INTRODUCTION

As genes with major effects are discovered, it is possible to directly select on the genotype for genetic improvement of traits. For example, Rothschild et al. (1994) discovered a major gene for litter size in pigs and was reported to resulting in 1 pig more born alive. However, the major question commercial breeders are asking is how much advantage will this new technology offer over existing technology. Few scientists fully understand what advantages molecular genetics offers to breeding programs. To further complicate the issue, some quantitative geneticists have presented an overly optimistic scenario for marker assisted selection (MAS). For example, Lande and Thompson (1990), using unrealistic assumptions, theoretically show that selection efficiency can be increased over 400% with MAS.

Assuming that technology will eventually overcome all barriers, and that we will eventually have a complete understanding of the molecular basis for all traits, there are two important question which can be addressed from a theoretical basis bases before that state of knowledge is achieved: 1) What is the optimum method for incorporating major genes in a breeding program?; and 2) What is the relative advantage of directly incorporating genes in a breeding program as compared to conventional phenotypic selection?

With respect to the first question, Hillel (1994) states that "once a linked marker has been identified, efforts should be made to increase the frequency of the desired allele in the breeding stock or even bring it to fixation following a single or few steps." Is this an optimum strategy? Should one fix each favorable allele as fast as possible? With respect to the second question, Gibson (1994) conducted simulations using an additive infinitesimal model plus a single bi-allelic locus. His results showed that the answer depended on the time horizon. In the short term, greatest response resulted from inclusion of molecular information while in the long term phenotypic selection gave the greatest response.

The objective of this research was to examine, through use of a Monte Carlo gene level simulation program, the potential relative advantage that direct selection on the genotype adds to a breeding program. Because selection is directly on the gene of interest, recombination with a marker is not a concern and, as such, these simulations should be viewed as the upper limit of what is possible with MAS.

MATERIALS AND METHODS

A gene level simulation program was developed that included 500 quantitative trait loci (QTL), each with two alleles, randomly distributed over 5 chromosome pairs, including the sex chromosomes, with a total genomic size of 8000 centiMorgans. Sex was assumed to be determined by the usual manner for mammals with the male as the hemizygous sex. Chromosome lengths were randomly set at initiation. Recombination was based on Haldane's (1919) formula as a function of map distance. Distribution of gene effects (a_i) was assumed exponential, i.e. a large number of loci with small effects and a few with large effects. Allele frequencies (p_i) at each locus was randomly set based on the uniform distribution. The base population was constructed such that it was in both gametic and zygotic phase equilibrium. Thus, additive genetic variance in the base population was computed as Óp_i(1-p_i)á_i² , where á_i= (a_1i-a_2i)/2 for the two alleles at the ith locus. The phenotype was determined by adding the allelic effects over all loci plus a random environmental effect. Environmental effects were normally distributed with a variance set to give a desired heritability. Three heritabilities were examined, .01, .1 and .4. A population of size 4,800 (equal numbers of each sex) was sampled from the base population. The trait of selection was assumed to be measurable on both sexes. Each generation 60 males and 240 females were selected to produce the next generation, mating each selected male to 4 females at random. Two selection schemes were examined: Phenotypic (P), and a combination (B) using an index of the phenotypic and the major gene, giving a relative weight of 100 to the major gene. Choice of which allele to select as the candidate gene was determined by that which would result in the greatest change in the mean, i.e. (1- p_i)a_i. With selection method B, the candidate allele was selected to fixation, after which the next ranked candidate gene, at that generation, was chosen as the next major allele. This process continued for 30 generations. The entire simulation was replicated 40 times, starting over with a new distribution of allele effects, chromosome positions, chromosome lengths, and allele frequencies.

RESULTS AND DISCUSSION

Results show that with a very low heritability (1%), B was superior to P in all generations. With a heritability of 10% and higher, P was superior to B. These results clearly show that placing too much emphasis on major gene was detrimental to the breeding program. Comparison of polygenic frequencies showed that for all heritabilities, P increased average polygene gene frequency faster than B. In addition, comparison of number of favorable alleles lost due to random genetic drift was always greatest with B. Finally, for heritabilities of 1% and 10%, the rate of inbreeding was always greater for B than P. At a heritability of 40%, differences in inbreeding were minor between methods. This study indicates that when major genes are identified, the optimal program is not to fix the gene as rapidly as possible. In doing so, animals with many favorable alleles will also be discarded.

The procedure to find optimal weights the major genes has yet to be demonstrated, however, by trial and error it was possible to find the optimal weight. Even still, the relative advantage of combined selection was less than 5% with heritabilities greater than 10% and diminished to less than 1% as the heritability approaches 40%.

