ruminal, à savoir la diffusion passive ou le transport facilité. Cependant, le modèle ne tient compte que d’une petite fraction (45%) des proportions molaires d’AGV.
Néanmoins, les modèles existants ne permettent pas de prédire quantitativement tous les nutriments apparaissant en veine porte. Il apparaît clair aujourd’hui que le développement conjoint des modèles empiriques et des modèles mécanistes à compartiments est prometteur pour progresser et relever les défis actuels (Martin et Sauvant., 2007).
La première partie de mon travail de thèse a ainsi consisté à établir des modèles de prédiction de l’apparition nette en veine porte (ANP) de l’ensemble des nutriments énergétiques (AGV totaux, profil en AGV, glucose, β -hydroxybutyrate, L-lactate et azote-α- aminé). Nous avons pour cela recherché les meilleurs prédicteurs parmi les critères de caractérisation des aliments du système INRA (2007) pour ruminants. Nous nous sommes attachés à mettre en évidence des critères facilement mesurables et suffisamment génériques pour pouvoir intégrés, à terme, dans les pratiques de rationnement.
Pour caller la démarche, il a été choisi dans un premier temps de modéliser l’ANP du glucose. En effet, les travaux d'Offner et Sauvant (2004) avaient montré qu'il était possible à partir des critères INRA de prédire la part d'AMNdR. Ce travail a donné lieu à une communication orale au congrès international ISEP (International Symposium on Energy and Protein Metabolism) à Vichy en 2007, dont le texte est repris en annexe (Communication n°4).
Dans un deuxième temps, nous avons appliqué une démarche similaire pour développer des modèles de prédiction de l’ANP des AGV totaux, du profil en AGV, du glucose, du β - hydroxybutyrate et du L-lactate, à partir de critères issus du système INRA de caractérisation des aliments pour ruminants. Ce travail a fait l’objet d’une publication scientifique de rang A, qui a été publiée dans J. Anim. Sci. et constitue la première partie des résultats présentés ici (Publication n°1).
Afin de prédire l’ensemble des nutriments énergétiques apparaissant en veine, il était
des nutriments, communément évaluée d'après la consommation d’oxygène des tissus digestifs. La prédiction de la consommation d’oxygène par les tissus drainés par la veine porte (Communication n°2) a été réalisée lors du stage de Master 2 (Statistiques et traitement des données, Université Blaise Pascal, Clermond Ferrand) de Sylvie Amblard, que j’ai encadrée.
Ce stage s’est étalé de mars à septembre 2008. Il a permis de proposer une communication qui a été acceptée sous forme de poster au congrès organisé par l’ASAS ADSA CSAS (Joint Annual Meeting) qui s’est tenu en 2009 à Montréal.
Enfin, dans le but de réaliser une évaluation quantitative des différents modèles établis et de s'assurer qu'il était possible de faire évoluer le concept d'Energie Métabolisable vers la notion de Nutriments, nous avons cherché à évaluer la cohérence quantitative entre le concept de Nutriments Absorbés (la somme des nutriments énergétiques apparaissant en veine porte et oxydés par les tissus drainés par la veine porte) et l’Energie Métabolisable. Ce travail a fait l’objet d’une communication orale au 7e Workshop International: Modelling Nutrient Digestion and Utilization in Farm Animals, qui s’est tenu à Paris (2009) et qui est présenté en Communication n°3.
Ces différentes étapes ont donc permis de mettre en évidence qu’il était possible de prédire de manière satisfaisante la quantité et la nature des nutriments énergétiques apparaissant en veine porte à partir des critères issus du système INRA. L’évolution et le rayonnement international du système américain d’évaluation des aliments pour ruminants nous a conduit à réaliser le même travail de prédiction de l’ANP des nutriments à partir des critères de caractérisation des aliments issus du NRC (National Research Council, 2001). Ceci a permis dans un premier temps de vérifier si les modèles de prédiction obtenus à partir des critères INRA pouvaient être améliorés par l’emploi de critères issus d’autres systèmes, en particulier celui du NRC. D’autre part, ce travail a permis de comparer la capacité des systèmes INRA et NRC d’évaluation des aliments à prédire les flux de nutriments apparaissant en veine porte. Compte tenu du temps disponible pour ma thèse, seuls les modèles d’ANP des AGV ont été établis (Publication n°2, en cours de correction).
