Heterosis for Viability, Fecundity, and Male Fertility in Drosophila melanogaster : Comparison of Mutational and Standing Variation

Corresponding author: James D. Fry, Department of Biology, Utah State University, Logan, UT 84322-5305, jdfry{at}biology.usu.edu (E-mail).

Communicating editor: A. G. CLARK

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

If genetic variation for fitness traits in natural populations ("standing" variation) is maintained by recurrent mutation, then quantitative-genetic properties of standing variation should resemble those of newly arisen mutations. One well-known property of standing variation for fitness traits is inbreeding depression, with its converse of heterosis or hybrid vigor. We measured heterosis for three fitness traits, pre-adult viability, female fecundity, and male fertility, among a set of inbred Drosophilia melanogaster lines recently derived from the wild, and also among a set of lines that had been allowed to accumulate spontaneous mutations for over 200 generations. The inbred lines but not the mutation-accumulation (MA) lines showed heterosis for pre-adult viability. Both sets of lines showed heterosis for female fecundity, but heterosis for male fertility was weak or absent. Crosses among a subset of the MA lines showed that they were strongly differentiated for male fertility, with the differences inherited in autosomal fashion; the absence of heterosis for male fertility among the MA lines was therefore not caused by an absence of mutations affecting this trait. Crosses among the inbred lines also gave some, albeit equivocal, evidence for male fertility variation. The contrast between the results for female fecundity and those for male fertility suggests that mutations affecting different fitness traits may differ in their average dominance properties, and that such differences may be reflected in properties of standing variation. The strong differentiation among the MA lines in male fertility further suggests that mutations affecting this trait occur at a high rate.

THE maintenance of heritable variation in fitness traits is a fundamental problem in evolutionary quantitative genetics. The naive expectation is that selection should erode genetic variation for traits closely related to fitness, so that little heritable variation for such traits should be observed in populations at equilibrium. Genetic variation in basic fitness traits (e.g., fecundity, egg-to-adult survival) appears to be ubiquitous in natural populations (MOUSSEAU and ROFF 1987 ; SHAW et al. 1995 ; VIA 1991 ), however, and by some measures exceeds that observed for morphometric traits (HOULE 1992 ).

Hypothesis for the maintenance of genetic variation for fitness traits may be divided into two broad categories, mutation-selection equilibrium and balancing selection. Under the mutation-selection equilibrium (MSE) hypothesis, variation is maintained by a constant input of mostly deleterious mutations. Here, we use the term "balancing selection" to refer to any type of selection that maintains genetic variation; examples include negative frequency-dependent selection, overdominance, and certain types of environmental heterogeneity (WRIGHT 1969 ; HEDRICK 1986 ). The relative importance of MSE and balancing selection in maintaining genetic variation for fitness traits has important implications for a number of evolutionary and ecological issues, including the maintenance of sexual reproduction (KONDRASHOV 1988 ; MAYNARD SMITH 1978 ; CHARLESWORTH 1990 ; HAMILTON et al. 1990 ), the evolution of mating systems and mate choice (CHARLESWORTH 1987 ; RICE 1988 ; CHARLESWORTH and CHARLESWORTH 1990 ; CHARLESWORTH et al. 1990 ; MAYNARD SMITH 1991 ), the nature of constraints on adaptive evolution (HOULE et al. 1996), and the effectiveness of different strategies in conservation biology (LANDE 1994 ; FRANKHAM 1995 ).

One approach to evaluating the MSE hypothesis is to study variation in fitness traits created by newly arisen mutations. To the extent that genetic variation in fitness traits in natural populations ("standing" variation) is maintained by MSE, then properties of the standing variation should be predictable from properties of variation created by new mutations. One well-known property of standing genetic variation for fitness traits is inbreeding depression, with its converse of heterosis or hybrid vigor. The magnitude of inbreeding depression for a fitness trait can be predicted from a simple model of MSE in an effectively infinite population (CHARLESWORTH and HUGHES 1998 ):

(1)

Here, B is the logarithm of the ratio of the trait mean of outbred genotypes to that of inbred genotypes, u_i is the mutation rate at the ith locus, t_i and s_i are the effects of homozygosity for a mutation at the ith locus on the trait and on total fitness, respectively, and h_i is the dominance coefficient (h = 0.5 for mutations with additive effects). This equation assumes that mutations are not completely recessive (h > 0), and that the effects of different mutations combine in multiplicative fashion.

In Drosophila melanogaster, inbreeding depression for egg-to-adult viability has been well-characterized, and the decline in viability caused by rendering second chromosomes from natural populations homozygous agrees roughly with that predicted from estimates of the rates and dominance properties of second chromosome viability mutations (reviewed in SIMMONS and CROW 1977 ; CHARLESWORTH and CHARLESWORTH 1987 ). There is much less information on mutation rates, dominance of mutations, and inbreeding depression for other fitness components in D. melanogaster. A prediction of Equation 1 is that if some fitness traits have higher mutation rates or lower average dominance of mutations than others, those fitness traits should exhibit greater inbreeding depression, all other things being equal. Here, we report the results of experiments that give a rough test of this prediction. We measured the degree of inbreeding depression for three fitness components, egg-to-adult viability, female fecundity, and male fertility, in a set of D. melanogaster lines recently derived from the wild. Inbreeding depression was measured by its converse, the degree of heterosis (hybrid vigor) when highly inbred lines are crossed. For comparison, we also measured heterosis for the same fitness components among a set of lines in which all or most variation resulted from spontaneous mutations that had been allowed to accumulate for over 200 generations (MACKAY et al. 1992 ; FRY et al. 1996 ). By similar reasoning as for Equation 1, the logarithm of the ratio of the trait mean among hybrids between different mutation-accumulation (MA) lines to that of the pure MA lines is expected to be

(2)

The absence of heterosis among a set of lines for a particular fitness component could arise for two quite different reasons: either the lines do not have different alleles at loci affecting the fitness component, or they do have different alleles, but the alleles act in additive fashion (or, in the case of male-limited traits, are X- or Y-linked). To distinguish between these possibilities for fitness components showing no heterosis in our study, we used an indirect approach to infer the relative contributions of the three fitness components to variation among each set of lines in a composite fitness index, productivity (roughly, the number of adult offspring resulting from a fixed number of mated females laying eggs for a fixed time period). The results showed that although male fertility showed no heterosis in either set of lines, male fertility made a strong contribution to the productivity variation among the mutation accumulation lines, and also may have contributed to the productivity variation among the inbred lines. Further genetic analysis was conducted on the MA lines to determine whether the absence of heterosis for male fertility was caused by additivity of autosomal mutations affecting the trait, or alternatively by X- or Y-linkage of the mutations.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Culture conditions:
Fly lines were reared in shell vials (2.5 cm diameter) with ~10 ml of cornmeal-molasses-agar medium at 25°. Unless otherwise indicated, the parental density was eight females and five males per vial, with the flies being discarded after 5 days of egg-laying.

