Deliverable D5.2 New traits to select for feed efficiency

Objectives

Selection in monogastric species is applied to pure lines in selection farms, while commercial animals are crossbreds raised in diverse conditions. Improving feed efficiency in these populations is a key to enhance the productivity and limit the environmental footprint of livestock production. However, recording feed efficiency is costly in most species because it requires measuring feed intake, and this trait is affected by genotype by environment interactions (GxE) that makes it a different trait in selection and commercial populations. Major gains in genetic progress could thus be achieved if more animals had records for feed intake or feed efficiency, and if these measurements could be obtained from any farm. Identifying new traits for selection of feed efficiency is thus crucial to improve the prediction accuracy of breeding values in livestock populations.

Depending on the species, measuring feed efficiency on-farm is a difficult issue: individual feeders for pigs have been available on-farm for long. They are costly to acquire and maintain, but at least they provide reference measurements in most populations. Poultry and rabbits still rely on measurements made in individual cages. This type of measurement is not representative for the performance of animals raised in groups, and is questioned in terms of welfare. Thus, our objectives were:

1. To test direct measures of feed intake and feed efficiency for genetic designs after the development of electronic feeders in WP2 for rabbits
2. To evaluate measures of components of feed efficiency (i.e., digestibility, activity and behaviour, robustness), which could be used to select more efficiently individuals dedicated to different breeding conditions when GxE is large. A major effort was undertaken to understand the contribution of the gut microbiota to feed efficiency and its potential as a criterion for selection, which is reported in a separate deliverable (D5.1)
3. To identify biological markers of feed efficiency and their components that could be measured on a large number of individuals at a moderate cost, potentially on production farms, so that selection accuracy for production conditions could be improved.

Rationale

To respond to these objectives, data and technologies from WP2 (i.e., new traits for feed efficiency) and new trials were combined to evaluate feed efficiency under a wide range of conditions, including different feed resources, different breeding systems, and different physiological stages of the animal. Indeed, reproduction has long been ignored when considering feed efficiency issues, while it has a major impact on management of body reserves and on female longevity. To ensure that the proposed solutions would not have a negative impact on other production traits of interest, indicators of robustness and product quality were recorded.

Classical genetic methodologies have been applied, either by comparing genetic lines selected for the trait of interest for multiple generations so that the genetic difference between animals for this trait has been established (i.e., the residual feed intake (RFI) lines in pigs, rabbits and layers) and the correlated response on other traits or indicators can be measured, or by measuring the traits in large cohorts of conventional populations (i.e., Large White and crossbred pigs, Caldes rabbits) or alternative lines (i.e., Duroc and Iberian pigs). When using selected lines, the genetic aspect of the response can be observed directly by comparing the mean line responses, whereas in large pedigreed cohorts,animal linear mixed models were applied to estimate genetic variances and heritabilities. To detect genomic markers associated with the traits of interest, the same animal linear mixed models can be used, including a SNP effect in iterative tests along the genome. Because feed intake is sometimes not available at the individual level, an original model based on a bivariate development of the same linear model was tested in rabbits to detect associated SNP with traits recorded in groups. Finally, multivariate models dedicated to the simultaneous analysis of large number of variables were used in transcriptomic studies to account for the number of repeated tests and the specificity of these data.

In a first set of analyses, new traits could be validated as heritable in the tested populations. For growing animals, these traits include measures of components of feed efficiency, such as feed intake records measured by automatic feeders in rabbits and digestibility indicators measured in group-housed pigs (i.e., through direct NIRS prediction) and in poultry (i.e., though indirect prediction via serum absorbance). The digestive efficiency in pigs was tested with a conventional and with a high dietary fibre diet, and the analysis showed that within the range of digestibility values explored, no strong genotype-by-diet interaction was observed for digestibility. Although digestive efficiency was strongly correlated with feed efficiency, some moderate adverse correlations were estimated with other production traits (i.e., carcass yield and meat quality traits). In reproductive females, using individual feed intake data from gestating sows appeared to be difficult in genetic studies, especially in relation with different management systems of the sows. In one study, a reasonable variability seemed to be available (Large White pigs in a French farm) whereas very little variability was observed in a second dataset for this period (Duroc pigs in a Spanish farm). Larger and more diverse datasets would be necessary to explore how and when management limits the expression of genetic variability in this period, so a more complete analysis could be envisaged. However in Duroc sows, records of lactation traits led to estimations of the genetic variability of feed intake and feed efficiency during this period. Despite a limited number of feed intake records, the estimates were high enough to envisage selection on these traits with a limited additional phenotyping effort. Additionally and for the first time, an estimation of the genetic variability of feed intake and feed efficiency during lactation in Iberian sows was provided. Finally, some components of feed efficiency, such as behaviour, activity, welfare, and robustness were also considered, as they can positively or negatively contribute to feed efficiency. Because direct measures of activity were not available, indirect indicators were considered. The first type of indicators focused on traces of interactions on the animal’s body. However, only few traits had high enough heritabilities and correlation with feed efficiency to be used to refine the accuracy of actual estimations. The second type of indicators were derived from automatic feeder records of animal activity: feeding behaviour traits were shown to be heritable (e.g., number of visits and feeding rate) and they had some genetic correlations with production traits. In the two datasets explored, correlations were higher with feed intake than with feed efficiency. In addition, feeding patterns could be used, either empirically or via a ranking approach, to propose indicators of the animal hierarchy in the pens. Interestingly, the more dominant animals are not necessarily more efficient. This novel aspect needs further analysis to be used in selection. Finally, welfare indicators were measured in the blood and in pig hair. Blood cell counts seemed to have promising genetic correlations with feed efficiency traits, which need to be explored further. Robustness indicators were tested in divergent lines, following the hypothesis that more efficient individuals would be less robust. The hypothesis was not sustained by the experiment, which was consolidated by a mirror experiment in which divergent animals for robustness were compared for their feed efficiency, with no deleterious effect of selection for robustness on the production traits.

In a second set of analyses, biological markers of feed efficiency at the genomic and the transcriptomic levels were identified. A first strategy, based on the sequencing of divergent layer lines, allowed the identification of 145 SNP differing between lines and candidates to be associated with feed efficiency. In rabbits, first analyses of a recently available SNP chip were run in two different populations. Four to five genomic regions were associated to the trait variability in each population, with no common region. In broilers, the genomic associations with digestibility traits indicated 12 significant SNPs. A few genes were identified as potential candidates for these regions, which needs further validation. Finally, expression studies were run between divergent lines to identify the biological pathways involved in the line differences in response to different treatments, as well as to identify biomarkers in layers and in pigs. In layers, the animals were slaughtered and multiple tissues with a potential impact on feed efficiency were sampled. In pigs, serial measurements were applied to blood samples. In both cases, some genes were identified as responsible for the differences between the lines. However, the genes were partly diet- or time-dependent and the way they contribute to the base difference versus the treatment difference needs to be explored further to propose biomarkers dedicated to specific situations.