No evidence of positive assortative mating for genetic quality in fruit flies

Abstract

In sexual populations, the effectiveness of selection will depend on how gametes combine with respect to genetic quality. If gametes with deleterious alleles are likely to combine with one another, deleterious genetic variation can be more easily purged by selection. Assortative mating, where there is a positive correlation between parents in a phenotype of interest such as body size, is often observed in nature, but does not necessarily reveal how gametes ultimately combine with respect to genetic quality itself. We manipulated genetic quality in fruit fly populations using an inbreeding scheme designed to provide an unbiased measure of mating patterns. While inbred flies had substantially reduced reproductive success, their gametes did not combine with those of other inbred flies more often than expected by chance, indicating a lack of positive assortative mating. Instead, we detected a negative correlation in genetic quality between parents, i.e. disassortative mating, which diminished with age. This pattern is expected to reduce the genetic variance for fitness, diminishing the effectiveness of selection. We discuss how mechanisms of sexual selection could produce a pattern of disassortative mating. Our study highlights that sexual selection has the potential to either increase or decrease genetic load.

1. Introduction

Genetic load owing to deleterious alleles at mutation–selection balance is thought to be an important determinant of mean fitness in many populations [1]. There is evidence from several model systems that sexual selection acts more strongly against deleterious alleles in males than in females, which is expected to reduce the genetic load experienced by females, thereby potentially protecting population productivity [2–5].

These previous studies clearly show that many deleterious alleles reduce reproductive success in both sexes, while affecting males more strongly, at least in simple laboratory environments. Given this genetic variance for adult reproductive success in each sex, we should also consider how these genetically variable males and females will combine their gametes when producing offspring, i.e. the correlation in genetic quality between parents [6], where genetic quality is defined as the breeding value of an individual for total fitness [7].

Note that while sexual selection can reduce the relative contribution that low-quality individuals make to the gamete pool, the gametes in this pool could still combine with one another at random [8]. For instance, low-quality males might be less likely to obtain a mate than high-quality males, regardless of whether their potential partners are of high or low quality. A useful way to describe this scenario, where gametes combine at random, is to say that there is no correlation in genetic quality between the male and female parents.

Alternatively, various mechanisms of sexual selection could lead to a correlation in genetic quality between reproductive partners. Positive assortative mating by individual quality could arise if males compete more intensely for access to high-quality females relative to low-quality females, or if high-quality females exhibit stronger mate preferences than low-quality females; theoretical models predict the evolution of such behaviour when mate competition is costly [9].

If gametes combine non-randomly with respect to genetic quality, this will alter the pattern of genetic variance in the following generation. Positive assortative mating for genetic quality will increase genetic variance in the next generation relative to the case of random mating; by aggregating deleterious alleles in a subset of the population, positive assortative mating is expected to improve the response to selection and reduce mutation load [10]. Negative assortative mating, also called disassortative mating, will have the opposite effect.

There are many observations of positive assortative mating occurring in natural populations [11]. These studies were not designed to test the possible role of non-random mating in altering mutation load, and they have important limitations in that context. First, they generally consider only copulation, i.e. mating per se, as opposed to reproduction. Post-copulatory sexual selection, such as sperm competition, could be relevant to how gametes ultimately combine [6]. Second, these observations generally relate to readily measured phenotypes like body size, which are not necessarily strongly correlated with genetic values for fitness. For example, even if larger individuals tend to have greater reproductive success than smaller individuals, this difference is only relevant to the extent that body size is correlated with the genetic value for expected fitness.

The pattern of reproduction, as opposed to copulation only, can feasibly be addressed by using genetic markers to determine parentage. However, the requirement that we consider genetic quality as opposed to some other trait presents a challenge, because offspring inherit the genetic quality of their parents to some extent, and will be subject to viability selection. As a consequence, a particular pattern of genetic marker combinations among offspring could indicate non-random reproduction among parents, as intended, but is also likely to suffer from bias owing to non-random probabilities of survival among offspring genotypes.

