genotype by environment interaction

Introduction

The primary poultry breeding industry is under a continuously concentration process and nowadays there are only a few breeding companies which supply the world market. With the worldwide distribution of stocks, the major poultry breeders have to breed commercial layers that perform adequately in a wide variety of systems, ranging from large-intensive cage units to free- range management under different climatic and environmental conditions (Preisinger and Flock, 2000). One of the main important questions for a breeding company is whether different environments in the world can be supplied with one optimum global genotype or, on the contrary, specialised genotypes needs to be developed for each environment (Mulder et al., 2006).

The term genotype by environment (G x E) interaction is most commonly used to describe situations where different genotypes (e.g. breeds, lines, strains) respond differently to different environments. These differences in genotype response not only include changes in mean performance, but also include variability in the performance of different genotypes (Sheridan, 1990). In the presence of significant G x E interactions, some genotypes might be more affected by the environment than others, leading to a change in the ranking of the genotypes from one environment to the other (Mathur, 2003). One stress factor which may cause G x E interaction is hot climate, which reduces the feed intake of the birds thereby compromising their laying performance and the egg quality. The greatest effect can be observed for productive performance. This is reflected by a very low correlation between the breeding value estimates in the two environments, indicating that low indirect selection response may occur if selection is carried out in only one temperate environment (Horst 1985).

G x E might not be of much practical concern if special attention is paid to ensure good, or at least not extreme, environmental conditions (Hull and Gowe, 1962). In some cases, it is possible to adjust the environment conditions to the requirements of the desired genotype and reduce the effect of G x E interaction. In many cases, however, such adjustments are either not possible or simply not cost effective (Mathur, 2003). Therefore, poultry tend to be kept under environmental conditions that vary widely world-wide, such as a broad range of temperatures and housing systems (Sheridan 1990).

In poultry breeding programs, selection is done within closed populations housed under optimal conditions, which on the one hand, permits the optimal expression of the different traits to be recorded and on the other hand, guarantees that the genetic material supplied to different countries, have been produced under optimal hygiene and bio-security conditions, thereby eliminating the possibility of vertical transmission of diseases.

Hartmann (1990) stated that, in case of G x E interactions, particularly large effects have to be expected for traits with low heritability, e.g. egg production and viability. To deliver birds which perform to stringent commercial competitive standards and to reduce the possible impact of genotype x environment interactions on these traits, selection is not exclusively based on pure-line information due to the minimal disease environment of a typical breeding farm. On the contrary, most major poultry breeders recurrently tested pedigreed cross-line hens under typical commercial conditions, in a wide range of environments and housing systems (Preisinger and Flock, 2000; O´Sullivan et al., 2010). Almost all animal breeding programs are faced with the fact that nucleus pure-breds and commercial animals are raised in different environments (Wei, 1992).

Falconer (1952) indicated that to determine the importance of G x E interactions, the same traits recorded in two different environments (breeding and commercial farm) can be treated as different traits and genetic correlations between them can be estimated. According to Robertson (1959), the impact of G x E is economically important if the genetic correlation of a trait expressed in different environments falls below 0.8. Mulder et al. (2006) found that when genetic correlations between environments are higher than 0.50 to 0.70, a single breeding program with progeny testing bulls in different environments and applying index selection to simultaneously improve performance in different environments would be optimal to breed for general adaptability. When genetic correlations between environments are lower than 0.50 to 0.70, environment-specific breeding programs are necessary to breed for special adaptability. However, it is hard to justify specific breeding programs for environments of low importance.