Publication n°1 :
Prédiction empirique de l’apparition nette en veine porte
des acides gras volatils, du glucose et de leurs métabolites secondaires ( β -hydroxybutyrate, lactate) à partir des caractéristiques de l’alimentation
chez les ruminants : approche par méta-analyse
C. Loncke, I. Ortigues-Marty, J. Vernet, H. Lapierre, D. Sauvant and P. Nozière
Cet article a fait l’objet d’une publication dans Journal of Animal Science 87: 253-268.
Résumé:
L’évolution actuelle des systèmes d’alimentation énergétique pour les ruminants vers un système basé sur les nutriments exige de caractériser l’énergie apportée par l’alimentation en termes de quantité et de nature des nutriments absorbés. L'objectif de cette étude était d'établir des équations de réponse de l’apparition nette portale (ANP) des AGV et du glucose ainsi que de leur métabolites secondaires, β -hydroxybutyrate (BHBA) et lactate, aux variations du niveau d’alimentation et des caractéristiques chimiques des rations basées sur des critères de caractérisation des aliments issus du système INRA d'évaluation des aliments pour ruminants.
Des techniques de méta-analyse ont été appliquées sur les données de la base FLORA qui est une compilation exhaustive des publications internationales traitant des flux nets splanchniques de nutriments chez les ruminants multi-cathétérisés. Pour chaque nutriment plusieurs variables prédictives ont été testées. Nous avons obtenu des modèles robustes pour des quantités ingérées allant jusqu’à 30 g MS/j/kg PV et des rations comprenant moins de 70 g de concentrés/100 g MS. Ces modèles ont permis de prédire l’ANP (mmol/h/kg PV) des AGV totaux à partir des quantités de matière organique fermentescible dans le rumen (MOF) ingérées [R² ajusté (R² adj) = 0,95; erreur résiduelle du modèle (RMSE) = 0,24], de prédire le profil en AGV (mol/100 mol AGV totaux) à partir de la nature de la MOF ingérée (acétate : R² adj = 0,85; RMSE = 2,2; propionate : R² adj = 0,76, RMSE = 2,2; butyrate : R² adj = 0,76 ; RMSE = 1,09), et de prédire l’ANP (mmol/h/kg PV) de glucose à partir des quantités d’amidon digéré dans l’intestin grêle indépendamment de l’espèce, et avec des modèles ne présentant pas de facteurs interférents sur les pentes individuelles.
Le modèle permettant de prédire l’ANP (mmol/h/kg PV) de BHBA à partir des quantités de MOF ingérées (R² adj = 0,91; RMSE = 0,036) était dépendant de l’espèce, et le modèle prédisant l’ANP (mmol/h/kg PV) du lactate à partir de l’amidon digéré dans le rumen (R² adj
= 0,77; RMSE = 0,042) présentait une large dispersion. Cependant, l’ANP (mmol/h/kg PV) de BHBA était reliée à l’ANP de butyrate (R²adj = 0,85; RMSE = 0,054) et d’acétate (R² adj
= 0,85; RMSE = 0,052), et l’ANP (mmol/h/kg PV) du lactate était reliée à l’ANP du propionate (R² adj = 0,51; RMSE = 0,096). Cette étude a montré qu’il était possible de prédire de manière satisfaisante la quantité et la nature des flux de nutriments à partir des caractéristiques des rations chez les moutons et les bovins. Ce travail a pour ambition de quantifier les conséquences de la digestion et du métabolisme des tissus drainés par la veine porte sur la disponibilité des nutriments. Ces résultats peuvent augmenter les connaissances sur les processus biologiques et participer au développement de nouveaux outils de formulation des rations.
Empirical prediction of net portal appearance of volatile fatty acids, glucose, and their secondary metabolites (beta-hydroxybutyrate, lactate) from dietary characteristics in
ruminants: a meta-analyses approach1
C. Loncke,* I. Ortigues-Marty,* J. Vernet,* H. Lapierre, † D. Sauvant, § P. Nozière, *2
* Institut National de la Recherche Agronomique, UR 1213, Unité de Recherches sur les Herbivores, Theix, 63122 St Genès Champanelle, France
†Dairy and Swine Research and Development Centre, Agriculture and Agri-Food Canada, Sherbrooke, Quebec JIMIZ3, Canada
§UMR 791, Physiologie de la Nutrition et Alimentation, INRA-AgroParisTech, 75231 Paris, France
1The authors wish to thank INZO and LIMAGRAIN as well as the Association Nationale de la Recherche Technique for funding this project.