In an effort to address these issues, Sharp & Agrawal [6] studied reproductive patterns in small groups of marked Drosophila melanogaster, where males and females of the parental generation varied in condition owing to a diet quality treatment, but all offspring were reared on a high-quality diet. Manipulating condition as a proxy for genetic quality in this way should reduce or eliminate any bias owing to viability selection on offspring. This study found evidence for a positive assortative pattern of copulation among virgins, but no evidence for non-random reproduction considering several days of mating and re-mating. The appropriateness of considering condition instead of genetic quality in this context is unknown, but there is evidence that diet manipulations and deleterious alleles can have distinct effects on courtship signals [12], and post-copulatory traits [13], so using one as a proxy for the other may not be ideal.

To explore this issue further, here we introduce an improved method to test for non-random reproduction with respect to genetic quality in D. melanogaster, and we examine reproductive patterns on a greater spatial and temporal scale than previous laboratory studies. Our approach relies on the fact that one dimension of genetic quality—the degree of homozygosity for partially recessive deleterious alleles—can be manipulated in parents without altering the genetic quality of offspring. Based on the reproductive patterns of thousands of flies across dozens of experimental populations, we find no evidence of positive assortative reproduction. Instead, we detect a pattern indicating disassortative reproduction.

2. Methods

We first describe our general approach before describing each step in detail. Two types of male and two types of female were created with genetic markers that allow the parentage of all offspring to be established following a period of mating and oviposition. In half of these mating trials, one type of male and one type of female had inbred genotypes. Importantly, inbred males and females in a given mating trial were always derived from different families, such that the offspring of all parental combinations were outbred. In the other half of the mating trials, all flies were outbred; this was done to establish and account for any effect that the genetic markers themselves might have on the pattern of reproduction. This approach is described in more detail below and a graphical illustration is shown in figure 1.

Figure 1.

Experimental design; example of four cages. Each cage (rectangles) contained focal flies and standard flies of both sexes, for a total of 60 flies. In the inbreeding treatment (top row), focal flies were obtained by crossing parents from the same inbred line (top left), and were therefore inbred (red). In the control treatment (bottom row), focal flies were obtained by crossing parents from different inbred lines (middle left), and were therefore outbred (blue). To ensure that the offspring of focal flies were always outbred, focal males and females in a given cage always originated from different lines, or combinations of lines, from one another. The males and females obtained from each pair of parental crosses were used in a pair of ‘converse’ cages (arrows). Standard flies were obtained from large outbred populations (not shown). Chromosomes 2 and 3 are depicted for each genotype, which were either wild-type (+), or marked with bw, or marked with e. Line numbers used in these examples are arbitrary; for actual lines used, see the electronic supplementary material, table S1. (Online version in colour.)

We began by backcrossing the autosomal recessive phenotypic markers bw¹ (brown; chromosome 2; manifests as brown eyes) and e¹ (ebony; chromosome 3; manifests as dark body colour) for three generations into an outbred population, Dahomey, which has been maintained under laboratory conditions for many years. Flies were reared at 25°C, 70% relative humidity and a 12 h L : 12 h D cycle. Virgin flies were used for crosses and assays whenever necessary to control offspring genotypes, including virgin males and females in the primary assay.

Starting with outbred flies homozygous for both markers (bw/bw; e/e), we generated 48 independent inbred lines through two consecutive rounds of mating between full siblings, completing several generations prior to the primary assay. A random set of 24 inbred lines was used to generate inbred flies for the primary assay, by crossing flies from within a given inbred line to one another (electronic supplementary material, table S1).

All 48 lines were used to generate outbred flies for the primary assay, by crossing flies from two different inbred lines to each other (electronic supplementary material, table S1). In all cases, these crosses were conducted in both directions, and consisted of four vials each, with each vial containing two or three flies of each sex. From these vials, we collected ‘focal’ flies of both sexes for the primary assay (figure 1).

At the same time, we used the backcrossed marker strains to set up 40 vials of bw/bw; +/+ flies (i.e. marked with bw only) and 40 vials of +/+; e/e flies (i.e. marked with e only), with each vial containing three flies of each sex. From these vials, we collected ‘standard’ flies for the primary assay (bw/bw; +/+ females and +/+; e/e males; figure 1).

