The equilibrium between As and Ans . . . . . . . . . . . . . . . 95

5.5 Pancreatic and biliary secretions . . . . . . . . . . . . . . . . . . 96

5.6 Absorption through intestinal wall . . . . . . . . . . . . . . . . . 96

5.7 Water equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . 96

6 Numerical Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

6.1 Influence of dietary fibre on intestinal absorption . . . . . . . . . 97

6.2 Water associated to dry matter . . . . . . . . . . . . . . . . . . . 101

6.3 The variation of the water ratio inside the bolus . . . . . . . . . . 104

7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

8 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

8.1 The Model equations . . . . . . . . . . . . . . . . . . . . . . . . 106

Conclusion et perspectives 109

Bibliographie 116

15
TABLE DES MATIÈRES

16
Liste des tableaux

1 La composition du bol alimentaire, les enzymes inclus et les réactions enzyma-

tiques au sein du bol alimentaire dans les différents modèles présentés dans le
Chapitre I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

I.1 Digestion of starch in pigs by modeling : Comparison between simulated and

experimental data by [1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

III.1 The proportion of absorbed nutrients to DM for different scenarios of solubili-

zation in the bolus at the entry of duodenum. . . . . . . . . . . . . . . . . . . 98

III.2 The relation between the absorbed dry nutrients at the end of digestion and the
dm . . . . . . . . . . . . . . . . . . . . . . . . .
different initial values of Bint 98

III.3 Intestinal transit time for the different quantities of F = Fsol or Finsol in the
bolus at the entry of duodenum . . . . . . . . . . . . . . . . . . . . . . . . 100

III.4 Transit time and absorbed dry nutrients at the end of the small intestine depend
on the values of α, β and γ. . . . . . . . . . . . . . . . . . . . . . . . . . . 103

III.5 Dependence of absorbed dry nutrients to the water ratio. DM stands for the
amount of dry matter in the bolus at the entry of bolus. . . . . . . . . . . . . 104

17
LISTE DES TABLEAUX

18
Table des figures

1 Schéma général de l’intestin grêle (U.S. National Library of Medicine). . . . . . 26

2 Différentes échelles spatiales de l’intestin grêle et du côlon. . . . . . . . . . . 27

3 La différence entre la solution théorique en verte et la solution numérique en

bleu. L’axe des abscisses représente la longueur de l’intestin grêle et l’axe des
ordonnées la concentration de la matière. Les quantités ∆t et ∆x représentent le
pas du temps et de l’espace dans le schéma numérique. . . . . . . . . . . . . . 32

4 Représentation des ondes péristaltiques générées à x = 0 par une fonction pério-

dique W (t). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

I.1 The cylindric bolus with its different characteristics . . . . . . . . . . . . . . 41

I.2 Digestion through Model 4 . . . . . . . . . . . . . . . . . . . . . . . . . . 50

I.3 Velocity of the bolus versus Time . . . . . . . . . . . . . . . . . . . . . . . 51

I.4 Relative variation of As regarding to C, Cabs . . . . . . . . . . . . . . . . . . 53

I.5 Relative variation of Babs regarding to Cabs , Ciabs and kt . . . . . . . . . . . 54

II.1 Periodically oscillating velocity and averaged one. . . . . . . . . . . . . . . . 62

II.2 The different scales on intestinal anatomy relevant to our model. . . . . . . . . 68

II.3 A simple example of the domain Ω . The oscillations on the villi represents the
microvilli. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

III.1 Physical and chemical transformations inside the bolus and included in the model
are represented in this scheme. “B.B“ stands for brush border enzymes and “pcr”
stands for pancreatic ones. . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

19
TABLE DES FIGURES

III.2 The distribution of insoluble fibre in the bolus as assumed in the model. The
apparent volume Vapp (see Definition 5.3) is the white part of the cylinder. . . . 92

III.3 The equilibrium Adm

s − Ans is reached quickly for different initial value of As . . 99

III.4 Change in the final mass of bolus and absorbed dry nutrients for different amount
of DF in the bolus at the entry of duodenum. . . . . . . . . . . . . . . . . . . 100

III.5 Dependence of the absorbed dry nutrients and the final mass of bolus at the end
of the small intestine on the value of α, β and γ. . . . . . . . . . . . . . . . . 101

III.6 The evolution of the equilibrium Adm

s − Ans depends strongly on the quantity of
α, β and γ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

III.7 The equilibrium Adm

s − Ans for different quantity of α, β and γ. . . . . . . . . 103

20
Nomenclature

Symboles Unités Définitions

And g Matière non-dégradable
Ans g Matière non-solubilisée
As g Matière solubilisée
Adm
s g Proportion de la matière sèche présente dans As
Bint g Matière intermédiaire (non-absorbable)
dm
Bint g Proportion de la matière sèche présente dans Bint
Babs g Matière absorbable par la paroi intestinale
dm
Babs g Proportion de la matière sèche présente dans Babs
Fsol g Fibres solubles
dm
Fsol g Proportion de la matière sèche présente dans Fsol
Finsol g Fibres non-solubles
dm
Finsol g Proportion de la matière sèche présente dans Finsol
W g Quantité d’eau présente dans le bol alimentaire
M g Masse totale du bol alimentaire
V m3 Volume du bol alimentaire
e — Enzymes pancréatiques et stomacaux
eexo — Enzymes exogènes
R m Rayon
` m Longueur du bol alimentaire
x m Distance
Was g Quantité d’eau associée à Adm
s
Wint g dm
Quantité d’eau associée à Bint
Wabs g dm
Quantité d’eau associée à Babs

21
NOMENCLATURE

22
Introduction générale

En recherche, les disciplines biologiques s’appuient souvent sur l’expérimentation

animale. En effet, les phénomènes naturels sont complexes et rarement strictement iden-
tiques d’une espèce à l’autre. La démarche expérimentale permet le recueil des données
concernant chaque espèce et dans des conditions variées, ces données peuvent être ensuite
analysées, traitées et comparées permettant la compréhension de mécanismes d’intérêt.
Pourtant les interrogations de la société concernant l’utilisation des animaux à des fins
de recherche ont nécessité la mis en œuvre d’alternatives à l’expérimentation animales.
Ainsi, Russel et Burch [2], ont proposé la règle des “3R“ de l’utilisation des animaux dans
les recherches biologique : remplacer, raffiner et réduire.

Les modèles mathématiques sont souvent cités comme une alternative à l’expérimen-
tation animale et sont utilisés notamment quand la complexité du phénomène observé
ne permet pas à l’intuition ou à des approches analytiques de comprendre son fonction-
nement et son évolution (J. VanMilgen et P. Lescoat [3]). La modélisation mathématique
consiste à tenter de comprendre les lois régissant les phénomènes naturels du vivant à
travers leur représentation par des équations mathématiques. Mais, une condition sine qua
non est que la complexité des phénomènes biologiques implique l’adoption d’une ap-
proche réductionniste par le modélisateur. En d’autres termes en fonction de l’objectif,
le modélisateur choisit de négliger dans le modèle les variables d’état qui n’interviennent
que très peu dans l’évolution du système biologique tout du moins au regard de la question
traitée (J. Clairambault. [4]).

a. Pourquoi un modèle mathématique ?

La contribution de la modélisation aux ”3R” de l’expérimentation animale révèle

son intérêt dans les recherches biologiques. Ainsi, les modèles mathématiques, par leur
rôle dans l’orientation des recherches biologiques (raffinement) ont pu contribuer à la
réduction de l’expérimentation animale. De plus, les différents logiciels informatiques
permettent des simulations numériques du modèle construisant des expérimentations vir-
tuelles (in-silico), ce qui pourrait remplacer (ou au moins limiter) des redondances expé-
rimentales.

23
INTRODUCTION GÉNÉRALE

D’autre part, la complexité des phénomènes naturels peut rendre l’interprétation des
expérimentations “in-vivo” difficile et même impossible. Or, après la construction du mo-
dèle, son analyse mathématique, permet de vérifier l’existence et l’unicité des solutions,
l’analyse de leur stabilité ainsi qu’une analyse qualitative sur le comportement du sys-
tème [4]. Mais cela ne doit pas faire oublier que les résultats ainsi obtenus par modé-
lisation ne sont valides que dans le cadre des hypothèses posées initialement. En outre,
la mise en œuvre de théorèmes mathématiques rend valide les résultats, ceci soulignant
l’importance en modélisation appliquée à la biologie de l’utilisation des connaissances
mathématiques.

Par exemple, les systèmes du vivant ont souvent plusieurs échelles emboîtées, de la
molécule à la cellule, de la cellule à l’organe ou tissu, de l’organe à l’individu, etc. Les
propriétés macroscopiques de l’individu sont souvent liées au fonctionnement à l’échelle
inférieure (ex. influence de la quantité des cellules matures des villosités intestinales sur
l’absorption) et inversement, l’évolution à l’échelle microscopique du système dépend de
son état à l’échelle supérieure (ex. influence de la composition du bol alimentaire dans la
lumière intestinale sur l’utilisation des transporteurs pour l’absorption active). Un modèle
mathématique peut décrire l’évolution du système dans l’un de ces niveaux, mais il peut
aussi intégrer plusieurs échelles simultanément (ex. modèle d’évolution de la digestion
en fonction de la composition du bol alimentaire ainsi que la maturité des cellules sur les
villosités). Cette approche par modélisation permet aussi d’étudier l’influence respective
de modification des propriétés à chaque échelle sur l’évolution du système.

Suite à la phase de modélisation mathématique, la simulation numérique du modèle à

l’aide de logiciels appropriés est indispensable pour permettre à l’utilisateur d’effectuer
ses propres expériences “in-silico“. Cela permet d’étudier différentes hypothèses simulta-
nément et de s’affranchir de certaines limites existant dans les approches expérimentales,
notamment en termes de nombre de combinaisons et de domaines explorés par le modèle.

b. Limites d’utilisation d’un modèle mathématique

Un modèle mathématique contient très souvent de nombreux paramètres dont les va-
leurs sont fixées grâce aux données expérimentales ou à dire d’expert par absence de
mesures existantes ou disponibles. Ce grand nombre et la difficulté de les obtenir sont des
questions clés quant à la possibilité d’utiliser les modèles et quant à leur portée. D’autre
part, une étape importante qui suit la modélisation est la confrontation des sorties du mo-
dèle à des données issues de l’expérimentation afin de valider le modèle, ou pour tout le
moins son comportement.

En conclusion, de la même façon que l’expérimentation est nécessaire pour valider la

modélisation, un modèle devrait permettre de mieux définir des plans d’expérimentation.
Il est toutefois illusoire de croire que les phénomènes du vivant, d’une complexité hors de

24
1. INTESTIN GRÊLE ET DIGESTION DES ALIMENTS

notre portée, peuvent se réduire à des équations mathématiques, aussi complexes soient-
elles. C’est un débat depuis les origines de la connaissance scientifique [3].

c. Modélisation pour comprendre la digestion

L’intestin grêle est la partie la plus longue du tube digestif. Elle a des conditions fa-
vorables pour la dégradation des macromolécules (motricité, activité enzymatique, condi-
tions physico-chimiques, etc). La connaissance de la dynamique de la digestion dans l’in-
testin grêle des différentes espèces permettrait de mieux valoriser les aliments fournis et
donc d’optimiser les intrants et de limiter les rejets dans l’environnement. Or, les méthodes
expérimentales et les mesures in-vivo provoquent des perturbations dans le fonctionne-
ment de l’intestin grêle et ne permettent pas d’étudier la digestion en prenant simultané-
ment en compte les différentes échelles présentes dans l’intestin grêle. La modélisation
permet de surmonter certaines des difficultés qui existent dans des approches expérimen-
tales sous réserve d’une prise en compte argumentée des phénomènes biologiques connus
et pertinents pour l’objet étudié. .

L’objectif de ce travail est de construire un modèle générique de la digestion dans l’in-

testin grêle c’est-à-dire le transport des aliments le long de l’intestin grêle, la dégradation
des macromolécules au sein du bol alimentaire et l’absorption des nutriments à travers de
la paroi intestinale chez les monogastriques.

Nous rappellerons d’abord les principaux éléments de la digestion dans l’intestin grêle
chez les monogastriques, puis abordera des différentes méthodes de la modélisation que
nous avons testées. Cette thèse est organisée en trois chapitres dont les résumés sont pré-
sentés à la fin de l’introduction.

1 Intestin grêle et digestion des aliments

L’intestin grêle est un organe multifonctionnel, il est d’une part responsable du transit
des aliments dans la lumière intestinale et de leur mélange avec des enzymes et, d’autre
part, il est un lieu idéal pour la dégradation par l’hydrolyse enzymatique et l’absorption
des nutriments.

L’intestin grêle comporte trois segments : le duodénum qui suit le pylore, le jéjunum
qui est le lieu principal de l’absorption à cause de la présence d’une quantité importante
des villosités intestinales et l’iléon. La Figure 1 présente un schéma de l’intestin grêle.
La longueur de l’intestin grêle varie selon les espèces ; par exemple pour le porc 1 en
1. Toutes le mesures utilisées dans cette thèse correspondent à celles du porc en croissance.

25
1. INTESTIN GRÊLE ET DIGESTION DES ALIMENTS

croissance la longueur de l’intestin grêle est de 17 à 20 m et le diamètre est de 4 à 5 cm.

F IGURE 1: Schéma général de l’intestin grêle (U.S. National Library of Medicine).

La digestion est le processus par lequel les molécules organiques sont réduites pour
être rendues absorbables à travers la paroi du tube digestif. L’intestin grêle est l’organe
majeur de la dégradation des aliments et de l’absorption des nutriments. Dans le cadre
de ce travail, la digestion est représentée par trois phénomènes principaux : le transport
des aliments le long de l’intestin grêle par les ondes péristaltiques, la dégradation des ma-
cromolécules par l’hydrolyse enzymatique et l’absorption des nutriments par absorption
active et passive [5].

L’étude de la digestion nécessite la connaissance des divers caractéristiques anato-

miques de l’intestin grêle qui se représentent selon quatre échelles (Figure 2) :

• Échelle macroscopique ( 17 m) qui correspond à la longueur de l’intestin grêle.

• Échelle mésoscopique ( 1 cm) qui correspond à la longueur des villosités intesti-
nales (leur rôle étant d’augmenter la surface d’absorption) et la muqueuse intesti-
nale.
• Échelle sous mésoscopique ( 1 mm) qui correspond aux micro-villosités et aux
cryptes.
• Échelle microscopique ( 10 µm) qui correspond aux cellules épithéliales.

Par la suite, nous expliquons brièvement chacun de ces phénomènes.

26
1. INTESTIN GRÊLE ET DIGESTION DES ALIMENTS

F IGURE 2: Différentes échelles spatiales de l’intestin grêle et du côlon.

Dégradation

La dégradation dans l’intestin grêle est principalement due à l’hydrolyse enzymatique.

L’hydrolyse est une réaction spontanée très lente qui est accélérée par des enzymes. L’in-
testin grêle reçoit les sécrétions digestives du foie et du pancréas qui contiennent des
enzymes nécessaires pour les diverses réactions d’hydrolyse : des enzymes protéolytiques
(peptidases, trypsine et chymotrypsine), des amylases et des lipases. Ces sécrétions ont
notamment pour rôle de neutraliser le contenu acide provenant de l’estomac et d’émul-
sifier des lipides par les sels biliaires. En outre les enzymes intestinaux, les enzymes de
la bordure en brosse interviennent dans le processus final de la digestion, en complétant
l’hydrolyse commencée par les enzymes gastriques et pancréatiques. Parmi ces enzymes
citons la lactase, la maltase, les aminopeptidases A et N, la phsosphatase alcaline, etc
(cours de L. Montagne [6]).

27
2. MÉTHODES MATHÉMATIQUES

Absorption

L’absorption est le passage des nutriments à travers la paroi intestinale et les cellules
absorbantes. (P. Meunier [5]). L’absorption dans l’intestin grêle se fait selon deux méca-
nismes : (i) la diffusion passive et le transport facilité (par un transporteur membranaire)
selon un gradient de concentration, (ii) le transport actif permet l’absorption contre un
gradient chimique ou électrique, ce processus nécessite la consommation d’énergie. Dans
le cadre de ce travail nous nous intéressons à la diffusion passive et au transport actif.
Cette étude nécessitera l’intégration des propriétés du système à l’échelle microscopique
dans le modèle.

Transport

Le mouvement de l’intestin grêle a pour objectif de mélanger son contenu ainsi que
de le propulser. En période inter-digestive, ce mouvement se propage le long de l’intes-
tin grêle selon un cycle régulier. Ce phénomène est appelé le complexe moteur migrant
(CMM) et son rôle est de vider l’intestin grêle des résidus alimentaires et d’éviter sa co-
lonisation par des bactéries coliques. Chez toutes les espèces la durée de la propagation
de CMM est de 90 à 120 minutes, ce qui veut dire une propagation plus rapide chez les
espèces avec un intestin plus long. En période digestive le CMM s’interrompt et c’est
une activité contractile continue et irrégulière, un mélange de contractions segmentaires
et péristaltiques, qui prend sa place et permet le brassage des aliments et leur propulsion
aborale.

Dans le cadre de ce travail, nous modéliserons la propagation des ondes péristaltiques

et leur influence sur l’avancement du bol alimentaire. Ces ondes sont générées à l’entrée
de l’intestin grêle avec une période de l’ordre de la seconde. Leur intégration dans le
modèle de digestion implique une étude à deux échelles du temps :

• Échelle de la génération des ondes péristaltiques ( 12 sec)

• Échelle de la durée de la digestion intestinale (4 à 5 h).

2 Méthodes mathématiques

Nous présentons ici des méthodes utilisées pour modéliser la digestion. Ces méthodes
sont basées sur deux approches souvent utilisées pour étudier des modèles en physique ou
en chimie :

• l’approche eulérienne : en adoptant cette approche, l’observateur est dans une posi-

28
2. MÉTHODES MATHÉMATIQUES

tion fixe et observe l’évolution du bol alimentaire. Dans la Section 2.1, un modèle
de la digestion basé sur les équations aux dérivées partielles (EDP) est présenté.
Ce modèle est un exemple d’approche eulérienne en modélisation. Nous étudions
ensuite des avantages et des inconvénients de cette méthode pour modéliser la di-
gestion.
• l’approche lagrangienne : en adoptant cette approche, l’observateur se déplace avec
le même mouvement que les bols alimentaires. C’est l’approche que nous avons
adapté dans cette thèse pour modéliser la digestion.

2.1 Équations de transport-dégradation-absorption

Un premier modèle basé sur des équations aux dérivées partielles a été construit par
l’équipe MODINGRE 2 . Les équations de type transport-dégradation- absorption ont été
utilisées.

Dans ce modèle, l’intestin grêle est représenté par l’intervalle ]0, L[ et un flux de nut-
riments est injecté en x = 0 qui représente le pylore. Les notions suivantes sont utilisées
afin de représenter le contenu de l’intestin grêle :
(a) x ∈]0, L[ représente la position dans l’intestin grêle.
(b) A(x, t) est la concentration des aliments à l’instant t et à la position x.
(c) Ces nutriments sont digérés grâce à une enzyme de concentration e(x, t).
(d) B(x, t) est la concentration en aliments digérés, donc assimilables par l’organisme.
La réaction de dégradation de A en B le long de l’intestin grêle est catalysée par
l’enzyme e, ce phénomène est représenté par une réaction du type (A + e 7→ B + e).
L’efficacité de ces enzymes dépend du pH de l’intestin grêle à la position x.

Le mélange des aliments (A + B) et les enzymes (e) sont transportés à la vitesse c qui
peut dépendre de x et t :
∂e ∂e
= −c − k̄e , (1)
∂t ∂x
le premier terme du membre de droite, décrit le transport par la vitesse c et le deuxième
explique la décroissance exponentielle des enzymes par un taux de dégradation k̄.

Les aliments sont transportés par une vitesse c et sont dégradés par les enzymes

∂A ∂A
= −c − k(pH(x))Ae , (2)
∂t ∂x
2. Équipe qui regroupe des chercheurs de l’INRA et du Laboratoire de Mathématiques et Physique
Théorique (LMPT).

29
2. MÉTHODES MATHÉMATIQUES

le premier terme explique l’avancement de la matière A avec une vitesse c, le deuxième

terme décrit la dégradation de A qui est proportionnelle à A (loi d’action de masse) et à la
quantité d’enzymes e avec une efficacité qui dépend du pH de l’intestin grêle en position
x.

L’évolution de la matière absorbable (par la paroi intestinale) est décrite par

∂B ∂B
= −c + k(pH(x))Ae − k̃B , (3)
∂t ∂x
comme précédemment le premier terme représente le transport, le deuxième terme repré-
sente la production de B par A et le dernier terme l’absorption de B par la paroi intestinale
au taux k̃.

Les condition initiales sont A(x, 0) = B(x, 0) = e(x, 0) = 0. Les conditions aux
limites en x = 0 sont

• A(0, t) = φ(t) est une fonction périodique (ex. une indicatrice ou une fonction
constante par morceau valant 0 et 1) modélisant la vidange gastrique.
• B(0, t) = 0 est la quantité de matière dégradée à la sortie de l’estomac. En effet, la
dégradation au niveau de l’estomac est considérée comme négligeable.
• e(0, t) = ēφ(t), par cette hypothèse, nous supposons que les enzymes sont injectés
avec A dans l’intestin grêle, et de plus leur quantité est proportionnelle à celle de
A.

Comme le système qui représente la digestion est basé sur les équations du transport
avec un terme de réaction-absorption, il est classique d’utiliser la méthode des carac-
téristiques pour calculer sa solution théorique. Nous présentons ici l’utilisation de cette
méthode pour calculer la solution théorique à travers un exemple. Nous comparons ensuite
cette solution avec la solution numérique obtenue par des méthodes numériques.

Courbes caractéristiques

Pour k̄ = 0 et c constante dans l’équation (2), nous avons

∂A ∂A
+c = −ke0 A . (4)
∂t ∂x
Nous introduisons
a(t) = A(x(t), t) ,
ce qui implique
da(t) ∂A(x(t), t) ∂A(x(t), t)
= + ẋ(t) = −ke0 A(x(t), t) = −ke0 a(t) , (5)
dt ∂t ∂x
30
2. MÉTHODES MATHÉMATIQUES

donc par identification, nous avons

ẋ(t) = c . (6)

Les solutions de cette équation différentielle sont les courbes caractéristiques associées à
l’équation (4). Pour ces courbes la solution A de l’équation (4) est déterminée par l’équa-
tion différentielle ordinaire (EDO) (5).