Origin of mutation-accumulation lines:
A single subline of the Harwich strain was inbred by full-sib mating for 41 generations (MACKAY et al. 1992 ). The strain was then expanded for one generation, after which it was subdivided into 25 independent MA lines. These were subsequently kept separately by mass matings of 10 randomly chosen males and females per 2-week generation. At generation 202 of MA, 20 lines were chosen for the work reported here. To control for common environment effects, each line was divided into two sublines two generations before the work reported here was begun. The sublines were maintained separately for the duration of the experiments (about 20 fly generations), with one or more rearing vials per subline.

Examination of insertion sites of retrotransposable elements on larval polytene chromosomes, on slides made between generations 170 and 180, revealed that 19 of the 20 chosen lines were free from contamination (NUZHDIN and MACKAY 1994 ). A subsequent check of copia insertion sites after the work reported here was completed revealed that these 19 lines were still uncontaminated (S. NUZHDIN, personal communication). The 19 lines shared a common set of 101 insertion sites of the elements 297, mdg3, Doc, and copia (NUZHDIN and MACKAY 1994 ). The remaining line (line 14 in this paper; line 32 in NUZHDIN and MACKAY 1994 ) was missing four of these sites; this could have been the result of contamination, or of unusual transposable element activity in this line. If contamination was the cause, it appears that only a small percent of the genome of line 14 derived from the contaminating strain, because 97 of the 101 common sites were still present. Most of these sites would not likely to have been shared by an unrelated contaminating strain. Where appropriate below, we report results of analyses with line 14 excluded.

Origin of inbred lines:
In July 1994, 33 isofemale lines were established from flies collected at the Raleigh, NC, Farmers' Market. Inbred lines were established from each isofemale line by 14 consecutive generations of full-sib mating. In each generation of inbreeding, three crosses per line were set up; if all three crosses failed to produce progeny, the line was discarded. Twenty-four lines survived the inbreeding process; of these, 16 were judged to be healthy enough to be used for the experiments reported here. Possible biases introduced by selection in the inbreeding process, and by the final choice of lines, are discussed below. Each inbred line was divided into two sublines, which were maintained separately until the work with them was complete (about 10 fly generations), with three rearing vials per subline.

Tests for heterosis in egg-to-adult viability, female fecundity, and male fertility:
We used a "round-robin" crossing scheme to test for heterosis in egg-to-adult viability, female fecundity, and male fertility (Table 1). Five types of crosses were conducted to produce different levels of heterozygosity in the offspring, female parents, and male parents, and the productivity of the crosses (number of adult flies emerging) was measured under competitive and noncompetitive conditions. For example, to test for heterosis in egg-to-adult viability, the mean productivity of crosses between females and males of different lines was compared to that of crosses between females and males of the same lines (groups II and I, respectively, in Table 1). In this comparison, there is no difference in the mean heterozygosity of the parental females and males, but there is a difference in the heterozygosity of the offspring. Similarly, to test for heterosis in female fecundity, the mean productivity of crosses between F1 hybrid females and pure line males (from a line not used as one of the parents of the hybrid females) was compared to that of crosses between females and males of different pure lines (groups IV and II, respectively, in Table 1), and the mean productivity of the hybrid female x hybrid male group (V) was compared to that of the line female x hybrid male group (III). In both of these comparisons, the offspring and the parental males have the same average heterozygosity, but the females differ in heterozygosity. Similarly, to test for heterosis in male fertility, the productivity of group III was compared to that of group II, and the productivity of group V was compared to that of group IV. Note that each line contributes 1/nth of the genetic material of the progeny, female parents and male parents of each group in Table 1, where n is the number of lines; therefore under additive inheritance, no differences among the groups in productivity would be expected.

To measure productivity under competitive conditions, 11 virgin females and 7 males of the appropriate genotypes (pure line or hybrid) were first placed together in vials. After 3 days, the males were discarded, and females from the cross were placed in a new vial, along with mated females from an unrelated stock bearing the yellow body color marker. For tests involving the MA lines, eight line females were placed with six yellow females; for the inbred lines, the corresponding numbers were nine and five. These numbers were chosen to produce an average proportion of wild-type (MA or inbred line) flies of about 50% among the progeny of group I crosses, based on preliminary tests. The females were removed after four days, and emerging wild-type and yellow adults were counted until day 19 after initiation of the vials. This competitive assay is similar to methods used by LATTER and ROBERTSON 1962 , KONDRASHOV and HOULE 1994 , and LATTER and MULLEY 1995 . To measure productivity under noncompetitive conditions, two virgin females and two males of the appropriate genotypes were placed in a vial and discarded after five days, with progeny counted again until day 19. This assay differs from the noncompetitive productivity assay that HAYMER and HARTL 1983 showed to be uncorrelated with competitive fitness measures. Our competitive and noncompetitive assay in fact generally give consistent results, with little genotype-environment interaction across the two assay types (FRY et al. 1996 ; J. D. FRY, S. L. HEINSOHN and T. F. C. MACKAY, unpublished data).