For each focal genotype, the males collected were used to set up one mating assay and the females were used to set up a separate mating assay (figure 1). This step is important because it ensures that flies will not produce offspring with any immediate relatives, such that all offspring combinations are outbred, eliminating any bias that might arise from selection against low-quality offspring prior to scoring. We will refer to these pairs of cages as being the ‘converse’ cages of one another. Apart from the paired nature of the assays, which is owing to the fact that we used both males and females from each genotype, we ensured that all combinations of genotypes in inbred and control assays were unique (electronic supplementary material, table S1).

The primary assay involved 48 cages (22 × 25 × 33 cm), each containing 60 flies: 15 focal males, 15 focal females, 15 standard males and 15 standard females. In 24 cages (the ‘inbreeding’ treatment), focal flies were inbred as described above. In the other 24 cages (the ‘control’ treatment), focal flies were outbred as described above. In all cases, we can determine the parental type of all resulting offspring based on marker phenotypes. For example, offspring expressing the bw phenotype but not the e phenotype must have a female parent of the ‘standard’ type and a male parent of the ‘focal’ type (figure 1).

There is evidence that the spatial environment in which mating takes place can influence relative reproductive success, the degree of sexual harassment and selection against deleterious alleles [14–16]. In an attempt to provide some spatial complexity in the assay environment, we included at the bottom of each cage a grid-like cardboard divider often used in trays of Drosophila vials.

Upon cage set-up (day 1), five oviposition vials, seeded with live yeast, were available in each cage. Vials were changed each day, alternating between five yeasted oviposition vials or three yeasted holding vials. Offspring from oviposition vials (representing eggs laid on days 1, 3, 5 and 7) were later scored, whereas holding vials were discarded.

Scoring took place on days 12 and 15 post-oviposition. Each set of five oviposition vials was estimated to contain approximately 500 offspring; to save time, only a random subset of these offspring was scored, following the mixing of offspring from the five vials.

(a). Viability assay

In addition to these assays of reproductive success, we performed an additional test to estimate the relative viability of the marker genotypes that will appear among the offspring in the primary assays. We set up 32 vials with six bw/+; e/+ females and six bw/bw; e/e males each (the same density of parents per vial expected, on average, in the primary assay). After 1 day of oviposition, we transferred flies to new vials to oviposit for an additional day. This cross is expected to produce four offspring genotypes in equal numbers in the absence of viability selection. In this assay, we scored 11 295 offspring in total. The results are shown in the electronic supplementary material, table S2. There was no evidence for variation in viability among genotypes (goodness-of-fit test, χ₃² = 1.72, p = 0.63) nor an interaction between the two markers (2 × 2 contingency test, χ₁² = 0.24, p = 0.63). Therefore, we can most likely interpret the results of the primary assays solely in terms of differences in adult reproductive success, rather than viability effects of the markers.

(b). Statistical analysis

To quantify the reproductive pattern of parents in a given cage population, we assigned all offspring a value of X ∈ {0,1} and Y ∈ {0,1} based on their phenotype, where X is the type of male parent (0 for standard, 1 for focal) and Y is the type of female parent (0 for standard, 1 for focal). For example, offspring expressing the bw phenotype but not the e phenotype would be assigned values of X = 1 and Y = 0 to represent a focal male parent and a standard female parent. For each cage, we calculated E[X], the proportion of offspring sired by focal males (equivalently, the proportion of offspring expressing the bw phenotype), and E[Y], the proportion of offspring due to focal females (equivalently, the proportion of offspring expressing the e phenotype). The expected values of X and Y are both 0.5 if the focal flies and their standard competitors have equal reproductive success. The markers themselves could affect reproductive success, and so we tested for an effect of inbreeding by comparing the success of focal males or females between treatment and control cages; to test for effects of inbreeding and assay day, we used generalized linear mixed-effect models with binomial error distribution, with random effects of parental genotype combination and an individual-level random effect to account for overdispersion. Note that the measured effect of inbreeding will include any interaction between the marker alleles and the inbred genetic background. We also tested whether males and females of the same genotype tended to have similar reproductive success by examining the Pearson correlation between E[X] from a given cage and E[Y] from the converse cage. These correlations will reflect the intersexual genetic correlation for fitness but will be attenuated by sampling error, making these tests conservative.

