Selection adjustments for matching genotypes to herds environments
An obvious selection adjustment to overcome the limits encountered with single-selection is to select on more than one trait (i.e. multi-trait selection). Multi-trait selection index are used to balance the selection emphasis on others traits of economic importance than solely production traits (Groen et al., 1997; Miglior et al., 2005). For this purpose, there has been a growing interest to record new traits and to develop reliable methods for estimating the parameters needed to predict selection response. Another possible selection adjustment is to define the environment and the selection criteria at a much finer-grained level of aggregation, especially because it can change the relative importance of traits. Once the environment has been characterized, customized indices can be used to indicate to farmers the best sires for their own herd conditions (Bowman et al., 1996). Although practical interest in using such indices instead of global or national indices is typically limited, it could become more attractive with the possibility to estimate more and more accurately environment-specific breeding values, based on animal trait measures and herd descriptors
Adjustments in the statistical methods used for genetic selection are likely to be effective for matching genotypes to herd environment. Part of their effectiveness, however, implies a trade-off between the accuracy of the estimated parameters used to make calculations (which generally require a substantial data set), and the customization of the selection index to accommodate the environmental particularities in which selection response takes place (Bourdon, 1998). The progress in computational power and the increasing quantity of information available on farm will probably help to overcome these difficulties in the future.
This opens promising perspectives for their integration into decision support tools that can help farmers. From this perspective, it would be crucial to carefully consider the role of the manager can who creatests the local selection environment scriteria and that on match-frm strategy,animal genetic modulates expression or not throughe herd time environment. and
Nutritional insights on adaptation
Short-term efficiency gains are largely achieved by an increase in production relative to non-productive functions (Veerkamp, 1998), i.e. there is an increased partition of nutrients towards milk. However, this change in partition has been shown to have unfavourable consequences on both reproduction and health status, in both cases to a large extent mediated by a decrease in body reserves (Rauw et al., 1998). The follow-on consequence of this is a decrease in longevity, and hence the length of the productive life of the animal. Thus, lifetime efficiency is decreased (because the productive portion of the lifespan declines). In this context, the adaptive capacity of animals and the role of nutrient partitioning are clearly important, especially with increasing environmental variability.
Two types of regulations in nutrient partitioning are now recognized (Friggens et al., 2013). Homeostasis refers to the regulations involved in the maintenance of the internal environment (Chilliard et al., 2000). They are directed to the survival of individual when they face some environmental disturbance, e.g. body reserves mobilization during underfeeding. Others regulations that have been less obvious among animal scientists are those referring to teleophoresis (Bauman and Currie, 1980; Chilliard, 1986; Friggens and Newbold, 2007). These regulations drive orchestrated changes of priority for the nutrient partitioning between functions and are directed towards the perpetuation of the specie. They are part of the genetic make-up of the animal, and their expression often explain why genetic correlation change for instance for body condition score during the stage of lactation (Pryce et al., 2001; Friggens, 2003). This clearly stresses an important connection between selection and nutrition. The modeling of teleophoretic and homeostatic regulations led to major advances in the prediction of multiple nutritional responses at the animal level (Sauvant, 1994; Martin and Sauvant, 2010). It represents also a strong conceptual basis to tackle the issue, as yet unexplored (Bryant et al., 2005), of integrating such animal model into a population/herd structure so as to test responses to selection.
Modeling as a tool for dealing with complexity
Modeling has been identified as a powerful tool for integrating complex relationships between genetic factors, the nutritional environment and time (McNamara, 2012). However, animal nutrition and genetic selection are classically viewed as separated models (Figure 1-2), probably because as they involve processes at different spatial and temporal scales. As previously mentioned, genetic improvement generally occurs at the level of national or regional populations. Based on the observed phenotypic traits of interest and on pedigree information, breeders select superior genotypes as parents (Figure 1-2A) and iterate the process over successive generations. By contrast, animal nutrition is practiced in individual herds with a wide range of different feeding systems often related to the local farm environment. Moreover, herd managers often adjust animal nutrition according to their short-term response and their reproductive stage as they know that animals change their nutrients partition between functions (Figure 1-2B).