De la même manière, la solution B de l’équation (3) et la solution e de l’équation (1)

peuvent être déterminées par une EDO. En conclusion, nous obtenons un système d’EDO
(à résoudre, suivant les cas, explicitement ou numériquement) qui permet de calculer la
solution du système d’EDP.

Nous avons aussi utilisé la méthode des différences finies afin de calculer la solution
numérique de l’équation (4). La Figure 3 montre la différence entre la solution théorique
de l’Équation (4) et sa solution calculée par des méthodes numériques 3 . Dans la première
courbe (Figure 3.a), avec le pas de l’espace ∆x = 5cm et le pas du temps ∆t = 4s, la
différence entre les deux solutions théorique et numérique est plus que décevante. Les pas
du temps et de l’espace ont été affinés (∆x = .25cm et ∆t = 1s) afin d’améliorer le
résultat, mais la différence entre les deux solutions théorique et numérique, affichée dans
le Figure 3.b, n’est toujours pas acceptable.

Cet exemple met en évidence des difficultés de l’approche eulérienne pour modéliser
la digestion. Nous avons constaté que la différence entre la solution obtenue par la mé-
thode des différences finies et la solution théorique n’est pas acceptable. Ce problème est
lié au fait que d’une part, l’équation du transport est connue pour être intrinsèquement
délicate à résoudre et d’autre part, les échelles multiples à prendre en compte aggravent
la situation.

Le développement d’un modèle edp ne semble donc pas être une approche adéquate
pour modéliser la digestion. Nous avons donc décidé de changer l’approche et d’adopter
une approche lagrangienne qui est inspirée par la méthode des caractéristiques. Dans la
nouvelle approche, un modèle basé sur des équations différentielles couplées (l’analogue
des équations de la méthode des caractéristiques couplées avec celles de la dégradation
et de l’absorption) a été mis au point. Ce modèle a permis dans un premier temps, de
prendre en compte un aspect plus sophistiqué de la digestion comme l’effet des ondes
péristaltiques ainsi que la dégradation par des différents types d’enzymes.

3. Illustrations par Isaline Aubert, Stagiaire du magistère de l’ENS de Cachan

31
2. MÉTHODES MATHÉMATIQUES

(a) ∆x = 5cm, ∆t = 4s

(b) ∆x = 0.25cm, ∆t = 1s

F IGURE 3: La différence entre la solution théorique en verte et la solution numérique en bleu.

L’axe des abscisses représente la longueur de l’intestin grêle et l’axe des ordonnées la concen-
tration de la matière. Les quantités ∆t et ∆x représentent le pas du temps et de l’espace dans le
schéma numérique.

2.2 Système d’équations différentielles ordinaires couplées

Dans cette approche, les aliments dans la lumière intestinale forment des paquets que
nous appelons bolus (ici des cylindres de rayon R(t), de longueur ` et de centre x(t)) qui
ont des évolutions indépendantes les uns par rapport aux autres. Les inconnues sont la
position x(t), le rayon R(t) et la quantité d’enzymes e(t). Ici, nous étudions l’évolution
d’un seul bolus dans la lumière intestinale.

Dans ce premier modèle, la digestion se fait via la surface de contact, en plus nous
supposons que la vitesse de disparition de A est proportionnelle à la surface du cylindre

32
2. MÉTHODES MATHÉMATIQUES

et à l’activité enzymatique :
dA
(t) = −k(pH(x(t)), e(t)).2πR(t)`A(t) ,
dt
de
où e est la concentration en enzyme au point x(t) qui suit toujours la loi (t) = −k̄e(t).
dt
Bien évidemment, il y a la création de nutriments assimilables B(t) qui sont absorbés par
la paroi avec un taux k̃ :
dB
(t) = k(pH(x(t)), e(t)).2πR(t)`A(t) − k̃B(t) .
dt

Le transport du bol alimentaire est assuré par des ondes péristaltiques qui sont géné-
rées périodiquement. La Figure 4 montre un exemple d’ondes générées toutes les minutes,
L’efficacité de ces ondes est proportionnelle à la taille du bolus

Les ondes périst alt iques

600

500

400

300

200

100

0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10

Tim e (hour)

F IGURE 4: Représentation des ondes péristaltiques générées à x = 0 par une fonction périodique
W (t).

dx
(t) ∝ (A + B)(t)
dt
pour (A + B)(t) = πR2 (t)`ρ , où ρ représente la masse volumique du bolus. D’autre part,
l’efficacité du transport décroît avec la distance
dx 1
∝ .
dt x
Enfin, la vitesse est atténuée par un terme constante K de viscosité. L’équation suivante
modélise alors l’avancement du bol alimentaire le long de l’intestin grêle :
dv c0 + c1 (A + B)(t)/ρ
(t) = (1 − v(t)/c)W (1 − x(t)/c) − Kv(t) .
dt a + bx(t)

33
2. MÉTHODES MATHÉMATIQUES

dx
la fonction v(t) = (t) représente la vitesse d’avancement du bol alimentaire. La vitesse
dt
moyenne des ondes péristaltiques est représentée par c.

Ce modèle, basé sur une approche lagrangienne, nous permet de calculer la solution
numérique avec moins d’erreurs et d’autre part de prendre en compte conjointement la
complexité du bolus et celle du milieu intestinal.

Nous avons donc retenu et développé ce modèle pour obtenir un modèle plus réaliste
qui permet de tenir compte des phénomènes qui jouent un rôle important dans la digestion
intestinale (viscosité du bol alimentaire, interactions entre les différentes composantes du
bol, effet de l’eau sur la digestion, etc). Le modèle initial ne prend pas en compte la
présence des villosités intestinales ainsi que les absorptions active et passive. Cependant,
au Chapitre II, nous démontrons que les propriétés à l’échelle microscopique peuvent être
moyennées et insérées dans un modèle de digestion à l’échelle macroscopique.

Ce travail est organisé de la manière suivante :

Chapitre I Les trois étapes clés qui guident la première partie de ce travail sont : le
transport du bol alimentaire par les ondes péristaltiques, la dégradation des aliments par
de nombreuses réactions biochimiques et l’absorption des nutriments via des processus
actifs ou passifs au sein de la paroi intestinale

Dans le premier chapitre, une succession de modèles génériques de la digestion dans

l’intestin grêle a été construit. Le bol alimentaire est assimilé à un cylindre homogène
avec une longueur fixe ` et rayon variable R(t). Il se déplace dans la lumière intestinale à
l’aide des ondes péristaltiques. La dégradation des aliments est le résultat de l’hydrolyse
enzymatique par des transformations volumiques et surfaciques. Ces phénomènes sont
présentés par un système d’équations différentielles ordinaires couplées qui se résout par
les méthodes classiques d’intégration numérique de Runge-Kutta.

Nous détaillons les différentes étapes de la modélisation de la digestion sous forme de

quatre modèles, chaque modèle étant une version “améliorée“ du précédent. Ainsi, chaque
nouveau modèle contient plus de phénomènes biologiques que le précèdent. Le Tableau
1 représente les hypothèses de trois premiers modèles. Le modèle 4 est une simplifica-
tion du modèle 3 par des méthodes mathématiques d’homogénéisation. Nous démontrons
mathématiquement que l’accélération due aux ondes péristaltiques peut être moyennée de
façon à obtenir une variation plus lente de la vitesse.

Chapitre II L’une des difficultés de l’étude de la digestion dans l’intestin grêle vient
du fait que l’environnement du système est complexe. Ainsi la paroi intestinale joue un
rôle clé dans le transfert des nutriments dans le sang et dans la dégradation du bolus tout
autant que dans le transit du bol alimentaire par les ondes péristaltiques. Afin de prendre en
compte cette complexité tout en la réduisant partiellement, la seconde partie de ce travail

34
2. MÉTHODES MATHÉMATIQUES

Modèles Composition Enzymes Réactions

(forme de matière) enzymatiques
Modèle 1 Solubilisé As Gastriques As → Babs
Absorbable Babs
Modèle 2 Solubilisé As Gastriques As → Bint
Intermédiaire Bint Pancréatiques As → Babs
Absorbable Babs Brush-border Bint → Babs
Modèle 3 Non dégradable And Gastriques As ↔ Ans
Non solubilisé Ans Pancréatiques As → Bint
Solubilisé As Brush-border As → Babs
Intermédiaire Bint Bint → Babs
Absorbable Babs
Eau W
TABLE 1: La composition du bol alimentaire, les enzymes inclus et les réactions enzymatiques
au sein du bol alimentaire dans les différents modèles présentés dans le Chapitre I.

consiste à utiliser une méthode d’homogénéisation pour simplifier le modèle du transport

initial et à analyser et justifier le choix du taux d’absorption par la paroi intestinale.

D’abord, le transport des bolus à l’intérieur de l’intestin grêle est induit par des im-
pulsions à haute fréquence. Ces impulsions entraînent une variation rapide de la vitesse
du bolus dans l’intestin grêle. Nous démontrons mathématiquement que les impulsions
peuvent être moyennées d’une manière appropriée. Par conséquent, la vitesse à variation
rapide dans le modèle peut être remplacée par cette nouvelle vitesse moyennée.

Ensuite, afin de prendre en compte l’influence des villosités intestinale sur la diges-
tion à l’échelle macroscopique (intestin grêle), nous définissons un taux effectif moyen
d’absorption à l’aide des méthodes mathématiques d’homogénéisation.

L’homogénéisation est une méthode rigoureuse pour démontrer des propriétés de mo-
yennisation. En d’autres termes, en prenant en compte le comportement du système à
l’échelle microscopique, l’homogénéisation permet de décrire son comportement à l’échel-
le macroscopique. Dans l’étude de la digestion, au lieu d’étudier un seul problème à
l’échelle microscopique de taille   1, nous étudions une suite de problèmes pour 
qui tend vers zéro. La question est donc de trouver la limite de cette suite des problèmes
(G. Allaire [7]).

Dans ce but, nous supposons que les villosités intestinales sont distribuées périodi-
quement avec la période   1 le long de l’intestin grêle, une équation aux dérivées par-
tielles tri-dimensionnelle de diffusion-transport-réaction pour modélise la digestion dans
ses différentes échelles. Les conditions aux limites de type Neumann représentent le taux
d’absorption à travers la paroi de l’intestin grêle.

35
2. MÉTHODES MATHÉMATIQUES

Notre méthode consiste à un passage à la limite de ce modèle tri-dimensionnel pour

obtenir une équation de transport uni-dimensionnelle avec un taux d’absorption moyen.
Ce taux d’absorption tiendra compte de l’influence des villosités intestinales sur l’ab-
sorption. En effet, nous démontrons qu’un modèle complexe tri-dimensionnel peut être
simplifié à un modèle uni-dimensionnel sans perdre des propriétés importantes du sys-
tème.

Chapitre III Le passage d’un bolus homogène à un mélange de nutriments en inter-

action au sein du bol alimentaire est une étape importante à intégrer dans notre modèle.
Afin d’étudier l’influence des propriétés physicochimiques de l’aliment sur la digestion,
la présence des fibres alimentaires dans la composition du bolus a été prise en compte.

Les objectifs sont de modéliser l’impact de l’enrichissement de l’aliment en fibres sur,

d’une part, la digestibilité des nutriments et d’autre part, la vitesse de transit du bolus. Le
modèle tient compte de deux catégories principales des fibres, solubles et non solubles.
Les résultats des expériences numériques en présence des différents pourcentages des
fibres apparaissent cohérents avec les études bibliographiques. Ces dernières ne précisent
pas la raison des effets observés mais proposent des hypothèses que nous avons mises en
œuvre dans le modèle.

Enfin, une dernière partie de la thèse abordera quelques éléments de la discussion et

les principales perspectives de ce travail.

36
Chapitre I

Mathematical Modeling of Transport

and Degradation of Feedstuffs in the
Small Intestine

We describe a mathematical model of digestion in the small intestine. The main inter-
est of our work is to consider simultaneously the different aspects of digestion i.e. trans-
port of the bolus all along the intestine, feedstuffs degradation according to the enzymes
and local physical conditions, and nutrients absorption. A system of coupled ordinary dif-
ferential equations is used to model these phenomena. The major unknowns of this system
are the position of the bolus and its composition. This system of equations is solved nu-
merically. We present several numerical computations for the degradation, absorption and
transport of the bolus with acceptable accuracy regarding the overall behavior of the mo-
del and also when challenged versus experimental data. The main feature and interest of
this model are its genericity. Even if we are at an early stage of development, our approach
can be adapted to deal with contrasted feedstuffs in non-ruminant animal to predict the
composition and velocity of bolus in the small intestine 1 .

1. Ce chapitre a fait l’objet d’une publication au journal of Theoretical Biology [8] :

Mathematical modeling of transport and degradation of feedstuffs in the small intestine
Masoomeh Taghipoor, Philippe Lescoat, Jean-René Licois, Christine Georgelin, Guy Barles
Received 15 April 2011. Revised 3 October 2011. Accepted 18 October 2011. Available online 10 Novem-
ber 2011.

37
1 Introduction

The main step of digestion and absorption along the gastrointestinal tract takes place in
the small intestine for non ruminant animals. To reach an optimized composition of avai-
lable nutrients due to their behavior in the digestive system, it is necessary to understand
and predict the digestion and absorption of the ingested feedstuffs in the small intestine
[9, 10, 11]. It is also now well-known that the use of implanted experimental devices may
modify the dynamic of digestion in the small intestine [12, 13].

Several models have been developed representing the digestion and transport of bolus
in the small intestine. In the model of [14] digestion and absorption are integrated and
represented considering only the polymers and individual absorbable end products. The
transit through the small intestine is modeled mainly as a result of gastric emptying. No
peristaltic wave is taken into account, and the bolus contained only the dry matter. [15]
describe the digestion and absorption using the plug flow reactors to encapsulate com-
plex digestion phenomena in a simple set of equations. Different rate of absorption and
degradation are involved : first order kinetics, Michaelis-Menten and the sigmoid ones. A
detailed model of the intestinal propulsion is provided by [16, 17]. However, these mo-
dels portray the transport of bolus simplistically, or they represent only a limited number
of different processes involved in digestion.

This article tries to go further in the modeling of digestion in the small intestine by
considering the different steps of digestion i.e. the transport of the bolus all along the
intestine, feedstuffs degradation according to the enzymes and local physical conditions,
and nutrient absorption. Therefore a system of coupled ordinary differential equations is
used. The major unknowns of this system are the position of the bolus and its composition.

In fact, several models are presented reflecting the modeling process at its different
stages with our attempts to make it more realistic by inclusion of more sophisticated and
relevant biological phenomena and chemical transformations. We decided to describe the
different steps with the assumptions leading us to our choices instead of presenting only
the last model since the whole process by itself may help to underline relevant questions
to be further discussed. Of course, this modeling process is an iterative one and is still
going ahead in directions which are described in Section 6.

Our models intend to be a mechanistic approach of feedstuffs digestion even though

simplifications were included according to participatory approaches between biologists
and mathematicians. However they involve a lot of different unknowns and parameters,
and require a numerical software to obtain suitable approximation of the solutions since
it is hopeless to obtain explicit ones. Scilab software was used to perform these numerical
computations 2 .

2. The reader can perform its own numerical experiments, with the possibility of changing the parame-
2. GENERAL HYPOTHESIS AND SYNTHETIC PRESENTATION OF THE
DIFFERENT MODELS

In all our models, we try to estimate the parameters using data from scientific litera-
ture. When these data are not available, we assume the reasonable values for the parame-
ters.

The article is organized as follows : Section 2 is devoted to present the main assump-
tions of our models and most of our notations. In Section 3, we describe the transport
equations. In our four different models, we point out that there are only two different
ways of modeling the transport of the bolus in the intestinal tract. The main differences
concern the degradation itself, with several possibilities for the composition of the bolus,
for the enzymatic reactions and the water influence. The outcoming stages (4 different
models) are presented in Section 4, with the key assumptions and characteristics of each
model. Section 5 is a comparison of these four models and of the numerical results of the
most sophisticated model (Model 3) versus some experimental data from the literature.
Finally, in Section 6, we criticize our models and describe the perspective.

2 General Hypothesis and Synthetic Presentation of the

Different Models

Common assumptions to all models are the following ones.

(i) The first simplification concerns the small intestine representation itself. Instead
of taking into account its complex geometry, it is represented as a one-dimensional
interval [0, L]. The position of the bolus in the small intestine at time t is given by
x(t) ∈ [0, L] (cf. Figure I.1). The equation of the transport of bolus along the small
intestine is described on x(t).
(ii) The bolus is treated as a homogeneous cylinder with a fixed length ` and variable
radius R(t) which is a function of time t. To locate this cylinder, we use the position
x(t) of its center. This assumption is justified by the general shape of the small
intestine’s segments as well as the observation of the real bolus in animals’ small
intestines. As the length of bolus is assumed fixed, the degradation only changes the
radius. Degradation of substrates is obtained by enzymatic reactions with possible
subsequent absorption by the intestinal wall [5].
(iii) The enzymes which participate in enzymatic hydrolysis included in our models
consist systematically in pancreatic and brush-border ones with the possibility of
including exogenous and gastric ones. The enzymes’ activity depends on the pH of
the small intestine at each point along its length. The brush-border enzymes on the
intestinal wall are assumed to be always in excess [5].
The bolus moves through the intestinal tract because of the pulses resulting from per-
istaltic waves and gastric emptying which gives an initial velocity to the bolus [5]. Per-
istalsis are series of wavelike contractions occurring in the smooth muscle layer of the
ters, using our Scilab software online at the URL : [Link]

40
2. GENERAL HYPOTHESIS AND SYNTHETIC PRESENTATION OF THE
DIFFERENT MODELS

F IGURE I.1: The cylindric bolus with its different characteristics

gastrointestinal tract. It is a physiological process that results in intestinal motility and

propulsion of ingested food along the intestine. It starts as a ring-like constriction initially
which later moves mostly forward along the intestine. Moreover it might be assumed that
it helps the bolus to be digested by spreading the food particles along the intestinal wall
for effective digestion and absorption. Therefore to model the transport, bolus movement
is connected to pulses all along the intestine with the initial velocity coming from the gas-
tric emptying effect. Efficiency of pulses is proportional to the volume of the bolus and
it is inversely correlated to the distance between the bolus and the pylorus [18, 19]. The
bolus speed is assumed to be slowed down by the local conditions in the small intestine
lumen (friction on the borders, viscosity effects, etc). The effects of these different local
factors depends on the composition of the bolus, and in particular its dilution.

For the bolus content, the following assumptions and notations are used throughout
this paper
• The bolus includes a single species whose total mass is denoted by A. In the most
sophisticated model presented in this work, A is composed of As , Ans and And , in
other words A =As +Ans + And . As is the mass of solubilized fraction of A which
can be hydrolyzed in the presence of the enzymes. The index s stands for “solubili-
zed“. Ans is the mass of “non-solubilized“ fraction of A, for example dry starch or
the non-emulsified lipids. Transformation of Ans into As requires a sufficient quan-
tity of water. Regarding lipids emulsion, we assume that the bile salts are in excess.
The mechanism Ans ↔ As is described through an equilibrium property depending
on the water quantity in the bolus. Finally And is the mass of non-degradable A,
which enters and leaves the small intestine without any change. For example the
vegetal fiber in feed are poorly digestible. Moreover the fiber matrix of feedstuffs
or the anti-nutritional factor content can be responsible for a reduction in the diges-

41
2. GENERAL HYPOTHESIS AND SYNTHETIC PRESENTATION OF THE
DIFFERENT MODELS

tibility of some amino acids in some feedstuffs [20].

• The quantity B is the mass of product obtained from As by enzymatic reactions, it
is composed of Bint and Babs
enzymes
enzymes
As −→ Bint −→ Babs .

The quantity Bint is the product of hydrolysis due to gastric and pancreatic enzymes,
the index int stands for ”intermediate” substrate which is not yet absorbable since
it is not fully degraded. This transformation has to be completed by a second one at
the border of the small intestine via the brush-border enzymes (e.g. : proteins being
degraded to polypeptides and afterwards to dipeptide or amino acids, which are
absorbable). This second transformation is also able to give Babs directly from A.
The quantity Babs is the absorbable fraction with index abs indicating “absorbable“.
• The quantity e represents the gastric and pancreatic enzymes.
• The quantity W is the mass of water in the bolus and [W ] indicates the proportion
of water in the bolus : W/(A + B + W ).
• The quantity V (t) denotes the volume of bolus which is equal to (A + B + W )/ρ,
where ρ denotes the density. For the sake of simplicity, we assume that all the sub-
strates of bolus have the same density ρ. The total mass of the bolus is (A + B +
W )(t) at each moment.
Digestion consists in the transformation of digesta to absorbable nutrients through enzy-
matic hydrolysis. Volumic transformation is the degradation of As into Bint inside the
bolus and transformation on the bolus surface is the degradation of both As and Bint into
Babs on a thin layer around the bolus. The following hypothesis are added progressively
with upgraded versions of the model.

Model 1. In the first model, the whole bolus is considered to be solubilized (A =As ).
A is hydrolyzed thanks to gastric enzymes and becomes directly absorbable nu-
trients (B = Babs ). In this model, brush-border or pancreatic ones are not taken
into account. Such mechanisms are associated for example with the consumption of
disaccharides (resp. monosaccharides) such as milk sugar(resp. glucose).
Model 2. This model is an attempt to have a more realistic modeling of degradation.
The bolus is still assumed to be completely solubilized. The absorbable nutrients
can be obtained by two ways : either by a direct surfacic transformation A → Babs
or through a first volumic degradation A → Bin followed by a second one Bint →
Babs at the bolus surface by brush-border enzymes.
Model 3. This model includes the solubilization of the bolus in presence of water.
A is splitted into As , Ans , And . Equations are added to express the equilibrium
As ↔ Ans which depends on the quantity of water. The non-degradable part of bo-
lus enters and leaves the small intestine without any mechanical or chemical change
in its initial form. A key feature of this model concerns the transport of bolus along
the small intestine since it is connected to the quantity of water through lubrification
effects.