We then used X and Y values to calculate the Pearson correlation between parental types, r_X,Y, as a measure of assortative mating in each cage. In the absence of assortative mating, we expect r_X,Y = 0. Note that this is the case even when there is sexual selection in favour of focal or standard flies, i.e. E[X] ≠ 0.5 or E[Y] ≠ 0.5. A positive value of r_X,Y indicates positive assortative reproduction by type (focal versus standard), and a negative value indicates disassortative reproduction by type. Ideally, we would find r_X,Y = 0 in the control cages, where focal and standard flies are all outbred and there is low variance in genetic quality. In practice, the genetic markers could have some direct effect on r_X,Y, and so we compared the r_X,Y values of treatment cages with the r_X,Y values of control cages. We will refer to r_X,Y as the ‘type correlation’.

In summary, for each cage, we used the counts of four types of offspring to calculate three values of interest: the success of focal males, the success of focal females and the tendency of focal males to reproduce with focal females more or less often than expected. To account for marker effects, we compared these values between control cages, where all flies are outbred, and treatment cages, where focal flies are inbred.

3. Results

From the primary assays, we scored 61 268 offspring in total, or 1276 per cage (range: 977–1729). Examining the proportions of offspring attributable to focal individuals using generalized linear mixed models with a random effect of parental genotype (where converse cages are assigned the same genotype), there is strong evidence that inbred males and females suffer from reduced reproductive success compared with controls (figure 2). On average, focal control females produced 44% of the offspring in their respective cages, whereas focal inbred females produced 33% of the offspring in their respective cages (where the remainder of the offspring are from standard competitor females; effect of inbreeding: χ²₁ = 16.13, p = 5.93 × 10^–5). Similarly, control males sired 51% of offspring, whereas inbred males sired 30% of offspring (effect of inbreeding: χ²₁ = 23.05, p = 1.58 × 10^–6). These differences between inbred and control flies were consistent across all four scored timepoints of the experiment (figure 2). There was no evidence of interactions between inbreeding status and day (females: χ²₁ = 1.21, p = 0.27; males: χ²₁ = 0.51, p = 0.47).

Figure 2.

Reproductive success of focal flies over time, in competition with standard flies. Inbred flies generated fewer offspring than controls, which was the case for both (a) females and (b) males. The reduction in reproductive success owing to inbreeding was consistent across the four timepoints of the experiment, which span 7 days. Inbreeding caused a greater reduction in reproductive success in males than in females. (Online version in colour.)

We next examined the correlation between the female reproductive success of a given line (inbred or control) and the male reproductive success of the same line in the converse cage (figure 1). We find a positive correlation between male and female reproductive success for both inbred lines (r₂₂ = 0.42, p = 0.042; figure 3) and control lines (r₂₂ = 0.58, p = 0.003; figure 3), suggesting that the variation in genetic quality in our experiment is generally sexually concordant. However, because each cage contains a pair of focal genotypes (one for each sex), which are also found in the converse cage, the cages are not all strictly independent. Repeating the above correlation tests assuming half the sample size, the p-values are still suggestive of a positive intersexual correlation (inbred lines: r₁₀ = 0.42, p = 0.057; control lines: r₁₀ = 0.58, p = 0.007). Additionally, we find that the effect of inbreeding is stronger on males than on females (paired t-test, t₂₃ = 3.73, p = 0.001) consistent with previous observations of the male-biased effects of inbreeding [17,18] and deleterious alleles more generally [2–5].

Figure 3.

Male versus female reproductive success among genotypes. Points represent total reproductive success of a given male genotype plotted against total reproductive success of the same genotype when expressed in females. For both control and inbred genotypes, male and female reproductive success were positively correlated. Note that males and females of a given genotype were assayed in converse cages along with another focal genotype (figure 1), and therefore, these points are paired and not strictly independent. (Online version in colour.)