Apart from specific breeding programs for egg markets with preferences for white- or brown-shelled eggs, there has been little evidence so far of efforts in commercial breeding enterprises to stocks adapted particularly to a specific environment (e.g. type of management, feed composition or climatic conditions). This might be an indication of the good general adaptability of the presently available commercial stocks yielding high performance under intensive production systems (Hartmann, 1990). In addition, commercial layers are a four-way cross-bred, which explode the effect of heterosis, in other words, their greater tolerance and resistance to environment stress compared with their parents, thus further reducing the adverse impact of G x E interactions. (Sheridan, 1990)

While the genetic potential of the birds is improved, management and nutrition also have to be adapted to changing demands. The general goal for the future is to breed chickens with the ability to perform well within a wider range of production conditions and do not respond to the slightest stress (Preisinger and Flock, 2000).

Genotype by environment interaction due to housing system

In addition to conventional selection criteria like egg production, feed conversion and egg quality, traits related to animal welfare have become more important in Europe (for instance the ban of conventional cages in the European Union from 2012) and North America. It has been suggested that cage testing of pure-line and cross-line stocks results in birds that are specifically adapted to cages and less capable of adapting to alternative systems (Muir, 1996). This view is reinforced by the fact that alternative systems may be far more stressful to laying hens as compared to cages. To improve these traits and simultaneously capture hen-individual performance data in non-cage environments, the Weihenstephan Funnel Nest Box (FNB) was developed (see Icken et al., 2009).

Strains found to be superior in a cage environment; may not be able to maintain their superiority in the environment of floor system. Therefore, the magnitude of G x E interactions has to be estimated to optimise testing systems for within line selection and to the most suitable line combination for cross-line breeding. A comparison of performance parameters from full siblings, tested in single bird cages and the FNB, leads back to potential G x E interactions that will determine which testing system should be mainly used in future for continuous improvement of egg production and egg quality.

Table 1: Heritabilities and genetic correlations in cage (C) and floor system (F) (Icken et al., 2011)

Flock	Egg Number 20 to 27 weeks	Egg Number 28 to 39 weeks	Breaking Strength	Egg Weight
C	F	C	F	C	F	C	F
1a	C	0.26	0.97	0.10	0.44	0.32	0.37	0.72	0.78
1b	F		0.15		0.63		0.32		0.47
2a	C	0.29	0.56	0.14	0.18	0.41	0.79	0.67	0.99
2b	F		0.31		0.29		0.38		0.69
3a	C	0.39	0.94	0.11	0.22	0.32	0.68	0.71	1.00
3b	F		0.38		0.12		0.41		0.48

* Flock 1 and 2 are brown layers and flock 3 white layers

As shown in Table 1, genetic correlations between full-sibs tested in the floor (FNB) or cages were quite high for egg number at the beginning of lay and egg weight. This suggests that G x E interactions, due to housing system for these traits might not exist. On the contrary, lower genetic correlations were found for egg number on peak production, which might reflect the existence of G x E interactions. Breaking strength represents an intermediate case, where moderate but inconsistent genetic correlations were estimated. Therefore, in this case, it is not possible to extract a clear conclusion about G x E interactions in this study.

Only eggs which are laid in the nest can be recorded and allocated to the hen in this system. That means that the definition of laying performance is not exactly the same in both environments, which might have led to a lower genetic correlation between egg production traits in both housing systems. Moreover, the magnitude of the correlations in this study should be taken with care. Using simulated data, Simianer (1991) has shown that the estimates of genetic correlations are highly dependent on the sample size and heritability. Even when there are no G x E interactions and the true genetic correlation is exactly 1, very low estimates of r_g may be obtained if the sample size and heritability are small.

Pure-bred and cross-bred performance

Another type of G x E interaction can occur in the comparison between performance of a breeder´ pure-bred selected lines and that of their cross-bred progeny which are distributed to poultry producers. This comparison can involve the purebred parental and cross-bred progeny populations being tested under either the same or different environment conditions (Sheridan, 1990). Due to genetic differences between pure-breds and cross-breds and environmental differences between nucleus and field conditions, performance of pure-bred parents can be a poor predictor of performance of their cross-bred descendants (Dekkers, 2007).