Productivity assays with the MA lines were conducted in two sets per treatment; the first set compared groups I, II, and V, and the second set compared groups II–V. Two (three) replicate vials per cross were set up in each set of competitive (noncompetitive) assays. Productivity assays with the inbred lines were conducted in one set per treatment, comparing all five groups simultaneously, with three replicate vials per cross under both treatments. Within each set of assays, crosses were set up on the same day in a randomized order, using parental flies that had been collected from rearing vials set up on the same day. The age of the parental flies varied from 2 to 9 days after emergence, but within any one set of assays, most parental flies were within 3 days age of each other, and flies of different ages were distributed randomly among the treatments. The two sublines per line were pooled for these assays, because the goal was to test for differences in the means between the groups, not to partition the variance within groups into genetic and environmental components. Nonetheless, similar assays comparing pure lines in which the sublines were kept separate showed that line effects were much stronger than subline effects (FRY et al. 1996 ; J. D. Fry, S. L. Heinsohn and T. F. C. Mackay, unpublished data).

Statistical analysis of heterosis tests:
For the competitive assays, two variables were analyzed, the total number of wild-type flies emerging, and the proportion of emerging flies that were wild type. Angular transformation of the proportions had no effect on the conclusions, so analyses of the untransformed proportions are reported. For the noncompetitive assays, the total number of wild-type flies was the only analysis variable. Analysis was performed on the means of the two or three replicate vials per cross. Means of the groups were compared by analysis of variance. The assumption of independent errors is obviously violated in this case, but the following arguments suggest that this is not likely to have caused incorrect conclusions. (1) Positive correlations of errors within groups make ANOVA tests too liberal, whereas positive correlations of errors between groups make ANOVA tests too conservative. Because the same or related genotypes occur both within and between groups in this instance, the two effects should roughly cancel each other. (2) Group differences were often highly significant, and there was consistency between the results for the competitive and noncompetitive treatments, even though these were performed at different times. (3) A method was devised to compare groups I and II that avoided problems of nonindependence (see below), and this gave similar results as ANOVA. (4) Additional experiments described below permitted confirmation of the conclusions regarding heterosis for egg-to-adult viability in both sets of lines, and of those regarding heterosis for male fertility in the MA lines, with problems of nonindependence avoided.

A test of the difference between groups I and II that avoids problems of nonindependence was performed as follows. If P_ixj is the productivity of the cross between line i females and line j males, then, in the absence of heterosis for offspring viability, the differences D_i = P_ix(i+1) - should on average be zero. Although the differences D_i and D_i+1 are not statistically independent, the differences D_i and D_i+2 are independent, because they involve nonoverlapping sets of lines. Therefore the set of n differences, where n is the number of lines (n = 20 or 16), can be divided into two sets of n/2 differences that are mutually independent within sets, {D₁, D₃, D₅, ... D_n-1} and {D₂, D₄, D₆, ... D_n}. Because the independence assumption is satisfied within each set, a t-test can be used to compare the mean of each set to zero; by using a critical value of = 0.025, the probability of one or more type I errors in the two comparisons should not exceed 0.05. One-tailed tests were used, as there was a clear prediction that the mean difference should be either positive or zero.

The competitive assays were used to estimate the magnitude of heterosis for egg-to-adult viability, female fecundity, and male fertility. An appropriate measure of relative fitness of wild-type flies to yellow competitors in a cross is given by w = , where p_w and p_y are the proportions of wild-type and yellow flies emerging from the cross (p_w + p_y = 1), and n_w and n_y are the numbers of wild-type and yellow parents used (LATTER and SVED 1994 ). For each group in Table 1, the mean w was calculated. Heterosis for pre-adult viability was estimated as Heterosis for female fecundity was estimated by taking the average of and , and heterosis for male fertility was estimated by the average of and .

We report results of the heterosis tests with the possibly contaminated MA line included; conclusions of all significance tests were unaffected by deletion of crosses involving this line.

Diallel crosses to assess variation in egg-to-adult viability, female fecundity, and male fertility:
The absence of heterosis for a fitness component could have two explanations: there could be no genetic variation for the trait; or genetic variation for the trait is present, but acts additively on average. To help distinguish between these possibilities, we used a diallel crossing technique to infer which fitness components caused the productivity variation among the lines. From each set of lines, we chose three lines that consistently showed relatively high productivity, and three lines that consistently showed low productivity (see below for details), for a total of six lines per set. These six lines were crossed to each other in all 36 possible combinations, and the productivity of the crosses assayed under both the competitive and noncompetitive treatments described above. If variation among the lines in female fecundity was the main cause of the productivity variation, productivity should follow the female parent of the cross: i.e., all crosses involving females from the high (low) productivity lines should have high (low) productivity, regardless of the male of the cross. Similarly, if variation among the lines in male fertility was the main cause of the productivity variation, productivity should follow the male parent. Other possible results are less definitive: e.g., if approximately equal contributions of female and male parent to the productivity variation are observed, this could be explained by variation in egg-to-adult survival of the progeny, or by roughly equal contributions of female and male fertility.

The 20 MA lines were tested for productivity variation previously (FRY et al. 1996 ). The lines were ranked based on the mean number of progeny in the competitive and noncompetitive treatments in the first two blocks of that study, and the three with the highest and lowest values were chosen for the diallel crosses. One of the high productivity lines chosen was line 14, the possibly contaminated line (this line was not included in the analysis reported in FRY et al. 1996 ). When relevant, results of analyses below will be reported both with this line included and with it excluded. Twenty-one of the surviving inbred lines were also tested for productivity in two blocks (J. D. FRY, S. L. HEINSOHN, and T. F. C. MACKAY, unpublished data), using identical methods as reported for the MA lines (FRY et al. 1996 ; the 16 inbred lines used in the round-robin crosses described above are a subset of these 21 lines). The three lines with the highest mean productivity in the two treatments, and three of the five with the lowest mean productivity, were chosen for the diallel crosses (two of the other low productivity lines were deemed too unhealthy to work with). Methods for the productivity assays from the diallel crosses were the same as those described above, except in this case, the two sublines per line were kept separate. Within each combination of treatment and set of lines, there were two blocks set up at different times: in the first block, subline 1 of each of the six lines was used, and in the second block, subline 2 of each line was used. Within each block, four and six replicate vials per each of the 36 crosses were set up in the competitive and noncompetitive treatments, respectively.

Statistical analysis of diallel crosses:
Analyses were performed on the mean number of MA or inbred line flies emerging from the four or six replicate vials per cross-block combination, and on the mean proportion of flies that were wild-type in the competitive treatment.

