Is GxE a burden or a blessing? Opportunities for genomic selection and big data

Genomic selection and genomic prediction have been widely adopted in many livestock breeding programmes, including some fish schemes. Genomic prediction increases the accuracy of breeding values, especially for lowly heritable traits and traits that are difficult to measure. Genotype-by-environment interaction (GxE) is often seen as a burden for genetic improvement across multiple commercial environments, while genomic selection may be used to overcome some of the difficulties, because it is no longer required to record relatives of the selection candidate in all environments. Furthermore, we are in the era of big data. Sensors, electronic feeding stations, automatic milking systems or activity loggers generate voluminous or big phenotypic data per individual animal with multiple measurements from each day. The longitudinal phenotypic profiles together with longitudinal environmental profiles provide opportunities to develop indicators to increase the resilience of animals. I define resilience as “the ability to maintain productive and healthy life despite environmental perturbations.” The genetic differences in resilience are the result of some form of GxE and can be observed as genetic variation in macro-/micro-environmental sensitivity. Here, I will first discuss the opportunities of genomic selection to deal with animals performing in multiple environments in the presence of GxE, and secondly, how GxE observed as genetic variation in macro-/micro-environmental sensitivity can be exploited by genomic selection to improve resilience using big data. The GxE between environments leads to lower genetic gain if selection is performed in a different environment (e.g., a nucleus), than in which the commercial animals are performing. This situation is most obvious in pigs and poultry where selection is on information of pure-bred animals in the nucleus, while the breeding goal is to improve performance of cross-bred animals in commercial environments. The non-unity pure-bred–cross-bred genetic correlation is partly due to GxE, while it is largely due to the difference in genetic background between pure-bred and cross-bred animals (Wientjes and Calus, 2017 J. Anim. Sci. 95:3467–3478). Because cross-bred performance is a genetically different trait than pure-bred performance due to genetic background, GxE for cross-bred performance over environments is potentially different from GxE of pure-bred performance over environments. When the reference populations of cross-bred animals are made of animals raised in a wide range of environments, genomic selection for improved cross-bred performance is more efficient than with traditional selection relying on pedigree information. Such multi-environment reference populations could be easily combined with extended phenotyping of animals using, for instance, electronic devices, such as feeding stations and sensors. The produced big data would help to breed resilient animals across environments, because resilience may also be subject to GxE. Furthermore, a multi-environment reference population enables customized marketing of semen, embryos or live animals because accurate GEBV can be estimated for all genotyped selection candidates for performance and resilience in specific environments, which was not possible with pedigree-based EBV. If large numbers of selection candidates are genotyped, genomic selection increases the possibility to create specialized lines because break-even genetic correlations are expected to be higher with genomic selection than with traditional selection (Mulder, 2007, PhD-thesis, Wageningen University & Research). Creating specialized lines would increase genetic diversity. Genotype-by-macro-environment interactions mean genetic variance in macro-environmental sensitivity. Macro-environments are defined as environments that differ due to known environmental factors such as nutrition and climate. These macro-environments can appear among farms or across time periods within a farm. Here, I focus on GxE across time periods (within farms) as a source of genetic variation to increase resilience. Examples are daytime length, temperature or occurrence and severity of diseases. If the environments can be described by a continuous parameter such as temperature or daytime length, reaction norm models can be used to estimate genetic variation in the slope of the reaction norm. The EBV for the slope of the reaction norm could be used as a resilience indicator. Mulder (2016, Frontiers in Genetics 7:178) showed that genomic selection is more highly capable of exploiting genetic variation in macro-environmental sensitivity to increase resilience than traditional sib or progeny testing schemes, provided that the reference population is distributed along the environmental gradient. Big data helps to have phenotypic profiles of animals during periods with and without environmental challenges. Therefore, big data combined with genomic selection helps to further increase the response to selection in resilience exploiting genetic variance in macro-environmental sensitivity. Genotype-by-micro-environment interactions mean genetic variance in micro-environmental sensitivity. Micro-environments are unknown and may vary among animals. This could be related to the internal environment of the animal (e.g., metabolic state) or whether the animal is or was affected by an endemic disease, such as mastitis. Genetic variance in micro-environmental sensitivity can be observed as genetic variance in environmental variance (Ve). If animals are measured only once, the heritability of Ve, defined as the heritability of the squared phenotypic deviation, is often as low as 0.01–0.05 and often even lower, resulting in a very low accuracy of selection (Hill and Mulder, 2010 Genet. Res. Camb. 92:381–395). However, if many repeated records on an animal are available, for example for daily milk yield or colour of several eggs, the heritability of the within-individual variance is much higher, for example 0.1–0.2 (Mulder et al., 2017 Book of abstracts EAAP p462; Mulder et al., 2016 Genet Sel Evol 48:39). Therefore, big data combined with genomic selection further increases the response to selection in resilience exploiting genetic variance in micro-environmental sensitivity. For both macro- and micro-environmental sensitivities, genomic selection combined with big data increases resilience by lowering environmental sensitivity, for example a flatter slope of reaction norms and a lower within-individual variance. For increasing resilience, both types of GxE are a blessing, because without GxE, there would be no genetic variance. It is largely unknown to what extent the different types of GxE are determined by the same genes. In dairy cattle, Mulder et al. 2013 (Genet Sel Evol 45:23) found a positive genetic correlation of 0.77 between macro- and micro-environmental sensitivities. Furthermore, empirical evidence on the influence of selection for resilience on animal health and welfare is largely lacking, except a few studies relating Ve to health, longevity or stress-related traits. Mulder et al. (2017 Book of abstracts EAAP p462) found that the variance in milk yield was a good predictor for mastitis, ketosis and longevity in dairy cattle. In a selection experiment in rabbits, Blasco et al. (2017 Genet Sel Evol 49:48) showed that selection on low Ve of litter size increased resilience measured by immune and stress response in the lines selected on high and low Ve. In the coming years, big data from sensors, electronic feeding stations, automatic milking systems, etc., in combination with abundant genomic and sequence data is expected to increase the possibilities to study the genetic background of various resilience indicators, for instance by quantitative genetic analyses and genome-wide association studies. This knowledge will contribute to developing sustainable genomic selection strategies to increase resilience of animals exploiting genetic variation in macro- and micro-environmental sensitivity. Such sustainable breeding strategies aiming for high production combined with good health and welfare of animals will increase farming profits and farmers’ satisfaction, and also the societal acceptance of animal husbandry. Therefore, genomic selection and big data can largely turn the burden that GxE brings with respect to genetic improvement across environments into a blessing by increasing possibilities to increase resilience and genetic diversity.