New Technologies in Animal Breeding
Jack C. M. Dekkers
Department of Animal Science
Iowa State University
Ames, Iowa 50011, USA
Swine breeders, and livestock breeders in general, are continuously challenged by the need to evaluate new developments and new technologies; breeding is a business and just like any other business, only breeding organizations that;
1) stay up-to-date with new developments
2) can recognize developments and innovations that can aid their business
3) are able to implement such developments in an effective and rapid manner,
will be successful in tomorrow’s increasingly competitive market for supplying superior germ plasm.
Examples of important technologies that have impacted swine breeding programs over the past decade are the selection index, Best Linear Unbiased Prediction (BLUP) of breeding values, the halothane test and, more recently, the Ryanodine Receptor test, as well as artificial insemination, and ultra-sound technology. With advances in reproductive and, in particular, molecular genetic technology, the number of new developments that have the potential to impact swine breeding programs is increasing rapidly. The objective of this paper is to review the potential of some of these technologies, with specific emphasis on molecular genetic technologies.
Role of Molecular Genetics
To date, most genetic progress for quantitative traits in livestock has been made by selection on phenotype or on estimated breeding values (EBV) derived from phenotype, without knowledge of the number of genes that affect the trait or the effects of each gene. In this quantitative genetic approach to genetic improvement, the genetic architecture of traits has essentially been treated as a ‘black box’. Despite this, the substantial rates of improvement that have been and continue to be achieved in commercial populations is clear evidence of the power of these approaches.
However, genetic progress may be further enhanced if we could gain insight into the black box of quantitative traits. Molecular genetics allows for the study the genetic make-up of individuals at the DNA level and may provide the tools to make those opportunities a reality, either by direct selection on genes that affect traits of interest – major genes or quantitative trait loci (QTL) - or through selection on genetic markers linked to QTL. The main reasons why molecular genetic information can result in greater genetic gain than phenotypic information are:
· assuming no genotyping errors, molecular genetic information is not affected by environmental effects and, therefore, has heritability equal to 1.
· molecular genetic information can be available at an early age, in principle at the embryo stage, thereby allowing early selection and reduction of generation intervals.
· molecular genetic information can be obtained on all selection candidates, which is especially beneficial for sex-limited traits, traits that are expensive or difficult to record, or traits that require slaughter of the animal (carcass traits).
The eventual application of molecular genetics in breeding programs depends on developments in the following four key areas:
1) Molecular genetics: identification and mapping of genes and genetic polymorphisms
2) QTL detection: detection and estimation of associations of identified genes and genetic markers with economic traits
3) Genetic evaluation: integration of phenotypic and genotypic data in statistical methods to estimate breeding values of individual animals in a breeding population
4) Marker-assisted selection: development of breeding strategies and programs for the use of molecular genetic information in selection and mating programs.
Aspects of each of these areas of research will be reviewed in what follows, with main emphasis on strategies for gene- and marker-assisted selection.
1) Molecular Genetics
Through advances in molecular genetic technologies, the number of genes that has been mapped in the pig and other livestock species has increased exponentially during the past decade. The majority of mapped genes are so-called anonymous marker genes. These genes are not directly responsible for differences in economic traits but they may be linked to QTL. As such, they can be used to find QTL and to select for QTL through marker-assisted selection.
2) QTL Detection
Two approaches have been used to identify genes affecting traits of interest: the candidate gene approach and the genome scan approach (Haley 1999). In the candidate gene approach (Rothschild and Soller 1997), knowledge from species that are rich in genome information (e.g., human, mouse) and/or knowledge of the physiological basis of traits is used to identify genes that are thought to play a role in physiological mechanisms underlying economic traits. Using this information, candidate genes are identified in the species of interest and polymorphisms in the coding, but usually non-coding, regions of the gene are detected. Associations of these polymorphisms with the trait of interest are then identified using statistical analysis of phenotypic records of a sample of individuals from the population of interest. Using this approach, several genes with major effect have been identified, a prime example being the estrogen receptor gene (ESR) affecting litter size in pigs (Rothschild et al. 1991).
The genome scan approach to QTL detection uses anonymous genetic markers spread over the genome to identify QTL. Unless marker density is high, these studies cannot rely on population-wide linkage disequilibrium between markers and QTL. Instead, they rely on the linkage disequilibrium that exists within families in outbred populations or that is created in crosses between breeds or lines. Using statistical methods QTL can then be identified and their position and effect estimated by associating marker data to phenotypic records. The precision of, in particular, estimates of QTL position that can be obtained from these approaches is, however, limited, and large population sizes are needed.