As stated by Arthur (1986), all leading egg-type breeders today probably use a combination of cross-line and pure-line records for line improvement. Flock and Preisinger (1997) indicated that especially for traits with low heritability, such as liveability and egg production, it would not be recommended to exclusively on the basis of pure-line information from the minimal disease environment of a typical breeding farm. Usually, no environmental effects have been taken into account in the estimation of (r_pc) because this is usually confounded with the difference between pure-bred and cross-breds. Environmental variance increases in the less controlled environment and consequently decreases the heritability. Secondly, the genotypes may express themselves differently in different environments, i.e. the ranking of genotype values may change. (Wei, 1992)

Crossbreeding is a standard practice in poultry breeding programmes as a way of exploiting heterosis. However, there is no consensus on the most effective way to maximize the genetic response in cross-bred commercial animals (Besbes and Ducrocq, 2003). Furthermore, the goal of breeding is not to maximise heterosis, but to maximise overall profitability in the commercial cross, the parents and the pure-lines (Flock et al., 1991). Environmental differences exist between breeding farms (e.g. with single cages) and commercial farms. Thus, both environments should be taken into account in the breeding program, to minimise possible G x E interactions that can reduce the response to selection on the commercial level. Considering cross-bred and pure-bred performance as two separate but correlated traits, offers an elegant way to take these environmental effects into account. The cross-bred performance is captured in the environment in which the breeding goal is defined. In order to optimise a breeding program, ranking based on pure-bred performance has to be compared with ranking based on cross-bred performance on commercial farms. If there are significant differences, cross-bred information has to be included in the selection process.

Cavero et al. (2010) estimated genetic correlations between pure-bred and cross-bred performance in commercial white-egg and brown-egg lines. The pure-line data used for the analysis represent three generations of a White Leghorn (WL) and a Rhode Island Red (RIR) male line, recorded in breeding farms with single cages and high standards of bio-security, feed quality and environment control. The single-cross daughters of the same sires were housed in group cages (4 hens per cage) in four commercial farms each generation, two farms for each cross. The cross-bred hens had only sire pedigree. On average per generation, about 70 sires per line were used to produce 50 WL and 69 RIR pure-bred daughters and 30 cross-bred daughters per sire.

Table 2: Genetic parameters for egg production traits in pure-line and cross-line with cross-line data as cage average (Cavero et al., 2010)

	Pure-line	Cross-line (Cage average)
h²	h²	r_g
Egg Production	WL	0.40	0.36	0.63
(20 to 27 weeks)	RIR	0.37	0.16	0.83
Egg Production	WL	0.13	0.03	0.50
(28 to 47 weeks)	RIR	0.24	0.04	0.10
Egg Production	WL	0.17	0.06	0.27
(48 to 68 weeks)	RIR	0.30	0.07	0.33

The heritabilities were consistently higher based on full pedigree pure-line records from single cages than based on the cross-line data from commercial multiple bird cages (Table 2). The genetic correlations between pure-bred and cross-bred performances were moderate to high at sexual maturity (r_g= 0.63 and 0.83), whereas the correlations decreased to moderate to low levels in the other two stages of production (r_g between 0.10 and 0.50).

Table 3: Genetic parameters for egg quality traits in pure-lines and cross-lines, with cross-bred data as cage average and as individual records (Cavero et al., 2010)

	Pure-line	Cross-line Cage average	Cross-line Individual records
h²	h²	r_g	h²	r_g
Egg Weight	WL	0.67	0.95	0.93	0.66	0.90
RIR	0.64	0.99	0.84	0.59	0.73
Shell Strength	WL	0.38	0.25	0.85	0.32	0.84
RIR	0.43	0.18	0.83	0.29	0.73
Shell Colour	WL	0.70	0.92	0.83	0.53	0.82
RIR	0.49	0.96	0.89	0.61	0.88

Estimates of genetic parameters for egg quality traits are summarized in table 3. For the cross-breds, individual egg quality measurements were used. Additionally, cage averages were assigned to one of the hens in the cage and used for the analysis to underline the problem of using these traits as cage averages. The heritability estimates based on cage averages are obviously too high and reflect the reduced residual variance when using cage averages. Using the single egg records, the heritability estimates are slightly lower than in the pure-breds, which would be expected under "suboptimal" conditions.