42
3. TRANSPORT

Model 4. This model is a simplification of the previous one by mathematical ar-

guments. Through homogenization methods it is shown that the acceleration can
be averaged and an equation with this averaged acceleration is substituted for the
pulses in transport equation. Detailed models are described in Section 4.

3 Transport

We present a mathematical formulation of the transport of bolus in the small intestine.

It is based on the physiology of the pig’s small intestine to get consistent parameters.
The duodenum is characterized by oscillatory electrical events (slow waves) occurring
at a rate of 18/min. After food intake, some of these waves are associated with spikes
bursts which are responsible for contractions and therefore propelling the bolus through
the small intestine. We assume only 6 of the 18 slow waves by minute are followed by
the spikes. It leads to one efficient contraction every 10 seconds [18]. The mean transit
time of each peristaltic wave is assumed to be 150 minutes to move along small intestine
from duodenum to the end of ileum according to [21] and [22]. We assume also that the
pig’s small intestine is about 18 meters [23], hence the average velocity c of these waves
is 7, 2 m/h. Each peristaltic wave takes x(t)/c seconds to reach the bolus in position x(t),
therefore the pulse which pushes the bolus in time t is generated in duodenum at time
t − x(t)/c.
dx
If v(t) denotes the velocity of the bolus (v(t) = (t)), the effect of pulses is modeled
dt
through the following equation

d2 d d
2
x(t) = v(t) = y (t − x(t)/c) ,
dt dt dt

d
the term y represents the pulses which are defined as a periodic function of period 10
dt
seconds such that
Z 10
d
y(t)dt = 1
0 dt
0
and for t < 0, we assume y (t) = 0.

Over a period, each pulse is an approximation of a Dirac mass of the origin. Therefore
we define it as a function with the value 1/ during a very short interval of time  and 0 at
all other time.

According to [19] and [24] the efficiency of the peristaltic waves increases with the
size of the bolus and decreases with the distance from pylorus. We assume that all these

43
4. DIGESTION

dependences are affine, namely

d2 x d c0 + c1 V (t)
2
(t) = [y (t − x(t)/c)] ,
dt dt a + bx(t)
where, c0 and c1 are determined under the assumption that the acceleration depends li-
nearly on V (t). THe constants a and b are obtained from experimental data.

The intestinal lumen is a confined environment which prevents the bolus to move
perfectly according to the previous equation : the bolus has to work its way through the
small intestine and is also submitted to the friction with the intestinal wall. All these
friction effects are related to the “viscosity” of the bolus and we have two different ways
to model the friction term : either as a constant effect which is independent of the bolus
composition (models 1 & 2) or with a lubrification effect coming from the proportion
of water in the bolus (models 3 & 4). More specifically, in Equation (I.1) below, the
coefficient K(t) is either constant in models 1 & 2 or K̃/[W ] in models 3 & 4, where K̃
is a constant.

The final equation of transport is therefore the following one

d2 x d c0 + c1 V (t) dx
2
(t) = [y(t − x(t)/c)] − K(t) (t) (I.1)
dt dt a + bx(t) dt
dx
(0) = v0 , x(0) = 0
dt

4 Digestion

Digestion is a mechanical and chemical process by which the feedstuffs molecules

are broken down to the smaller ones by enzymes in order to to make them available for
absorption. The uptake of the obtained nutrients is mainly by absorption. In this section
the different steps of modeling are detailed.

4.1 Model 1

In this model the bolus is assumed to be completely solubilized (A=As ). We also

assume that the necessary enzymes for hydrolysis are mixed with the bolus in the stomach.
The product of following reaction is directly absorbable (B = Babs )
A + e → B.
e denotes the gastric enzymes. The first aim of this model is to define the variation of the
bolus which means the amount of the different substrates A, B and e at every time in the

44
4. DIGESTION

cylinder, and the second is to locate the bolus along the small intestine. We assume that,
the evolution of A or its volumic transformation depends on its mass at each moment and
the enzyme activity. This equation follows the law of mass action
dA
= −Ck(x, e)A
dt
where C denotes the degradation rate and, k(x, e) is the enzyme activity which depends
on the pH of the small intestine and the presence of the enzymes at each point along it.

The product B of the volumic transformation of A is absorbed by intestinal wall with

a constant rate kabs
dB
= Ck(x, e)A − kabs B.
dt
There are also the degradation and inactivation of the enzymes along the small intestine
de
= −ke e
dt
where ke is the rate of degradation of the enzymes which depends on their types. The
activity of each enzyme as a function of pH of small intestine is roughly known. We
know also the pH of each point along the small intestine. The composition of these two
functions gives the enzyme activity at each point x along it.

4.2 Model 2

In this second model, the presence of pancreatic enzymes in the small intestine as
well as the brush-border ones on its wall are considered. The pancreatic secretions help
neutralizing the stomach acid as they enter the small intestine. They also contain pan-
creatic enzymes. The level of the secretions is a function of volume and composition of
the bolus entered the small intestine. The brush border enzymes are the enzymes for the
terminal stage of digestion which is the surfacic hydrolysis. Contrary to the pancreatic
enzymes they are not free in the intestinal lumen, but rather, in the plasma membrane of
the enterocyte.

We assume that the bolus is completely solubilized. The product of the hydrolysis B
consists in Bint and Babs (B = Bint + Babs ).

The following scheme represents the chemical reactions of the bolus in this model

As → Bint , As → Babs , Bint → Babs .

The first reaction takes place inside the bolus by pancreatic and gastric enzymes, the
second and the third ones take place on the surface of the bolus.

45
4. DIGESTION

The degradation of A in this model is the result of the volumic hydrolysis of A as in

Model 1, and its surfacic hydrolysis by brush-border enzymes

dA A
= −Ck(x, e)A − Cabs (2πR`)
dt A + Bint + Babs
p A
= −Ck(x, e)A − 2Cabs πl/ρ ,
(A + Bint + Babs )1/2

the second term represents surfacic transformation of A to Babs . We recall that the mass
of the bolus in this model is

A(t) + Bint (t) + Babs (t) = ρV (t) = ρπR2 (t)l,

and therefore the lateral surface of the cylinder is given by

p
2πR` = 2 πl/ρ(A + Bint + Babs )1/2 .

This transformation depends on the fraction of A on the surface of the bolus which
A
is written by (2πR`) . The unit of the degradation coefficient per unit of
A + Bint + Babs
surface and time, Cabs , is g.m−2 .s−1 .

After a distance traveled by bolus of about 5% of the total length of the small intestine
which is approximatively 85 cm in an growing pigs, the input of secretions starts and it
stops after a distance of α meters traveled by bolus. We assume their mass is about β% of
the bolus mass. In the following equation, the effect of these secretions on the variation of
A is taken into account
dA 1 dx
= ... + ln(1.β) χ ((x(s) − 0.85) /α) A,
dt α dt
where χ is a localization function in the above equation which reflects the fact that secre-
tions arrive in the small segment of the intestine, say between 0.85 cm and 0.85 + α cm.

The product of volumic hydrolysis, Bint , participates in the creation of Babs on the
surface of the bolus. Therefore its variation is modeled by

dBint 1 dx
= Ck(x, e)A + ln(1.25) χ((x(s) − 0.85)/α)Bint
dt α dt
p Bint
−2Ciabs πl/ρ .
(A + Bint + Babs )1/2

46
4. DIGESTION

The absorbable nutrients on the bolus are not absorbed instantaneously [15]. In this model
we assume that the absorption rate follows Michaelis-Menten mechanism. The constant
kabs is the maximal rate of absorption at saturation, k is the Michaelis constant which is
half saturation
dBabs p Cabs A + Ciabs Bint Babs
= 2 πl/ρ 1/2
− kabs .
dt (A + Bint + Babs ) k + Babs

4.3 Model 3

In this model the ingested food consists in Ans , And , As and water (A = Ans + And +
As + W ). We incorporate two effects of water on digestion : the first one is the dilution
of the bolus and its impacts on degradation and absorption and the second one is the
lubrification and its consequences on the transport.

We assume that the evolution of As and Ans aims at reaching an equilibrium in which
the ratio between As and Ans is fixed and depends only on the proportion of water, namely
As = µ ([W ]) Ans stressing that solubilization of Ans depends on bolus dilution. From
the mathematical standpoint, we write this evolution as
 
dAns
= −ks µ ([W ]) Ans − As , (I.2)
dt
where µ is a linear function of water and the constant ks represents the return rate to
equilibrium.

The amount of water in the intestinal lumen is regulated by several complex biological
phenomena. In fact the proportion of water in the bolus aims at reaching [W0 ] in a rather
fast way which we translate it on a mathematical standpoint
d[W ] 1 dx
= −kw ([W ] − [W0 ]) + ln(1.β) χ((x(s) − 0.85)/α)[W ] (I.3)
dt α dt
where kw is large enough to reach the equilibrium in an adequate time. The second term
of above equation is the fraction of water in pancreatic secretions.

The variation of As depends on its degradation by volumic and surfacic hydrolysis,

and contribution of pancreatic secretions as in previous models. It also depends on the
equilibrium with Ans resulting from equation (I.2) which is the first term of equation be-
low
 
dAs
= ks µ ([W ]) Ans − As − Ck(x, e)As (t)
dt
p As
− 2Cabs πl/ρ [W ] (I.4)
(As + Ans + And + Bint + W + Babs )1/2
1 dx
+ ln(1.25) χ ((x(s) − 0.85) /α)As .
α dt
47
4. DIGESTION

The variation of absorbable nutrients depends on the creation of Babs by enzymatic hy-
drolysis of As and Bint and its absorption by intestinal wall

dBabs p Cabs A + Ciabs Bint

=2 πl/ρ [W ]
dt (As + Ans + And + Bint + W + Babs )1/2
Babs
− kabs . (I.5)
k + Babs

The non-degradable fraction of A, namely And , enters in the small intestine and leaves it
without any change in its structure.

As we already indicated in Section 5.1, lubrification of the bolus depends on the pre-
sence of water. For this model, the friction coefficient in equation (I.6) is written as

K̃
K(t) = .
[W ](t)

4.4 Model 4

This model is a mathematical simplification of the transport equation by means of

homogenization theory. Homogenization theory is concerned with equations with rapidly
oscillating coefficients and its aim is to provide an “homogenized” or “averaged” equa-
tion which is a limiting equation when the frequency of the oscillations tends to infinity.
Advantages of homogenization theory are clear : on most occasions, it is simpler to use
homogenized equations (for example to compute the solution) and, when the frequency
of oscillations is over a given value, this approximation of the real equation by the homo-
genized one may be rather accurate as seen in the next section.

Homogenization problems for ODEs were studied by [25] but it is worth pointing
out that our particular case does not fall into the theory described in [25]. Fortunately
the specific structure of the transport equation allows us to do a complete analysis of the
problem and even to compute explicitly the averaged equation.

More specifically, in the transport equation, pulses reach the bolus every 10 seconds
approximately. Compared to the time scale of digestion phenomena (the bolus stays in the
small intestine for several hours [13, 1]), this represents a very high frequency and causes
very rapid variations in the velocity of the bolus (see the velocity profile in Figure I.3).

We can prove mathematically that the pulses can be averaged out in an appropriate
way and we can replace the rapidly varying velocity by a slowly one.

48
5. RESULTS

In the simplest case, by normalizing the pulses, we assume that their mean effect over
a period is e(). Thus, over a time t = N , their mean effect is N e() = te()/. On the
other hand we assume
lim e()/ = τ
→0

the mean effect over a time t is therefore

lim N e() = te()/ = tτ.

→0

Inserting this equality in transport equation (I.1), the homogenized transport equation
reads

d2 x c0 + c1 V (t) K̃ dx
2
(t) = ā(t) − (t)
dt a + bx(t) [W ](t) dt
dx
(0) = v0 , x(0) = 0
dt
where ā is the averaged effect of the pulses. Its value is

1 dx
ā(t) := τ (1 − (t)).
c dt

5 Results

In the first part of this section, the graph of degradation of model 4, and the graph of
transport of model 3 and 4 are developed. The second part concerns the evaluation of the
last model by comparing its outputs with experimental data. Only a limited number of
outputs can be compared because of the lack of experimental data. However, the model is
evaluated in relation to our objective which is developing a mathematical model that takes
into account the physiology of the small intestine and process of digestion in it.

5.1 Digestion

The graph of digestion of model 4 is shown in figure I.2. We should at first initiate
the bolus composition. These initial conditions vary following the different types of feed-
stuffs. We fixed the initial value of Ans as three times that of As . We dilute Ans by two
times its volume water. Solubilized substrate As and non-solubilized one Ans reach a dy-
namic equilibrium all along the small intestine, as explained in Section 4. This balance
is reached rapidly at the beginning of the small intestine due to the large difference in
quantity between these two substrates. The result of this equilibrium is the increases of

49
5. RESULTS

the value of As and the decreases of the values of Ans , as seen in the graph of digestion.
The inverse process might take place by lack of water. The absorption curve corresponds
to the collected absorbable nutrients from x = 0 to x = x(t), where x indicates the lo-
cation of bolus in the small intestine. Obviously, the graph of the fourth model contains
more details about different steps of digestion than the first two graphs thanks to the mo-
del structures. The digestion graph of model 4 is similar that of model 3. The only change
in model 4 deals with the transport equation.

Digest ion vs dist ance

100
As
Babs
90 wat er
Ans

80

70
gram s of subst rat es

60

50

40

30

20

10

0
0 2 4 6 8 10 12 14 16 18
dist ance as m

F IGURE I.2: Digestion through Model 4

5.2 Velocity

Figure I.3 provides a numerical evidence of the homogenization phenomena. The

graph of transport resulting from models 3 and 4 is shown in figure I.3. The effect of
pulses on the curve of velocity is obvious. However, using the homogenization theory in
Model 4, we obtain a smooth graph of velocity which replaced that of Model 3.

50
5. RESULTS

F IGURE I.3: Velocity of the bolus versus Time

5.3 Model evaluation

For a specific family of nutrients, here starch, digestion is calculated using Model 4
and is compared to data reported by [1]. To parameterize adequately the model, we adapt
the enzyme activity of the last model to the activity of amylase in the small intestine. Amy-
lase is the enzyme required for degradation of starch. The optimal activity of pancreatic
amylase is in neutral pH [5].

Pancreatic secretions have no impact on the variation of As since there is no starch

from this source.

The inputs of model are only Ans and W which are Starch and Water. The outputs
are the values of these substrates at the end of ileum. The data in the article of [1] are for
purified protein free wheat starch, agreeing with our hypothesis for the composition of the
bolus 1. The outputs concerns the collected data after at the end of ileum.

51
5. RESULTS

TABLE I.1: Digestion of starch in pigs by modeling : Comparison between simulated and experi-
mental data by [1]

Experimentation Modeling
Input(g) Output(%) Input(g) Output(%)
wet digesta 2571 8 113.10 5.33
dry matter 688 0.50(g) 37.70 0.04

Regarding the data presented table I.1, percentages of dry matter and wet digesta collec-
ted at the end of the ileum are approximatively the same as the output of model 4. The
difference between inputs is due to the simulation calibration which takes into account
only one bolus i.e. a fraction of the daily meal. However, differences between outputs are
low in percentage enabling to conclude that the model can roughly simulate very simple
situations.

5.4 Sensitivity analysis

Sensitivity analysis is performed to identify the key parameters affecting the digestion
process. The chosen parameters are set at 5% and 50% of their original values.

Output Parameters
As C, Cabs
Babs Cabs , Ciabs , kabs
v a,b, c0 , c1 , K

Studied digestion parameters are C and Cabs for degradation of As , and Cabs , Ciabs
and kabs for the absorption of Babs .

If y is the output and θ the parameter, the relative variation of y can be expressed as
follows
|yθ − yθ+∆θ |
.
yθ

a. Influence on As

Both parameters C and Cabs are overestimated by 5 and 50%. The figure I.4 shows
the relative variation of As at each moment. The relative variation of As resulting from 5
and 50% values of C is not meaningful. The parameter Cabs has the largest effect on As
degradation. Observing the graph of relative variation of As , figure I.4, we conclude that
increasing the value of Cabs increases the relative variation value with time. .

52
6. CONCLUSION AND PERSPECTIVES

Variat ion of 5% Variat ion of 50%

100 100
C C
Cabs Cabs

80 80

Relat ive sensit ivit y(%)

relat ive sensit ivit y(%)

60 60

40 40

20 20

0 0

-20 -20
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Dist ance(m ) Dist ance(m )

F IGURE I.4: Relative variation of As regarding to C, Cabs .

b. Influence on Babs

The parameters kt , Cabs and Ciabs are overestimated by 5 and 50%. The figure I.5
shows the relative variation of Babs by time. The quantity of Bint being very small in the
model, the effect of changing the parameter Ciabs is neglectible on the relative variation of
Babs by time. The quantity Babs is very sensitive to the variation of parameter Cabs firstly
because of the high quantity of As , then its influence decreases because of decreasing
quantity of As over time. The quantity Babs is dependent on kt because of the large impact
of kt on the nutrient absorption rate.

6 Conclusion and Perspectives

This model is obtained from simplified biological assumptions and it can be used to
illustrate generically the rate of degradation and absorption all along the small intestine.
This is a global model of digestion of a bolus composed of one substrate and water. This
section is devoted to a discussion on the current state of our modeling, our assumptions,

53
6. CONCLUSION AND PERSPECTIVES

Variat ion of 5% Variat ion of 5%

100 100
Cabs Cabs
Ciabs Ciabs
kt kt
80 80
Relat ive sensit ivit y(%)

Relat ive sensit ivit y(%)

60 60

40 40

20 20

0 0

-20 -20
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Dist ance(m ) Dist ance(m )

F IGURE I.5: Relative variation of Babs regarding to Cabs , Ciabs and kt

difficulties and on the future development of the model.

The first assumption to be discussed is the “cylinder” one. As mentioned earlier, it

was introduced for technical reasons. Our aim was to solve a partial differential equation
with very different scales of times (pulses arising every 10 seconds while whole digestion
process in the small intestine lasts for several hours [13, 1]) and with an highly variable
domain with contrasted scales (few centimeters for the bolus compared to the 18 meters
of the small intestine). Solving this PDE seems unreasonable since it was leading to the
usual diffusive phenomena and large errors. We also notice that these 18 meters of the
small intestine were rather empty and therefore we were spending a lot of time to compute
functions which were very often 0.

From the transport standpoint, the “cylinder” assumption can be seen as a Lagrangian
method, the ordinary differential equations on x(·) being (essentially) the characteristic
curves of the transport equation. This is the first justification of this hypothesis, the second
being the direct observation of animals bolus which convinces us that it can be represented
as a cylinder, even if its geometrical characteristics could be more complicated. However
we have to work more on the evolution of the length of the cylinder.

54
7. ACKNOWLEDGEMENT

A more Eulerian “compartmental approach” is studied simultaneously but we are still

facing difficulties for modeling the transport and, specifically to reproduce some features
of the “cylinder” model.

The transport equation seems to take into account rather closely the phenomena which
c0 + c1 A
are described by the experts . It will be difficult to validate the term and to have
a + bx
a precise idea of the value of the different constants but such a modeling seems more
appropriate than trying to use a complicated fluid mechanics approach whose laws may
not be valid in this very confined domain. The same remarks hold for the effects of the
water : it seems correct even if a relevant validation will be difficult.

For food digestion and absorption, we are only at a first stage of modeling. The ab-
sorption phenomena were not studied explicitly leading to required further development
with a focus on the assumed interactions between the animal physiological status and
absorption. The spatial aspects (location of the absorption) were clearly neglected so far.

For digestion, the next step will be to mix different nutrients and adapt the enzyme
breakdown to each of them. We have also to examine more closely the respective effects
of the different categories of enzymes together with the role of the water. Moreover inter-
actions between nutrients on the digestion processes should be questioned.

As a conclusion of this first stage of modeling, consistent behaviors of the model were
reached. Moreover, the simplicity of the current model allows easy developments in any
directions. Our next target will therefore be to iterate the model development according to
the above proposed research areas.

7 Acknowledgement

The multidisciplinary collaboration on this research project between the INRA Center
of Nouzilly and the Laboratoire de Mathématiques et Physique Théorique was initiated
within and supported by the CaSciModOT program (CAlcul SCIentifique et MODélisa-
tion des universités d’Orléans et de Tours) which is now a Cluster of the french Region
Centre. This collaboration also takes place in a CNRS-INRA PEPS program “Compré-
hension et Modélisation du devenir de l’aliment dans le tube digestif“. This work is part
of the PhD thesis of Masoomeh Taghipoor, financed by CNRS and INRA.

55
Chapitre II

Mathematical Homogenization in the

Modelling of Digestion in the Small
Intestine

Digestion in the small intestine is the result of complex mechanical and biological
phenomena which can be modelled at different scales. In a previous article, we introduced
a system of ordinary differential equations for describing the transport and degradation-
absorption processes during the digestion. The present article sustains this simplified mo-
del by showing that it can be seen as a macroscopic version of more realistic models
including biological phenomena at lower scales. In other words, our simplified model can
be considered as a limit of more realistic ones by averaging-homogenization methods on
biological processes representation. 1

1. Ce chapitre a fait l’objet d’une publication au journal de MathSinAction [26] :

Mathematical Homogenization in the Modelling of Digestion in the Small Intestine,
Masoomeh Taghipoor, Guy Barles, Jean-René Licois, Christine Georgelin, Philippe Lescoat.

57
1 Introduction

When building a model for digestion in the small intestine, difficulties occur. The first
one is the extreme complexity of the mechanical/biological phenomena. Transport of the
bolus through the peristaltic waves, feedstuffs degradation by numerous enzymatic reac-
tions and the active/passive absorption of the nutrients by the intestinal wall are known
to be the key steps but they are not biologically nor fully understood and neither quanti-
tavely parameterized. Modelling approaches are a way to integrate complex mechanisms
representation of these phenomena helping to improve our understanding of them. Since
it is almost impossible to build direct experiments for studying the digestion in the small
intestine, modelling is a way to test in silico hypotheses that could be challenged through
limited in vivo experiments.

A second difficulty relies on the complex environment within the digestive tract. For
example, the intestinal wall plays a key role in the transfer of the digested food in the blood
and interferes in the degradation of the bolus via the brush-border enzymes and causes the
transit of the bolus by transmitting the pulses coming from the peristaltic waves.