Our primary goal was to test whether gametes combine in a non-random fashion with respect to the genetic quality of parents. To do this, we calculated the ‘type correlation’ for the offspring produced in each cage, where focal parents were assigned a value of 1 and standard parents were assigned a value of 0, and took the average over converse cages. If flies have random reproductive partners, the type correlation will be zero, even if focal flies have reduced reproductive success (e.g. owing to the presence of inbreeding depression). In control cages, the type correlation was slightly positive, on average, for most of the experiment (figure 4a). Recall that these cages serve as a control for any effects that the phenotypic markers bw and e might have on mating patterns. In treatment cages, where focal flies had reduced reproductive success owing to inbreeding depression, the average type correlation tended to be negative, and was always less than that of the control cages (figure 4a). Inbreeding had a significant negative effect on the type correlation on day 1 (t₂₂ = –2.19, p = 0.0397) and a non-significant negative effect on all other days (all t₂₂ > –1.21, all p > 0.24). Combining evidence across days, there is evidence that type correlations were consistently smaller under the inbreeding treatment (combined evidence from t-tests: z = –1.97, n = 4, p = 0.0483). Type correlations from cages in the inbreeding treatment were also more likely to be negative in sign than those from the control cages (combined evidence from Fisher's exact tests: z = –2.487, n = 4, p = 0.0129). The difference in type correlation between inbred cages and the respective control mean, shown in figure 4b, diminished over the course of the experiment (linear mixed model with random effect of genotype; effect of day: χ²₁ = 6.16, p = 0.0131). These results suggest that reproduction was disassortative with respect to genetic quality, particularly within the first day or so of reproduction.

Figure 4.

Type correlation (assortativeness) over the course of the experiment. (a) The type correlation was generally positive in cages from the control treatment, which was used to account for any marker effects, and generally negative in cages from the inbreeding treatment. (b) The difference in type correlation (inbred–control) was negative, indicating disassortative reproduction for genetic quality. (Online version in colour.)

4. Discussion

By pooling deleterious alleles, assortative reproduction could cause selection to be more effective. In fruit flies, there seems to be little support for assortative reproduction when variation in quality is owing to rearing conditions [6] or inbreeding (this study). If anything, we find evidence for disassortative mating, which will generate more genotypes of intermediate quality than expected by chance, hampering selection against deleterious alleles.

In their meta-analysis of mating patterns with respect to phenotypes such as body size, Jiang et al. [11] found that few studies (about 1 in 10) indicated disassortative mating. While these exceptions could represent sampling error, their data do not rule out the possibility of disassortative mating [11]. Unlike most other examples, our experiment focused on assortative reproduction (rather than mating) for genetic quality (rather than a trait like body size, where the relationship to fitness is unclear). So, while the conservative interpretation of our data is a lack of any correlation in the genetic quality of parents, we should also consider potential explanations for a genuine pattern of disassortative reproduction, given the unique nature of our dataset.

The signal of disassortative reproduction in our study was strongest at the earliest data collection point, where flies had been interacting for 0–24 h, and diminished thereafter (figure 4), implying that disassortative reproduction may have taken place among virgin flies. Although high-quality females are thought to be preferred by males for their high fecundity, they may also have a greater capacity to resist male harassment. If low-quality females are less able to resist mating attempts from high-quality males, a pattern of disassortative reproduction could be produced. However, note that fecundity of focal females was consistent over time (figure 2), and so, it is unlikely that high-quality females refrained from mating early in the experiment.

We can also consider our results in the context of post-copulatory sexual interactions, which can be influenced by both genetic and environmental variation. The ejaculates of male Drosophila contain accessory proteins that induce greater fecundity in females in the short term, but also result in decreased female lifespan [19]. Drosophila males show genetic variance in their effects on female fecundity [20], and the pattern of male investment in ejaculate components is expected to vary with condition and female receptivity [21]. In turn, the nature and extent of female responses to male manipulations have also been found to vary with female condition [21,22]. If the post-copulatory effects of high-quality male ejaculates are relatively greater when females are of low quality, perhaps because low-quality females are less resistant to their effects [22], we would observe a pattern of disassortative reproduction. If true, this might explain why the disassortative pattern diminished over time in our experiment: males are expected to invest more in ejaculate components that enhance female fecundity, and less in sperm, when a female has not yet mated and the risk of sperm competition is low [23]. Such post-copulatory interactions have been observed in the context of diet manipulations [22] and could also explain why flies in a previous experiment [6] showed no assortative reproduction for condition overall, despite the presence of assortative mate choice among virgins.