These results agree with Smith and Webb (1981) who compared the efficiency of direct selection on major genes, on the trait, and an index combining the two. They concluded that efficiency depends largely on the ratio of additive genetic variance due to the major locus to the total additive genetic variance. When phenotypic selection is effective (high heritability), further information on a major gene will add little to the rate of improvement. If phenotypic selection is not very effective, as for traits with low heritability or if indirect selection must be used, then selection on the candidate gene can add significantly to the rate of genetic improvement (Smith and Webb, 1981). These results do not take into account if the effects of the gene are antagonistic for different traits. In which case there may be undesirable correlated selection responses or small responses if selection for the two traits is balanced (Smith and Webb, 1981).

These results are in contrast to those of Gibson (1994). In those simulations, MAS gave the greatest response in the short term while phenotypic selection gave the greatest response in the long term. Our results, even when a single major gene was examined, gave the same result in both the short and long term. However, when the number of loci was greatly reduced, such that the major gene accounted for a much larger proportion of the genetic variance, the results were similar to that observed by Gibson (1994). However, the crossover in response between short and long term efficiency occurred after the major gene was fixed.

Van der Beck and Van Arendonk (1996) also noted that polygenic response was lower with MAS than Phenotypic selection and conversely major gene response was higher with MAS than phenotypic selection. Similarly Ruane and Colleua (1995) observed in their simulations that MAS resulted in substantially greater increase in the major gene than BLUP, but lower polygenic response. Meuwissen and Van Arendonk (1992) concluded that conventional selection would be superior to MAS in the long term because conventional selection allocates less selection differential to the fixation of major genes and more to selection differential of polygenes with small effects. These results support the conclusion of Meuwissen and Van Arendonk (1992).

It should be noted that many of the assumptions of this simulation would be impossible to achieve in an actual breeding program. Allele effects for all loci affecting a trait may never be known, and certainly they will not all act additively. Also, instead of phenotypic selection, BLUP or some other optimal selection scheme would be used, which would further erode the relative advantage of including candidate genes in the breeding program. In short, every advantage was given to using the candidate gene while phenotypic selection was handicapped. Even still, the advantage of including major genes in the breeding program would probably not be worth the cost of genotyping the animal.

This research should not be interpreted as meaning that molecular technology will not be of some advantage, but rather defines where it will not. For sex-limited traits, such as egg or milk production, or traits that cannot be measured in either sex, such as meat quality, direct selection on the genotype may be of great value.

CONCLUSIONS

This study shows that when major genes are identified, the optimal program is not to fix the gene as rapidly as possible. In doing so, animals with many favorable alleles will also be discarded to the overall detriment of the program. Rather, the gene should be incorporated into the population over a period of several generations using an index giving a relative weight to the gene proportional to its effect on the overall genetic variance for the traits. Even with optimal weighting, it was shown that for traits which can be measured in both sexes, the maximum advantage of combined selection was less than 5% with heritabilities greater than 10% and diminished to less than 1% as the heritability approached 40%. Clearly, the benefit of such a program could not justify the expense. However, this research should not be taken to mean that direct selection on the genotype will not be cost effective in some cases. For sex limited traits, such a milk or egg production, the only information on the male would come from relatives and direct genotyping. In which case, combined selection is expected to be of greater benefit. In the case where phenotypic information on either sex cannot be determined, such as with carcass quality and disease resistance, direct selection on the genotype may be the only viable alternative.

REFERENCES

Gibson, J.P. 1994. Short-term gain at the expense of long-term response with selection of identified loci. Proc. 5^th WCGALP. U. of Guelph, Guelph, Ontario CAN. 21:201-204.

Haldane, J.B.S. 1919. J. Genetics 8:299-309

Hillel, J. 1994. DNA markers for quantitative production traits. Proc. 5^th Workd Congress on Genetics Applied to Livestock Production. University of Guelph, Guelph, Ontation CA. 20:112-119.

Lande, R., and R. Thompson. 1990. Efficiency of marker-assisted selection in the improvement of quantitative traits. Genetics 124:743-756.

Meuwissen, T.H.E., and J.A.M. Van Arendonk. 1992. Potential impact in rate of genetic gain from marker-assisted selection in dairy cattle breeding. J. Dairy Sci. 75:1651-1659.

Rothschild, M.F., Jacobson, D.A., Vaske, C.K., Tuggle, T.H., Short, S., Sasaki, Eckardt, G.R., and McLaren, D.G. 1994. Proc. 5th WCGALP, U. of Guelph, Guelph, Ontario CAN. 21:225-227.

Ruane, J., and J.J. Colleua. 1995. Marker assisted selection for genetic improvement of animal populations when a single QTL is marked. Genet. Res. Camb. 66:71-83.

Smith, C., and Webb, A.J. 1981. Effects of major genes on animal breeding strategies Z. Tiezuchtg Zuchgsbiol. 98:161-169.

Van der Beck, A., and J.A.M. Van Arendonk. 1996. Marker-assisted selection in an outbred poultry breeding nucleus. Animal Science 62:171-180.