We start by considering a simple model of the productivity variation in the diallel crosses, ignoring block effects for the moment for simplicity. Let Y_ij be the number of progeny produced in the cross between females of line i and males of line j. Three types of genetic variation potentially contribute to variation in the Y_ij: variation among the female parents affecting the number of eggs that are laid; variation among the male parents affecting the number of eggs that are successfully fertilized; and variation among the progeny affecting their probability of surviving to the adult stage. A simple model that incorporates these three sources of variation is given by:

(3)

(4)

It would also be possible to add a term to Equation 4 that reflects the interaction between the fertility contributions, but this interaction would be confounded with the p_i*j.

We next consider estimation and significance testing of the parameters in Equation 2. Because the lines were not chosen at random, we consider line effects as fixed, so that:

(5)

Suppose the diallel data (cross means for productivity) are laid out in a 6 x 6 table with rows for the line of the female parent (i) and columns for the line of the male parent (j). Let F_i and M_j be the observed marginal means for females of line i and males of line j, respectively, calculated on data standardized to have mean zero. From Equation 2 and Equation 3, F_i estimates f_i + p_i , while M_j estimates m_j + p_j . We therefore used analysis of variance, with line of female parent and line of male parent as crossed, fixed factors, to test the null hypotheses that f_i + p_i 0 and m_j + p_j 0. A random block term was included in the ANOVAs, and the female and male parent main effects were tested over their respective interactions with block. The ANOVAs were also used to test the female parent x male parent interaction, i.e., whether the null hypothesis p_i*j 0 can be rejected.

Additional diallels to analyze genetic basis of male fertility variation among the MA lines:
The above diallel crosses revealed that the high and low productivity MA lines differed strongly in male fertility (see below). Additional diallel crosses were performed to analyze the genetic basis of the male fertility variation. The three high productivity and three low productivity MA lines were again crossed in all 36 possible combinations. Instead of analyzing the productivity of these crosses, F1 males for each cross were collected and crossed to the same genotype of female, and the productivity of these crosses analyzed. These "F1 diallels" were repeated under two sets of conditions, with two blocks each involving the independent sublines as above. In the competitive F1 diallel, the F1 males were crossed to virgin females from an average-productivity MA line (FRY et al. 1996 ); these females were then used for competitive productivity assays, using the same methods and sample sizes as for the competitive diallels describe above. In the "single-male" F1 diallel, single F1 males were crossed to single virgin females from the yellow stock (N = 7 replicate crosses x 36 combinations x 2 blocks). After 24 hr, the males were removed from the vials; the females were allowed to lay eggs for 3 more days, and then transferred to new vials for another 4 days. Adult progeny emerging from both sets of vials were counted, with the two counts pooled for analysis. An additional diallel was conducted using F2 males (progeny obtained by allowing F1 females and males from a given cross to mate) instead of F1 males. For the F2 diallel, single F2 males from each of the 36 combinations were again mated to single yellow females (N = 4 replicate crosses per combination and block); the males were discarded after 24 hr, and the females allowed to lay eggs for 3 more days, after which they also were discarded. The statistical analysis of the F1 and F2 diallels will be described along with the results.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Heterosis in the mutation-accumulation and inbred lines:
The first set of heterosis tests with the MA lines compared the productivity of lines crossed to themselves, lines crossed to different lines, and hybrids between lines crossed to hybrids between different pairs of lines (Figure 1). There was significant variation among these groups in the number of MA line flies emerging per vial in both the competitive and noncompetitive treatments, and in the proportion of wild-type (MA line) flies in the competitive treatment (ANOVA; P < 0.0005 in each case). Pairwise comparisons among the groups using Tukey's studentized range test (SAS INSTITUTE INC. 1989) showed that in each case, means of the hybrid x hybrid group were significantly greater than those of the first two groups, with no significant differences between the first two groups. The lack of difference in productivity between pure line and interline crosses was confirmed by the paired t-test method (see MATERIALS AND METHODS; P > 0.35 in each case).

View larger version (14K): In this window In a new window Download PPT slide   Figure 1. —Results of the first set of heterosis tests with the MA lines. The cross types are lines crossed to themselves (I), lines crossed to different lines (II), and hybrids between lines crossed to hybrids between different pairs of lines (V; see Table 1). Standard errors among the 20 cross means are shown. (A and B) competitive treatment. (C) noncompetitive treatment.

The second set of heterosis tests with the MA lines compared the productivity of four groups of crosses, all of which resulted in offspring that were heterozygous for mutations fixed in different lines, but which differed in the hybridization status of the female and male parents (Figure 2, Table 2). Two-way analysis of variance (Table 2) showed a significant effect of female hybridization status on productivity in both the competitive and noncompetitive treatments, with crosses involving hybrid females producing more offspring, and a greater proportion of MA line vs. competitor offspring, than crosses involving pure line females. In contrast, there was no significant effect of male hybridization status on productivity, nor were the interactions between female and male hybridization status significant (Figure 2, Table 2).

View larger version (16K): In this window In a new window Download PPT slide   Figure 2. —Results of the second set of heterosis tests with the MA lines. See Table 1 for definition of cross types. Standard errors among the 20 cross means are shown. (A and B) competitive treatment. (C) noncompetitive treatment.

The five types of crosses involving the inbred lines (Table 1) were compared in productivity simultaneously (Figure 3). The analysis was divided into two parts: the paired t-test method was used to compare groups I and II, and two-way ANOVA was used on the data from groups II to V to test for the effects of female and male hybridization status. The t-test method showed that, in contrast to the situation with the MA lines, crosses between different lines had significantly higher productivity than intra-line crosses. (In the first grouping of lines from the competitive treatment, P-values from the t-tests were 0.037 and 0.020 for the number and proportions of wild-type flies, respectively; in the second grouping of lines, the corresponding P-values were 0.006 and 0.004. In the noncompetitive treatment, P-values from comparing the number of flies emerging were 0.016 and 0.36 in the two groupings.) In cross types II through V (i.e., all but the pure line crosses), hybrid status of females significantly increased productivity in both the competitive and noncompetitive treatments (Table 3). In contrast, the effect of male hybridization status on the number of wild-type (inbred line) flies emerging was nonsignificant in both treatments. Hybridization of males was associated with a slight, marginally significant (P = 0.04) increase in the proportion of wild-type flies emerging in the competitive treatment. There were no significant interactions between hybridization status of females and that of males.