It can be concluded that combined cross-bred and pure-bred selection should be better than selection based only on pure-line performance, when improvement of cross-bred performance is the goal. This strategy is generally more efficient when genetic correlations are below 0.8, which is the case for laying rate. However, it is essential to find suitable farms with reliable data recording and flexible housing schedules to match well-designed tests to augment the basis for selection focused on the needs of the egg industry. If the residual variance is high due to uncontrolled factors which are not representative for other farms, the heritability will be low and the gain from using this information questionable.

The results of this study indicate that the genetic correlation between pure-line and cross-line egg production was moderate, whereas the genetic correlations for all egg quality traits were high. This confirms the working hypothesis that egg production data of cross-line relatives collected under commercial conditions can contribute significantly to improve total genetic merit under field conditions, while selection for egg quality traits can be limited to pure-line records. The loss of statistical information on individual variation due to working with cage mean is inevitable, but according to Hartmann (1990), testing of half-sibs under commercial situations is a good option to maintain genetic gain in the presence of G x E interactions.

Genome wide selection and future prospective

The availability of many thousands of single nucleotide polymorphism (SNP) markers spread across the genome and the possibility of genotyping many individuals for these markers, enables the opportunity to predict total breeding values on the basis of SNPs (Meuwisen et al., 2001; Goddard and Hayes, 2007), utilizing linkage disequilibrium (LD) between SNPs and the QTL, and consequently, to perform genomic selection.

Meuwisen et al. (2001) stated that selection on genetic values predicted from markers could substantially increase the rate of genetic gain in animals and plants, especially if combined with reproductive techniques to shorten the generation interval. As indicated by Calus (2009), the use of genomic selection reported accuracies that are up to 0.31 higher than those of pedigree indexes, without the need to phenotype the animals themselves, or close relatives thereof. In simulation studies, it has been calculated that breeding progress can be increased by 20 to 40 % if genomic selection is applied extensively in broiler breeding (Avendaño et al. 2009). The application of genomic data in commercial layer breeding value estimation has proven increased accuracies for selection at an early stage (Wolc et al., 2011).

Although traditional selection methods could be used, gains could be accelerated if the loci responsible for the genotype by environment interaction, or DNA markers in linkage disequilibrium with these loci, could be identified and then used in marker assisted selection (Hayes et al., 2009). Furthermore, Dekkers (2007) proposed to pure-breds for commercial crossbred performance using genomic selection. This approach involves estimating the effects of SNPs on cross-bred performance, using phenotypes and SNP genotypes evaluated on crossbreds, and applying the resulting estimates to SNP genotypes obtained on pure-breds. Ibañez-Escriche et al. (2009) showed the potential for genomic selection of pure-breds for commercial cross-bred performance in the field, without the need to track pedigrees through the system and with the advantage of reducing the rate of inbreeding and making accommodation of non-additive gene action easier.

However, this approach adds additional costs for the breeding companies to genotype a large number of crossbred birds, which are not candidates for selection. Beside the economical aspects, there is also a major obstacle for the applicability of this approach and thus that under field conditions, only multiple-bird systems are available. That means that there is no individual information, but only cage average information. As an exception, individual information might be captured for certain traits such as body weight, mortality and some individual evaluations for different conformation traits.

Furthermore, O´Sullivan et al. (2010) indicated the possibility to exploit the genotyping tools for multiple mating. Each sire and dam will have progenies from numerous different mates, with genetic diversity opportunities currently not available without genotyping. With the ability to genotype all progeny and assign sires post hatch, every dam will have the opportunity to be matched with multiple sires and still maintain accuracy of the relationship in the population.