2Corresponding author:
The current trend in energy feeding systems for ruminants toward a nutrient-based system requires dietary energy supply to be determined in terms of amount and nature of absorbed energy-yielding nutrients. The objective of this study was to establish response equations on the net portal appearance (NPA) of VFA and glucose, and their secondary metabolites β- hydroxybutyrate (BHBA) and lactate, to changes in intake level and chemical dietary characteristics based on the Institut National de la Recherche Agronomique Feed Evaluation System for Ruminants. Meta-analyses were applied on published data compiled from the FLORA database, which pools the results on net splanchnic nutrient fluxes in multi- catheterized ruminants from international publications. For each nutrient, several prediction variables were tested. We obtained robust models for intakes up to 30 g of DM·d−1·kg of BW−1 and diets containing less than 70 g of concentrate per 100 g of DM. These models were designed to predict the NPA (mmol·h−1·kg of BW−1) of total VFA based on the amount of ruminally fermented OM (RfOM) intake [adjusted R² (R² adj) = 0.95; residual means square errors (RMSE) = 0.24], to predict VFA profile (mol/100 mol of total VFA) based on type of RfOM intake (acetate: R² adj = 0.85, RMSE = 2.2; propionate: R² adj = 0.76, RMSE = 2.2;
butyrate: R² adj = 0.76, RMSE = 1.09), and to predict the NPA (mmol·h−1·kg of BW−1) of glucose based on the starch digested in the small intestine independent of ruminant species, and while presenting no interfering factors on the residuals and individual slopes. The model predicting the NPA (mmol·h−1·kg of BW−1) of BHBA based on the amount of RfOM intake (R² adj = 0.91; RMSE = 0.036) was species-dependent, and the model predicting NPA (mmol·h−1·kg of BW−1) of lactate based on starch digested in the rumen (R² adj = 0.77;
RMSE = 0.042) presented a wide dispersion.
However, the NPA (mmol·h−1·kg of BW−1) of BHBA was related to the NPA of both butyrate (R² adj = 0.85; RMSE = 0.054) and acetate (R² adj = 0.85; RMSE = 0.052), and the NPA (mmol·h−1·kg of BW−1) of lactate was related to the NPA of propionate (R² adj = 0.51; RMSE
= 0.096). This research showed that it is possible to accurately predict the amount and nature of absorbed nutrient fluxes based on dietary characteristics in both sheep and cattle. This work aims to quantify the consequences of digestion and portal-drained viscera metabolism on nutrient availability. These results can provide deeper insight into biological processes and help develop improved tools for dietary formulation.
Key words: diet composition, energy, intake level, meta-analyses, portal-drained viscera, ruminant
INTRODUCTION
The feeding systems developed for ruminants were designed to adjust feed allowances to fit the production potential of the animals (Vermorel and Coulon, 1998). Major progress has been made in feed characterization for predicting ME intake, but energy is still evaluated as an aggregated unit. The new challenges of ruminant nutrition require that feed evaluation systems evolve toward more precise and mechanistic nutrient-based systems (Reynolds, 2000). Improvements to diet formulation depend on our capacity to predict the amount and the nature of the nutrients absorbed by the portal-drained viscera (PDV;
Dijkstra et al., 2007). Bermingham et al. (2008) used meta-analyses to quantify the incremental responses of the net portal appearance (NPA) of energy-yielding nutrients to increases in DMI and digestible OM (dOM) intake at fixed diet composition, but the prediction of nutrient absorption after a change in diet composition was not addressed. The objectives of the present work were to compare several descriptors of ruminant diets for their ability to predict the NPA responses of nutrients derived from digestion (i.e., VFA and glucose) or from gut metabolism [i.e., β-hydroxybutyrate (BHBA) and lactate]. The response
MATERIALS AND METHODS
Animal Care and Use Committee approval was not obtained for this study because the data were obtained from an existing database (FLORA; Vernet and Ortigues- Marty, 2006).