View larger version (17K): In this window In a new window Download PPT slide   Figure 3. —Results of heterosis tests with the inbred lines. See Table 1 for definition of cross types. Standard errors among the 16 cross means are shown. (A and B) competitive treatment. (C) noncompetitive treatment.

Heterosis estimates for both sets of lines are shown in Table 4. In the MA lines, if offspring and both sexes of parents were hybrid, there was a 75% increase in relative fitness compared to the pure line. Hybrid status of the female parent alone increased fitness by 40%. These two figures come from the first and second sets of heterosis tests with the MA lines, respectively; the difference between them may be purely a sampling effect, as no significant effects of hybridization status of males or offspring were detected. In the inbred lines offspring had twofold higher fitness than pure line offspring, and hybrid status of female parents was associated with a fourfold fitness increase. The estimated heterosis for net fitness in the inbred lines was 14-fold.

Diallel crosses:
In crosses among six MA lines, three with relatively high productivity and three with relatively low productivity, ANOVA showed no significant effect of line of female parent on productivity (Table 5, Figure 4). In contrast, there was a significant effect of line of male parent on both measures of productivity in the competitive treatment, and a nearly significant effect in the noncompetitive treatment.

View larger version (16K): In this window In a new window Download PPT slide   Figure 4. —Results of diallel crossing experiments among three LP (low productivity) and three HP (high productivity) MA lines. Each point gives the mean number of MA line flies emerging from the cross between line i females and line j males. Within a column, the six points represent crosses to females of the same line; the six columns give results for the females of the different lines. Open circles, squares, and triangles correspond to LP lines 2, 3, and 21, respectively. Closed circles, squares, and triangles correspond to HP lines 11, 14, and 16. (A) competitive treatment. (B) noncompetitive treatment.

The significant effects of line of male parent in the above ANOVA could have resulted from either genetic variation in male fertility, or genetic variation in progeny survival, or both. Under the hypothesis that the significant male parent effects were entirely caused by variation in progeny survival, line of male parent and line of female parent should have contributed equally to the productivity variation, with any apparent differences explained by sampling error. (An exception occurs in the case of X-linkage of genes affecting progeny survival, which would cause line of female parent to contribute more to productivity variation than line of male parent; however, this is opposite to the pattern observed.) The hypothesis of equal maternal and paternal contributions can be tested using the differences between the female parent and male parent marginal means, F_i - M_i, which estimate f_i - m_i (i.e., female fecundity minus male fertility; see Equation 4). If the hypothesis f_i - m_i 0 can be rejected, there must be variation in female fecundity, male fertility, or both. The hypothesis can be tested by testing for a "maternal effect" using traditional analysis methods for diallel data. We used the program of SCHAFFER and USANIS 1969 ; the maternal effect sums of squares provided by this program reflects variation in the quantities (F_i - M_i)/2. An F-test was conducted by dividing the maternal effect mean square (5 d.f.) by the maternal effect x block mean square (5 d.f.). The F-test was significant for both measures in the competitive treatment (P = 0.01), and nearly significant in the noncompetitive treatment (P = 0.07). These results mean that line of male parent contributed significantly more to the productivity variation than line of female parent. If these analyses are repeated with the possibly contaminated line excluded, the effect of male parent remains significant in the competitive treatment, but becomes nonsignificant in the noncompetitive treatment (Table 5). Maternal effects also remain significant (P < 0.05) in the competitive treatment, and in fact become more significant (P = 0.02) in the noncompetitive treatment.

There were no significant interactions between line of female parent and line of male parent in either treatment (Table 5), confirming the absence of heterosis for egg-to-adult viability in the MA lines. The mean proportions of wild-type flies resulting from within- and between-line crosses in the competitive diallel can be used to calculate an additional heterosis estimate for egg-to-adult viability (Table 4).

The above analyses do not directly address the question of what fitness traits were responsible for the productivity difference between the high productivity (HP) and low productivity (LP) lines. We therefore used t -tests to compare the mean values of F, M, and F - M (averaged across blocks) between the three HP and three LP lines. If, for example, the productivity differences were caused primarily by differences in male fertility, we would expect differences in the marginal means by male parent, but little or no difference in marginal means by female parent. For F and M, we used one-tailed t-tests, because we expect values of f, m, or p to be either higher for HP line than for LP lines, or else not different between the groups. For comparing the quantities F - M, we used two-tailed t -tests, because differences in either direction might plausibly occur. Higher values of F - M for HP lines than for LP lines would be expected if _high > _low (i.e., if the high lines had higher female fecundity), with little or no difference between _high and _low. Similarly, lower values of F - M for HP lines than for LP lines would be expected if _high > _low (i.e., if the HP lines had higher male fertility), with little or no difference between _high and _low. One could also imagine that _high < _low and _high < _low, but these possibilities seem a priori unlikely.

In the competitive treatment, crosses involving a given line of female always produced more offspring on average when the male parent came from an HP line than when the male came from a LP line (Figure 4A). Not surprisingly, M values (i.e., means by male parent) were significantly higher for the HP lines than for the LP lines when compared by t-tests (Table 6). In the noncompetitive treatment, the effect of male parent was not as pronounced (Figure 4B), but M values were nonetheless still significantly higher for the HP lines than for the LP lines (Table 6). F-values (i.e., means by female parent) were significantly higher for the HP lines than for the LP lines in the competitive treatment, but no similar difference was observed in the noncompetitive treatment. Estimated "maternal" effects, calculated as F - M, were significantly lower for the HP lines than for the LP lines in the competitive treatment, with a similar difference that approached significance (P = 0.06) in the noncompetitive treatment. These differences give evidence that male fertility was greater in the HP lines than in the LP lines. If the possibly contaminated line is excluded, male parent marginal means in the competitive treatment remain significantly higher for the HP lines than for the LP lines, but the differences in female parent marginal means and estimated maternal effects become not quite significant (P between 0.068 and 0.094). Excluding the possibly contaminated line from the noncompetitive dataset causes the difference in male parent marginal means to become not quite significant, and also decreases the significance of the difference in maternal effects (Table 6).

