Assortative mating for relatedness in a large naturally occurring population of Drosophila melanogaster

Authors


Stephen P. Robinson, Centre for Evolutionary Biology, School of Animal Biology (M092), The University of Western Australia, Crawley, WA 6009, Australia.
Tel.: +61 8 6488 5824; fax: +61 8 6488 1029;
e-mail: stephen.robinson@graduate.uwa.edu.au

Abstract

New theoretical work on kin selection and inclusive fitness benefits predicts that individuals will sometimes choose close or intermediate relatives as mates to maximize their fitness. However, empirical examples supporting such predictions are rare. In this study, we look for such evidence in a natural population of Drosophila melanogaster. We compared mating and nonmating individuals to test whether mating was nonrandom with respect to relatedness. Consistent with optimal inbreeding, males were more closely related to their mate than to randomly sampled females. However, all individuals collected mating showed higher relatedness and males were not significantly more related to their mate than to other mating females. We also found a negative relationship between relatedness and fecundity. Our results are consistent with the hypothesis that inclusive fitness benefits may drive inbreeding tolerance despite direct costs to fitness; however, an experimental approach is needed to investigate the link between mate preference and relatedness.

Introduction

Inbreeding is well known for its severe negative consequences. Avoidance of close inbreeding is almost universal in human cultures (Brown, 1991). Indeed, Darwin recognized that a lack of genetic diversity between pairs was avoided in nature, writing ‘Nature… abhors perpetual self fertilization’ and speculating that ‘there must be something injurious in the process’ (Darwin, 1862). It is no surprise then that researchers have often considered inbreeding as purely negative. But are all the consequences of inbreeding negative? There is a growing body of theoretical and empirical work to indicate that the situation may be more complex than previously recognized. Parker (1979) first argued that gains in inclusive fitness could theoretically lead individuals to choose to mate with related individuals despite costs through inbreeding depression, if these related individuals would otherwise have low mating success. More recently, Kokko & Ots (2006) and Puurtinen (2011) have developed this theory further and present modelling which suggests that inbreeding tolerance, or even preferences for related mates, can be expected under a wide range of realistic scenarios. This could represent a major shift in the way we think about inbreeding and may have wide reaching implications for evolutionary processes, including the evolution of female preference, the evolution of display traits, the dynamics of adaptation and the purging of genetic load.

An individual’s inclusive fitness is made up of both its direct fitness (the number of offspring it produces) and also the indirect fitness gained when another individual, such as a relative who shares some proportion of genes in common, passes on those shared genes (Hamilton, 1964). From the viewpoint of the selfish gene (Dawkins, 1976), it makes no difference whether the gene is passed on directly by the individual or whether it is passed on by a relative, only that the gene becomes more common in the population as a result. Therefore, individuals who can help a relative to produce more offspring than they would otherwise have been able to produce may be able to increase their inclusive fitness. Inclusive fitness gains have been suggested to explain some cases of apparent altruistic nest helping behaviour in birds, where individuals may have higher fitness if they help a relative to produce a greater number of offspring than if they try to produce offspring of their own without help (Stacey & Koenig, 1990).

A gain in inclusive fitness from inbreeding can also occur when inbreeding results in offspring that are more closely related to the parents and therefore represent a greater genetic benefit. An individual’s inclusive fitness can be increased by inbreeding if, by mating with a close relative, that relative’s reproductive fitness is increased over what it would otherwise have been. However, it is still crucial that in order for mating with a relative to represent a net gain via inclusive fitness, the relative’s reproductive success must be higher as a consequence of the mating than it would otherwise have been. Any offspring produced by a relative, regardless of who they have mated with, will result in a benefit via inclusive fitness, as they will pass on genes shared through common decent. If, by mating with a relative, that relative loses another mating opportunity, then any benefit the individual receives through increased relatedness to it’s own offspring may be negated by the loss of inclusive fitness that the relative would have provided by passing on shared genes through mating with another individual (Dawkins, 1976). However, as Smith (1979) and Parker (1979) point out, if the mating that a sister gives to a brother is the only one that he is likely to receive, then she does not deprive herself of an independent nephew or niece, but gains an offspring that is very closely related to her and therefore more likely to pass on her genes. Therefore, the mating success of relatives, mate encounter rates and the time spent out of the mating pool due to each mating are all expected to have large effects on whether there is a net benefit to inclusive fitness from inbreeding. It is modelling of such factors that led Kokko & Ots (2006) to conclude that in many real-world situations, gains in inclusive fitness should be able to offset even substantial costs of inbreeding depression.

