Drosophi la melanogaster

We have explored in Drosophila melanogaster the fitness effccts of allelic variation at three enzyme loci: aGdh, Adh, and Acph. Viability and rate of development are studied at two densities, near-optimal and competitive. No genotypic effects could be demonstrated on rate of development at either density or on viability under optimal conditions. Small but significant effects on viability appear under competitive conditions. Fecundity is measured for all ninc possible mating combinations between the three female and thc three male genotypes at each locus. Female genotype has important fitness consequences; heterosis exists at every locus. Male genotype also contributes to fitness, but without heterosis. There are significant interactions between female and male genotypes, so that the fecundity of a mating combination cannot be determined from the average fitnesses of the female genotype and the male genotype involved.


Introduction
The neutrality theory of protein evolution propounds that most protein variants (allozymes) found in natural populations at nontrivial frequencies are adaptively equivalent. When investigating a particular protein polymorphism, the neutrality theory may be taken as the null hypothesis. Experiments are, then, designed to test whether the allozyme variants are selectively identical. If selection is demonstrated, the neutrality hypothesis is rejected; otherwise, neutrality remains the standing hypothesis for the investigated polymorphism.
The selective effects of allozyme variants have been investigated in very many ways, both in nature and in the laboratory. A particularly informative approach consists in examining whether allozyme variation affects one or other component of the life cycle. In the present paper we report the results of a study of three allozyme polymorphisms in Drosophila melanogaster. Three fitness components are examined for each gene locus: viability, rate of • development, and fecundity. The experiment is designed so that the fecundity of a female can be measured not only as a function of its own genotype but also as a function of the genotype of the male with which it mates.

Material and methods
The experimental strains were derived from a sample of l 500 Drosophila melanogaster females collected in the fields of the University of California, at Davis. Each female was placed in a half-pint bottle with medium and the progeny maintained by mass culture. Very numerous single-pair crosses were made with the progenies of the wild flies, in such a way that the two parents in a cross always descended from different wild females. Five days after a single-pair cross was made, the two parents were assayed by electrophoresis (procedures of Ayala et al., 1972). If both parents were identical homozygotes with the desired genotype, their progeny was identified as one of the experimcntal lines and maintained by mass culture. In order to obtain the desired number of strains with each genotype, single-pair crosses had to be made in some cases for more than one generation; but in every case, the two parents were derived from different wild females so that no inbreeding was involved.
The three loci studied are aGpdh (chromosome II, 17.8, coding for alpha-glycerophosphate dehydrogenase), Adh (II, 50.1, coding for alcohol dehydrogenase), and Acph (111, 101.1, coding for acid phosphatase). Each locus is polymorphic for two alleles, F and S, with some other alleles found in low frequencies. At all three loci,'the rarer of the two common alleles is S, which in the Davis natural population has a frequency of0.18 for aGpdh, 0.40 for Adh, and 0.11 for Acph.
Four sets of experimental lines were obtained, each consisting of 30 independent strains homozygous at all three loci under investigation. The genetic constitution of the four sets is, as follows: (1) (2) The effect of genotypic variation at aGpdh was studied by using sets (I) and (2); sets (1) and (3) were used for Adh; and sets (1) and (4) for Acph. It can be seen that the experimental flies are homozygous for the Fallele at the two loci other than the one under investigation. In order to obtain heterozygous flies for aGpdh, crosses were made between strains of set (1) and strains of set (2). Similarly, crosses between sets (I) and (3) produced the Adh heterozygores, and between sets (1) and (4) the Acph heterozygotes. Three fitness components were studied for each gene locus: fecundity (number of eggs laid by three females during four days), viability (percent of eggs developing into imagoes), and rate of development (number of days from egg to imago). Viability and rate of development were measured in two density conditions: quasi-optimal (40 eggs placed in a halfpint culture with 40 cc of food medium) and competitive (200 eggs in a 2 × 8 cm vial with 8 cc of medium). The eggs counted in the fecundity experiment, placed in the appropriate cultures and densities, served for measuring viability and rate of development.
Fecundity was measured by placing in a vial three virgin females of a given genotype with three males of the same or different genotype. A paper teaspoon with medium was replaced daily in the vial. The eggs laid between days 6-9 were counted. The flies introduced in the vials were obtained as follows. For each homozygous gcnotype, 15 crosses were made, each involving two of the original 30 experimental strains; for each heterozygous genotype, 15 crosses also were made, involving 15 randomly chosen strains with a given homozygous genotype and 15 strains with the other genotype. The three females placed in a vial for measuring fecundity always came from strains different from those producing the males introduced in the same vial. Thus, the genetic background was thoroughly randomized not only for the flies mating in the experimental vials, but for their progeny as well. All nine possible combinations between the three female genotypes and the three male genotypes (Q FF × (~ FF, g FF × S FS, Q FF X S SS, ~ FS × S FF, etc.) were made for each locus. Each combination was replicated 60 times. Hence, the total number of experimental vials for the fecundity experiment was: 3 loci × 9 genotypic combinations × 60 replicates = I 620, in each of which theeggs were counted on four consecutive days.
Eggs of the same age and genotype, counted in the fecundity experiment, were placed in groups of 40 in half-pint cultures ('optimal' conditions), or in groups of 200 in vials (competitive conditions). The flies emerging in these cultures were counted daily. Viability was measured simply by the percent of flies emerging. Rate of development was measured in days by the formula Ynidi/Yni, where ni is the number of flies emerging on the di day after the eggs were placed in the cultures. Each genotype was studied in 30 replicates at each density, so that the number of cultures counted was 7 genotypes × 2 densities × 30 replicates = 420.
All experiments were performed in a controlledtemperature incubator at 25 + 0.5 o C and ca. 70% relative humidity.
The effects of genotype and density on viability and rate of dcvelopment were first tested by a factorial analysis of variance. The percent viabilities were transformcd into arcsinex/percent for the analysis of variance. The genotypic effects on viability were significant and, hence, nested analyses of variance were made separately for each of the two densities. A posterior± comparisons between means were made by a Student-Newman-Keul test for least significant range (LSR) (Sokal & Rohlf, 1969, pp. 294-295).
The fecundity data were tested, first, by a factorial analysis of variance. Nested analyses of variance were also performed for the average fccundity of the three genotypcs at each locus, separately for females and for males. (Nested analyses of variance were made also for testing the effects of each male genotype on a given female gcnotype, and viceversa; but these will not be given here, although they are taken into account in the evaluation of the rcsults.) LSR tests also wcrc carricd out for comparisons between fecundity means. Table 1 gives the mean viability and rate of development for each genotype at each density. The table also shows by means of t-test comparisons that the viability is significantly lower, and the rate of development longer, under competitive conditions than when density is near optimal. The experiments were madc at two densities not in order to test whether a reduction in fitness would occur at competitive dcnsities, but rather in order to see whether the relative fitnesses of genotypes would change from one to another density, as it has been shown in other instances (e.g. Tosi~ & Ayala, 1981).