Y-linked effects on offspring survival provide an alternative explanation to male fertility variation for the strong effects of male parent on productivity variation seen in the MA lines. In the competitive diallels, the sexes of emerging wild-type flies were distinguished, allowing a check for this possibility. An ANOVA of the type shown in Table 5 revealed no significant effect of line of male parent on progeny sex ratio, and there was no tendency for crosses involving LP line males to produce a lower proportion of males than crosses involving HP line males (data not shown), thus ruling out the Y-linkage explanation.

In the crosses among six inbred lines with relatively high or low productivity, ANOVA showed significant effects of line of male parent on productivity in both the competitive and noncompetitive treatments (Figure 5, Table 7). The effect of line of female parent was significant in the noncompetitive treatment, and approached significance in the competitive treatment. The F-test for "maternal" effects was significant for the proportion of wild-type flies in the competitive treatment (P = 0.043), but not for the number of wild-type flies in either the competitive (P = 0.33) or noncompetitive treatments, although it approached significance (P = 0.097) in the latter case. In contrast to the results with the MA lines, there were highly significant interactions between line of female parent and line of male parent in the competitive treatment, with the interaction approaching significance in the noncompetitive treatment. A linear contrast between the six within-line crosses and the remaining 30 between-line crosses accounted for a large proportion of the interaction sums of squares in each case, and means of between-line crosses were higher than those of within-line crosses, a pattern discernible from Figure 5. These results confirm the presence of heterosis for egg-to-adult viability in the inbred lines. An estimate of the magnitude of the heterosis based on the diallel cross results is similar to that from the "round-robin" crosses (Table 4). In the competitive treatment, but not the noncompetitive treatment, significant interaction nonetheless remains after partitioning out the contrast between the within- and between-line crosses (Table 7). This remaining interaction appears to result partly from the unexpectedly high productivity of crosses between line 25 females and line 1 males (both HP lines; Figure 5).

View larger version (15K): In this window In a new window Download PPT slide   Figure 5. —Results of diallel crossing experiments among three LP and three HP inbred lines. See legend to Figure 4 for explanation. Open circles, squares, and triangles correspond to LP lines 18, 26, and 34, respectively. Closed circles, squares, and triangles correspond to HP lines 1, 20, and 25. (A) competitive treatment. (B) noncompetitive treatment.

When compared by t-tests, marginal means by both female and male parent were in all cases either significantly or nearly significantly (P = 0.055) higher in the HP inbred lines than in the LP lines (Table 8). In contrast, estimated "maternal" effects did not differ significantly between the two groups of lines.

Diallel crosses to analyze genetic basis of variation in male fertility among MA lines:
F1 and F2 males from all 36 possible crosses among the three high productivity and three low productivity MA lines were crossed to the same genotype of females in the competitive F1, single-male F1, and single-male F2 diallels. In the F1 diallels, X-linked or cytoplasmic contributions to male fertility variation should be detectable as a maternal effect: i.e., the productivity of a cross should be influenced more by the line of a male's mother than by the line of the male's father. In addition, Y-linked effects on fertility would produce an apparent "maternal effect" in opposite direction to the productivity differences between the lines (negative for high productivity lines, positive for low productivity lines). Therefore, we tested for maternal effects (variation among the differences F - M) by both the F-test and t-test methods described above. There were no significant maternal effects in either F1 diallel (Table 9). In addition, t-tests showed no significant differences in estimated maternal effects between the HP and LP lines in the F1 diallels (Table 10).

The F2 diallel was conducted to provide information on possible Y-linked or cytoplasmic effects on male fertility. The two groups of reciprocal F2 males should differ in the origin of their Y chromosome and their cytoplasm, but not in their autosomes or X chromosome. Either cytoplasmic or Y-linked effects should lead to a significant maternal effect by the F-test approach, provided that the two do not cancel. If the differences F - M are greater for HP lines than for LP lines, this would suggest the presence of cytoplasmic effects (because F2 males descending from crosses between HP females and LP males have higher fertility than males from the reciprocal cross). If instead the differences are lower for HP lines, this would suggest the presence of Y-linked effects. The F-test approach revealed no significant maternal effects in the F2 diallel (Table 9). In contrast, estimated maternal effects (F - M) were marginally significantly higher for the HP lines than the LP lines when compared by a t-test (Table 10). If real, this difference could be explained by a cytoplasmic effect, but given the lack of significant maternal effects in the F1 diallels, and the high P value (P = 0.87) from the F-test for maternal effects, the result seems more likely to be spurious.

In the absence of sex-linkage, cytoplasmic effects, and other possible sources of maternal effects, the quantities (F + M)/2 estimate general combining abilities (GCAs), which can be interpreted as the average effects of the haploid autosomal contribution of each line (cf. GRIFFING 1956 ). As maternal effects appeared to be absent in the F1 diallels, we tested for variation among the quantities (F + M)/2 by both an F-test, and by a one-tailed t-test. For the F-test, the GCA mean square provided by the program of SCHAFFER and USANIS 1969 was divided by the GCA x block interaction mean square (both 5 d.f.). GCAs did not vary significantly among the lines in the F1 diallels when tested by ANOVA, but did vary significantly or nearly significantly in the F2 diallel (Table 9), depending on whether the possibly contaminated line was included. In both F1 diallels and the F2 diallel, estimated GCAs were significantly higher for the HP lines than for the LP lines when compared by t-tests (Table 10), although the differences in the F1 competitive diallel become nonsignificant when the possibly contaminated line is excluded. These differences in GCAs, combined with the apparent absence of maternal effects, indicate that the male fertility variation was primarily autosomal.

