An Experimental Comparison of Selection Alternatives

Three a!teriiative selection methods for extending selection limits or breaking respome plateaus were compared over ten generations in a replicated model experiment using two unrelated populations of Drosophila melunoguster that no longer responded to purebred selection for high egg number, a heterotic pclygenic trait. The three methods were: ( 1 ) reciprocal recurrent selection (RRS) with selection within each of the plateaued populations based solely on crossbred performance, (2) a modification of reciprocal recurrent selection (MRRS) with selection within each population based on both purebred and crossbred performance, and (3) purebred selection within a new synthetic population formed by crossing the two plateaued populations.-Conflicting estimates were obtained for heritability of purebred egg number in each of the plateaued populations. The realized heritability values and estimates from diallel analyses indicated an absence of additive genetic variation for both populations; however, estimates from conventional intraclass correlation methods were positive. The diallel analyses revealed significant amounts of nonadditive gene effects for purebred egg number in each population, while the significant gene effects for crossbred egg numbers were additive. Estimates of the genetic correlation between purebred and crossbred egg number were negative (-0.85 iz 0.68 and -0.32 t 0.25) for the two base populations.-All three alternatives to continued purebred selection gave significant responses, with the average gain per generation from MRRS being significantly superior to the other two methods. Observed purebred and crossbred responses under RRS were in agreement with quantitative genetic theory. Such was not the case for MRRS, which suggested the possibility of major gene segregation.-Evidence supparting a negative genetic correlation between purebred and crossbred performance and the possibility of overdominance is presented and discussed. PLATEAUS, or limits in selection response, are observed frequently in longterm selection studies. The prevailing experimental material for such studies has been Drosophila melanagaster, and the selected traits, bristle number and wing length, are not directly associated with reproductive fitness (PAYNE 1920; MATHER 1941 ; ROBERTSON and REEVE 1952; SCOSSIROLI 1954; RASMUSON 1955; Journal Paper Number 7079 of Purdue University Agricultural Experiment Station. Supported by National Science Present address: Department of Biology, Marietta College, Marietta, Ohio. Foundation Grants G-5485, G-17649 and GB-4Q3. Genetics 94: 477496 February, 1980. 478 W. P. BROWN AND A. E. BELL CLAYTON and ROBERTSON 1957; LATTER 1970). In all of these reports, a limit or plateau in selection response was accompanied by reduced fitness. Components of fitness and their response to selection have not been studied extensively. In three independent model experiments with Drosophila in which artificial selection was practiced for the fitness trait “egg number” (BELL, MOORE and WARREN 1955; RASMUSON 1956; KOJIMA and KELLEHER 1963), limits in response were observed without a noticeable decline in fitness. The impact of natural selection for fitness in these experimental populations was not delineated. Further studies, both experimental and theoretical, on the interplay between artificial and natural selection is obviously needed (ROBERTSON 1969). The genetic bases for selection limits will undoubtedly vary, depending on the trait selected and the nature of the genetic variation. To avoid response plateaus or, at the least, to extend the limits of selection, it is important to discriminate among such limiting causes as physiological ceilings, genetic polymorphisms, negative genetic correlations between selected traits or between a selected trait and reproductive fitness, absence of genetic variation due to homozygosity and exhaustion of additive gene effects. The last was found to be the case for a Drosophila population plateaued for egg number, even though significant nonadditive gene effects were present (BROWN and BELL 1961). Selection methods designed specifically to exploit nonadditive genetic variation due to either dominance or epistasis (e.g., reciprocal recurrent selection. as proposed by COMSTOCK, ROBINSON and HARVEY 1949), provide a theoretical alternative to response plateaus. A recent review (BELL and MOORE 1972) of the experimental evaluation of such methods suggested the following alternatives to continued purebred selection within plateaued populations: ( 1 ) reciprocal recurrent selection between plateaued populations with selection based on crossbred performance, (2) modified reciprocal recurrent selection