If indeed reproduction in fruit flies is random with respect to genetic quality, we should consider how this finding might be reconciled with the ample evidence of assortative mating for various traits in other organisms [11]. One possibility is that assortative mating patterns typically do not reflect reproductive patterns, and that the random union of gametes is the norm, excluding a possible evolutionary benefit of sexual selection. Alternatively, random reproduction could be the norm in many organisms or environments, but not all. While there is evidence of variability in the degree of assortative mating among taxa and traits [11], few attempts have been made to predict these patterns based on population characteristics [24]. In any case, the issue will only be resolved with data on genotypic reproductive patterns from more systems, with particular attention paid to the potential impact of post-copulatory interactions.

Our experimental approach was designed to eliminate bias owing to the heritability of genetic quality and potential interaction effects on offspring viability. Parental quality may also have non-genetic impacts on offspring; on their own, such maternal and paternal effects would contribute to our estimates of male and female reproductive success but would not affect our estimates of assortativeness. However, an interaction between maternal and paternal effects could produce a spurious signal of non-random reproduction. For example, if zygotes whose parents are both low quality are less likely to survive to adulthood than expected given the maternal and paternal effects alone, we would see a signal of disassortative reproduction. Rather than hampering selection, a parental effect interaction could have the same effect as synergistic epistasis, speeding the removal of deleterious alleles. While parental effect interactions are known to occur [25], we are not aware of any evidence of synergistic parental effects on offspring viability.

Our experiment was conducted under conditions intended to allow for a broader range of fly reproductive behaviour when compared with a vial, i.e. dozens of potential mating partners, over several days, in a space that may allow females to escape male harassment. Nevertheless, the laboratory environment is relatively constant, and certain conditions that could create a pattern of ‘incidental’ assortative mating [11], such as phenological or geographical segregation of types, were absent. Still, the conditions of our experiment clearly permitted strong sexual selection (figure 2), and so we can conclude that the presence of genetic variation in reproductive fitness alone is not sufficient to produce a pattern of positive assortative reproduction in this organism.

The premise of our approach was that inbreeding would expose partially recessive deleterious alleles, reducing genetic quality, but, in principle, the inbreeding depression we observed could also be owing to segregating alleles conferring heterozygote advantage, maintained by balancing selection [26]. Alleles with intra-locus sexually antagonistic effects would be particularly important in the present context—if much of the genetic variation in quality were because of such alleles, then positive assortative mating for genetic quality would be irrelevant to population productivity, as this would create offspring with a mix of alleles that are favoured in males but disfavoured in females, and vice versa, with no improvement in the effectiveness of selection. Previous experiments with the Dahomey population have indicated that new mutations have recessive, sexually concordant effects [4,27], but also suggest that some segregating alleles affecting reproductive traits could be subject to balancing selection [27]. Our finding that male and female reproductive success were positively correlated suggests that sexually antagonistic alleles were not the primary source of variation in sexual fitness in our experimental populations.

Our approach can be used to test for assortative reproduction in model laboratory organisms, but it should also be possible to address this question in well-studied wild populations, where pedigrees can be used to determine inbreeding status, parental relatedness and reproductive success. Assortative mating for heterozygosity has been observed in birds [28–30] but could stem partly from heterozygote advantage at particular loci, e.g. the major histocompatibility complex. Nevertheless, whole-genome data coupled with pedigrees may be a promising avenue to test for assortative reproduction for genetic quality in the wild.

Data accessibility

Data from this research are available in the Dryad Digital Repository: https://dx.doi.org/10.5061/dryad.v40j7gf [31].

Authors' contributions

N.P.S. and M.C.W. designed experiments. N.P.S. performed experiments, analysed the data and wrote the paper. All authors gave final approval for publication.

Competing interests

The authors have no competing interests.

Funding

This work was supported by a Banting Postdoctoral Fellowship to N.P.S. and a Discovery Grant to M.C.W. from the Natural Sciences and Engineering Research Council of Canada.