The pooled specific combining ability (SCA) plus reciprocal effect was not significant in any case, although it approached significance in the F2 diallel (Table 9). The sum of squares for this effect is mathematically identical to the sum of squares for the female parent x male parent interaction in ANOVAs of the form given in Table 5 and Table 7. The lack of significant interaction is consistent with the absence of heterosis for male fertility in the round-robin crosses. In the F1 competitive diallel, the mean proportions of wild-type flies were nearly identical between within- and between-line crosses, so that an estimate of the magnitude of heterosis for male fertility from this diallel is 1.00 (Table 4). The mean number of wild-type flies produced in each of the F1 male diallels was slightly lower in between- than within-line crosses, and was only 2% higher in between- than within-line crosses in the F2 diallel (data not shown).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In the first part of the work reported here, we used a round-robin crossing scheme to measure heterosis among a set of inbred lines and a set of MA lines for three fitness components: preadult viability, female fecundity, and male fertility. Heterosis for preadult viability was present in the inbred lines but absent in the MA lines in both the competitive and noncompetitive treatments, and these results were confirmed in the diallel crossing experiments. Both sets of lines showed strong heterosis for female fecundity in both treatments. The inbred lines showed only marginally significant evidence for heterosis for male fertility by one of two measures in the competitive treatment, and no evidence for heterosis in the noncompetitive treatment. The MA lines showed no evidence for heterosis for male fertility in either treatment, a result that was later confirmed in the male diallels.

The absence of heterosis for male fertility in both sets of lines, and for egg-to-adult viability in the MA lines, could be explained if the lines were not genetically differentiated for these traits. In the diallel crosses among high and low productivity MA lines, however, productivity was influenced more by the line of the male parent than by the line of the female parent, indicating that the lines differed in male fertility. It is less clear whether genetic variation for male fertility was present among the inbred lines. In the diallel crosses among high and low productivity inbred lines, both male and female parent contributed to the productivity differences. In contrast to the results with the MA lines, there was no consistent difference between the HP and LP lines in estimated "maternal effects," which would include effects of male fertility. There was nonetheless some evidence that the line of the male parent may have had more influence on productivity than the line of the female parent. The ANOVA test for maternal effects was significant for the proportion of wild-type flies in the competitive diallel, and nearly significant for the number of wild-type flies in the noncompetitive diallel. This indicates that there was probably some variation among the six lines in the quantities f - m (see MATERIALS AND METHODS). Inspection of the estimates of these quantities (Table 8) indicate that in both cases where significant or nearly significant variation existed, the lowest estimate came from a high productivity line (line 1), and the highest estimate came from a low productivity line (either line 26 or line 18). This could be explained if line 1 had relatively high male fertility (high m) and lines 18 and 26 had relatively low male fertility. The alternative explanation is low female fecundity in line 1 and high fecundity in lines 18 and 26, but this seems less likely, as it requires the fecundity differences to be in opposite direction to the productivity differences. Nonetheless, it is clear that while differences in male fertility may have contributed to the productivity difference between the HP and LP inbred lines, differences in female fecundity and/or egg-to-adult viability also made a major contribution.

It is also not clear whether there was genetic variation in offspring viability among the MA lines. Crosses involving females from the HP lines produced significantly more progeny than crosses involving females from the LP lines in the competitive treatment diallel, but not in the noncompetitive treatment. The significant difference in the former case could be explained either by a difference in offspring viability, or a difference in female fecundity, or both. Because the MA lines showed heterosis for female fecundity, they had clearly accumulated different mutations affecting fecundity, but it is possible that these mutations contributed relatively little to productivity differences among the lines.

The results therefore do not rule out the possibility that the lack of heterosis for offspring viability in the MA lines, and the lack (or at least weakness) of heterosis for male fertility in the inbred lines, were caused simply by absence of genetic differentiation among the lines for these traits. Alternatively, genetic differentiation for the traits may have been present, but directional dominance absent. Some evidence in favor of this possibility for viability mutations in the MA lines comes from results showing that the dominance of mutations affecting viability is inversely correlated with their homozygous effect, the mildly deleterious mutations acting in a nearly additive manner (SIMMONS and CROW 1977 ; CROW and SIMMONS 1983 ). Because the MA lines were kept at a population size of 10 pairs, selection should have been strong enough to prevent fixation of mutations with large homozygous effects on viability, but not strong enough to prevent fixation of mutations with relatively mild homozygous effects. (Recall that Equation 2 ignores selection in the MA lines). In contrast, in the process of full-sib inbreeding, some of the inbred lines were likely to have become fixed for mutations with large homozygous viability effects that were segregating at low frequencies in the base population. This may explain why the inbred lines, but not the MA lines, showed heterosis for offspring viability. This hypothesis could be tested by subjecting sublines of the MA lines to full-sib inbreeding. If the hypothesis is correct, the sublines should show heterosis for viability on intercrossing as a result of having become fixed for mutations with large effects on viability that were segregating at low frequency. An implication of the hypothesis, together with our observation of heterosis for female fecundity among the MA lines, would be that mild-effect fecundity mutations are more recessive than mild-effect viability mutations, a possibility suggested by the analysis of CHARLESWORTH and HUGHES 1998 .

X- or Y-linkage of genes affecting male fertility could also explain the lack of heterosis for male fertility in the inbred lines. This possibility could be ruled out for the MA lines, in which no evidence for X- or Y-linkage of genes affecting male fertility was found in the male diallels (cf. CLARK 1990 ). The autosomal basis of the male fertility variation in the MA lines, coupled with the absence of heterosis for male fertility, indicates that mutations affecting this trait that became fixed in the MA lines acted in an additive manner on average. If this result holds in general, it could explain the lack or weakness of heterosis for male fertility in the inbred lines. Our results do not imply, however, that mutations with large effects on male fertility, such as those causing complete sterility, act in additive manner; as with viability mutations, only mutations with comparatively small effects on male fertility could have become fixed in the MA lines.

Although the results give clear evidence for male fertility variation in the MA lines, we do not know the mechanism of the variation. Males from the high productivity lines may have mated with higher frequency, transferred more viable sperm with each mating, and/or stimulated females to lay more eggs (cf. CHEN et al. 1988 ) than males from the low productivity lines. Some evidence to discriminate these possibilities comes from the F1 single-male diallels, in which females were transferred to a second vial 3 days after being exposed to a male for 24 hr. Of crosses that produced at least one offspring in the first vial, indicating that mating had occurred, 26% of the second vials produced offspring when the male resulted from a cross between two HP lines (N = 107), compared to 14% when the male resulted from a cross between and HP and an LP line (N = 210), and only 7% when the male resulted from a cross between two LP lines (N = 91). These results are presented without analysis, because of the lack of independence of crosses within each group. Nonetheless, they suggest that the differences in male fertility between the HP and LP lines were not simply the result of differences in the frequency of mating, because mating with males from the HP lines apparently resulted in females producing offspring for longer periods than mating with males from LP lines. Whatever the exact cause of the male fertility differences, it is interesting that the disproportionate effects of line of male parent on productivity were observed in both the competitive diallel, in which males were not present during the egg-laying period, and in the noncompetitive diallel, in which males were present.