Thirdly, digestion in the small intestine has contrasted but relevant macroscopic and
microscopic scales, both in space and time. To give few figures, the length of the small
intestine in a growing pig reaches 18 meters, which is a large figure compared to its radius
(2-3 centimeters) and even more compared to the size of the villi (around 1 millimeter).
In the same way, the bolus stays in the small intestine for several hours, while the efficient
peristaltic waves which ensure the transport of the bolus, start approximatively every 12
seconds from the pylorus.

Because of these different scales, a model based on partial differential equations and
capturing all the interesting phenomena, would be too complicated and impossible to solve
numerically. Therefore we have adopted in [8] a model based on ordinary differential
equations (ode in short) : each bolus of feedstuffs coming from the stomach is identified
as a cylinder and the odes describe the evolution of the position and composition of the
cylinder. Since digestion could be described by a transport equation (or a system of such
equations) with reacting terms, our strategy was essentially to use the Characteristics of
this equation. At least numerically this type of Lagrangian method appears to be more
efficient. We refer to [8] for details on our different models since several stages of the
modelling process were developped in this paper.

The aim of the present article is to provide mathematical justifications of some as-
sumptions of the modelling presented in [8]. We focus on the bolus transport and on the
effects related to absorption and enzymatic breakdown by the brush border enzymes, phe-
nomena which are related to averaging/homogenization type processes.

More precisely, in Section 2, we examine the effects of the pulses generated by the
1. INTRODUCTION

peristaltic waves. Considering that the time scale for these pulses is small compared to the
duration of the digestion i.e. that their frequency is high, we rigorously establish that their
effect is the same as the one of a constant driving force. This result is biologically very
interesting since it allows to get rid of this very small time scale and to do the numeri-
cal computations in a much more efficient way opening ways to alternative experimental
approaches on digestive tract studies. Related and more general results on the homogeni-
zation of odes can be found in L. C. Piccinini[25] but we point out that our case does not
fall into the scope of [25].

In Section 3, we consider the complex phenomena related to the villi and micro-villi :
the active/passive absorption by the intestinal wall and the brush border enzymatic reac-
tions. In order to study these phenomena, we introduce a 3-d model where we focus on
the boundary effects. As a consequence, the other phenomena are highly simplified. The
lumen of the small intestine is modelled as a cylindrical type, periodic domain whose axis
is Re1 , where e1 := (1, 0, 0). In order to model the villi, this domain has an highly oscilla-
tory boundary of order ε−1 while its radius is of order ε. In this domain, we have a system
of parabolic, transport-diffusion equations with oscillatory coefficients for the absorbable
and non-absorbable nutrients. The key feature is the Neumann boundary condition which
describes the phenomena on the intestinal wall : the effects of the brush-border enzymes
together with the active-passive absorption.

Using homogenization method, we prove that, when ε tends to 0, this problem converges
to a 1-d system of transport-reaction equation. The key issue is to show how the effects
of the diffusion and the degradation- absorption on the highly oscillatory boundary are
combined in order to produce the final reaction terms. For the readers convenience, we
provide both a formal and a rigorous proof of this result. The formal proof gives rather
explicit formulas which can easily be interpreted from the biological point of view. Mo-
reover we point out that, even if we are using a very simplified framework, we show that
it captures the key features of the absorption process.

The homogenization methods used in Section 3 are based on viscosity solutions’

theory and in particular the “perturbed test function method” of L. C. Evans [27, 28] :
we refer to [29] and references therein for the applications of such methods for problems
with Neumann boundary conditions and oscillatory boundary. To our knowledge, it is the
first time that such methods are used to obtain a convergence of a 3-d problem to a 1-d
problem.

Acknowledgement. The multidisciplinary collaboration on this research project between

the INRA Center of Nouzilly and the Laboratoire de Mathématiques et Physique Théo-
rique was initiated within and supported by the CaSciModOT program (CAlcul SCIenti-
fique et MODélisation des universités d’Orléans et de Tours) which is now a Cluster of the
french Region Centre. This collaboration also takes place in a CNRS-INRA PEPS pro-

60
2. TRANSPORT EQUATION

gram “Compréhension et Modélisation du devenir de l’aliment dans le tube digestif“. This

work is part of the PhD thesis of Masoomeh Taghipoor, financed by CNRS and INRA.

2 Transport Equation

Peristalsis is the phenomenon in which a progressive wave of contraction or expansion

(or both) propagates along a tube [30, 16, 31, 8]. The peristaltic waves are responsible for
the fluid dynamics of the contents of the small intestine and can be divided into segmen-
tation and propulsive contractions. The segmentation motion are responsible for mixing
the bolus. Propulsive contractions are responsible for transporting the bolus through the
small intestine. The effective peristaltic waves generated in the pylorus reach the bolus
approximatively every 12 seconds. This is very small compared to the time scale of di-
gestion phenomena which lasts several hours. This causes the observation of very rapid
variations in the velocity of the bolus.

In [8], the authors present a first simplified model of bolus transport along the small
intestine. We use Homogenization Theory to simplify this equation to replace the per-
iodically oscillating velocity by an averaged one (II.1). This section provides a rigorous
mathematical justification of this transport equation.

2.1 Position of the problem

In this section, we formulate a simplified version of the transport problem. The small
intestine is represented by the interval [0, +∞) and the position of the bolus at time t is
given by x(t) ∈ [0, +∞). Roughly speaking, x(t) is the distance between the center of
bolus and the pylorus.

The bolus is composed of different types of nutrients, say nutrients

1, 2, · · · , K and the quantity of nutrient i at time t is denoted by yi (t) for i = 1, 2, · · · , K
and we set y(t) := (y1 (t), y2 (t), · · · , yK (t)). The variation of the different yi depends on
its production and degradation rate which is summarized through the equation
ẏ(t) = d(x(t), y(t)) , (II.1)
where d := (d1 , d2 , ...dn ) with di : [0, +∞) × RK → [0, +∞) a Lipschitz continuous
function. Since we are mainly interested in the transport equation in this section, this
simple equation is written to fix ideas but also because the transport equation will depend
on the composition of the bolus y.

The peristaltic waves are created at the pylorus and they travel along the intestinal wall
at a quasi-constant velocity : the average wave velocity of each peristaltic wave is about

61
2. TRANSPORT EQUATION

5
Velocit y (m /h)

1
0 1 2 3 4 5 6 7

Tim e (hour)

F IGURE II.1: Periodically oscillating velocity and averaged one.

c ' 7.2m/h. These waves are periodic of period denoted by   1 and to model them, we
can say that at time t, an electric signal of size ψ(t/) starts from the pylorus and reaches
a point x of the small intestine at time t + x/c. Here we assume that ψ(s) ≡ 0 if s ≤ 0
and on [0, +∞), ψ is the restriction of a smooth, 1-periodic function on R.

At time t, the bolus is at the position x(t) and is reached by the wave generated at
time s = t − x(t)/c whose intensity is ψ(s/). we assume moreover that the impact of
this pulse on acceleration of the bolus is given by a smooth, positive function g (s, v, x, y)
where, as above, s is the time when the pulse was generated, v is the relative velocity of a
pulse with respect to the bolus velocity (v = (c − ẋ(s))/c ), x is the position of the bolus
and y its composition.

Indeed, according to [32] and [24] the efficiency of the peristaltic waves increases
with the size of the bolus which is roughly speaking the sum of the yi for 1 ≤ i ≤ K, and
decreases with the distance from pylorus x(t).

The function g is also -periodic in s, we emphasize this fact by writing

g (s, v, x, y) = g(s/, v, x, y) ,

62
2. TRANSPORT EQUATION

where g(s, v, x, y) is a smooth function which is 1-periodic in s for s > 0.

Taking into account the friction inside the small intestine as in [8] through a −k(t)ẋ(t)-
term where k(t) > 0 for any t, the equation for the transport of the bolus reads
ẍ(t) = g −1 (t − x(t)/c), 1 − ẋ(t)/c, x(t), y(t) − k(t)ẋ(t)


(II.2)
with x(0) = 0 and ẋ(0) = v0 where v0 < c.

Having in mind the example of a water wave in a channel, if the bolus velocity is the
same or is close to the wave one, then obviously the peristaltic wave will have either no
effect or at least a small effect on the bolus velocity. Translated in term of g, this means
that g(t, 0, x, y) = 0 and even g(t, v, x, y) = 0 if v ≤ 0. Thus there exist a smooth function
g̃ : R × R × [0, +∞) × RK → [0, +∞) such that g(s, v, x, y) = g̃(s, v, x, y)v. We notice
that, since we assume g to be positive, then g̃(s, v, x, y) ≥ 0 if v ≥ 0 while we have
g̃(s, v, x, y) ≡ 0 if v ≤ 0 .

Because of the dependence of Equation (II.2) on , we denote their solutions by x , y 

and our aim is to study the behavior of these solutions as  tends to 0, and in particular the
behavior of x .

2.2 The asymptotic behavior

We first rewrite the equation satisfied by x , y  . For Equation (II.2), we have

ẍ = (1 − ẋ /c)g̃(−1 (t − x /c), 1 − ẋ /c, x , y  ) − k(t)ẋ (II.3)
while Equation (II.1) reads
ẏ  = d(x , y  ). (II.4)
The initial conditions are
x (0) = 0, ẋ (0) = v0 , y  (0) = y0 , (II.5)
where v0 < c because of physiological reasons.

In order to formulate our result, we introduce the function F (t, V, X, Y ) given by

Z t
F (t, V, X, Y ) = g̃(s, V, X, Y )ds (II.6)
0

and, recalling that g̃(s, V, X, Y ) is 1-periodic for s ≥ 0, we denote the averaged of F over
a period by F̄ (V, X, Y ). Of course we have
Z 1
F̄ (V, X, Y ) = F (1, V, X, Y ) = g̃(s, V, X, Y )ds,
0

and F, F̄ are smooth functions since g̃ is a smooth function.

63
2. TRANSPORT EQUATION

Example 2.1. In [8], the authors introduce the the function g as

c0 + c1 y
g(s, v, x, y) := ψ̇(t − x/c) · v ,
a + bx
for the real non-negative values c0 , c1 , a and b. Where y = Σni=1 yi .

Our main result is the

Theorem 2.1. Let (x , y  ) the unique solution of equations (II.3)-(II.4)-(II.5), then the se-
quence (x , y  )>0 converges strongly in C 1 ([0, T ], [0, +∞)) to (x, y) the unique solution
of the averaged system of equations

c − ẋ(t)
ẍ(t) = F̄ (1 − ẋ(t)/c, x(t), y(t)) − k(t)ẋ(t) (II.7)
c
ẏ(t) =d(x(t), y(t))

with the initial conditions

x(0) = 0, ẋ(0) = v0 , y(0) = y0 . (II.8)

The key interpretation of this result is the following : the effect of frequent pulses on
the transport of the bolus is the same as the one obtained through an averaged constant
signal.

Proof of Theorem 2.1. We prove it in two steps : first we obtain various estimates sho-
wing that the sequences (x , y  )>0 converge strongly in C 1 (at least along subsequences)
and, then, in the second step, we prove that they converge to the unique solution of the
averaged system (II.7) (which will imply that the whole sequence converges by a standard
compactness argument).

The following lemma provides a proof of convergence of x and y  .

Lemma 2.1. Let (x , y  )>0 the unique solution of (II.3)-(II.4)-(II.5). Then x , y  are uni-
formly bounded in C 2 and therefore there exists a subsequence which is converging stron-
gly in C 1 and such that ẍ is converging in L∞ weak-∗.

Proof of Lemma. We first prove that ẋ (t) ≤ c. To this aim, we define the positive function
φ(t) as follows
(
+ ẋ (t) − c if ẋ (t) − c > 0
φ(t) = (ẋ (t) − c) = (II.9)
0 otherwise,

64
2. TRANSPORT EQUATION

then multiply the both sides of equation (II.3) by φ(t)

c − ẋ tc − x
 
  +

+
ẍ (ẋ (t)−c) = g̃ , 1 − ẋ /c, x , y (ẋ (t)−c)+ −k(t)ẋ ẋ (t) − c .
  
c c
The right-hand side of this equation is negative since g̃(s, V, X, Y ) is non negative if
Y ≥ 0, and k(t) > 0, therefore

ẍ (t)(ẋ (t) − c)+ ≤ 0

which is equivalent to
1d 2
(φ (t)) ≤ 0 .
2 dt
The function φ2 (t) is therefore a decreasing function. Furthermore, since v0 < c, we have

φ2 (t) ≤ φ2 (0) = [(v0 − c)+ ]2 = 0 ,

which yields the result.

Using the same method with (ẋ)− = max(−ẋ, 0), we can prove that ẋ is a non-
negative function.

Gathering these information, we obtain that, for any t

0 ≤ x (t) ≤ ct , 0 ≤ ẋ (t) ≤ c ,

and therefore the sequence (x )>0 is uniformly bounded and equicontinuous on [0, T ].
Using these informations and the equations for the y  , we also see that the y  are also
uniformly bounded in C 1 (and even in C 2 ) and coming back to the x equation we see
also that the x are also uniformly bounded in C 2 .

Consequently the Arzela-Ascoli compactness criterion ensures that there exists a sub-
sequence (xj , y j ) which converges in C 1 . Moreover, since ẍ is bounded in L∞ , we can
also extract a subsequence such that ẍj converges in the L∞ weak-∗ topology.

We return now to the proof of Theorem 2.1. To simplify the exposure, we still denote
by (x , y  ) the converging subsequence (xj , xj ) and we denote by (x, y) the limit. By
inserting the Definition (II.6) into Equation (II.3), we get
∂F −1
ẍ (t) = (1 − ẋ /c) ( (t − x /c), 1 − ẋ /c, x , y  ) − k(t)ẋ (t)
∂t
therefore, using the notation v  = 1 − ẋ /c and dropping most of the variables to simplify
the expressions, we have
d  ∂F ∂F ∂F
ẍ (t) =  F (−1 (s − x /c), v  , x , y  ) − v˙ − ẋ − ẏ  − k ẋ

dt ∂V ∂X ∂Y
65
2. TRANSPORT EQUATION

and then integrate the both sides of equation over [0, t]

Z t Z t
d

ẍ ds =  (F (−1 (s − x /c), v  , x , y  ))ds
0 0 ds
Z t Z t
∂F  ∂F  ∂F
−  (v˙  + ẋ + ẏ )ds − k ẋ ds
0 ∂V ∂X ∂Y 0

which leads to
ẋ (t) − v0 = F (−1 (t − x /c), 1 − ẋ /c, x , y  )
Z t Z t
∂F  ∂F  ∂F
−  (v ˙ + ẋ + ẏ )ds − k ẋ ds
0 ∂V ∂X ∂Y 0

since F (0, V, X, Y ) = 0 for any V, X ∈ R and Y ∈ RK .

Now we have to let  tend to 0. First, since g̃ is periodic, it is standard to prove that
F (−1 t, V, X, Y ) → F̄ (V, X, Y )t ,
locally uniformly. Especially, it is easy to see that if n ≤ −1 t < n + 1, therefore
F (−1 t, V, X, Y ) → [nF̄ (V, X, Y ) + O(1)].

In the same way, because of the definition of F and the regularity properties of g̃, for
ξ = V, X, Y we also have
∂F −1 ∂ F̄
 ( t, V, X, Y ) → (V, X, Y )t locally uniformly.
∂ξ ∂ξ
As a consequence, since x and y  are converging respectively to x and y in C 1 , we have
also
F (−1 (s − x (s)/c), v  (s), x (s), y  (s)) → F̄ (v(s), x(s), y(s))(s − x(s)/c),
∂F
uniformly on [0, T ], where v = 1 − ẋ/c. And the same is true, replacing F by and F̄
∂ξ
∂ F̄
by .
∂ξ
From these properties, it is easy to deduce that
Z t Z t
 ∂F  ∂F ∂ F̄ ∂ F̄
 (ẋ + ẏ )ds → (s − x/c)(ẋ + ẏ )ds as  → 0 ,
0 ∂X ∂Y 0 ∂X ∂Y
for any t ∈ [0, T ].

On the other hand, v˙ = −ẍ /c converges in the L∞ weak-∗ topology to v̇ and there-
fore Z t Z t
∂F ∂ F̄
 v˙ ds → (s − x/c)v̇ ds ,
0 ∂V 0 ∂V

66
3. ON THE EFFECTS OF INTESTINAL VILLI

for any t ∈ [0, T ].

Gathering all these informations, we finally obtain

ẋ(t) − v0 =(t − x(t)/c)F̄ (1 − ẋ(t)/c, x(t), y(t))
Z t
∂ F̄ ∂F ∂F
− (s − x/c)(v̇ ds + ẋ + ẏ )ds
0 ∂V ∂X ∂Y
Z t
− k(s)ẋ(s)ds (II.10)
0

The right-hand side being C 1 in t, we deduce that x is a C 2 -function and derivating

the both side of (II.10), we have the equation
ẍ(t) = (1 − ẋ(t)/c)F̄ (1 − ẋ(t)/c, x(t), y(t)) − k(t)ẋ(t).

3 On the Effects of Intestinal Villi

As mentioned in the introduction, in the 1-d model of digestion presented in [8], we

take into account the absorption effect by a simple absorption term through a Michaelis-
Menten type nonlinearity and this can be assumed unreasonable when compared to the
complexity of the involved phenomena. The same can be said for the enzymatic break-
down by the brush-border enzymes. Consequently the first aim of this section consists in
giving some rigorous justification of these choices.

Our effort is therefore to find an appropriate system of equations describing the dif-
ferent effects of the structure and the spatial distribution of intestinal villi on these key
phenomena of digestion. Therefore we introduce a 3-d toy model which takes into ac-
count the complex geometry of the small intestine as well as all these boundary effects,
but this implies unavoidable simplifications on the transport process.

We start by a short presentation of the small intestine anatomy followed by introducing

the three dimensional toy model of digestion.

A large number of villi and micro-villi are present on the surface of the small intes-
tine. Their role is to enlarge the digestive and absorptive area in the small intestine. They
increase the area of the small intestine at least 500 times ([31]). The absorptive surface of
the villi contains the brush border enzymes which are responsible of the final step of de-
gradation (surfacic degradation) for some nutrients. This increase is therefore a key issue
in the process of nutrients degradation and absorption.([33]).

67
3. ON THE EFFECTS OF INTESTINAL VILLI

These finger like villi are covered by epithelial cells. They consist of absorptive, goblet
and entero-endocrine cells. The epithelial cells are produced in crypts, they migrate and
become mature from the crypts to the tips of the villi([34]). More precisely, the absorption
rate is also proportional to the distance of each of the villi from its tip.

A microscopic observation of the small intestine surface is necessary in order to give

realistic absorption and degradation shapes. The spatial aspect of absorption related to
the distribution of villi and their absorption capacity is often neglected in modelling of
digestion. In these models, absorption and degradation are modelled by a constant rate or
a Michaelis-Menten process (see [8], Logan [15], ...).

As shown in figure (II.2) the size of the period is small compared to the size of the
unfold small intestine which is around 18 meters. We consider, for the sake of simplicity,
that the villi are distributed periodically in the inner surface of the small intestine.

F IGURE II.2: The different scales on intestinal anatomy relevant to our model.

In this section, we seek a macroscopic description of digestion in the small intestine

by taking into account all the effects of the presence of the villi in microscopic scale. Our
approach is based on an asymptotic analysis, as  goes to zero. The absorption rate of the
limit problem is said to be the homogenized absorption rate.

68
3. ON THE EFFECTS OF INTESTINAL VILLI

3.1 Position of problem

The small intestine is assumed to be an axisymmetric cylindrical tube with a rapidly

varying cross section. In order to describe it, we first introduce an axisymmetric, smooth
domain Ω which is confined in a cylinder of radius r > 0. More precisely, we assume

{(x1 , x2 , x3 ) ∈ R3 | x2 = x3 = 0} ⊂ Ω ⊂ {(x1 , x2 , x3 ) ∈ R3 | x22 + x23 = r2 } .

In addition, we assume that Ω is periodic in the x1 -direction (say 1-periodic), namely

(x1 + 1, x2 , x3 ) ∈ Ω if (x1 , x2 , x3 ) ∈ Ω.

The small intestine is represented, for some 0 <   1 by the domain Ω given by

Ω = Ω ∩ {x1 ≥ 0} (II.11)

The figure II.3 is a simple representation of this domain :

microvilli

ε
rε villi

x1

Ωε

F IGURE II.3: A simple example of the domain Ω . The oscillations on the villi represents the
microvilli.

In this definition, the small intestine has an infinite length. However this assumption
is not a real restriction, since we focus on the local absorption-degradation processes.
Moreover, the x1 = 0 part of the boundary corresponds to the pylorus and ∂Ω to the
villi. It is worth pointing out that Ω is -periodic in the x1 direction, the parameter 
characterizes the distance between the villi and thus it is natural to assume it to be very
small.

69
3. ON THE EFFECTS OF INTESTINAL VILLI

A simple example of Ω in cylindrical coordinates, can be the following

Ω = {(r, θ, z) s.t | r |≤  + ψ(z/, θ), z ≥ 0, θ ∈ [0, 2π]}

where z plays here the role of x1 and ψ(z, θ) is a 1-periodic function of z.

We introduce two functions u , v  : Ω × [0, T ] → R for describing the evolution of

the concentration of feedstuffs in the small intestine. For x ∈ Ω and t ∈ [0, T ], v  (x, t)
denotes the concentration of the large feedstuffs molecules which are transformed into
absorbable nutrients after different enzymatic reactions. The quantity u (x, t) represents
the concentration of produced nutrients at position x at each time t.

The evolution of substrates u and v  in the intestinal lumen is due to (i) their dif-
fusion by Fick’s law, (ii) their propagation through intestinal lumen by a given velocity
coming from the peristaltic waves and (iii) the enzymatic reactions which transform v 
to u both inside the intestinal lumen but also on the intestinal wall by the brush-border
enzymes. When these reactions take place in the intestinal lumen, we call them volumic
transformation, while we talk about surfacic transformation when they take place on the
villi.

The rate of the volumic reactions depends on the concentration of feedstuffs and also
enzymes activity at time t and at x, namely ζ(x, t), where ζ : [0, ∞) × [0, T ] → R is
a continuous, positive and bounded function. There is a limitation in the transformation
which is described by ϕ : R → R, which is a bounded, increasing and Lipschitz conti-
nuous function such that ϕ(s) = 0 if s ≤ 0. These assumptions on ζ and ϕ are denoted
by (T1) in the sequel.