Selection and nutrition operate thus at very different scales. Though, nutritional responses are well recognized as being regulated by genetic drives (Bauman and Currie, 1980; Chilliard et al., 2000). Theoretically, this component should be thus partly inherited from parents to offspring. In turn, within-life performances shape the response to selection, in particular through their effects on reproduction and survival rates. This suggests that the selection and the nutrition processes interact. A view that brings them together would be thus valuable at the herd level to provide more insights on how to take advantage of G × E interactions by the means of genetic selection and herd management (Figure 1-2C). This is clearly not a simple issue to address. Modeling represents a powerful tool to deal with such complexity; however it requires an appropriate theoretical basis to do so.
Managing G × E interactions at the herd level
The resource allocation approach
The integration of selective breeding and nutrient partitioning approaches to be able to manage and exploit G×E could benefit from other theoretical developments in biology. Resource allocation principles have been developed by evolutionary biologists who were interested in why some wild species are found in particular ecosystems, and how can it result from the interactions between natural selection and ecological processes (Williams, 1966; van Noordwijk reflected and deJoing,the1986;terms Stearns ,allocation 1992) and.One partition in gaspect of , the which an are logy effectively wih farm synonymous animal sciences.Asubtleis difference however is that allocation refers to the action of assigning resource for a particular purpose whereas partitioning refers to the division into parts, without any notion of achieving ofspecificevolutionaryfunctionsgoals.This differenceforanimalsunderlinthatseektheto explicitincrea consideration, by life history scientists, possibly various strategies . se their natural fitness by the mean of This conceptual distinction with a passive view of animal functioning traditionally held in animal nutrition has also emerged in animal science (Friggens and Newbold, 2007). The remaining aspect of the analogy lies in the parallel between individual natural fitness and the selection index used in selective breeding, as both reflect an adaptation measure of individuals to their environment but with a different– weighting of traits. A central tenet of– this thesis is that both aspects of the framework resource allocation and– natural fitness can be applied and adapted– to existing concepts in farm animal science nutrient partitioning and selection index , and thus provide an attractive way for connecting genetic selection and animal nutrition in a same view (Beilharz et al., 1993; Rauw et al., 1998; Friggens and Newbold, 2007; Rauw, 2009).
In life-history biology, resource allocation principles have been applied in a number of models to contribute insight into how the development of associations between traits can emanate from physiological constraints. The basic constraint which is postulated is that resources obtained from the environment are limited so that animals should find an optimal allocation between growth, reproduction and survival in order to maximize their fitness (Williams, 1966; Stearns, 1992). An interesting model outcome is that such physiological constraints are not always reflected in the observed phenotypes, and are therefore sometimes counter-intuitive (van Noordwijk and de Jong, 1986). This is illustrated in Figure 1-3 with a basic model of resource allocation, in which individual animals acquire a limited amount of resource (here called R) that they– allocate in a given proportion (here called c) to one trait 1 and in the remaining proportion (1 c) to another trait 2 (right corner). The line in panel A represents a same level R of resource acquired at the population level. Different positions on this line reflect different allocation c between traits 1 and 2. Then, if animals vary more in their resource allocation than in their acquisition, traits compete for using the resource acquired and a negative correlation emerges (A). Conversely if animals vary more in the quantity of the resource acquired than in the way they allocate it between traits, a positive relationship between traits emerges (B). The important assumption here is the existence of a genetic variation in c which makes the link between the individual model and the population response across generations. Such a genetic variation is the basis of the selection response (Reznick, 1985).
The previous finding is particularly useful for breeding situations where one is interested in either selecting animals with favourable allocation or manipulating the nutritional environment in which the resource is obtained. From these theoretical considerations, it would be expected that a trade-off between traits would emerge as response to selection for trait 1 or 2, only if the nutritional environment limits the resource acquired (Beilharz et al., 1993). However it would require the quantification of genetic and environmental effects on the observed variations in acquisition and allocation parameters. A first step towards this objective has been applied with a similar model (Van Der Waaij, 2004), using a time-step of a generation. It has been found theoretically relevant regarding to selection (Bijma, 2009) and recognized as a good starting point for integrating genetic and nutritional factors (Friggens and Van Der Waaij, 2009).
Figure 1-3: An individual model of resource allocation between two traits (A and B) and its consequences, at the population level, on the relations between traits and under two scenarios of variation in acquisition (R) and allocation (c). In population responses, each point represents one individual. Adapted from Van Noordwijk and De Jong (1986).