Results
The analyses of variance shown in Table 2 manifest the large differences between the two densities but no significant interactions between density and genotype for either viability or rate of development. Genotypic differences have no significant effects on rate of development, but they are marginally significant for viability. Nested analyses of variance (Table 3) show that the genotypic effects on viability Table 1. Viability (in percent) and rate of development (in days) of individuals with different genotype at one of three loci coding for enzymes. The FFindivid uals have this genotype at all three loci; individuals shown with the FS or SS genotypc at a givcn locus, have the /"F genotype at the other t~o loci. The mean and standard error are based on 30 replicates. Student's t is for pairwise comparisons between optimal and competitive dens±tics. arc not significantly hcterogeneous at optimal density, but they are at the competitive density. It can be seen in Table I that the greatest differences occur at the Adh locus. Indeed, the LSR test shows that only one significant diffcrence exists between the means, namely that betwecn the viability of the SS and FS genotypes of Adh at high density. Tables 4-6 give the mean number of eggs laid by each mating combination. Clearly, genotypc has a substantial effect on fecundity. This is cxamined by factorial analyses of variance (Table 7) showing that both the female and the male genotype affect fecundity, and that there is significant interaction between the two. The contributions of the two sexes are examined by separate nested analyses of variance (Tables 8 10). The femalc genotypcs are (highly) significantly heterogeneous at all three loci. The male genotypes are significantly heterogeneous only for Acph. Yet therc are significant interactions between female and male genotype in all cases. That is, the performance of a given female genotype depends on thc male with which she mates; and the performance of a given male genotype depends on the female with which he mates. In spite of these interactions, it appears that for aGpdh and Adh a male genotype that increases the fecundity of a given female gcnotype must have compensatory effects on the other female genotypes, since the male genotypes do not differ significantly in mean performance.
The most striking effect when the individual gen- Table 7. Factorial analysis of varlancc for fecundity. The interactions between female and male genotype are explored in Table l l, where the difference between the observed and the expected fecundity is given for each mating combination. The expected fecundity of a mating type is simply estimated by the product of the average fecundity of the two genotypes involved divided by the grand average. One discernable pattern appears at the c~Gpdh and Adh loci; namely the fecundity of a homozygote is greater when it mates with a homozygote for the alternative allele than when it mates with a like homozygote. The departures from expectation for thesc combinations are highlighted in italics in "fable 11. The evolutionary significance of this pattern of interactions is that it will contribute to maintaining genetic polymorphism (Serradilla & Ayala, 1983;Hadeler & Liberman, 1975;Cockeram et al., 1972). Figure 1 summarizes the fecundity data. It clearly shows the overdominance of the female heterozygotes as well as the complexity of the female-male interaction (e.g., the SS male genotype at the aGpdh locus yields the highest fecundity of the three male genotypes when mating with the FF females, but the lowest when mating with the FS females). in the middle represent the fecundity of the heterozygous female; the two sets on the sides show the fecundity of the homozygous females. For each female genotype, the middle bar represents the contribution of the heterozygous males; the two bars on each side, the contribution of the homozygous males. The heterozygous females exhibit overdominance at all three loci, but there is no male heterosis, l,arge interactions between female and male genotype are manifest.