Selection of Publications from the FLORA Database
The FLORA database was used. It is an exhaustive base from approximately 200 international publications reporting results of net splanchnic nutrient fluxes in multi-catheterized ruminants. The Scopus and Current Contents Connect online bibliographic databases were queried. The structure of FLORA has been described in detail by Vernet and Ortigues-Marty (2006). Animals were identified according to species (sheep, cattle) and physiological status (nonproductive adult, growing, gestating, or lactating animals). Chemical composition and calculated nutritional value of feeds and diets, plus catheter location, blood sampling, metabolite analyses methods, and results for blood or plasma flows, metabolite concentrations, and fluxes were entered into the FLORA database after systematic verification steps. In the present study, only a limited number of publications reported results for each target nutrient after a dietary treatment. There were 29, 55, 38, and 29 publications for total VFA, glucose, lactate, and BHBA, respectively. These publications were selected as being relevant to studying the response of nutrient NPA to changes in dietary intake and composition.
Description and Calculation of the Nutritional Values of Diets
Publications featuring net splanchnic nutrient fluxes in ruminants do not systematically report the chemical composition and nutritional value of feeds and diets, making it necessary to produce a consistent description of the chemical composition and nutritional values of the feeds and diets used in all relevant published reports. We chose to describe all feeds according to the INRA Feed Tables (INRA, 2007), which give detailed chemical composition and nutritional values for about 160 concentrates and 1,260 forages. After selection of the most representative feed ingredients, the estimated chemical composition and nutritional values of the diets were calculated assuming additivity. The estimated values were validated by comparison with the values reported by the authors. Comparisons focused primarily on OM digestibility, ME and CP concentrations, and subsequently on crude fiber concentrations in DM, which were reported in 18, 45, 60, and 15% of the selected publications, respectively. If a publication did not report at least one of these criteria, it was excluded from the models.
Differences between the estimated values and those reported by the authors were analyzed by GLM analyses including a publication fixed effect. If the difference was greater than the measurement uncertainties for OM digestibility (±2.5 g/100 g), ME concentration (±1.0 MJ/kg or DM), or CP concentration (±12.5 g/kg of DM), then the publication was excluded from the models. If the difference was greater than the measurement uncertainty for crude fiber concentration (±17.5 g/kg of DM), then the publication was only excluded from the models if it was subsequently identified as an outlier.
Choice of Explanatory Variables to Predict Variations in Nutrient NPA
Considering the importance of digestibility in the nutritive value of ruminant diets, dOM intake level was considered as a predictor of the variations in NPA for all studied nutrients (VFA, BHBA, glucose, and l-lactate). The proportion of concentrate in the diet (DM basis), albeit partly linked to dOM intake, was also factored in. Other potential predictors were identified for each individual nutrient, based on current knowledge of digestion and PDV metabolism, as described below. The first criterion for predicting the NPA of total VFA [i.e., acetate (C2) plus propionate (C3) plus butyrate (C4)] was the amount of ruminally fermentable OM (RfOM) intake. Ruminally digestible NDF (RdNDF) intake and ruminally digestible starch (RdS) intake were also tested as predictors of total VFA NPA. Variations in molar proportions of individual VFA and the C2:C3 ratio in the net portal flux were predicted from the dietary concentration of RdNDF and RdS, with their respective contribution to RfOM (RdNDF:RfOM, RdS:RfOM) used as indexes of the type of fermentable energy because of the absence of link between the molar percentage of VFA and DMI (Bermingham et al., 2008). Soluble sugars in feeds could not be taken into account, because they are highly variable and insufficiently reported (INRA, 2007). Ruminally fermentable OM concentrations were calculated according to INRA (2007; i.e., RfOM = dOM − fat − undegradable CP − undegradable starch − ferment ation products, where undegradable CP and undegradable
OM – undigestible CP – undigestible fat – undigestible starch)]. Values for undigestible OM, undigestible CP, undigestible fat, and undigestible starch were based on the INRA Feed System (INRA, 2007), assuming a constant fat digestibility of 80% (Doreau and Ferlay, 1994). These calculations were validated by comparing estimated with measured data (Sauvant et al., 2008a). Concentrates accounted for less than 20% of dietary digestible NDF.