Genomic selection provides precise tools which can already be used in growing animals without performance testing. This increases the speed and accuracy of selection decisions. However, as the estimation of the effects for predicting genomic selection values is based on phenotypes of the training data set, it is evident that extensive testing in different housing systems and challenge situations, as well as the recording of new traits, serve as a base to maximise the use of this new tool. Many of these traits will involve bird behaviour and traits related with animal welfare, as selection for nesting or ranging behaviour (Icken et al. 2008; Icken et al., 2009).

Conclusion

Breeding companies are involved in developing stocks to serve many markets worldwide. Changing consumer demands will drive the adjustment of future breeding goals. We are looking for possibilities to breed more robust chickens with a capacity to adapt to different climatic conditions, feed sources, housing and management.

Extensive field testing of pedigreed cross-line sibs is necessary to improve egg production under a wide range of sub-optimal conditions. Egg production data of extensive field testing of cross-line relatives collected under commercial conditions, can contribute significantly to genetic improvement, while the selection for egg quality traits can be limited to pure-line records. Increased disease exposure under field conditions with larger number of birds tested, also permits breeders to improve liveability. Special attention is being directed to animal welfare. In this context, testing pedigreed cross-lines in alternative housing systems, will support selection for improved adaptability to alternative housing systems.

Parallel to improvement in the genetic potential of the birds, standardizing the management and dietary conditions according to breeder´s recommendation should even further reduce the incidence of G x E interactions.

References

Arthur J.A. 1986. An evaluation of industry breeding programs for egg type chickens. Page 325-336 in: Proc. 3^rdWCGALP Lincoln, NE, USA.

Avendaño S., Watson K., Kranis A. 2009. Genomics in Poultry Breeding: the cost of doing business. Page 212 in: Proc. of the 60^th Annual Meeting of the EAAP in Barcelona 24-27 August.

Besbes B., Ducrocq V. 2003. Use of mixed model methodology in breeding strategies for layers. Pages 127-146 in: Poultry Genetics, Breeding and Biotechnology. Eds. W.M. Muir and S.E. Aggrey. CABI Publishing, Wallingford, UK.

Calus M.P.L. 2009. Genomic breeding value prediction: methods and procedures. Animal 4, 1-8.

Cavero D., Schmutz M., Preisinger R. 2010. Genetic evaluation for pure-line and cross-line performance in layers. Page 54 in: Proc. of 9^th WCGALP, August 1-6 in Leipzig, Germany.

Dekkers J.C.M. 2007. Marker-assisted selection for commercial cross-bred performance. J. Anim. Sci, 85: 2104-2114.

Falconer D.S. 1952. The problem of environment and selection. Am. Nat. 86:293-298.

Flock D.K., Ameli H., Glodek P. 1991. Inbreeding and heterosis effects on quantitative traits in a white leghorn population under longterm reciprocal recurrent selection. Brit Poult. Sci. 32: 451-462.

Flock, D.K., Preisinger R. 1997. Commercial breeding of egg-type chickens to maximize egg income over feed cost. Pages 63-81 in: Proc. National Breeders Roundtable, St. Louis, USA.

Goddard M.E., Hayes B.J. 2007. Genomic selection. J. Anim. Breed. Genet. 124: 323-330.

Hartmann W. 1990. Implications of genotype-environment interactions in animal breeding: genotype-location interactions in poultry. World´s Poult. Sci. J. 46: 197-210.

Hayes B.J., Bowman P.J., Chamberlain A.J., Savin K., van Tassell C.P., et al. 2009. A Validated Genome Wide Association Study to Breed Cattle Adapted to an Environment Altered by Climate Change. PLoS ONE 4(8): e6676. doi:10.1371/journal.pone.0006676.