with selection based on both purebred and crossbred performance, and (3) selection on purebred performance within new heterogeneous populations originated by the crossing of existing plateaued populations. The present study evaluates these three alternatives over ten generations in a replicated experiment with two unrelated populations of Drosophila that had plateaued in response to purebred selection for high egg number. MATERIALS AND METHODS The basic genetic material for this study was two unrelated plateaued populations (R and T) of Drosophila melanogaster developed in our laboratory by some 4 0 generations of withinline selection for high individual egg number. Genetic markers, ebony (e) for R and poliert (spapol) for T, had been incorporated in the base populations to protect the integrity of pedigree in both purebred and crossbred matings. Inbreeding had been minimized to some degree by reproducing each population from ten or more families each generation and by the avoidance of full and half-sib matings. Population R was described in our earlier studies (BELL, MOORE and WARREN 1955; BROWN and BELL 1961). Population T was synthesized from four laboratory stocks unrelated to R and was developed specifically for the present study by individual selection for high egg number. In order to eliminate lethal and sterility genes from each population and to estimate genetic parameters with diallel systems, isogenic lines were extracted by the marked SELECTION IN PLATEAUED POPULATIONS 479 inversion technique (BROWN and BELL 1961). Prior to the selection phase, six random isogenic lines within each set ( R or 7') were systematically intercrossed to reconstitute the R and T populatj ons . The measured variable throughout this study was the number of eggs oviposited by a single, mated female in two consecutive 24-hour periods. A half-pint bottle served as the oviposition chamber with the female laying her eggs on a button of medium (water, agar, banana and powdered charcoal) attached to the bottle cap and seeded with yeast. The measurements were taken during the peak of oviposition, beginning on the fifth day and continuing through the seventh day after eclosion. All means and standard deviations are presented on the basis of daily (24-hour) egg number. The experimental flies were held in a climate-controlled room at 25", with 12 hours of white fluorescent light daily and at approximately 70 percent relative humidity. PARAMETERS OF BASE POPULATIONS Pullebred and crossbred performance The levels of performance for populations R and T during the 14 generations immediately preceding the present study are shown in Figure 1. During this preliminary period, mass selection was applied to maintain the populations in equilibrium state. Effective selection differentials (8) each generation were as shown graphically. Control populations were not observed during this phase. Overall, R showed a slightly positive but nonsigrdicant trend (0.58 eggs per generation), while T exhibited a nonsignificant negative response (-0.15). Since R had been plateaued for an earlier period of 30 generations under controlled conditions (BELL, MOORE and WARREN 1955), the nonsignificant positive trend most likely reflects chance environmental fluctuations. These static responses to long-term selection, even though selection differentials were large, typify plateaus for fitness traits. The mean performances during these preliminary generations were 87.4 and 80.6 eggs for R and T , respectively. Reciprocal crossbreds of the two populations were on the average 21 .O percent superior to the midparental value. Eight replications of concurrent measurements in purebreds and crossbreds are given in Table 1. This significant heterosis observed for crossbred Performance reflects the genetic diversity existing between these populations in terms of nonadditive gene loci influencing egg number. Heritability of purebred egg number, h,2 Estimates of hp2 for R and T were obtained by: (1 ) a modified diallel analysis using single crosses among the isogenic lines from which each base population was reconstituted, (2) analysis of variance within and among isogenic lines, and (3) conventional intraclass correlations among fulland half-sibs. Modified diallel analysis: GRIFFING (1956) presented a generalized treatment for the use of diallel crosses for obtaining unbiased estimates of genotypic variance in a random mating population. In the absence of epistatis effects, the component of variance for general combining ability, u ' ~ . ~ . ~ . , is unbiased and equals half of the additive genetic variance, uA2. Thus. 480 W. P. BROWN A N D A. E. BELL BASE POPULATION R . A I20 I " ' " " " " ' I BASE POPULATION T C " " " " " " -12 -10 8 6 -4 -2 0