Taking into account the three above-mentioned phenomena, the equation for the evo-
lution of concentration of the non-absorbable feedstuffs molecules in the intestinal lumen
reads
∂v 
= ω ∆v  − c(x1 , x/, t)Dv  − ζ(x1 , t)ϕ(v  ) in Ω × (0, T ) (II.12)
∂t
while for the absorbable nutrients, we have

∂u
= χ ∆u − c(x1 , x/, t)Du + ζ(x1 , t)ϕ(v  ) in Ω × (0, T ). (II.13)
∂t

The first terms of the right-hand-sides of the above equations, where ∆ denotes the
usual Laplacian 2 , are diffusion terms. The diffusion coefficients of large molecules of
feedstuffs and small molecules of nutrients are denoted by ω and χ respectively.
∂2φ ∂2φ ∂2φ
2. If φ is a smooth function, ∆φ = + +
∂x21 ∂x22 ∂x23

70
3. ON THE EFFECTS OF INTESTINAL VILLI

It is shown that, for a fixed temperature, the diffusion coefficient d is inversely propor-
tional to the molecular weight, to be more precise for a spherical molecule we have
kT ρ 1/3
d= ( )
3µ 6πM
in which k is Boltzmann constant, T is the intestinal temperature, µ the viscosity of the
the intestinal liquid, ρ the molecule density and M the molecular mass. For fixed T and
µ, this constant is very small because of the very small value of kM −1/3 [33].

For reasons explained in section 3.3, we assume that

ω := ω and χ := χ ,

for some constants ω, χ > 0. Since the nutrients molecules are smaller than feedstuffs
particles, we also have ω ≤ χ. The second terms of the right-hand sides are transport
terms. The C 1 -function c : [0, +∞) × Ω × [0, T ] → R3 is modelling the velocity of
substrates which comes from the peristaltic waves. The effect of the peristaltic waves
is known to depend on the position in the small intestine and on time, this justifies the
dependence of c(x1 , X, t) on x1 and t, while the dependence on X = (X1 , X2 , X3 ) takes
into account the local effects at a lower scale.

A priori the diffusion of bolus is small compared to its velocity through the small
intestine and therefore ω and χ are expected to be smaller than c(x1 , X, t).

We assume the function c to satisfy the following properties :

(C1) The function c(x1 , X, t) is a Lipschitz continuous function which is 1-periodic in X1
and, if e1 = (1, 0, 0), then, for any x1 ∈ [0, +∞), X ∈ Ω and t ∈ [0, T ],
Z
c(x1 , X, t) · e1 ≥ 0 and c(x1 , X, t).e1 dX > 0 ,
P

where P is a period in Ω, say P := {X ∈ Ω ; 0 ≤ X1 ≤ 1}.

In addition to the regularity properties of c, this assumption means that the effect of
the peristaltic waves is to move ahead the bolus in the small intestine.
(C2) For any x1 ∈ [0, +∞), X ∈ Ω and t ∈ [0, T ], divX (c) = 0 where divX denotes the
divergence operator in the X-variable only.
This second assumption is justified by the incompressibility of the bolus at the microsco-
pic level.
(C3) For any x1 ∈ [0, +∞), X ∈ ∂Ω and t ∈ [0, T ], c(x1 , X, t) · N (X) = 0, where N (X)
denotes the outward, unit normal to ∂Ω at X.

This last assumption means that the velocity vector is always tangent to the boundary.
It is worth pointing out that, if X = x/ then N (X) = n(x), therefore it is true both for X

71
3. ON THE EFFECTS OF INTESTINAL VILLI

in Ω and for x in Ω . As a consequence of this property, the nutrients reach the boundary
only because of the diffusion effects.

Once they reach the boundary, the large particles of feedstuffs can change of chemi-
cal structure because of the presence of brush-border enzymes. As we already mentioned
above, this effect is called surfacic degradation of feedstuffs and the result is the pro-
duction of the smaller absorbable molecules of nutrients u . We assume moreover that a
portion 0 ≤ β < 1 of these nutrients is absorbed instantaneously while the remaining part
(α := 1 − β) diffuses in the small intestine. The surfacic degradation is modelled by the
Neumann boundary condition
∂v 
ω = −%(x1 , X)v  on ∂Ω × (0, T ) (II.14)
∂n
where, % is a continuous, positive and X1 -periodic function which represents the rate of
surfacic degradation.

On the boundary of the small intestine, there are two main effects for the nutrients u .
We already describe the first one which is a production of nutrients by the surfacic de-
gradation. The second one is the active and passive absorption of nutrients, namely their
transport across the intestinal wall to the blood circulation. An active process requires the
expenditure of energy, while a passive process results from the inherent, random move-
ment of molecules [33]. These different categories of absorption as well as the production
of u from v  on the boundary construct the boundary condition of Equation (II.12)
∂u α
χ = −ηp (x1 , x/)u − ηa (x1 , x/, t)ga (u ) + %(x1 , x/)v  . (II.15)
∂n ω

The functions ηp and ηa denote respectively the passive and active absorption rates.
Both of them depend on the global position in the small intestine x1 and the local one
x/, by which we take into account the effect of the special physiology of the villi on the
absorption rate which has been described at the beginning of this section. The dependence
in time in the active absorption ηa , describes the presence of energy at time t. The function
ga governs the active absorption and depends on the nutrients categories. Typically, it is
assumed to be the Michaelis Menten and therefore, it is a bounded, continuous, increasing
function.

We formulate the key assumptions on the functions ηp , ηa and ga

(T2) The functions ηp (x1 , X), ηa (x1 , X, t) are bounded, continuous, positive, 1-periodic
functions in X1 , and the function ga is a bounded, continuous, increasing function with
ga (s) = 0 if s ≤ 0. Moreover, there exists η > 0 such that ηp (x1 , X) ≥ η for any
x1 ∈ [0, +∞), X ∈ Ω.

Finally, we complement the equations with the initial conditions

u (x, 0) = 0, v  (x, 0) = 0 in Ω , (II.16)

72
3. ON THE EFFECTS OF INTESTINAL VILLI

which means that the small intestine is empty at time t = 0 and by a Dirichlet boundary
condition at x1 = 0, modelling the gastric emptying, namely
v  (x, t) = v0 (t) for x1 = 0, t ∈ (0, T ) (II.17)
u (x, t) = u0 (t) for x1 = 0, t ∈ (0, T ) , (II.18)
where u0 and v0 are bounded continuous functions on [0, T ] with u0 (0) = 0 and v0 (0) = 0.

3.2 Formal asymptotic

In order to study the limit as  → 0 of the system (II.12)-(II.18), we first argue for-
mally : we consider the following expansions (called ansatz) for the solutions u and v 
x
u (x, t) = u(x1 , t) + u1 (x1 , , t) + o() (II.19)

 x
v (x, t) = v(x1 , t) + v1 (x1 , , t) + o() (II.20)

x x
where u1 (x1 , , t) and v1 (x1 , , t) are 1-periodic functions in second variable.
 
From now on, in order to simplify the notations, we systematically denote by X the
fast variable x/. On the other hand, the above system can be decoupled and we can
first study the asymptotics of v  , namely only the initial-boundary value problem (II.12)-
(II.14)-(II.16)-(II.17) and then use the result for studying the behavior of u through
(II.13)-(II.15)-(II.16)-(II.18). Since we use the same methods in both cases to obtain the
homogenization results, we present the details only for the equation of nutrients u while
we only give the results for v  .

We first plug these expressions of v  and u into (II.13), and then examine the higher
order terms in . We find
∂u ∂ 2u 1 ∂u
= χ ( 2 + ∆X u1 ) − c(x1 , X, t)( e1 + DX u1 ) + ζ(x1 , t)ϕ(v) + o(1) (II.21)
∂t ∂x1  ∂x1
At this stage, we notice that the relevant choice for observing the effects of villi is indeed
χ = χ, for some positive constant χ. With this choice, we obtain
∂u ∂u
= χ∆X u1 − c(x1 , X, t)( e1 + DX u1 ) + ζ(x1 , t)ϕ(v) + o(1). (II.22)
∂t ∂x1
The equation for the first corrector u1 (the “cell problem”) is an equation in the fast va-
riable X, i.e. for the functions X 7→ u1 (x1 , X, t), x1 , t playing the role of parameters.
Setting
∂u ∂u
p := (x1 , t)e1 , λ := − (x1 , t) and δ := ζ(x1 , t)ϕ(v),
∂x1 ∂t

73
3. ON THE EFFECTS OF INTESTINAL VILLI

and substituting p and λ in (II.22), we obtain the equation on Ω

−χ∆u1 + c(x1 , X, t)[p + DX u1 ] = λ + δ in Ω. (II.23)

We argue in the same way for the boundary condition : plugging (II.19) and (II.20)
into (II.15), we obtain
∂u  α 
χ( e1 + DX u1 ).n = − ηp (x1 , X)u + ηa (x1 , X, t)ga (u) − %(x1 , X)v + o(1) .
∂x1 ω
(II.24)
Using the introduced notations and recalling that N (X) = n(x), the above equation gives
1 α 
(p + DX u1 ).N = − ηp (x1 , X)u + ηa (x1 , X, t)ga (u) − %(x1 , X)v + o(1) . (II.25)
χ ω
Introducing the notations µ := u(x1 , t) and ν = v(x1 , t) and
α
Θ(x1 , X, t, u, v) := ηp (x1 , X)u + ηa (x1 , X, t)ga (u) − %(x1 , X)v ,
ω
the complete cell problem reads

 −χ∆u1 + c(x1 , X, t)[p + DX u1 ] = λ + δ in Ω
1 (II.26)
 (p + DX u1 ) · N = − Θ(x1 , X, t, µ, ν) on ∂Ω
χ
We assume that this problem has indeed a smooth solution u1 which is 1-periodic in X1 .
Recalling that Ω is 1 periodic in the X1 direction and integrating (II.26) over a period P
(remarking also that ∆X u1 = ∆X (u1 + p · X)), we obtain
Z Z
(λ + δ)|P | = χ −∆X (u1 + p · X)dX + c(x1 , X, t)[p + DX u1 ])dX (II.27)
P P

where |P | denotes the Lebesgue measure of P . By using Green Formula

Z Z
−χ ∆X (u1 + p · X)dX = −χ (DX u1 + p).Ñ dσ
P ∂P

where Ñ denotes the outward, unit normal to ∂P and where

∂P = (∂P ∩ ∂Ω) ∪ (∂P ∩ Ω) .

We first point out that, because of the periodicity of u1 and the opposite orientation of the
normal vector on both side of the cell
Z
χ (DX u1 + p) · Ñ dσ = 0. (II.28)
(∂P ∩Ω)

74
3. ON THE EFFECTS OF INTESTINAL VILLI

On the other hand, recalling the boundary condition of (II.26)

Z Z
−χ (DX u1 + p).Ñ dσ = Θ(x1 , X, t, µ, ν)dσ.
∂P ∩∂Ω ∂P ∩∂Ω

Next we consider the c-term : by integration by parts

Z Z Z
c(x1 , X, t)DX u1 dX = u1 c(x1 , X, t).Ñ dσ − u1 divX (c)dX.
P ∂P P

Because of (C2), the last integral of the right-hand side vanishes, while, for the first one,
we use similar argument as above : because of the periodicity properties of the velocity
function c, the integral over ∂P ∩ Ω is 0 (the same reasons as for (II.28)) and by (C3), it
is also the case for the integral over ∂P ∩ ∂Ω.

Gathering these informations, inserting them in (II.27) and recalling the definition of
Θ, one gets
Z
α
(λ + δ)|P | = [ηp (x1 , X)µ+ηa (x1 , X, t)ga (µ) − %(x1 , X)ν]dσ
∂P ∩∂Ω ω
Z
+p · c(x1 , X, t)dX. (II.29)
P

In order to obtain the homogenized equation, we introduce

Z
1
c̄(x1 , t) = c(x1 , X, t)dX,
|P | P
Z
1
η̄p (x1 ) = ηp (x1 , X)dσ (II.30)
|∂P ∩ ∂Ω| ∂P ∩∂Ω
Z
1
η̄a (x1 , t) = ηa (x1 , X, t)dσ
|∂P ∩ ∂Ω| ∂P ∩∂Ω
Z
1
%̄(x1 ) = %(x1 , X)dσ
|∂P ∩ ∂Ω| ∂P ∩∂Ω

where |∂P ∩ ∂Ω| denotes the area of the surface ∂P ∩ ∂Ω. With the notation
α
Θ̄(x1 , t, u, v) := η̄p (x1 , X)u + η̄a (x1 , X, t)ga (u) + %̄(x1 , X)v ,
ω
|∂P ∩ ∂Ω|
R(P ) := ,
|P |
we get
λ = R(P )Θ̄(x1 , t, µ, ν) + c̄(x1 , t) · p − δ, (II.31)

75
3. ON THE EFFECTS OF INTESTINAL VILLI

The one dimensional averaged equation of transport and absorption of nutrients is thus
obtained by inserting the value of λ and p in the equation (II.31)
∂u ∂u
+ c̄(x1 , t) · e1 = ζ(x1 , t)ϕ(v) − R(P )Θ̄(x1 , t, u, v) (II.32)
∂t ∂x1
The term R(P )Θ̄(x1 , t, u, v) represents the global result of the different phenomena on the
boundary of the small intestine : production of nutrients by surfacic degradation, active
and passive absorption. The interesting feature in this term comes from the coefficient
R(P ) which measures the ratio between the large surface of the villi compared to the
relatively small volume of each cell. It therefore describes the effect of the geometry of
the villi on the absorption and degradation processes.

The term R(P )[η̄p (x1 )u + η̄a (x1 , t)g(u)] gives an averaged value of absorption by
intestinal wall, which takes into account the effect of villi folds as well as the differences
between passive and active absorption.

In the same way as for the nutrients u , we may obtain the one dimensional homoge-
nized equation for feedstuffs v 
∂v ∂v 1
+ c̄(x1 , t) · e1 = −ζ(x1 , t)ϕ(v) − R(P ) %̄(x1 )v. (II.33)
∂t ∂x1 ω

In order to compare the homogenized equations (II.32)-(II.33) with the models pre-
sented in [8], we recall that, roughly speaking, in these models, the bolus is identified as
a cylinder of fixed length and variable radius r, composed of a single feedstuff A which
is transformed into an absorbable nutrient B through different types of enzymatic degra-
dations. In fact, the main model is more sophisticated since A and B can appear under
several forms (typically for A a solubilized and a non-solubilized form).

Two degradation mechanisms are taken into account : a “volumic” one taking place
inside the bolus and resulting from the action of pancreatic and gastric enzymes and a
“surfacic” one taking place on the villi and resulting from the action of the brush-border
enzymes. Then, once the absorbable nutrient B reaches the surface of bolus, hence the
intestinal wall, the absorption is ensured by a Michaelis- Menten mechanism. Therefore,
even if the above 3-d model is very simplified, the functions v and A have the same nature
and represent the large particles of feedstuffs, as well as the functions u and B represent
the absorbable nutrients. Furthermore the 3-d model described the same phenomena, at
least on the boundary.

Therefore, as we already mentioned it in the introduction, the above homogenization

process equations (II.32)-(II.33) justifies the rather simple form of the equations presented
in [8] : as long as we are just interested in “macroscopic” phenomena, it is reasonnable to
describe the effects of the complex geometry of the villi, the different types of degradation
and the absorption process by these odes.

76
3. ON THE EFFECTS OF INTESTINAL VILLI

Remark 3.1. In the above analysis, the effects of villi is summarized and measured by
the (a priori large) R(P )-coefficient which described the consequences of their particular
finger-like geometry. It is worth pointing out that this type of analysis can be used as well
to understand the effects of villi in the intestinal tract but also the effects of micro-villi
inside the villi.

3.3 The rigorous result and proof

We are now in position to state the rigorous result.

Theorem 3.1. Assume that Ω is a C 2 -domain satisfying the properties described in Sec-
tion 3.1, that (C1)-(C3), (T1)-(T2) holds and that u0 , v0 are continuous functions such
that u0 (0) = v0 (0) = 0. Then the sequences (u , v  ) converge locally uniformly, as
 → 0, to the unique (viscosity) solution (u, v) of the system

∂u ∂u

 + c̄(x1 , t) · e1 =ζ(x1 , t)ϕ(v) − R(P )Θ̄(x1 , t, u, v) in QT
∂t ∂x1




 ∂v

∂v 1
+ c̄(x1 , t) · e1 = − ζ(x1 , t)ϕ(v) − R(P ) %̄(x1 )v in QT (II.34)

 ∂t ∂x1 ω
u(0, t) = u0 (t) and v(0, t) = v0 (t) on ∂QT





u(x1 , 0) = v(x1 , 0) = 0 in [0, +∞)


where QT = (0, +∞) × (0, T ) and ∂QT = {x1 = 0, t ∈ (0, T )}.

The averaged problem (II.34) can be seen as a simplified version of the more com-
plicated initial-boundary value problem (II.12)-(II.18) : it is clearly easier to compute the
solution of (II.34) than to take into account the complex geometry and boundary condition
of (II.12)-(II.18).

Proof of Theorem 3.1. Before providing the proof, we make some remarks about the exis-
tence and uniqueness of u and v  . The system II.12)-(II.18) is in fact decoupled and
therefore we prove (by similar methods) the existence and uniqueness of v  and then of
u .

The initial-boundary value problem (II.12)-(II.14)-(II.16)-(II.17) is a classical parabo-

lic problem with Dirichlet and Neumann boundary conditions : it therefore admits smooth
solutions. If one does not insist on proving the existence of smooth solutions, the existence
and uniqueness of a viscosity solution of this problem can also be obtained by easier vis-
cosity solutions arguments, using Perron’s method (cf. [35], [36]) and comparison results
([37], [36]). Of course, the result for (II.13)-(II.15)-(II.16)-(II.18) follows from the same
arguments.

77
3. ON THE EFFECTS OF INTESTINAL VILLI

Applying the Maximum Principle (or a comparison result for viscosity solutions), it is
easy to prove that 0 ≤ v  (x, t) ≤ ||v0 ||∞ in Ω × [0, T ] since 0 and ||v0 ||∞ are respectively
subsolution and supersolution of (II.12)-(II.14)-(II.16)-(II.17) . In particular, the v  ’s are
uniformly bounded. For the u , the situation is unfortunately a little bit more complicated :
since 0 is a subsolution of (II.13)-(II.15)-(II.16)-(II.18), we have u (x, t) ≥ 0 on Ω ×[0, T ]
but it is not obvious at all to get an upper bound. For the time being, we assume that the
u ’s are uniformly bounded and we will come back on this point at the end of the proof.

We provide the full convergence proof only in the case of the u ’s, the one for the
v  being obtained by similar and even simpler argument. In this proof, because of the
decoupling of our system, we assume that we already know that the v  ’s are converging
uniformly.

In order to prove the convergence of u towards u, we use the standard method in

such problems : we combine the half-relaxed limit method [38, 36] with the Perturbed
Test-Function method introduced by L. C. Evans [27]. It is worth pointing anyway that
the non-classical feature in our result and proof comes from the 3-d to 1-d passage to the
limit and the change in the nature of the problem.

To this end, we introduce

ū(x, t) = lim sup u (y, s) , u(x, t) = lim inf u (y, s) .

→0,y→x →0,y→x
s→t s→t

We have to prove that ū is a subsolution of (II.34) and u is a supersolution of (II.34) ;

since the proofs for the sub and supersolution cases are similar, we only present the argu-
ments for the subsolution case.

Let φ : [0, +∞) × [0, T ] → R be a smooth test-function and (x01 , t0 ) be a strict

maximum point of ū − φ. In order to prove that ū is a subsolution of (II.34), we first
consider the case when x01 > 0, t0 > 0 where we have to prove
∂φ 0 ∂φ 0
(x1 , t0 )+c̄(x01 , t0 ) · e1 (x , t0 ) ≤
∂t ∂x1 1
ζ(x01 , t0 )φ(v) − R(P )Θ̄(x01 , t0 , u(x01 , t0 ), v(x01 , t0 )) (II.35)

To apply the perturbed test-function method, we need the

Lemma 3.1. The cell problem (II.26) has a X1 -periodic solution u1 if and only if the
parameters λ, p, µ, ν, δ, x1 , t satisfy Equation (II.31). Moreover this solution is unique up
to an additive constant.

Démonstration. The proof is standard and relies on the Fredholm alternative. By (C2), the
operator L := −χ∆ − c · D together with Neumann boundary conditions, is self-adjoint

78
3. ON THE EFFECTS OF INTESTINAL VILLI

and the Strong Maximum Principle shows that the kernel of this operator only contains
the constant functions. On the other hand, the computations of the previous subsection
ensure that the right hand side of equation (II.26) (including the boundary condition) is
orthogonal to the constant functions, i.e. the kernel of L, if and only if Equation (II.31)
holds. Therefore this condition implies the existence of a solution of (II.26), which is C 2
by using standard elliptic regularity. This solution is of course unique up to an additive
constant because of the structure of the kernel.

We pick some constant 0 < γ  1. In view of Lemma 3.1, for the choice of the
parameters x01 , t0 , δ := ζ(x01 , t0 )ϕ(v(x01 , t0 ))

∂φ 0
ν = v(x01 , t0 ) , µ = u(x01 , t0 ) − γ , p = (x , t0 )e1 (II.36)
∂x1 1

and if we choose λ given by Equation (II.31), there exists a smooth solution u1 (X) of
(II.26) associated to these parameters.

We use this function to introduce the perturbed test-function φ

x
φ (x, t) = φ(x1 , t) + u1 ( ) .


By standard results ([38], p.88), for  small enough, there exists a maximum point
(x , t ) of u − φ near ((x01 , 0, 0), t0 ). Moreover

x → (x01 , 0, 0), t → t0 as →0 (II.37)

u (x , t ) → ū(x01 , t0 ) as →0 (II.38)

First, we prove that the maximum point (x , t ) can not be on the boundary for  small
enough. Otherwise, if (x , t ) ∈ ∂Ω × (0, T ), then, by the maximum point property on
the boundary
∂   
(u (x , t ) − φ(x1 , t ) − u1 (X  )) ≥ 0
∂n
 
where X = x /, thus

∂u   ∂φ  
(x , t ) − [ (x , t ).e1 + DX u1 (X  )].n ≥ 0.
∂n ∂x1 1

Using the smoothness of φ and recalling that n(x ) = N (X  ), we can write this inequality
as
∂u  
(x , t ) − [p + DX u1 (X  )] · N ≥ o(1), (II.39)
∂n
79
3. ON THE EFFECTS OF INTESTINAL VILLI

where, here and below, o(1) denotes a quantity which goes to 0 as  tends to 0.