Dietary digestible NDF was then calculated by additivity. Variations in BHBA NPA were predicted from the same explanatory variables as for VFA (i.e., RfOM intake, RdNDF intake, and RdS intake). Furthermore, the responses of the C4:BHBA ratio to the percentage of concentrate and RdS, RdNDF, RdNDF:RfOM, and RdS:RfOM contents were tested.
Variations in glucose NPA were predicted by calculating the amount of starch digested in the small intestine (SIdS). Because the meta-analyses approach requires a minimum variation in the explanatory variable, the model based on SIdS only included the experiments with at least 1 starch-containing diet. The amount of SIdS was predicted from in sacco starch degradation (INRA, 2007) and the empirical model of ruminal and intestinal starch digestion (Offner and Sauvant, 2004). For diets containing no starch, potential predictors of glucose NPA were RdNDF intake, RfOM intake, and DMI. Given that lactate NPA results from both ruminal fermentation and net production (from C3, valerate, or glucose) by the PDV (Leng et al., 1967; Weeks and Webster, 1975; Weigand et al., 1975), we used the same explanatory variables as for the NPA of VFA and glucose (i.e., RfOM intake, RdNDF intake, RdS intake, and SIdS). Relationships between the NPA of lactate (Y) and the NPA of glucose or C3 (X) were tested on all the publications reporting results on the effects of diet on the relevant NPA.
Similarly, relationships between the NPA of BHBA (Y) and the NPA of C4 and C2 (X) were also tested with a variance-covariance model.
Meta-Analyses
Data Coding. To select relevant experiments for subsequent meta-analyses, all publications within FLORA were coded in several steps (Sauvant et al., 2005). A first coding step dealt with the experiments and the type of experimental factors within each publication. Then, within an experiment, experimental groups were defined as the groups of treatments that changed due to the variable of interest and were coded accordingly (Sauvant et al., 2005;
Vernet et al., 2005).
Data Expression and Minimum Between-Treatment Variation on the Explanatory Variables. Nutrient intake and NPA were expressed as a function of BW1.0. Indeed, although across-species metabolic processes are best described as a function of BW0.75 (Brody, 1964),
intake as well as nutrient digestion and absorption have been shown to be similar between sheep and cattle when expressed as a function of BW1.0 (Vernet et al., 2005; Sauvant et al., 2006a). Calculation of reliable within-experiment responses requires a minimum variation on each explanatory variable (X) within that experiment (Sauvant et al., 2005). Therefore, the minimum acceptable variation of X (Δ Xmin) was determined from all the selected publications for each target nutrient, as follows:
Δ Xmin = μ (Δ Xij) − | 2 × SD Δ X |;
with Δ Xij = |Xtreatment i – Xtreatment j| and SD Δ X = SD of Δ Xij.
Thus, variations of X were considered relevant for the subsequent statistical analyses when the variation between within-experiment control and experimental treatments was greater than 10 g of concentrate DM/100 g of total DM, 30 g of RdNDF/100 g of DM, 25 g of RdS/100 g of DM, 1.2 g of dOM intake·d−1·kg of BW−1, 1 g of RfOM intake· d−1·kg of BW−1, 0.8 g of RdNDF intake· d−1·kg of BW−1, 0.5 g of RdS intake· d−1·kg of BW−1, 0.2 g of SIdS· d−1·kg of BW−1, 0.1 g of RdNDF/g of RfOM, and 0.1 g of RdS/g of RfOM. This selection led to the elimination of only 3 to 15% of the treatments depending on the explanatory variable.
Meta-Design, Variance-Covariance Models, and Post-Optimization Analyses. Descriptive statistics (μ, SD, range of values) were generated for each variable in the selected data sets as well as matrix of correlation between all the nutritional parameters. Normal distribution of data and homogeneity of variances were tested by Shapiro-Wilk and Levene tests, respectively. Relationships between Y (nutrient NPA, molar percentage of VFA, C2:C3, or C4:BHBA in net portal flux) and the explanatory variable X were studied with a variance- covariance model: Y = α + βX + species + αi (species) + species × X + e,
where α = the overall intercept and αi = the effect of the experimental group i on the intercept α, nested within species, and β is the slope of the relationship. If appropriate, quadratic models were also fitted to the data and compared with the linear model. The fit of each relationship was then examined by studying the Studentized residuals of the model and