Discussion
Genotypic variation associated with the three loci examined (ceGpdh, Adh, and Acph) in this experiment is effectively neutral with respect to rate of development and with respect to viability under nearly optimal conditions; when there is competition for resources, the Adh genotype has small effects on viability. All three gene loci have significant effects on fecundity. These effects include overdominance of the heterozygous female at each locus, as well as increased fecundity when different homozygotes mate -both of which are processes that will contribute to maintain genetic polymorphism. Whether the allozyme polymorphisms found at these loci are maintained in nature by these or other processes cannot be decided by our experiments. These experiments, however, establish that, at least under certain conditions, fitness differences exist among individuals having different genotypes at the three loci.
The question arises whether the fitness differences observed are due to the loci examined or to other loci that might be associated in linkage disequilibrium with the target loci. The following observations are relevant. The two genomes of each fly used in the fecundity experiments come from separate wild-collected flies; thus, the experimental flies have levels of heterozygosity comparable to wild flies. The same can be said of the parents of the experimental flies (which come from the four sets of 30 experimental strains), all containing two genomes descended from two separate wild-collected flies. And the same is also the case for the zygotes in which viability and rate of development are measured, because their male and female parents derive in each case from different sets of wild-collected flies. The sampling variance of D~i , the measure of linkage disequilibrium between loci i and j, can be calculated using the equation given by Laurie- Ahlberg andWier (1979, p. 1309). Following the reasoning of "losi~: and Ayala (1981), the sampling variance of D,3 in the present experiments can be calculated as, approximately, I / 240 (four sets of 30 experimental strains are used, each strain having no less than two wild genomes). The very small sampling variance justifies the conclusion that no linkage disequilibrium has been created in the laboratory experiments. In addition, the use of considerably more than 100 different wild chromosomes in the experiments ensures that more than 95% of the genetic variation present in the natural population is represented in the experiments.
That no linkage disequilibrium associations have been created in the laboratory, and that most of the natural genetic variation is represented in the experiment, do not warrant that the fecundity (and viability) variation observed in the experiments is due to the loci examined alone. If alleles at the experimental loci are nonrandomly associated in nature with alleles at other tightly linked loci, then the genetic effects observed might be due to such linked loci and/or to the target loci. But it should be noticed that if linkage disequilibrium exists in the experiments because it exists in nature, the observed effects would also affect the target loci in nature and thus contribute to maintaining the polymorphisms.
Our results are consistent with a number of studies in the last two decades showing that fertility may be a major contributor to fitness differences, often substantially more important than viability or rate of development. Sved and Ayala (1970) found that, in population cages, fertility ('late' fitness component) had greater importance than viability ('early' fitness component) in determining the fitness of D. melanogaster flies homozygous for full chromosomes. Anderson and Watanabe (1974) suggested that female fecundity and male mating capacity might be main contributors to the maintenance of chromosomal polymorphisms in D. pseudoobscura. Allozymes have been associated with fertility differences in a number of cases (e.g., Gilbert et al., 1982;Marinkovi6 & Ayala, 1975a, b).
It has become increasingly apparent that male genotype contributes significantly to fitness. It has been known for a number of years that the mating success of a male is affected by its genotype (Petit & Ehrman, 1969; for allozyme genotypes see Knoppien et at., 1980;McKenzie & Fegent, 1980;. Maynard Smith (1956) has shown that females inseminated by ontbred males produce more offspring than those inseminated by inbred males. Brittnacher (1981) has found in D. melanogaster that malc mating capacity ('virility') has a very large effect on fitness variation, it being perhaps the single most important fitness component. Gilbert et al. (1981) and Gilbert and Richmond (1982) have demonstrated in D. melanogaster that the enzyme esterase-6 is transmitted to the females 145 in the male ejaculate during insemination. Genetically determined variation in this enzyme affects the time that a female will take from one to another mating and thc number of progeny produced by the female.
The role of heterozygous superiority ('overdominance') in maintaining genetic polymorphisms has been discussed in a number of places and needs not be reviewed here. The particular case of allozymc overdominance has also been extensively considered (e.g., Milkman, 1966;Marinkovi~5 & Ayala, 1975a, b;Berger, 1976: Kirpichnikov & Muske, 1980. A more distinctive phenomenon is that matings between dissimilar homozygotes have greater fecundity than matings between like homozygotes. This mode of selection in which association of opposites increases their fitness has been named alloprocoptic selection (Serradilla & Ayala, 1983). How large a role it may play in preserving allozymc and other polymorphisms must wait further investigation. The generality of the interactions between male and female genotypes with respect to fecundity warrants in any case the design of experiments such as those reported in Tables 3-6, in which the fitness of all possible male:'female genotypic combinations can be evaluated. Few like experiments have been made (Bundgaard & Christiansen, 1972).