Herd management: a possible lever to alleviate trade-offs among traits
Because of the natural selection context for which they have been originally developed, resource allocation models generally have a coarse grain definition of animal environment both in space and time (Stearns, 1992). Seasonal variations of the environment are often assumed to limit the available resource for animals, which seems reasonable and apparently consistent with evolutionary strategies that can be found in wild species such as hibernation or migration. Domestication has considerably alleviated extrinsic limitations of the resource obtained (Beilharz and Nitter, 1998), by both improving the resource availability from the environment and by reducing the foraging effort. Moreover, for farm animals herd management determines the controlled environment in which animal performs and thereby may complicate the global herd response compared to the simple view proposed in Figure 1-3 (A and B).
Two main mechanisms influence the herd response. The first one concerns the emergence of herd properties from a variability of individual responses which can complement each other. This variability can directly result from voluntary decisions of the manager, for instance in terms of herd configuration. Dairy farmers can thus stabilize herd milk supply throughout the year by managing a diversity of lactation stages. With seasonal species such as goat, this can be done in spite of the innate reproductive seasonality, in particular through the use of daylight treatments. In others cases, the variability is an involuntarily consequence of herd management, mediated by the interaction with animal biology. For instance, the feeding plan, decided at a herd or group level, can create discrepancies between animal requirements and nutrient supply at the individual level. Interestingly, different feeding options can lead to the same level of efficiency but it involves different levels of biological solicitation (Puillet et al., 2011), thereby increasing the risk of involuntarily culling. In the past, individual variability was frequently considered as a positive attribute only for extensive systems, in particular with the use of different species (Tichit et al., 2004). Nowadays, there is a growing interest to study the advantages of managing individual variability in a larger spectrum of management situations (Tichit et al., 2011), including mono-specific intensive systems (André et al., 2010). The study of individual variability may thus benefit from the considerations of genetic differences and their change by the mean of genetic selection. A second mechanism concerns the cumulative long-term effects. It underpins the fact that effects of short-term decisions may not be reflected instantaneously in animal responses but only after certain time decay. Body reserves have an important role in these effects as they constitute the energetic capital that animals can either use to buffer nutrient variations in the environment or build to anticipate its use during high energy-demanding periods of their reproductive cycle, like in early lactation. Therefore, a strong biological investment during the current reproductive cycle, in terms of high mobilization or a low deposition of body reserves, can impair the performances of the future reproductive cycle (Walsh et al., 2011). There are feeding options which can take advantage of this capacity, especially for improving lifetime productivity (Rufino et al., 2009).
Sometimes there are good reasons to consider separately genetics and nutrition. The complexity of animal adaptation in herd systems suggests that there are also reasons to see how they interact (Figure 1-4). Herd management is more than just a provision of nutritional resource and the previous part suggests that some management opportunities could exist to alleviate trade-off between traits (due to the summed effects of individual responses or due the cumulative effects through time, or both). However, investigating these opportunities would require representing the biological variability which is generated within individual across lifetime (in response to feeding and reproduction management practices), and between successive generations (in response to the selection of individuals which produce and reproduce within the herd). To do so, a representation of the resource allocation process would require i) the incorporation of heritable traits that respond to selection, ii) the teleophoretic and homeostatic regulations that drive the resource allocation, and iii) the possible variations in the resource obtained, including those from the body reserves. )nfluence of herd management on animal s resource allocation between the traits of performanceFigure1-4:. Animal s response to reproductive inputs activates the genetic drives of resource acquisition (R) and allocation (c) throughout its life. This is expressed in the time-profile of performance (P1 and P2), as a result of the genotype (G) interaction with the herd environment (E). Depending on their performance, some genotypes are selected to produce further in the herd whereas others are culled. Offspring of the best animal can also be selected to bring about genetic progress within the herd.
Interest for the design of sustainable systems
An important issue that arises with the idea of combining selection and herd management is why particular selection objectives would be suitable or not for a giving herd system. Herd systems exhibit a large diversity not only in terms of production environment (e.g. feeding system, geography), but equally in terms of objectives. Moreover, the global environment in which farms evolve may require global adjustment of the selection objectives for instance to accommodate a global change in climate (Nardone et al., 2010). Giving these conditions, an important practical issue would also be to determine the relevant scale for the design of sustainable selection objectives.