Although the candidate gene and the genome scan approach are often viewed as alternate approaches for identifying genes of interest, it is clear that they can be complementary, with a genome scan identifying regions of the genome that harbor potential QTL, followed by further investigation of genes known to be located in that region using the candidate gene approach.
3) Use of Gene Marker Information in Genetic Evaluation
Although candidate gene and QTL mapping experiments can result in identification of genes of interest, the use of these genes in genetic improvement programs requires estimation of the effects of these genes in commercial breeding populations. In particular with the use of anonymous markers, marker and QTL effects must be estimated and re-estimated on a within-family basis. This requires routine systems for DNA collection and marker genotyping. Even when the actual gene has been identified, there will be a need to re-estimate gene effects on a regular basis to improve accuracy and to guard against unfavorable associations with other traits and against interactions with the background genome or environment.
Ideally, estimation of QTL effects is incorporated in routine animal model genetic evaluations, providing BLUP EBV for identified QTL and for the collective effect of all other genes that have not (yet) been identified (polygenes). Such BLUP QTL methods have been developed by Fernando and Grossman (1989) and others.
4) Use of Gene or Marker Information in Within-Breed Selection
Once a QTL or a marker closely linked to it has been identified, the important question that remains is how to use this information in selection. In this regard it is important to recognize that, although an identified QTL provides (accurate) information on the animal’s genetic value for one of the genes that affect the trait, there are many other genes that affect performance. An individual’s phenotypic performance can be used to estimate an animal’s breeding value for the collective effect of all other genes, like animal breeders have always used phenotypic data. Although the estimate of this ‘polygenic’ breeding value may not be as accurate as the estimate of the breeding value for the QTL, maximum genetic progress will be made by using both estimates, rather than basing selection on only one of them. The question then becomes how we can best combine the effect of the QTL with the EBV for the polygenes into a selection criterion.
For selection on a known QTL, the following strategies can be distinguished:
i) Two-stage selection, with selection on the QTL in the first stage, followed by selection on polygenic EBV among selected animals with the poorest QTL genotype.
ii) Standard QTL index selection, with selection on I = g + EBV, where g is the breeding value for the QTL and EBV refers to the polygenic EBV.
Although the index in strategy ii) provides the best estimate of the animal’s total breeding value, several studies (e.g. Gibson, 1994) have shown that selection on this index may not maximize response to selection over multiple generations (Dekkers and van Arendonk, 1998). Therefore, a third selection strategy must be considered also:
iii) Optimal QTL index selection, with selection on I = b g + EBV, where b is an index weight on the QTL breeding value, which can be optimized following procedures outlined by Dekkers and van Arendonk (1998), with the aim to maximize cumulative response after a given number of generations.
The impact of these three selection strategies is illustrated in Figures 1 and 2 for selection for litter size based on ESR. The three QTL selection strategies outlined above were compared to phenotypic BLUP selection, without use of ESR. Parameters for the effect and frequency of ESR were taken from Short et al. (1997) based on first parity results, which showed an additive effect of 0.39 and a dominance effect of 0.05 (i.e. BB sows average 2x0.39 = 0.78 piglets more than AA sows and AB sows average 0.39+0.05 = 0.44 piglets more than AA sows). Estimates of polygenic breeding values were derived from an index of pedigree information (selection was prior to first parity), resulting in an accuracy of polygenic EBV of 0.30 for both males and females. Selected proportions were 10% for boars and 25% for gilts. Initial frequency of the favorable (B) ESR allele was 50%, similar to what was found in commercial Yorkshire lines by Short et al. (1997).
Figure 1 shows the increase in frequency of the favorable ESR allele for the alternative selection strategies. Although phenotypic BLUP selection did not put explicit emphasis on ESR, it did increase the frequency of ESR because animals with favorable ESR genotypes (BB or AB) tend to have greater litter size and, therefore, higher EBV for litter size. The increase in frequency was, however, very gradual and did not reach fixation after 10 generations of selection. With pre-selection on ESR, the gene frequency was fixed after 1 generation; with a starting frequency of 0.5, the frequency of BB individuals is 0.25. With 10% selection among males, all selected boars were BB, leaving some room for selection on polygenic EBV in the second selection stage. With 25% selection among females, all selected sows were also BB. However, because the frequency of BB was also 25%, this left no room for selection on polygenic EBV among sows.