The rate of mutations affecting viability on the D. melanogaster second chromosome has been estimated as at least 0.1–0.2 per generation (MUKAI 1964 ; MUKAI et al. 1972 ; OHNISHI 1977 ), but there is less information on rates of mutations affecting other fitness components. Our results suggest that mutations affecting male fertility may have occurred at a high rate in the MA lines relative to mutations affecting other fitness components. If this conclusion holds in general, it would mean that the overall rate of deleterious mutations in D. melanogaster is much higher than that for viability alone. Nonetheless, it is possible that the base inbred population for the MA experiment had some peculiarity that made it particularly susceptible to mutations affecting male fertility. Work on the effects of spontaneous mutations on different fitness components using other base population is needed to determine whether male fertility provides a particularly large mutational "target." Supporting this possibility, mutations causing male sterility in homozygous condition have considerably outnumbered those causing female sterility in some mutational screens (LINDSLEY and TOKUYASU 1980 ), although different results are sometimes obtained (COOLEY et al. 1988 ). Additional indirect evidence comes from the repeated observation that second chromosomes causing sterility in males in homozygous condition are about twice as frequent in D. melanogaster populations as those causing sterility in females (TEMIN 1966 ; WATANABE and OSHIMA 1973 ; WATANABE and OHNISHI 1975 ).

Comparison to previous work:
Although many studies have found heterosis for female fecundity in crosses among inbred lines in D. melanogaster (e.g., GOWEN 1952 ; BARNES 1968 ; WATANABE and OHINISHI 1975; DOMINGUEZ and ALBORNOZ 1987 ; SANTIAGO et al. 1989 ; EHIOBU and GODDARD 1990; EHIOBU et al. 1990 ), we know of only two previous studies that have examined the dominance properties of new, spontaneous mutations affecting fecundity. FERNANDEZ and LOPEZ-FANJUL 1996 examined fecundity variation among MA lines maintained for over 100 generations by full-sib mating. Extreme lines for fecundity, both high and low, were crossed to a control line that had been maintained at a large population size, and fecundity of the hybrids examined. Variable patterns of dominance were observed, including apparent strong underdominance in a few cases, with no overall tendency towards dominance of high fecundity. The results of FERNÁNDEZ and LÓPEZ-FANJUL therefore appear at odds with our finding of heterosis for fecundity among MA lines. Some aspects of their results, however, suggest that the authors may not have been able to distinguish genetic variation among the lines from variation caused by common-environment effects. Of six extreme fecundity lines identified based on assays performed in generations 104–106, one of the four high lines showed no difference from the control when retested a few generations later, and one of the low lines had higher fecundity than the control by about 10 standard errors (FERNANDEZ and LOPEZ-FANJUL 1996 , Table 4). In addition, fecundity differences between the reciprocal F1's of several standard errors were observed in two of the six crosses, which is inexplicable on genetic grounds.

In a recent study, HOULE et al. 1997 crossed second-chromosome mutation accumulation lines to each other, and examined fecundity of both the pure lines and the hybrids. Hybrids had higher means for early and late fecundity than the homozygotes, with the difference significant for late fecundity. HOULE et al.'s results are therefore consistent with the heterosis for fecundity among lines bearing new mutations observed in this study. One difference between HOULE et al.'s approach and the one used here is that HOULE et al. made direct egg counts, while we inferred heterosis for female fecundity by comparing the mean productivity of crosses involving heterozygous females with that of crosses involving homozygous females. Our finding of heterosis for female fecundity by this method may reflect not only differences in the numbers of eggs laid, but differences in the hatch rate of those eggs. KEARSEY and KOJIMA 1967 showed that the hatch rate of eggs is strongly influenced by the genotype of the mother, with chromosomal heterozygotes laying eggs that hatched at a higher rate than chromosomal homozygotes.

While several studies have found heterosis for competitive mating success of males and/or mating speed in D. melanogaster (FULKER 1966 ; PENDLEBURY and KIDWELL 1974 ; BRITTNACHER 1981 ; SHARP 1984 ; MILLER and HEDRICK 1993 ; HUGHES 1995 ), much less information is available on heterosis for male fertility in the absence of mating competition. If new mutations with relatively mild effects on male fertility act in additive manner on average, as our results suggest, we would expect heterosis for male fertility to be less prevalent than for female fecundity. Our results with the inbred lines are consistent with this prediction, but the few other studies examining possible heterosis for male fertility used diverse methods and do not give a consistent picture. In an impressively large-scale study, TEMIN 1966 measured male fertility of over 200 second chromosome homozygotes and heterozygotes. Fertility was measured as the number of crosses producing offspring out of 10 single pair matings to females of a standard stock, an assay quite different from the one used here. Heterozygotes had 13% higher fertility than homozygotes, but this was mostly caused by completely sterile homozygous lines (about 8% of the total). When these were excluded, the fertility superiority of heterozygotes was only 4%. PARTRIDGE et al. 1985 measured male fertility of 40 third chromosome homozygotes, by counting the number of adult progeny produced in matings of single males to females of a standard stock. Mean fertility by this assay was only half that of a single heterozygous stock formed by pooling the homozygous lines. This difference was partly caused by 12 lines that were completely or almost completely male-sterile; nonetheless, the heterozygous stock had higher fertility than most of the remaining homozygous lines (Figure 2 and Table 2 in PARTRIDGE et al. 1985 ). Although this result suggests strong heterosis for male fertility among chromosomes with relatively mild effects on fertility, it was not clear whether the heterozygous stock contained equal representation of all the original third chromosomes when it was tested; if not, it would not have been a completely valid control for comparison with the homozygotes. Using a somewhat similar assay, HUGHES 1995 also compared male fertility of third chromosome homozygotes and heterozygotes; no evidence of heterosis was found, although the sample sizes were small (two males per line tested versus 20 by PARTRIDGE et al. 1985 ). In addition, HUGHES excluded from analysis homozygotes with fertility less tha