To address previous questions, an approach is to simulate the effects of farmer s management decisions on herd performance. However, this requires a credible farm decision model that is structured so as to be linked with an animal model. In particular, the animal model should be able to represent the possible constraints that emerge from the interaction between animal s biology and the herd production environment. These interactions have to be modeled over a span of time which makes sense with regard to sustainability criteria.
Research objective and case study
In this thesis, weanimalpropos eperformancethatthemanthatgemintegratesof Gboth× Etheat effectstheherdof selectionlevelrequiresandthea representation of effects of the herd production environment. The introduction above pointed out the potential interest of such integration for both genetics and animal production scientists but equally the conceptual challenge that it represents. Accordingly, the objective of this thesis was to describe the elaboration of animal performance so as to explore the long-term consequences of the interaction between genetic selection and management within a herd.
We previously pointed out the relevance of modeling as a tool to deal with the complexity of articulating the physiological and the genetic dimensions of animal performance. To tackle the issue, as yet unexplored, of incorporating genetic parameters within a nutritional model (Friggens and Newbold, 2007; Friggens et al., 2013), we elaborated on considerations from the resource allocation theory. Choosing this theory was motivated by the considerable available knowledge in the field of evolutionary biology, its importation and explanatory potential in livestock science (Beilharz et al., 1993; Rauw, 2009) and by the relevance of a previous model application (Van Der Waaij, 2004) as a starting point.
From a resource allocation perspective, a negative relationships between traits during selection results from functional relationship between these traits (Zera and Harshman, 2001). If they share a common pool of internal resources and if this pool is limited then a trade-off should emerge during selection for one trait because an increment of resource allocated for this necessitates a decrement of the resources allocated to the other trait. Clearly, others causes of negative relationships may exist such as genetic linkage or phylogenetic constraints (Stearns, 1992; Zera and Harshman, 2001). However, this is not under the scope of the resource allocation theory and the present thesis. Neither are the physiological causes (i.e. hormonal control) underlying the functional relationships between traits considered. Instead, the resource allocation theory used in this thesis focuses on illuminating the mechanisms of animal adaptation during the selection process. It aims to emphasize the functional relationships that constrain adaptation without, however, making explicit the underlying physiology.
Table of contents :
Chapter 1: INTRODUCTION
1 Context of the thesis
2 On the relation between genetic improvement and herd environments
2.1 Success of adapting herd environments to improved genotypes
2.2 Environmental limits and challenges emerging in the context of genetic improvement
2.3 Selection adjustments for matching genotypes to herds environments
2.4 Nutritional insights on adaptation
2.5 Modeling as a tool for dealing with complexity
3 Managing G × E interactions at the herd level
3.1 The resource allocation approach
3.2 Herd management: a possible lever to alleviate trade-offs among traits
3.3 Interest for the design of sustainable systems
4 Research question and case study
Chapter 2: GENERAL APPROACH
1 EXPLAIN: development of a herd simulation model
1.1 Model overview
1.2 Animal sub-model of resource acquisition – allocation
2 DESCRIBE: a time-profile of dairy goat performance
2.1 Model calibration on a meta-profile
2.2 Description of extended lactation from real data
3 EXPLORE: Simulation experiments
3.1 Description of G and E in the scenarios
3.2 Scenario simulated of G × E
Chapter 3: RESULTS
1 Exploring different selection strategies in an abundant environment
1.1 Contrasted strategies tailored different resource allocation
1.2 Consequences of a varying managers priority for milk production efficiency
2 Description of extended lactation (EL)
3 Exploration of selection strategies allowing for the use of EL in a variable environment
Chapter 4: DISCUSSION
1 An application of the resource allocation approach to the herd context
1.1 Outcomes from the resource allocation approach
1.2 The desired resource intake
1.3 Environmental variability reveals the evolutionary importance of body reserves
1.4 Extended lactation as an animal tactic to deal with the cost of reproduction
1.5 Implications for the management of G × E: from constraints to opportunities
2 The framework
2.1 The emergence of G × E interactions at the herd level
2.2 Management influences in the local herd environment
2.3 Distinguish the selection practice from the selection action
2.4 Model limitations and evaluation
3 Future prospects
3.1 Towards a generic model to manage G × E interactions: the problem of estimating the c allocation components
3.2 Scale issue for the design of sustainable management systems
3.3 Towards another research approach iteration: describe the herd environment and explain its dynamic