Recalling Equation (II.15) permits to obtain

 
1          
− Θ (x1 , X , t , u (x1 , t ), v (x1 , t )) − p + DX u1 (X ) · N ≥ o(1). (II.40)
χ

Because of (II.37)-(II.38) and the continuity properties of the functions ηp , ηa , % and ga

and because of the convergence of v  , Equation (II.40) gives
 
0  0 0 0 0 


−Θ x1 , X , t0 , ū(x1 , t ), v(x1 , t )) − χ p + DX u1 (X ) · N ≥ o(1).

Now we replace ū(x01 , t0 ) by µ + γ

 
x01 , X  , t0 , µ γ, v(x01 , t0 )) 


Θ + − χ p + DX u1 (X ) · N ≥ o(1).

Because of the properties of ηp , ηa , % and ga , Θ(x1 , X, t, u, v) is a strictly increasing func-

tion in u (uniformly wrt the other parameters) ; by using (II.26) together with (T2), we
obtain

−ηγ ≥ −Θ x01 , X  , t0 , µ + γ, v(x01 , t0 )) + Θ x01 , X  , t0 , µ, v(x01 , t0 )) ≥ o(1) (II.41)

which yields the contradiction.

Therefore the maximum point (x , t ) of u −φ is in Ω ×(0, T ). Since u is a solution

of (II.13), classical properties yield to the inequality
 
∂φ       ∂φ   
(x , t ) − χ∆X u1 (X )+c(x , X , t ) (x , t )e1 + DX u1 (X )
∂t 1 ∂x1 1
− ζ(x1 , t )ϕ(v) ≤ o(1). (II.42)

For  small enough, using (II.37) and the regularity of φ imply

∂φ 0 0
(x , t ) − χ∆X u1 (X  ) + c(x01 , X  , t0 )(p + DX u1 (X  )) − δ ≤ o(1). (II.43)
∂t 1
Furthermore, by Equation (II.26),

λ = −χ∆u1 + c(x01 , X  , t0 )(p + DX u1 ) − δ

which yields to the inequality

∂φ 0
(x , t0 ) + λ ≤ o(1).
∂t 1
80
3. ON THE EFFECTS OF INTESTINAL VILLI

As we already mentioned it above, this inequality is equivalent to the inequality (II.35) by

just inserting the value of λ from Lemma 3.1 into the above equation and letting  tend to
0. And the proof of this first case is complete.

We should now consider the cases when the maximum point is achieved either for
t = 0 or at x1 = 0 to complete the proof.

For the initial condition (t = 0), a combination of the above proof and classical argu-
ments shows that we have the viscosity inequality
∂ ū ∂ ū
min{ + c̄(0, t) · e1 − ζ(0, t)ϕ(v) + R(P )Θ(x1 , t, u, v), ū} ≤ 0 ,
∂t ∂x1
if x1 > 0, while, for x1 = 0, t > 0, one has
∂ ū ∂ ū
min{ + c̄(x1 , t) · e1 − ζ(x1 , t)ϕ(v) + R(P )Θ(x1 , t, u, v)), ū − u0 (t)} ≤ 0 ,
∂t ∂x1
and for the case x1 = 0, t = 0 –which is a priori a particular case–, since u0 (0) = 0, we
can still use one of these inequalities which are the same.

It is proved (cf. ([38]p.99)[36]) that, if ū is a subsolution of (II.34), then the above

initial condition in viscosity sense reduces in fact to a classical one
ū(x01 , 0) ≤ 0 on [0, +∞).
Since c̄(x1 , t) · e1 > 0, the generalized Dirichlet condition reduces also to a classical one
(cf. [38](cor 4.1 in p.169)), namely
ū(0, t) ≤ u0 (t) on [0, T ),
as a consequence of the fact that the characteristic are pointing outward the domain on the
boundary.

To conclude the proof, we invoke a (strong) comparison result for (II.34) : such result
is classical and it yields ū ≤ u on [0, +∞) × [0, T ], implying the desired convergence
result.

It remains to prove that the u ’s are indeed uniformly bounded. To this aim, we recall
that Ω is a C 2 -domain and therefore there exists a x1 -periodic, C 2 -function d : Ω̄ →
[0, ∞) such that
Dd(x) · N ≤ −1 on ∂Ω.
Because of the particular form of Ω, the function d, as well as its first and second deriva-
tives, are also bounded functions.

Now, we introduce the functions w : Ω̄ × [0, T ] → [0, ∞) given by

w (x, t) = k1 + k2 t + k3 (||d||∞ − d(x/)) ,

81
3. ON THE EFFECTS OF INTESTINAL VILLI

for some constants k1 , k2 , k3 ≥ 0. We first plug these functions into the boundary condi-
tion (II.15) : using that the v  ’s are bounded and that the absorption terms are positive, the
supersolution condition is satisfied if we choose k3 large enough. Then we consider Equa-
tion (II.13) : since d has bounded first and second derivatives, the supersolution condition
is also satisfied by choosing k2 large enough. Finally we choose k1 large enough to treat
the boundary condition (II.17).

Applying the Maximum Principle (or a comparison result for viscosity solutions) gives
u (x, t) ≤ w (x, t) in Ω × [0, T ] and the proof is complete.


82
Chapitre III

Digestion Modelling in the Small

Intestine : Impact of Dietary Fibre

In this work, we continue the modelling of the digestion in the small intestine, star-
ted in a previous article, by investigating the effects of dietary fibre. We recall that this
model aims at taking into account the three main phenomena of the digestion, namely the
transit of the bolus, the degradation of feedstuffs and the absorption through the intestinal
wall. In order to study the role of dietary fibre on digestion, we model their two principal
physiochemical characteristics which interact with the function of the small intestine, i.e.
viscosity and water holding capacity. This leads us to consider some features of diges-
tion which have not been taken into account previously, in particular the interrelationship
between the evolution of dry matter and water in the bolus. The numerical results are
in agreement with the positive effect of insoluble dietary fibre on the velocity of bolus
along the small intestine and on its degradation. These results highlight the negative ef-
fect of soluble dietary fibre on digestion. Although, this model is generic and contains a
large number of parameters, to our knowledge, it is among the first qualitative dynamical
modelling of fibre influence on intestinal digestion 1 .

1. Ce chapitre fait l’objet d’un preprint accessible en ligne sur le serveur HAL de CNRS,
Digestion modelling in the small intestine : impact of dietary fibre,
Masoomeh Taghipoor and Guy Barles and Christine Georgelin and Jean-René Licois and Philippe Lescoat.

83
1 Introduction

Digestion in the small intestine can be described through three main phenomena :
transit of the bolus along the small intestine, degradation of macromolecules into smaller
ones and absorption through intestinal wall. Taking into account these phenomena, the
authors have presented in [8] a generic model of digestion in which the bolus include only
one category of macromolecules (carbohydrates, proteins or lipids) and water.

However mixing these nutrients influences the digestion process through interactions
between molecules. In order to improve this model and to make it more realistic, we
should consider the effects of such interactions, and we have decided to do so by first
including dietary fibre in the bolus because of their significant role on the digestion. One
of the key properties of fibre is its water holding capacity and this leads us to investigate
the role of water in the digestion processes. To do so, we distinguish dry matter and water
in each substrate, we model water kinetic in correlation with the dry matter one and we
take into account the impact of water in all the aspects of digestion.

To be more precise on our approach, we first describe the main underlying assumptions
which guide our modelling, then we show how they are translated into equations and
finally numerical tests are performed for examining the effects of several hypothesis. In
addition, we point out known or assumed mechanisms relating dietary fibre and digestion
process.

This paper is organized as follows : physiochemical characteristics of soluble and

insoluble dietary fibre are introduced in Section 2. In Section 3, main assumptions of the
model are presented. Section 4 is devoted to the description of the composition of bolus,
the chemical transformations of macromolecules and all the notations. Digestion in the
presence of dietary fibre includes the modification of transport as well as degradation and
absorption equations as described in Section 5. Section 6 is a comparison of the numerical
results considered as in silico experiments. Finally, in Section 7, the model is discussed
and perspectives are proposed.

2 Biological Background on Water and Dietary Fibre

Digestion modelling requires the knowledges of the physiochemical properties of ma-

cromolecules concerned by this phenomenon as well as mechanical and biochemical reac-
tions observed for their degradation.

Dietary fibre (DF) is usually defined as the sum of plant non-starch polysaccharides
and lignin that are not hydrolysed by the enzymes secreted by the non-ruminant diges-
2. BIOLOGICAL BACKGROUND ON WATER AND DIETARY FIBRE

tive system, but that can be partially digested by microflora in the gut. A main effect
of fibre is to regulate intestinal degradation and absorption of nutrients as well as their
transit along the gut. Physiochemical characteristics of fibre include viscosity, hydration,
fermentability (mostly in the large intestine), adsorption or entrapment of nutrients and
bulking effect. Each of these characteristics affects meaningfully the function of the gas-
trointestinal tract [39, 40]. These characteristics depend on the polysaccharides chemistry.
One way to classify dietary fibre is based on their water solubility. Insoluble dietary fibre
include cellulose, some hemicelluloses and lignin. The other is soluble dietary fibre such
as viscous fibre which includes beta-glucans, pectins, gums, mucilages and some hemi-
celluloses [41, 42].

For monogastrics, most available nutrients are degraded and absorbed in the small
intestine. At the beginning of duodenum the bolus consists of partially degraded feedstuffs
and water. Once in the small intestine, mechanical and chemical digestion of feedstuffs
make the nutrients available to the organism. Enzymatic hydrolysis is the most important
chemical reaction in digestion, which takes place in aqueous solution. Enough water is
required for an efficient digestion process eventhough water/nutrient ratio are not precisely
known. Furthermore, classification of dietary fibre through their water solubility and the
impact of Water Holding Capacity (WHC) of DF on digestion reveals the key-role of
water on digestion. WHC is defined by B. Shneeman [40] as the ability of fibre source to
swell when mixed with water and to hold water within its matrix.

Soluble dietary fibre

Soluble DF are believed to impact significantly digestion and absorption as well as

bolus transport in the small intestine. The main physiochemical properties of soluble DF
are viscosity, water holding capacity (WHC) and organic compound entrapment [43].
Soluble DF, because of its high viscosity, is generally associated with slow transit through
the stomach and increasing of the small intestinal transit time [44].

Insoluble dietary fibre

Insoluble DF acts primarily in the large intestine where, due to its WHC, increases
faecal bulk, dilutes colonic contents and decreases mouth-to-anus transit time [43]. Ho-
wever, its effects on digestion and transit in the small intestine can not be neglected since
insoluble DF affects the transit time in the small intestine through its laxative property.
Recent studies have shown that the inclusion of a moderate level of dietary fibre improves
the digestibility in chicks [45]. Therefore to obtain an optimal efficiency in nutrient uti-
lization, Burhalter et al. [46] proposed to increase the ratio of insoluble to soluble DF.
Moreover, the use of insoluble fibre in commercial broiler chicks improves the intestine

86
3. KEY MODEL ASSUMPTIONS

morphological parameters and result in a better performance assumed to be connected to

more efficient digestion and absorption processes [47]. Two hypothesis are proposed in
order to study the influence of insoluble DF on nutrients digestibility in the small intes-
tine : (i) insoluble DF increases the retention time in the stomach changing the nutrient
profile of the bolus at the entry of the small intestine which could lead to a higher digestion
and absorption. (ii) Physical characteristics of insoluble DF change the digestion process
mostly through their capacity of swelling water and nutrients. These both hypothesis are
either tested in the in silico experiments (cf. Section 6.1) or included in the equations (cf.
Section 5).

Water

To have a better understanding of the role of dietary fibre, it is therefore required to

study more precisely the evolution of water during digestion in the small intestine. For
example, depending on the bolus composition, water absorption through intestinal wall or
its secretion into the small intestine lumen could be observed. However, the evolution of
water amount in the small intestine depends also on other components’ kinetics within the
bolus as described by (Reaction 1), (Reaction 2) and (Reaction 3) in Section 4.

3 Key Model Assumptions

In this section, the key assumptions for the model are presented.
H1 : Each component of the bolus (macromolecules, partially degraded macromole-
cules, nutrients and fibre) is represented mathematically as a portion of dry matter
and a characteristic proportion of water.
For example, “starch in a bolus” includes both dry starch and water used to maintain
starch molecules in aqueous solution. The same is observed for the “disaccharides in
a bolus” and “glucose in a bolus” combining smaller molecules resulting from starch
hydrolysis associated with a specific level of water. In other words, a component C of
bolus is represented as C dm + WC where C dm denotes the dry matter of C and WC is the
necessary amount of water to maintain it in a solution state. Moreover, we have assumed
that the mass of WC is proportional to the mass of C dm , i.e. equal to c C dm for some
characteristic number c ≥ 0 whic represents the necessary amount of water proportion to
maintain C in solubilized phase. Despite the presence of water in the bolus, a little amount
of non-solubilized dry matter may be included in bolus, which is (of course) associated
with c = 0.
H2 : Without DF, the bolus contains a single macromolecule and water. It is repre-
sented by a homogeneous cylinder with the constant length `. Including insoluble
DF transforms this homogeneous bolus into an heterogeneous one by modifying the

87
3. KEY MODEL ASSUMPTIONS

concentrations of feedstuffs and nutrients.

A bolus in the small intestine is a viscous solution of dry matter and water : we assume that
its volume is very close to the volume occupied by the water in the bolus. In other words,
our assumption is that DM does not fill any volume (e.g. : solubilized sugar + water
does fill the same volume as the water alone). Once the volume is known, [H2] allows
to compute the radius of the bolus r(t), which impacts its movement along the small
intestine. For a bolus with constant mass at the beginning of duodenum, [H2] influences
the degradation by increasing the concentration of nutrients in the homogeneous part of
bolus. Indeed the space occupied by insoluble dietary fibre is unavailable for the other
macromolecules.
H3 : Digestion in the small intestine is due to volumic and surfacic transformations.
Volumic degradation is the enzymatic hydrolysis of bolus components by pancreatic
and exogenous enzymes inside the bolus while the surfacic one is the degradation
by brush border enzymes on the internal wall of small intestine.
Some additional facts should be pointed out : water facilitates the contact of the macro-
molecules with the brush border enzymes, enhancing the surfacic degradation. Increasing
the water to dry matter ratio dilutes the bolus and decreases the volumic degradation. Both
of these reactions are proportional to the mass ratio of concerned substrates.
H4 : The bolus movement along the small intestine is due to peristaltic waves. The
efficiency of these waves are proportional to the radius of bolus, and inversely pro-
portional to the distance from the pylorus.
We use, in this article, the model of transport with an averaged velocity function as pre-
sented in [26]. The movement of the bolus in the model by this equation is due to a homo-
genized acceleration caused by the average effect of the peristaltic waves. Of course, there
would no difficulty to come back to the model which takes into account all the frequent
peristaltic waves.

H5 : The water in the bolus which is not hold by the macromolecules and DF through
WHC i.e. “available water”, decreases the viscosity of the bolus and facilitates its
movement. Due to osmotic type equilibrium, the concentration of this “available
water” tends to reach a fixed ratio.

In other words, “Available water” reduces the friction caused by the bolus contact with
the intestinal wall.
H6 : Dietary fibre modifies the bolus evolution through its WHC by holding the water
in its matrix, and therefore modifying the volume of the bolus. Soluble DF decreases
the efficiency of peristaltic waves.
By their WHC, dietary fibre holds a significant quantity of water in the bolus and therefore
keep the volume and the radius of the bolus higher. It is worth pointing out that soluble
dietary fibre change the consistence of the bolus in the sense of making it more jelly,
implying the decreases of efficiency of peristaltic waves.

88
4. PHYSIOLOGICAL ASPECTS AND BOLUS COMPOSITION

4 Physiological Aspects and Bolus Composition

Different steps of mechanical and chemical transformations are detailed below. We

present also the composition of the bolus and their interactions.

• Non degradable substrate And : The quantity And represents the mass of macro-
molecules which is not degradable by endogenous enzymes of the digestive tract.
• Non solubilized substrate Ans : The quantity Ans represents the mass of macromo-
lecules which is not accessible to enzymatic hydrolysis. In presence of a sufficient
quantity of water, Ans is transformed into As .
• Solubilized substrate Adm s : The quantity Adms is the mass of dry substrate in so-
lution state. It is called solubilized substrate and it is assumed that one unit of Adms
requires Ws units of water to remain solubilized. Recalling [H1], Ws represents the
required mass of water to solubilize Adm s . This quantity depends on the properties
of each macromolecules. The mass of Ws is assumed to be equal to the mass of
αAdm dm
s where α represents the ratio of water associated with As . For simplification
purposes, the mix of Adm s and water is represented by As .
dm dm
• Intermediate substrate Bint : The quantity Bint is the mass of dry intermediate
substrate obtained from the degradation of As by volumic transformation [H3]. It
is solubilized and Wint represents the required amount of water to maintain solu-
dm
bilization. The mass of Wint is assumed to be equal to the mass of βBint where β
dm dm
represents the ratio of water associated with Bint . For Bint = Bint + Wint , volumic
transformation is represented as

As + enzymes → Bint + (Ws − Wint ) (Reaction 1)

Depending on the value of Ws and Wint , the amount of Ws − Wint of water can be
released or hold in the bolus.
dm dm
• Absorbable nutrients Babs : The quantity Babs is the mass of dry absorbable nu-
dm
trients obtained from surfacic reactions (cf. [H3]). For Babs = Babs + Wabs , the
surfacic transformation is defined as

As + enzymes → Babs + (Ws − Wabs ) (Reaction 2)

Bint + enzymes → Babs + (Wint − Wabs ) (Reaction 3)

Wabs is the required amount of water to maintain solubilization and its mass is
dm
assumed to be equal to γ Babs where γ represents the ratio of water associated with
dm
the Babs .
dm dm
• Soluble and insoluble dietary fibre :Fsol and Finsol represent the soluble and non-
soluble dry dietary fibre respectively. The main property of dietary fibre presented
in Section 1 and hypothesis [H5] is its water holding capacity.
dm
Fsol = Wsol + Fsol

89
4. PHYSIOLOGICAL ASPECTS AND BOLUS COMPOSITION

dm
Finsol = Winsol + Finsol
dm dm
where the mass of Wsol (Winsol ) is assumed to be equal to λs Fsol (λi Finsol ) for λs
dm dm
and λi which represent the ratio of water associated with Fsol and Finsol respectively
. As described in Section 1, DF is not degradable by endogenous enzymes of the
small intestine.
The following diagram shows the different transformations inside the bolus.

Model with Fibre

Fs

And
W
Fins

pcr Bint
Ans As
B.B

Babs

F IGURE III.1: Physical and chemical transformations inside the bolus and included in the mo-
del are represented in this scheme. “B.B“ stands for brush border enzymes and “pcr” stands for
pancreatic ones.

• Dry Matter : Total amount of Dry Matter substrate in the bolus is therefore

DM = And + Ans + Adm dm dm dm dm

s + Bint + Babs + Fsol + Finsol .

• Water : Impact of dietary fibre on digestion is closely linked to their WHC capacity.
Though, water evolution in the bolus has to be described to understand effects of
DF on digestion.
Total water Wtot in the bolus comes from three main sources
(i) Wf eed : water incorporated naturally in feedstuffs (e.g. : one gram of wheat
contains 12 % of water). The amount of Wf eed coming out of the stomach is
assumed to be proportional to the ingested dry matter DM , Wf eed = Kf eed ·DM .
(ii) Wsec : Water included in the endogenous secretions of saliva and stomach
which is also assumed to be proportional to the ingested dry matter, Wsec =
Ksec · DM .
(iii) Wdrink : Drunk water is assumed to be independent to the quantity of DM .

90
5. MODEL EQUATIONS

Total water included in the bolus is therefore defined as the sum of Wf eed , Wsec and
Wdrink
Wtot = Wf eed + Wsec + Wdrink .
Thereby, “available water”, W , as presented in [H6], is defined as the difference
between Wtot and the water associated with DM for maintaining the solution state.
In term of mass, the quantity of “available water” in the bolus at each time is
W = Wtot − Ws − Wint − Wabs − Wsol − Winsol .
• Mass of bolus M : The total mass of bolus, M , is given by M = DM + Wtot .
• Volume of bolus V : To define the bolus volume, as explained in [H2], the volume
of each substrate in solution is assumed to be the same as the volume filled by
water associated with that substrate, i.e. the volume of As is equal to Ws /ρw , where
ρw = 1 is the water density. The volume of bolus is therefore represented as
V = Wtot = πr2 `ρw = πr2 `.
Since the length of the bolus is assumed to be fixed, the volume evolution leads to
compute the radius r(t) of bolus at each time. Consequently its surface is written as
S = 2πr`.
In the following section, the different properties of dietary fibre on the model of
digestion are taken into account.

5 Model Equations
dm
To include WHC property of soluble DF in the digestion model, the mix of Fsol and
Wsol is assumed to form a viscous gel in the bolus. Therefore in our model the mass
of “available water” in the bolus is reduced. Moreover, viscous fibre enhances motility
but decreases transit rate, since it resists propulsive contractions [48]. This resistance to
peristaltic waves is described through a new notion called efficient radius of bolus called
rsol . As described in [H2], the volume filled by soluble DF is Wsol = λs Fsol , then
Definition 5.1. The efficient radius of bolus is defined as
p
rsol = (Wtot − Wsol )/2π`.
In the same way, the efficient surface of bolus is described as Ssol = 2πrsol `.

This definition is used to describe the decrease of the surfacic degradation and absorp-
tion caused by soluble DF in the model.