With standard QTL index selection, the frequency of ESR also increased rapidly, with fixation in generation 4. For the optimal selection strategies, ESR also reached near fixation by the end of the planning horizon (2, 3, 4, 5 or 10 generations) but the increase in gene frequency was almost linear. This in contrast to standard QTL index selection, which resulted in a very rapid increase in frequency in the first two generations, followed by much smaller increases to reach fixation. Thus, optimal selection resulted in a much more balanced increase in gene frequency than standard index selection and it achieved that by reducing the index weight on the QTL (b<1).
Figure 2 compares cumulative responses to selection for the alternative selection strategies. All responses are expressed relative to response for phenotypic BLUP selection, which is on the zero axis. For example, as indicated by the arrows, the response to standard QTL index selection was 35.5% greater than response to phenotypic BLUP selection after 1 generation and 15.5% greater after 2 generations of selection.
Standard index selection was superior to phenotypic BLUP selection for the first 5 generations. After 6 generations, however, cumulative response with selection ignoring ESR was greater than response from standard index selection. This is similar to results observed by others (e.g. Gibson, 1994) and is due to the reduced selection emphasis on other genes that affect litter size when selection emphasis is placed on the QTL. Standard index selection was, however, superior to two-stage QTL selection for all generations. The reason for this is that two-stage selection on ESR allowed selection of individuals with inferior polygenic EBV which, although they had the favorable ESR genotype, did not have superior overall breeding values for litter size because of their inferior polygenic EBV. Index selection results in a balance between selection on ESR and polygenic EBV, which allows selection of individuals with the unfavorable ESR genotype if the superiority of their polygenic EBV compensates for the unfavorable ESR genotype. This result clearly demonstrates the importance of balancing selection on a QTL with selection on polygenic EBV.
An even better balance between the QTL and polygenic EBV is achieved by the optimal selection strategies. Although the optimal strategies resulted in less response in initial generations than standard index selection (Figure 2), responses were 3 to 5% greater in the final generation of the respective planning horizons. Optimal selection over 10 generations also resulted in slightly greater response to selection than phenotypic BLUP selection.
The previous applies to direct selection on known QTL rather than indirect selection through linked markers. Several studies have investigated the extra response to selection that can be achieved with selection on linked markers (e.g. Meuwisen and Goddard 1996). Extra responses depend on the amount of genetic variance explained by the marked QTL, the ability of markers to trace segregation of QTL (linkage and marker informativeness), and the efficiency of selection without marker information. In general, benefits have been found to be greatest in situations where regular selection is limited or inefficient. This includes selection for traits with low heritability and traits with restrictions on phenotypic recording, such as sex limited traits, traits recorded after selection, carcass traits, and traits that are expensive or difficult to evaluate (Meuwissen and Goddard 1996).
Most studies on marker-assisted selection used as selection criterion the simple sum of EBV for marked QTL and unknown polygenes, similar to the standard QTL index selection strategy described above. Opportunities to optimize selection on genetic markers, similar to the optimization of selection on known QTL described above, requires further investigation.
Integrating Molecular and Reproductive Technologies
The previous indicates that one of the main challenges for the use of molecular genetic information in selection is that it reduces selection on polygenes and, unless selection on the QTL is properly balanced against lost response in polygenes, QTL selection can be detrimental in the long-term and suboptimal in the short term. In the cases considered, however, QTL selection was incorporated into traditional selection stages, where QTL selection competes with polygenic selection. Several authors have suggested that the most effective manner to capitalize on QTL information is by incorporating it at stages, where selection was not possible previously, in particular prior to availability of phenotypic information. Soller and Medjugorac (1999) referred to this as finding and creating selection space for QTL selection. Increasing the reproductive rate of particular females through reproductive technology and, thereby open opportunities for selection on marker information without sacrificing traditional selection response can enlarge this selection space. Within this realm great opportunities exist for the development of enhanced breeding programs based on integrating molecular and reproductive technologies.