Horst P. 1985. Effects of genotype x environment interactions on efficiency of improvement of egg production. Pages 147-156 in: Poultry Genetics and Breeding. Eds. W.G. Hill, J.M. Manson and D. Hewitt. Br. Poult. Sci. Ltd, Longman Group, Harlow, UK.

Hull P., Gowe R.S. 1962. The importance of interactions detected between genotype and environmental factors for characters of economic significance in poultry. Genetics 47: 143-159.

Ibánẽz-Escriche N., Fernando R.L., Toosi A., Dekkers J.C.M. 2009. Genomic selection of purebreds for crossbred performance. Genet. Sel. Evol. 41:12.

Icken W., Thurner, S., Cavero, D., Schmutz, M., Wendl, G., Preisinger R. 2008. Analysis of the free range behaviour of laying hens and the genetic and phenotypic relationships with laying performance. Br Poultry Sci. 49: 533-541.

Icken W., Schmutz, M., Preisinger R. 2011. Genotype environment interaction: Breeding layers with different requirements for varying housing systems. Chapter 17 in: Alternative Systems for Poultry Health, Welfare and Productivity. Eds P. Hocking. CABI Publishing, Wallingford, UK. (In press)

Icken W., Thurner, S., Cavero, D., Schmutz, M., Wendl, G., Preisinger R. 2009. Analysis of the nesting behaviour from laying hens in a floor system. Arch. Geflügelk. 73: 102-109.

Mathur P.K. 2003. Genotype-environment interactions: Problems associated with selection for increased production. Pages 83-99 in: Poultry Genetics, Breeding and Biotechnology. Eds. W.M. Muir and S.E. Aggrey. CABI Publishing, Wallingford, UK.

Meuwissen T.H.E., Hayes B.J., Goddard M.E. 2001. Prediction of total genetic value using genome-wide dense marker maps. Genetics 157, 1819-1829.

Muir, W.M. 1996. Group Selection for Adaptation to Multiple-Hen Cages: Selection Program and Direct Responses. Poult. Sci. 75, 447-458.

Mulder H.A., Veerkamp R.F., Ducro B.J., van Arendonk J.A.M, Bijma P. 2006. Optimization of Dairy Cattle Breeding Programs for Different Environments with Genotype by Environment Interaction. J. Dairy Sci. 89: 1740-1752.

O´ullivan N., Preisinger R., Koerhuis A. 2010. Combining pure-line and cross-bred in poultry breeding. Page 158 in: Proc. of 9^th WCGALP, August 1-6 in Leipzig, Germany.

Preisinger R., Flock D.K. 2000. Genetic changes in layer breeding: historical trends and future prospects. Pages 20-28 in: The challenge of genetic change in animal production. Eds. W.G. Hill, S.C. Bishop, B. McGuirk, J.C. McKay, G. Simm and A.J. Webb. British Society of Animal Science, Edinburgh. Occasional publication no. 27.

Robertson A. 1959. The sampling variance of the genetic correlation coefficient. Biometrics 15: 469-485.

Sheridan A. K. 1990. Genotype x environment interaction. Pages 897-912 in: Poultry Breeding and Genetics. Ed. R. D. Crawford. Elsevier, Amsterdam, The Netherlands.

Simianer H. 1991. Test for genotype x environment interaction expressed as a genetic correlations. 42^nd Annual Meeting of the EAAP 8-12 September in Berlin, Germany.

Wei, M. 1992. Combined crossbred and purebred selection in animal breeding. Ph.D thesis. Department of Animal Breeding, Wageningen Agricultural University. Wageningen, The Netherlands.

Wolc A., Stricker C., Arango J., Settar P., Fulton J.E., O´Sullivan N.P., Preisinger R., Habier D., Fernando R., Garrick D.J., Lamont S.J., Dekkers J.C.M. 2011. Breeding value prediction for production traits in layer chickens using pedigree or genomic relationships in a reduced animal model. Genet. Sel. Evol. 43: 5.