One of the hypothesis in the first model of digestion in [8] is the bolus homogeneity.
The mass concentration of each component of the bolus is assumed to be its mass divided

91
5. MODEL EQUATIONS

by the total mass M of the bolus. To model the digestion in presence of insoluble DF, new
notions are defined because of heterogeneity of bolus as described in [H2]. The volume
dm
filled by insoluble DF (the mix of Finsol and Winsol ), is assumed to be unavailable to the
macromolecules of feedstuffs in the bolus. In the digestion model, this hypothesis is taken
into account by the following definition.
Definition 5.2. The apparent concentration of different substrates in the bolus is repre-
sented as
Adm
s
dm
Bint dm
Babs
[Adm
s ] = , dm
[Bint ]= dm
, [Babs ]=
M − Finsol M − Finsol M − Finsol
Ans W And
[Ans ] = , [W ] = , [And ] =
M − Finsol M − Finsol M − Finsol

The degradation of macromolecules As and Bint as well as the absorption of the nu-
trients Babs are affected by WHC property of the insoluble DF through this definition.
Figure III.2 shows the regions of bolus which are filled by insoluble DF and therefore
unreachable by the macromolecules and nutrients.

F IGURE III.2: The distribution of insoluble fibre in the bolus as assumed in the model. The
apparent volume Vapp (see Definition 5.3) is the white part of the cylinder.

Integrating the insoluble DF in the bolus changes also the region reachable to volumic
degradation (see Figure III.2). According to hypothesis [H2], Winsol is the volume filled
by insoluble fibre.
Definition 5.3. The apparent volume of bolus Vapp is therefore defined as

Vapp = Wtot − Winsol .

The above considerations are taken into account in the following steps of digestion
described below.

92
5. MODEL EQUATIONS

5.1 Transport of bolus

The averaged equation of transport of bolus introduced in [26] reads

d2 x dx c0 + c1 r Kvisco dx
2
= τ (1 − c−1 ) − ,
dt dt a + bx [W ] dt

where τ is the mean effect of the pulses by unit of time, c is mean velocity of peristaltic
waves, x the position in the small intestine. Taking into account the properties of dietary
fibre, this equation changes to

d2 x −1 dx c0 + c1 rsol Kvisco dx
= τ (1 − c ) − . (III.1)
dt2 dt a + bx [W ] dt

The bolus movement described by this equation depends on its position in the small intes-
tine and on its efficient radius. Moreover, the acceleration is slowed down by a viscosity
term which depends on the available water.

5.2 Volumic transformation

Volumic transformation presented by (Reaction 1) in Section 4 is the degradation of

Adm
s
dm
due to pancreatic and exogenous enzymes resulting in production of Bint . Evolution
dm
of As by this transformation is represented by

dAdm
s
= −kvol (x)[Adm
s ]Vapp
dt

where kvol (x) takes into account the enzymatic activity which is a function of bolus posi-
tion at each time t. The term [Adm
s ]Vapp describes the dependence of volumic degradation
dm
on the concentration of As at each unit of apparent volume i.e. the volume filled by the
insoluble DF is not accessible to the enzymes and macromolecules.

Consequently, integrating the insoluble fibre in the bolus influences the volumic trans-
formation by increasing the substrates concentration via the Definition 5.2 and by chan-
ging the volume and using Vapp introduced in Definition 5.3.

As described in Section 5.1, integrating the soluble fibre modifies the velocity of bolus
along the small intestine and therefore the distance travelled at each time x(t). Conse-
quently, it influences the volumic degradation through the function kvol (x(t)).

This reaction takes place in a solution and each unit of Adm

s degraded by this equation
dm
causes the release of “available water” Ws = αAs .

93
5. MODEL EQUATIONS

dm
The volumic production of intermediate substrate Bint is the result of degradation of
Adm
s
dm
dBint
= kvol (x)[Adm
s ]Vapp .
dt
dm dm
Each unit of produced Bint requires and uses Wint = βBint to maintain the solution
state. According to (Reaction 1), the result of volumic transformation is the consumption
or release of “available water” W . Thereby, volumic evolution of water is represented as
dW
= k̃vol (x)[Adm
s ]Vapp , (III.2)
dt
where k̃vol = (α − β)kvol .

Soluble fibre can be hydrolysed by exogenous enzymes eexo (cellulase, hemicellulases,

...) ingested by food
Fsol + eexo → Bint + (Wol − Wint ). (Reaction 4)

Hydrolysis of Fsol by exogenous enzymes follows the same evolution as the volumic
transformation of Adms
dm
dFsol ˜ dm
= −ks eexo ph(x)[F sol ]Vapp
dt
dm
where [Fsol ˜
] = Fsol /(M − Finsol ) and ph(x) is the exogenous enzyme activity along the
dm
small intestine. This reaction produces the intermediate substrate Bint

dm
dBint ˜ dm
= ks eexo ph(x)[F sol ]Vapp .
dt
The amount (Wsol − Wint ) is released by (Reaction 4) and modifies the evolution of water
dW ˜ dm
= ... + k̃s eexo ph(x)[F sol ]Vapp (III.3)
dt
where k̃s = (λs − β)ks .

5.3 Surfacic transformation

Surfacic degradation is the last step of transformation of macromolecules in the small

intestine. The produced nutrients by this degradation are then absorbed through intesti-
nal wall. Surfacic degradation depends on the fraction of As on the surface of the bolus
represented by [As ]Ssol therefore

dAdm
s
= −ksurf [As ][W ]Ssol
dt
94
5. MODEL EQUATIONS

where ksurf stands for the rate of surfacic degradation of As and the efficient surface Ssol
has been defined by Definition 5.1. Moreover, it is assumed that the brush-border enzymes
dm
are always in excess in the small intestine. Surfacic degradation of Bint follows the same
dm
process as for As . Therefore for [Bint ] defined by Definition 5.2, we have
dm
dBint
= −k̃surf [Bint ][W ]Ssol .
dt
where ksurf stands for the rate of surfacic degradation of Bint . Evolution of water in the
bolus is influenced by the surfacic degradation, i.e. the quantity of water consumed (or
released) by (Reaction 2). Therefore the surfacic evolution of water is
 
dW
= ... + (β − γ)k̃surf [Bint ] + (α − γ)ksurf [As ] [W ]Ssol . (III.4)
dt

5.4 The equilibrium between As and Ans

Modifications of feedstuffs in the stomach by the enzymes and water change most of
the Ans into As = Adm
s + Ws and makes them accessible to intestinal enzymes.

However, for some feedstuffs, the bolus may contain Ans at the beginning of small
intestine. In this case, the digestion of bolus contains also the transformation of Ans into
As . Mixing with bile acid for lipids and producing the micelles, denaturing for the proteins
and adding water and solubilization for the dry starch are examples of the transformation
of Ans into As in the small intestine.

The solubilization of Ans which takes place in the presence of enough quantity of W
and results in the production of As , is a phenomenon taken into account in the model.
Solubilization is reversible and lack of water may cause production of Ans releasing W in
the bolus.

Thereby, the balance is assumed to be reached when

As = µ([W ])Ans
for µ which is an increasing function of [W ]. If kequi stands for the rate of turning back to
equilibrium then the dynamical equilibrium may be defined
dAns

= −kequi µ([W ])Ans − As (III.5)
dt
and therefore
dAdms


= kequi µ([W ])Ans − As .
dt
The variation of water quantity caused by the equilibrium may be represented as
dW

= ... + αkequi µ([W ])Ans − As . (III.6)
dt

95
5. MODEL EQUATIONS

5.5 Pancreatic and biliary secretions

Pancreatic and biliary secretions consist of a solution of nutrients and enzymes which
do not contain available water W . In fact, water included in this solution is assumed to
be associated with nutrients and enzymes to keep them solubilized. Modelling details
on these secretions could be seen in [8]. Adding dietary fibre increases the quantity of
pancreatic secretions. However this point is not yet included in the model.

5.6 Absorption through intestinal wall

Absorption of nutrients through intestinal wall depends on their concentration on the

inner surface of intestinal wall
dm
dBabs dm
= ... − kabs [Babs ]Ssol
dt
The passage of nutrient through intestinal wall releases the associated water, thus
dW dm
= ... + γkabs [Babs ]Ssol (III.7)
dt
where kabs represents the rate of absorption through intestinal wall.

5.7 Water equilibrium

Water equilibrium was already taken into account in [8]. The assumption was that [W ]
tends to reach a fixed ratio (10%), suggesting the equation
d[W ]
= −kw ([W ] − 0.1) (III.8)
dt
where [W ] = W (t)/M (t), M (t) representing the bolus mass. The superposition of the
equations (III.2), (III.3), (III.4), (III.6), (III.7) and (III.8) provides the equation describing
the evolution of W along the small intestine.

The evolution of bolus mass is represented as

dM dAdm dB dm dB dm dF dm dAns dW
= (α + 1) s + (β + 1) int + (γ + 1) abs + (λsol + 1) sol + + ,
dt dt dt dt dt dt dt
each term of the above equation is replaced by its expression, therefore we obtain
dM M dm
= (−kw (W − 0.1M ) − kabs [Babs ]Ssol ). (III.9)
dt M −W

96
6. NUMERICAL SIMULATIONS

The variation of bolus volume depends on the absorption or secretion of “available water”
and endogenous secretions in the small intestine i.e.

dV dW
= , (III.10)
dt dt
therefore
dV
= −kw (W − 0.1M ) + Secretions.
dt

6 Numerical Simulations

A thorough examination of the effects of the different parameters of the model on

transport, degradation and absorption is carried out by Scilab software. In these in silico
experiences, the presence of dietary fibre Fsol and Finsol in the bolus, the variation of the
initial values of bolus and the sensitivity of the model on the parameters α, β and γ (cf.
Section 4) are investigated.

The mass of the bolus at the entry of the small intestine is assumed to be fixed in all
our following experiences. The bolus at the entry of the small intestine contains And , Ans ,
As , Bint , Babs and W . When studying the influence of dietary fibre on digestion, the non
degradable substrate And is replaced by Fsol or Finsol .

6.1 Influence of dietary fibre on intestinal absorption

As described in Section 1, the positive effect of insoluble DF in digestion may be due

to two main reasons : (i) modification of the composition of bolus due to the increase of
retention time in the stomach, (ii) modification of the bolus physical characteristics.

The effects of these two cases on the digestion model are studied in this section.

Influence of the modification of the bolus in the stomach

Including insoluble DF in the bolus delays gastric emptying. The direct effect of this
phenomenon is to increase the solubilization in the bolus and to start partially the degra-
dation. We studied the effects of such a change in the initial conditions for our model. To
this aim, two numerical experiences are carried out : (a) the increase in the ratio of As to
Ans and (b) the increase in the ratio of Bint to As when Ans = 0.

97
6. NUMERICAL SIMULATIONS

(a) Our first experience consists in increasing the ratio of As to Ans in the bolus at
the entry of the small intestine. The value of absorbed dry nutrients at the end of
the small intestine does not vary meaningfully. Table III.1 shows the absorbed dry
nutrients to DM ratio at the end of ileum x = 17 for different for the different
Adm
s to DM ratio at the beginning of the small intestine. Despite the variations in

x=0 End of the small intestine x = 17

−1
Adm
s .DM (%) Ans .DM −1 (%) Absorbed dry nutrients to DM ratio (%)
0 85 56
42 42 57
85 0 58
TABLE III.1: The proportion of absorbed nutrients to DM for different scenarios of solubilization
in the bolus at the entry of duodenum.

the ratio of As to Ans , the equilibrium between As and Ans defined by Equation
(III.5) is reached quickly (see Figure III.3). The sensitivity analysis shows also that
the value of absorbed dry nutrients is almost independent to the variations in the
ratio of As to Ans . However, this results depend on the choice of equilibrium rate
kequi , a small value of kequi may decrease the difference between the final values of
absorbed nutrients.
(b) In the second experience, the modification in the stomach is assumed to result in
the transformation of all Ans in the bolus into As and additionally the production of
Bint . Variations in the ratio of As to Bint inside the bolus at the entry of the small
intestine are tested. Table III.2 shows the variation of absorbed dry nutrients at the
end of the small intestine as a function of the initial value of Bint .
Numerical results shows the increase in absorbed dry nutrients when the ratio of
Bint to As is increased.
x=0 End of the small intestine x = 17
dm
Bint .DM −1 (%) As .DM −1 (%) Absorbed dry nutrients to DM ratio (%)
0 85 57
42 42 61
85 0 64
TABLE III.2: The relation between the absorbed dry nutrients at the end of digestion and the
dm .
different initial values of Bint

Direct effect of DF on the function of the small intestine

Besides the modification of the bolus in the stomach, presence of insoluble DF changes
also the physiochemical characteristics of bolus (Section 5).

98
6. NUMERICAL SIMULATIONS

40 40

35 35

30 30
Mass (g)

Mass (g)
25 25

20 20

15 15

10 10

5 5

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Tim e (hour) Tim e (hour)

(a) Adm
s = 36, Ans = 0 (b) Adm
s = 18, Ans = 18

40

35

30
Mass (g)

25

20

15

10

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Tim e (hour)

(c) Adm
s = 0, Ans = 36

F IGURE III.3: The equilibrium Adm

s − Ans is reached quickly for different initial value of As .

To observe the effect of DF in the model of digestion, value of insoluble and soluble
DF was increased from 1 g to 5 g in a bolus of 120 g.

Figure III.4 shows that the presence of insoluble DF promotes intestinal absorption,
however this increase in absorbed dry nutrients is not meaningful. The results in this
figure, show that the increasing of the value of soluble DF decreases the quantity of ab-
sorbed dry nutrients and increases the final total mass.

Therefore, numerical simulation shows that the positive effect of insoluble DF on

the amount of absorbed dry nutrients is mainly due to the modification of the bolus in
the stomach. However, the effect of the interaction of bolus with the small intestine as
described in Figure III.4 may not be neglected.

99
6. NUMERICAL SIMULATIONS

120 120

100 100

80 80

Mass (g)
Mass (g)

60 60
Absorbed nutrients Absorbed nutrients

40 Final mass 40 Final Mass

20 20

0 4 0 5
0 2 4 6 8 0 2 4 6 8
Babs Water Soluble DF (g)
Insoluble DF (g)

(a) Absorption versus Fins (b) Absorption versus Fs

F IGURE III.4: Change in the final mass of bolus and absorbed dry nutrients for different amount
of DF in the bolus at the entry of duodenum.

Time of intestinal transit in the presence of DF

Numerical results of transit time in presence of DF are presente in Table III.3.

These results show that integrating insoluble DF in the bolus decreases the time of
intestinal transit from 5 h for a bolus of 120 g without insoluble DF to 3, 9 h for a bolus
of the same mass which contains 5 g of DF. These results are consistent with published
values. The experiences done by Wilfart et al. [13] have shown that increasing dietary
fibre content reduced or tended to reduce the mean retention time in the small intestine.

These results show that integrating soluble DF in the bolus increases the intestinal
digestion time from 5 h to 6, 7 h illustrating the effect of viscosity due to soluble DF on
transit time.

x=0 Intestinal transit time (hour)

F · DM −1 (%) bolus containing Finsol bolus containing Fsol
0 5 5
2 4, 7 5, 4
7 4, 3 5, 9
11 4 6, 5
14 3, 9 6, 7
TABLE III.3: Intestinal transit time for the different quantities of F = Fsol or Finsol in the bolus
at the entry of duodenum

100
6. NUMERICAL SIMULATIONS

6.2 Water associated to dry matter

To study the influence of the quantity of associated water on digestion, variation of

the values of α, β and γ have been tested. We observe their influence on digestion and
specifically on the absorbed dry nutrients and As -Ans equilibrium. Two main simulations
are carried out : an uniform water content for As , Bint and Babs i.e. α = β = γ and a a
non uniform one.

The objective of these experiences is to understand how the different values of α, β

and γ influence the digestion in our modelling.

a. Uniform water content for As , Bint and Babs

The value of α = β = γ varied from 1 to 4 in the model presented in section 5. Our

objective is to observe its effect on the value of absorbed dry nutrients as well as on the
final mass of bolus.

Numerical results presented in Figure III.5 show the negative effect of this increase
on the absorbed dry nutrients. Increasing the quantity of water (α, β and γ) associated
with the dry feedstuffs (Adm m dm
s , Bint and Babs ) in our model, dilutes the bolus and decreases
the volumic degradation, it decreases also the quantity of dry nutrients in contact with the
internal surface of the bolus. These results seems to be consistent with the reality, in fact,

90
80
70
60
Mass (g)

50
40 Absorbed nutrients
30 Final mass
20
10
0
0 1 2 3 4 5

α=β=γ

F IGURE III.5: Dependence of the absorbed dry nutrients and the final mass of bolus at the end of
the small intestine on the value of α, β and γ.

the more water is presented in the bolus, the less (pancreatic and brush border) enzymes

101
6. NUMERICAL SIMULATIONS

and molecules are in contact.

We are also interested by the effect of these variations on the Adm

s − Ans equilibrium
defined by Equation (III.5). The results are shown in Figure III.6. The equilibrium is al-
most reached in the four experiences, however the choice of kequi can change the necessary
time to reach the equilibrium. A significant production of Ans is observed in Figure III.6.d
because of the lack of the available water at the beginning of the small intestine.

25 25

20 20
Mass (g)

15 Mass (g) 15

10 10

5 5

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Tim e (hour) Tim e (hour)

(a) α = β = γ = 1 (b) α = β = γ = 2

25 25

20 20
Mass (g)

Mass (g)

15 15

10 10

5 5

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0 1 2 3 4 5 6

Tim e (hour) Tim e (hour)

(c) α = β = γ = 3 (d) α = β = γ = 4

F IGURE III.6: The evolution of the equilibrium Adm

s − Ans depends strongly on the quantity of
α, β and γ.

b. Non-uniform water content for As , Bint and Babs

In the second experience, the quantity of absorbed dry nutrients at the end of the small
intestine and the numerical results of As -Ans equilibrium for different values of α, β and
γ were observed. The choice of the values of As , Bint and Babs is based on the hypothesis
that the value of β is always between the values of α and γ because of the molecule

102
6. NUMERICAL SIMULATIONS

dm
size of Bint . Even if the longest transit time was observed for α < β < γ, it shows the

Absorbed dry nutrients to Retention time in the small

DM ratio (%) intestine (h)
α = 1, β = 2, γ = 3 45 5, 2
α = 2 β = 2, γ = 2 50 4, 9
α = 3, β = 2, γ = 1 54 4, 6
TABLE III.4: Transit time and absorbed dry nutrients at the end of the small intestine depend on
the values of α, β and γ.

lowest level of dry absorbed nutrients. This stresses the key-role of “available water” W
on digestion.

As − Ans equilibrium is observed in two cases :

(i) Firstly, it was assumed that α > β > γ. There is a little production of Ans from
As (desoubilization) because of the lack of water at the entry of the small intestine
followed by the degradation of As .
(ii) Secondly, it was assumed that α < β < γ. There is a rapid solubilization of Ans
followed by degradation of As into Bint . The released water by degradation of As is
not sufficient to maintain the solution state of Bint since α < β. The required water
is provided by the transformation of As into Ans as it can be seen in the Figure III.7.

25 25

20 20

15 15
Mass (g)

Mass (g)

10 10

5 5

0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0 1 2 3 4 5 6
Tim e (hour) Tim e (hour)

(a) α > β > γ (b) α < β < γ

F IGURE III.7: The equilibrium Adm

s − Ans for different quantity of α, β and γ.

103
7. DISCUSSION

6.3 The variation of the water ratio inside the bolus

We are interested by the change in the ratio of W inside the bolus and its influence
on the absorption in the small intestine. For the sake of simplicity, it was assumed that
Wf eed = Wsec = 0.

In order to study this effect, the value of dry matter in the bolus is assumed fixed at
42 g while the value of water increases in each experience.

In the numerical simulations, the ratio of water included in the bolus represented 55%,
60% and 66% of bolus. The ratio of absorbed dry nutrients to the total absorption (water
and nutrients) is collected. Results are presented in Table III.5. The numerical results of

(%) Amount of (%) Ratio of (g) Mass of Transit time (hour)

water at x = 0 absorbed dry Absorbed dry
inside bolus nutrients at x = 17 nutrients
50 50 25 5, 7
60 35 26 5, 2
66 28 27 4, 5
TABLE III.5: Dependence of absorbed dry nutrients to the water ratio. DM stands for the amount
of dry matter in the bolus at the entry of bolus.

Table III.5 shows that increasing the value of water in the bolus decreases the ratio of
absorbed dry nutrient to the total absorbed matter (water+dry nutrients), even though
the value of absorbed dry nutrients increases.

In fact, in our model, at the end of each experience the value of Wt is approximatively
55% of the total mass of bolus and this equilibrium is achieved almost quickly (because
of the choice of kw ). Therefore, increasing the value of water in this model does not have
a meaningful effect on the final absorption and the slight increase in the absorbed mass
of dry nutrients is due to the change of the volume of bolus in each experience which
promotes the access of nutrients to intestinal wall for absorption, although this increase
dilutes the bolus and decreases the volumic degradation. Here again, this is the direct result
the choice of the parameters (rate of surfacic (ksurf ) and volumic (kvol ) degradation ).

7 Discussion

In this paper, we have continued the modelling of the digestion in the small intestine
started in [8]. The objective was to obtain a more realistic model of digestion process
by including new phenomena and completing the others : DM and water are treated se-

104
7. DISCUSSION

parately, water evolution is influenced by the degradation of other molecules, the effects
of DF on the digestion are taken into account which is also a first step toward a non-
homogeneous model with different types of feedstuffs.

The advantages and the defects of this model as well as the perspectives are outlined
in the following paragraphs.

One of the main aspect of this model remains its genericity, we have tried to iden-
tify and model the main generic phenomena of digestion and ignored or implicitly taken
into account in the parameters the ones which required the specific properties of feed-
stuffs’ molecules (effects on gastric emptying, on viscosity,...etc.). The different steps of
digestion (equilibrium between Ans and As , successive transformation of As into Bint and
Babs ), the effects of physical characteristic of bolus (surface and volume) on its degrada-
tion and transit, the interaction between DF and feedstuffs molecules (among others) have
been considered while some other phenomena like the separation between the enzyme
activities of different feedstuffs’ molecules, dependence of the enzyme activity with res-
pect to the dilution of bolus, the different substrates density, the impact of DF on initial
condition (and others) have been ignored.

To our point of view, we obtained a more realistic model by integrating these new
phenomena in the model of digestion. In particular, the WHC of dietary fibre which in turn
interferes the digestion of other feedstuffs molecules, leads us to introduce the separation
between DM and water and consider all the effects of water.