Use of Molecular Information in Crossbreeding Programs
Although breeding programs primarily rely on selection within purebred populations, in many cases the objective is to improve crossbred performance. This raises important additional questions on how to incorporate molecular genetic information in selection programs within pure breeds that contribute to crossbreeding programs, in particular for QTL that exhibit non-additive effects. Alternative scenarios for selection on a single identified QTL within sire and dam breeds for a two-way cross are presented in Table 1. From these simplified scenarios, it is clear that emphasis on the QTL may not be the same for sire and dam breeds and that there will be a need to simultaneously optimize selection on identified QTL within both breeds in order to maximize both the rate of improvement within the pure breeds and the level of performance in the crossbreds. While table 1 considers only a single QTL, it is clear that design of selection criteria and strategies will be further complicated when multiple QTL are available, each with their own mode of action and epistatic interactions.
Table 1. Possible directions for selection within sire and dam breeds that contribute to a two-way cross on an identified QTL for a trait expressed in crossbreds, depending on mode of action at the QTL. In all cases, selection emphasis on the QTL must be optimized against emphasis on polygenes.
1Choice of selection strategy within sire versus dam breeds may depend on the accuracy of estimates of polygenic breeding values within each breed, population size, inbreeding, etc.
The contribution of molecular genetics to enhance the knowledge on genetics of economic traits will not cease with identification of genes based on QTL or candidate gene searches. Although substantial additional genetic gains can be achieved with selection on linked genetic markers, the ultimate aim will be to directly identify the main genes involved in a trait and to elucidate the function of each of these genes. Technologies and approaches to advance molecular genetic knowledge to this level are being developed and applied within the human genome project and for model organisms (e.g. Schmitt, 1999). With the wealth of information that is generated in this research, bio-informatics plays an increasingly important role to organize, analyze, and interpret this information (Sobral, 1999). Although research on gene identification, function, expression and regulation in livestock can greatly benefit from genome research in humans and model organisms, much additional work will be needed elucidate the specific role of each gene involved in traits of economic importance in livestock and their interactions with other genes and the environment. Ultimately, however, this research will provide additional knowledge that can be used to enhance selection programs. And with the additional information on genetic control of traits will come additional knowledge on how traits can be controlled through external controls (e.g. feeding). The interaction between environment, management and genetics will become more important. Opportunities to select and manage for niche markets will increase.
Financial support from the Iowa Pork Producers Association through the National Pork Producers Council, the U.S.D.A. N.R.I., and the Pig Improvement Company for aspects of research topics discussed herein is gratefully acknowledged.
Dekkers, J.C.M. and van Arendonk, J.A.M. (1998) Genetical Research 71:257.
Fernando, R.L. and Grossman, M. (1989) Genet. Sel. Evol. 21:467.
Gibson, J.P. (1994). Proc. 5th Wld. Congr. Genet. Appl. Livest. Prod. 21:201.
Haley, C. (1999) In “From Jay Lush to Genomics: Visions for Animal Breeding and Genetics” p.47, editors J.C.M. Dekkers, S. Lamont and M.F. Rothschild, Iowa State University, Ames.
Meuwissen, T.H.E. and Goddard, M. E. (1996) Genet. Sel. Evol. 28:161.
Rothschild, M.F. (1998) Proc. 6th Wld. Congr. Genet. Appl. Livest. Prod. 26: 403.
Rothschild, M.F. and Soller, M. (1997) Probe 8:13
Rothschild, M.F., Jacobson, C., Vaske, D.A., Tuggle, C., Wang, L., Short, T., Eckardt, G., Sasaki, S., Vincent, A., McLaren, D.G., Southwood, O., van der Steen, H., Mileham, A. and Plastow, G. (1991) Proc. Natl. Acad. Sci. (USA) 93:201.
Schmitt, K. (1999) In “From Jay Lush to Genomics: Visions for Animal Breeding and Genetics” p.97, editors J.C.M. Dekkers, S. Lamont and M.F. Rothschild, Iowa State University, Ames.
Short, T. H., M. F. Rothschild, O. I. Southwood, D. G. McLaren, A. de Vries, H. van der Steen, G. R. Eckardt, C. K. Tuggle, J. Helm, D. A. Vaske, A. J. Mileham, and G. S. Plastow (1997) J. Animal Sci. 75: 3138.
Sobral, B.W.S. (1999) In “From Jay Lush to Genomics: Visions for Animal Breeding and Genetics” p.115, editors J.C.M. Dekkers, S. Lamont and M.F. Rothschild, Iowa State University, Ames.
Soller, M. and Medjugorac, I. (1999) In “From Jay Lush to Genomics: Visions for
Animal Breeding and Genetics” p.85, editors J.C.M. Dekkers, S. Lamont and
M.F. Rothschild, Iowa State University, Ames.