However, some other phenomena are still ignored either because of the lack of infor-
mation concerning their effects or because of their supposed little impact on digestion at
this scale. Of course, It would be interesting to include in the model the phenomena like
interaction between different categories of feedstuffs and then define the specific enzyme
activity for different cases to exploit their potential to impact the digestion. The future
development of the model will be based on these new objectives.

Modelling the influence of soluble and insoluble DF on the initial condition and on de-
gradation of other feedstuffs’ molecules as well as on the movement of bolus (Experience
6.1) is the first try to model a more realistic non-homogeneous bolus. DF have normally a
high WHC which increases the volume of aqueous phase in the bolus and therefore dilutes
the solution of nutrients and enzymes [49]. This is known to influence the volumic and
surfacic reactions. However, these effects depend highly to the choice of Wsol and Winsol .
Another aspect of the model is the water equilibrium and its impact on absorption, here
again the choice of parameters (Was , Wint and Wabs ) plays significant role on the final
results of digestion.

On the other hand, Experience 6.3 reveals the role of other parameters of model (kw ,
kequi , ksurf , kvol ). As it has been described in this experience, the change of water ab-
sorption rate kw , can change ( even inverse) the numerical results. However, it is worth

105
8. APPENDIX

pointing that the choice of the model parameters are based on the observed behaviours
(literature), the results are therefore consistent qualitatively with the reality (positive in-
fluence of insoluble DF on digestion, negative effect of soluble DF, ...).

Taking into account these new phenomena requires the introduction of new parameters
which can be identified with the help of existing experimental data. To the best of our
knowledge, some of these parameters like the water associated to dry nutrients (Wsol ,
Winsol , Was , Wint and Wabs ) are introduced for the first time and they should be identified

One of the advantages of this model is its capacity to be reduced and to be adapt to the
new experiences which makes the parameters identification possible. Reduction consists
in the decreasing the number of equations of system or the number of parameters without
affecting its genericity (e.g. a bolus which does not contain the DF, results in a more
simplified digestion process which in turn caused decreasing the equations (parameters)
involved in the digestion model).

It is also worth pointing that the value of most of the parameters depends to the special
category of feedstuffs. The close collaborations between biologists and mathematicians is
therefore required to identify these new parameters (literature data in biology, define the
new experimentations, ...). This reveals one of the main interest of modelling which is to
ask the precise questions about the modeled phenomenon. In fact, this approach allows
to use all the existing data and limit the new animal experimentation to the special cases
(when the existing data are not sufficient).

8 Appendix

8.1 The Model equations

 Equation of transport
d2 x dx c0 + c1 rsol Kvisco dx
2
= τ (1 − C −1 ) − .
dt dt a + bx [W ] dt
 Non solubilized substrate Ans
dAns

= −kequi µ([W ])Ans − As
dt
 Soluble DF Fsol
dm
dFsol ˜ dm
= −ks eexo ph(x)[Fsol ]Vapp
dt
 Solubilized substrate Adm
s

dAdm
s
= −kvol (x)[Adm


s ]Vapp − ksurf [As ][W ]Ssol + kequi µ([W ])A ns − As + secretions
dt

106
8. APPENDIX

dm
 Intermediate substrate Bint
dm
dBint ˜
=kvol (t)[Adm dm
s ]Vapp + ks eexo ph(x)[Fsol ]Vapp − k̃surf [Bint ][W ]Ssol + secretions
dt
dm
 Absorbable nutrients Babs
dm
 
dBabs dm
= k̃surf [Bint ] + ksurf [As ] [W ]Ssol − kabs [Babs ]Ssol
dt

 Exogenous enzymes e(t)

de
= −ke e
dt
 Free Water Evolution W(t)
dW
= − αkequi µ([W ])Ans − As + (α − β)kvol (t)[Adm


s ]Vapp
dt  
+ [W ] (β − γ)k̃surf [Bint ] + (α − γ)ksurf [As ]
˜
+ (λs − β)ks eexo ph(x)[F dm dm
sol ]Vapp + γkabs [Babs ]Ssol − kw (W − 0.1M ).

 Evolution of the bolus mass M(t)

dM M dm


= − kw (W − 0.1M ) − kabs [Babs ]Ssol + Secretions
dt M −W
 Evolution of bolus volume V(t)
dV
= −kw (W − 0.1M ) + Secretions
dt

107
8. APPENDIX

108
Conclusion et perspectives

Dans le premier chapitre, un modèle générique de la digestion a été présenté. Un sys-

tème d’équations différentielles couplées décrit les phénomènes principaux de la digestion
pour un bolus homogène : le transport du bolus par les ondes péristaltiques, la dégrada-
tion volumique et surfacique du bolus par les enzymes pancréatiques et bordure en brosse
et ensuite l’absorption des nutriments. Le modèle étant mécaniste, il a contribué à une
meilleure compréhension de ces phénomènes et des lois qui les régissent.

Dans le deuxième chapitre, nous avons justifié par des méthodes d’homogénéisation
mathématique le choix du modèle basé sur d’équations différentielles ordinaires pour dé-
crire la digestion en démontrant que ce modèle tient compte d’une façon moyennée de la
présence des différentes échelles du temps et de l’espace dans la digestion intestinale : (i)
l’influence des ondes péristaltiques sur l’avancement du bol alimentaire, (ii) l’influence
des villosités intestinales sur l’augmentation de la surface de contact des aliments avec la
paroi intestinale, (iii) l’absorption passive et active à travers la paroi intestinale.

Dans le troisième chapitre, l’influence du changement de la structure du bolus a été

étudiée. Le bolus homogène du Chapitre 1 a été remplacé par un bolus non-homogène qui
contient des fibres alimentaires. Des propriétés des fibres alimentaires comme la capacité
de rétention d’eau ainsi que les nutriments ont été pris en compte. Les résultats numé-
riques sont cohérents avec des données de la littérature : les fibres solubles et insolubles
diminuent la digestibilité totale, les fibres solubles diminuent aussi la vitesse du transit du
bolus alors que les fibres insolubles l’augmentent.

109
CONCLUSION ET PERSPECTIVES

Les thèmes suivants peuvent contribuer à rendre le modèle plus pertinent pour ouvrir
des pistes de recherche :

Nécessité d’identification des paramètres

Une étape importante afin de rendre ce modèle opérationnel est l’identification de ses
paramètres. En effet, les expériences du chapitre 1 et le chapitre 3 ont montré l’influence
des paramètres sur des résultats finaux de la digestion. Pour identifier les paramètres, des
données issues des expérimentations seront nécessaires. Dans un premier temps, afin de
rendre l’identification de ces paramètres possible, nous pourrons réduire les équations du
modèle en choisissant un bolus suffisamment simplifié (homogène, solubilisé, enzymes
endogènes, ...) pour ensuite complexifier le bolus au fur et à mesure. Cependant, l’identi-
fication de certains paramètres comme le coefficient d’équilibre As ↔ Ans ou la quantité
d’eau associée à la matière sèche (Was , Wint , ...) reste compliquée du point de vue expé-
rimental et nécessitera des approches d’expert de ce domaine.

Détermination de la condition initiale

La condition initiale du modèle actuel est choisie en se basant sur le ratio de la matière
sèche dans le bolus à l’entrée de l’intestin grêle et elle est indépendante du prétraitement
des aliments dans l’estomac, car notre modèle ne s’intéresse qu’à la digestion intestinale.
Cependant, la prise en compte de ce dernier permettra de traiter des bols avec une matrice
alimentaire plus complexe. Par exemple, dans notre modèle présenté au Chapitre 3, nous
avons introduit dans la condition initiale, la quantité d’eau bue (Wdrink ), de l’eau secrétée
dans l’estomac (Wsec ) et de l’eau associée aux aliments (Wf eed ), les vraies valeurs de ces
quantités ne peuvent être définies qu’en prenant en compte le prétraitement de l’aliment
dans l’estomac (la digestion partielle des aliments, l’absorption d’eau et les sécrétions,
etc). Nous pourrons ensuite avoir une estimation raisonnable du volume et la composition
du bolus à l’entrée de l’intestin grêle.

Ainsi, il est envisageable de coupler ce modèle avec un modèle de la digestion dans

l’estomac pour lequel, la condition initiale sera la composition de l’aliment ingéré. Après
la première étape de la digestion et de la solubilisation dans l’estomac, l’aliment entre
dans l’intestin grêle par la vidange gastrique. La vidange gastrique peut être représentée
par une fonction dont la période et l’efficacité dépendent de la composition de l’aliment
dans l’estomac.

110
CONCLUSION ET PERSPECTIVES

Bolus et sa composition

1. Passage à un bolus hétérogène : nous avons étudié la digestion pour un bolus ho-
mogène et ensuite pour un bolus hétérogène en intégrant les fibres alimentaires dans
sa composition. Il est donc intéressant de prendre en compte les trois catégories de
nutriments au sein du bolus : les glucides, les lipides et les protides. L’objectif est
d’étudier leur cinétique de digestion en présence des éventuelles interactions. En ef-
fet, les interactions entre les aliments variées au sein d’un bolus sont connues pour
être des étapes impactantes de la digestion. Ainsi, certains aliments peuvent limi-
ter la digestion des autres ou au contraire la favoriser. L’intégration de ces données
dans le modèle le rendrait évidemment plus compliqué (augmentation de nombre
des équations et des paramètres), mais peut permettre d’obtenir un modèle plus
opérationnel.
2. Changement de la longueur du cylindre : dans le modèle actuel, un cylindre de la
longueur fixe ` et du rayon variable R(t) représente le bol alimentaire. Le volume
du bolus à l’entrée de l’intestin grêle définit son rayon R(0). Cette quantité a un
maximum que nous appelons rmax . Dans le modèle actuel, la condition d’entrée
est toujours définie pour satisfaire cette contrainte. Cependant, il est clair que le
volume du bolus à l’entrée de l’intestin grêle varie en fonction de la composition
de l’aliment dans l’estomac, nous pouvons ainsi envisager d’étaler (augmenter ` )
le bolus lorsque R(0) > rmax .
3. Bolus consécutifs : l’évolution d’un seul bolus dans la lumière intestinale a été étu-
diée dans cette thèse. De la même manière, l’évolution de plusieurs bolus indé-
pendants (qui ne se rejoignent pas) peut être étudiée et modélisée. Il est toutefois
possible d’envisager un regroupement éventuel de ces bolus dans l’iléon et mo-
déliser l’influence de ce regroupement sur la digestibilité totale. Cependant, nous
ne pensons pas que cet aspect soit un point majeur comparé aux autres voies pour
développer le modèle.

111
CONCLUSION ET PERSPECTIVES

112
Bibliographie

[1] B. Darcy, J.P. Laplace, and P.A. Villiers. Digestion dans l’intestin grêle chez le porc.
Ann. Zootech, 30 :31–62, 1981.
[2] W.M.S. Russel and R.L. Burch. The principles of humane experimental technique.
Methuen & Co Ltd, London., 1959.
[3] J. Vanmilgen and P. Lescoat. Modelling digestion and metabolism in the
pig : an alternative to animal esperimentation ? Bull. Acad. Vét. France - N˚5
http ://[Link]/, Tome 161- N˚5 :435–440, 2008.
[4] J. Clairambault. Modélisation dans le domaine biomédical. ECP 2009-10 Modeling
in Biomedical, 2009-10.
[5] R. Lambert P. Meunier, Y. Minaire. La digestion. 1988.
[6] L. Montagne and J. Flament. La digestion des aliments. Agrocampus ouest.
[7] G. Allaire. The theory of periodic homogenization.
[8] M. Taghipoor, G. Barles, C. Georgelin, J. R. Licois, and P. Lescoat. Mathematical
modeling of transport and degradation of feedstuffs in the small intestine. Journal
of Theoretical Biology, 294 :114–121, October 2011.
[9] W. H. Karasov. Optimal digestive responses to changing diet and foraging costs. In
Adams, N. & Slotow, R. (Eds), Proc. 22 Int. Ornithol. Congr. Durban, University of
Natal, 1999.
[10] S. R. McWilliams and W. H. Karasov. Phenotypic flexibility in digestive system
structure and function in migratory birds and its ecological significance. Comp Bio-
chem Physiol A Mol Integr Physiol, 128(3) :579–593, Mar 2001.
[11] J.D. Logan, A. Joern, and W. Wolesensky. Chemical reactor models of optimal di-
gestion efficiency with constant foraging costs. Ecological Modelling, 168(1–2) :25–
38, 2003.
[12] V. P. Srivastava. Effects of an inserted endoscope on chyme movement in small
intestine– a theoretical model. Applications and Applied Mathematics, 2 :79–91,
2007.
[13] A. Wilfart, L. Montagne, H. Simmins, J. Noblet, and [Link] Milgen. Digesta transit in
different segments of the gastrointestinal tract of pigs as affected by insoluble fibre
supplied by wheat bran. Br J Nutr, 98(1) :54–62, Jul 2007.

113
BIBLIOGRAPHIE

[14] D. Bastianelli, D. Sauvant, and A. Rérat. Mathematical modeling of digestion and

nutrient absorption in pigs. J Anim Sci, 74(8) :1873–1887, Aug 1996.
[15] J.D. Logan, A. Joern, and W. Wolesensky. Location, time, and temperature depen-
dence of digestion in simple animal tracts. Journal of Theoretical Biology, 216 :5–
18, January 2002.
[16] R. Miftahof and N. Akhmadeev. Dynamics of intestinal propulsion. Journal of
Theoretical Biology, 246(2) :377 – 393, 2007.
[17] A. Shu-Han Lin, M. L. Buist, N. P. Smith, and A. J. Pullan. Modelling slow wave
activity in the small intestine. Journal of Theoretical Biology, 242(2) :356 – 362,
2006.
[18] L. Bueno, J. Fioramonti, and Y. Ruckebusch. Mechanisms of propulsion in the small
intestine. Ann Rech Vet, 8(3) :293–301, 1977.
[19] X. T. Zhao, M. A. McCamish, R. H. Miller, L. Wang, and H. C. Lin. Intestinal
transit and absorption of soy protein in dogs depend on load and degree of protein
hydrolysis. J Nutr, 127(12) :2350–2356, Dec 1997.
[20] B. Séve. Ileal standardised digestibility of amino acids in feedstuffs for pigs.
[21] J. P. Laplace. The transit of digesta in the different parts of the digestive tract of the
pig. Prog Clin Biol Res, 77 :847–872, 1981.
[22] V. Rayner and G. Wenham. Small intestinal motility and transit by electromyography
and radiology in the fasted and fed pig. J Physiol, 379 :245–256, Oct 1986.
[23] C. Gariepy D. S.A. Pommier, C. Pomar. Petitclerc, influence du genotype, de la cas-
tration, de la somatocrinine et déunerestriction energetique sur la qualité des viandes
et les proprietés de transformation des carcasses de porcs. rapport final d l’entente
auxiliaire canada-quebec sur le developpement agro-alimentaire. Quebec, Canada
1993.
[24] J. Rivest, J. F. Bernier, and C. Pomar. A dynamic model of protein digestion in the
small intestine of pigs. J Anim Sci, 78(2) :328–340, February 2000.
[25] Livio C. Piccinini. Homogeneization problems for ordinary differential equations.
Rend. Circ. Mat. Palermo (2), 27(1) :95–112, 1978.
[26] M. Taghipoor, G. Barles, C. Georgelin, J. R. Licois, and P. Lescoat. Mathematical
homogenization in the modelling of digestion in the small intestine. September
2011.
[27] L.C. Evans. The perturbed test function method for viscosity solutions of nonlinear
PDE. Proc. Roy. Soc. Edinburgh Sect. A, 111(3-4) :359–375, 1989.
[28] L.C. Evans. Periodic homogenisation of certain fully nonlinear partial differential
equations. Proc. Roy. Soc. Edinburgh Sect. A, 120(3-4) :245–265, 1992.
[29] G. Barles, F. Da Lio, P.-L. Lions, and P. E. Souganidis. Ergodic problems and per-
iodic homogenization for fully nonlinear equations in half-space type domains with
Neumann boundary conditions. ArXiv e-prints, October 2009.

114
BIBLIOGRAPHIE

[30] A. V. Mernone, J. N. Mazumdar, and S. K. Lucas. A mathematical study of peristaltic

transport of a casson fluid. Mathematical and Computer Modelling, 35(7-8) :895–
912, 2002.
[31] D. Randall, W. Burggren, K. French, and R. Eckert. Eckert Animal Physiology :
Mechanisms and Adaptations. W.H. Freeman & Company, 4th edition, 1997.
[32] X. T. Zhao, M. A. McCamish, R. H. Miller, L. Wang, and H. C. Lin. Intestinal
transit and absorption of soy protein in dogs depend on load and degree of protein
hydrolysis. J Nutr, 127(12) :2350–2356, December 1997.
[33] J. Keener and J. Sneyd. Mathematical Physiology (Interdisciplinary Applied Mathe-
matics) 2 Vol Set. Springer, 2nd edition, December 2008.
[34] K.e. Yamauchi. Review of a histological intestinal approach to assessing the intesti-
nal function in chickens and pigs. Animal Science Journal, 78 :356–370, 2007.
[35] H. Ishii. Perron’s method for Hamilton-Jacobi Equations,. Duke Math. J, 55 :369–
384., 1987.
[36] M. G. Crandall, H. Ishii, and P. L. Lions. user’s guide to viscosity solutions
of second order partial differential equations. Bull. Amer. Math. Soc. (N.S.),
Bull. :[Link].27 1–67, 1992.
[37] G. Barles. Nonlinear neumann boundary conditions for quasilinear degenerate ellip-
tic equations and applications. Journal of Diff. Eqns, 154 :191–224, 1999.
[38] G. Barles. Solutions de viscosité des équations de Hamilton-Jacobi (Mathématiques
et Applications) (French Edition). Springer, 1 edition, March 1995.
[39] C. Wenk. The role of dietary fibre in the digestive physiology of the pig. Animal
Feed Science and Technology, 90(1–2) :21 – 33, 2001. <ce :title>The Role of Dietary
Fibre in Pig Production</ce :title>.
[40] B. McCleary and L. Prosky, editors. Advanced Dietary Fibre Technology. Wiley-
Blackwell, 1 edition, 2 2001.
[41] J. W Anderson, P. Baird, R. H. Davis, S. Ferreri, M. Knudtson, A. Koraym, V. Wa-
ters, and C. L. Williams. Health benefits of dietary fiber. Nutr Rev, 67(4) :188–205,
Apr 2009.
[42] A. Tharakan. Modelling of physical and chemical processes in the small intestine.
PhD thesis, University of Birmingham, December 2009.
[43] K.E. and Bach Knudsen. The nutritional significance of "dietary fibre" analysis.
Animal Feed Science and Technology, 90(1-2) :3 – 20, 2001. The Role of Dietary
Fibre in Pig Production.
[44] M. Siljeström, E. Westerlund, I. Björck, J. Holm, N.-G. Asp, and O. Theander. The
effects of various thermal processes on dietary fibre and starch content of whole
grain wheat and white flour. Journal of Cereal Science, 4(4) :315 – 323, 1986.

115
BIBLIOGRAPHIE

[45] E. Jiménez-Moreno, J. M. Gonzàlez-Alvarado, R. Làzaro, and G. G. Mateos. Effects

of type of cereal, heat processing of the cereal, and fiber inclusion in the diet on giz-
zard ph and nutrient utilization in broilers at different ages. Poult Sci, 88(9) :1925–
1933, Sep 2009.
[46] T. M. Burkhalter, N. R. Merchen, L. L. Bauer, S. M. Murray, A. R. Patil, J. L. Brent,
and G. C. Fahey. The ratio of insoluble to soluble fiber components in soybean hulls
affects ileal and total-tract nutrient digestibilities and fecal characteristics of dogs. J
Nutr, 131(7) :1978–1985, Jul 2001.
[47] M. Rezaei, M. A. Karimi Torshizi, and Y. Rouzbehan. The influence of different
levels of micronized insoluble fiber on broiler performance and litter moisture. Poult
Sci, 90(9) :2008–2012, Sep 2011.
[48] FAO, Rome : Food Report of the Joint FAO/WHO Expert Consultation. Paper 66,
and Agriculture Organization, editors. Carbohydrates in human nutrition. FAO
Food and Nutrition, 1998.
[49] Barbara O. and Schneeman. Dietary fiber and gastrointestinal function. Nutrition
Research, 18(4) :625 – 632, 1998.

116
Résumé :

L’objectif de cette étude est de modéliser la digestion dans l’intestin grêle : le trans-
port des aliments par les ondes péristaltiques, la dégradation par les enzymes endogènes
et exogènes et l’absorption active et passive. Un modèle mécaniste basé sur les équa-
tions différentielles ordinaires a été utilisé pour représenter la digestion. Les équations
décrivent l’évolution de la position et de la composition du bolus provenant de l’estomac.
Nous montrons ensuite par les méthodes d’homogénéisation mathématiques que ce mo-
dèle peut être considéré comme une version macroscopique des modèles plus réalistes,
qui contiennent des phénomènes biologiques à des échelles inférieures de l’intestin grêle.
Enfin, nous étudions l’influence du changement de la structure de bolus sur la digestion
en intégrant les fibres alimentaires dans sa composition. Les deux principales caractéris-
tiques des fibres alimentaires qui interagissent avec la fonction de l’intestin grêle, à savoir,
la viscosité et la capacité de rétention d’eau ont été modélisées.

Mots clés :
Modélisation, EDO, Digestion, Intestin grêle, Péristaltique, Homogénéisation, Solu-
tions de viscosité, Fibres alimentaires.

Abstract :

The purpose of this study is to model the digestion in the small intestine : transport
of the the bolus by the peristaltic waves, feedstuffs degradation according to the endo-
genous and exogenous enzymes and nutrients absorption. A mechanistic model based on
ordinary differential equations is used to represent the digestion. The equations describe
the evolution of the position and composition of the bolus of feedstuffs coming from the
stomach. We prove by using the homogenization methods, that this model can be consi-
dered as a macroscopic version of more realistic models which contain the biological
phenomena at lower scales of the small intestine. Finally, we investigate the digestion of a
non-homogeneous feedstuffs matrix by integrating the dietary fibre in the bolus. The two
main physiochemical characteristics of dietary fibre which interact with the function of
the small intestine, i.e. viscosity and water holding capacity are modelled.
Keywords :
Modelling, ODE, Digestion, small intestine, peristaltic, Homogenization, Viscosity
solutions, Dietary fibres.