Puurtinen (2011) took this concept a step further, examining what the optimal level of inbreeding would be if individuals were to choose mates based on relatedness in order to maximize their fitness. Puurtinen (2011) demonstrates not only that inbreeding can be expected to evolve as a strategy under a wide range of strengths of inbreeding depression, but also that, for females, there should often exist an optimum level of mate relatedness that maximizes fitness. Puurtinen (2011) suggests that for a range of realistic values of inbreeding depression, positive values of the inbreeding coefficient (FIS) should be observed in populations if they are choosing mates based on relatedness in order to maximize fitness. However, empirical data in support of such a situation are somewhat lacking. Kokko & Ots (2006) provide a number of explanations for this mismatch between theory and empirical results and suggest that the evidence may be present if we look for it. Puurtinen (2011) points out that there are indeed examples in the literature that are consistent with this new theory, but states that more evidence is needed. There is now a growing list of studies that demonstrate classical mate choice for close or intermediate relatives under controlled experimental conditions (e.g. Bateson, 1982; Barnard & Fitzsimons, 1988, 1989; Keane, 1990; Schjorring & Jager, 2007; Thunken et al., 2007; Richard et al., 2009). However, such preferences have been demonstrated under natural conditions in only a handful of studies, and much of this evidence is largely restricted to birds (e.g. Ratti et al., 1995; Krokene & Lifjeld, 2000; Cohen & Derhorn, 2004; Barber et al., 2005; Kleven et al., 2005; Wang & Lu, 2011), although Sherman et al. (2008) found evidence of cryptic choice in a natural population of Person’s tree frog, Litoria peronii, with males of high genetic similarity to females having greater success at sperm competition.

Our goal was to gather empirical data on the level of inbreeding present in a natural population of the fruit fly Drosophila melanogaster. We chose D. melanogaster because this species has a number of attributes lending itself to such a study. Previous work means that we have a good idea of the morphological traits that are subject to sexual selection in this species, and there is a rich diversity of microsatellite markers available for genetic assays. They can also be easily collected copulating in the field, and their high mobility and population density make inbreeding due to a lack of unrelated mates less likely (if mating opportunities are restricted, individuals may pair with related mates more often than expected by chance, simply because they are the only mates available). There is also a mechanism that may allow Drosophila to discriminate between potential mates based on relatedness. Cuticular hydrocarbons (CHCs) have long been known to be important in sexual signalling in D. melanogaster (Jallon, 1984) and have been suggested to be important in species, sex and kin recognition in insects (Singer, 1998). CHCs have also been implicated in mediating assortative mating between ‘countryside’ and ‘urban’ strains of D. melanogaster in the Congo (Haerty et al., 2002).

Although mating is often considered as being driven by either female mate choice or male–male competition, the role of male mate choice should not be discounted (Edward & Chapman, 2011). In the closely related species, Drosophilila serata,Chenoweth & Blows (2005) found sexual selection acting on cuticular hydrocarbons via both male mate choice and female mate choice. The form of this selection differed between the sexes and suggested that males and females showed different mate preferences for this trait. Because optimal inbreeding theory predicts differing benefits for males and females from a given level of inbreeding, mate preference for relatedness is not expected to be symmetrical between the sexes (Parker, 1979; Kokko & Ots, 2006; Puurtinen, 2011). Although we were primarily interested in the overall outcome of male and female mate preference in this study, there may be subtle differences in the relatedness of the potential mates that males and females had access to. We therefore considered males and females in separate analyses.

We set out first to determine whether individuals collected mating in the field were more closely related to each other than to other potential mating partners. Approaches such as this, which consider relatedness in isolation, have been used in previous studies investigating the effects of relatedness on mate choice (e.g. Cohen & Derhorn, 2004). However, as Blows (2007) points out, selection does not act on traits in isolation and analysing individual components of fitness may give misleading results. Because an individual’s choice will be influenced by more than just their relatedness to a potential mate, we therefore included two morphological characters in our analysis, wing size and sex comb tooth number, which have previously been suggested to be under sexual selection (Taylor & Kekic, 1988; Ahuja & Singh, 2008). In using these analyses, our purpose is not to generate estimates of sexual selection, but rather to control for the potential confounding effects of morphological variation on mating success. Finally, to determine the effect of mate relatedness on offspring production, we determined the number of offspring surviving to adulthood that were produced by each mating pair under laboratory conditions.

Materials and methods

Field collection and offspring production

Pairs of copulating D. melanogaster were collected in April 2008 from a large, naturally occurring population at a winery located near Margaret River on the south-west coast of Western Australia (33.83°S, 115.06°E). All flies were collected from within a seven by seven metre area at our densely populated study site. Collections were made in the early morning between 6:30 am and 8:30 am during the peak of mating activity. Ninety-six copulating pairs, which will hereafter be referred to as the mating sample, were collected over three consecutive days by carefully coaxing them into individual vials where they were allowed to complete copulation. Collection vials were filled with 10 mL standard agar–maze–yeast-laying medium. Males and females were kept together on this medium and allowed to lay for 48 h before being stored in 100% ethanol. Sixteen days after the initial collection, the number of offspring emerging from each vial was recorded. All individuals were stored in 100% ethanol prior to processing. Due to the strong last male sperm precedence in D. melanogaster (Ram & Wolfner, 2007), the majority of these offspring are likely to be sired by the male collected with the female. Females had not previously been exposed to this artificial laying medium, so variation in oviposition preference is likely to have contributed random noise to this measure of offspring production. Directly after the collection of the last mating pair on each day, a random sample of individuals from the population was collected using a sweep net at the centre of the collection site. Of these randomly collected individuals, 71 males and 70 females, hereafter referred to as the random sample, were analysed along with the mating pairs. Offspring production was only assessed for females from the mating sample, because females from the random sample may not have mated, and if they had mated, the time of their last mating or how closely related they were to their last mate were unknown.

Morphological traits

Wing size

Wing size was measured because it has previously been found to be under directional sexual selection in D. melanogaster, with males having larger wings being more successful at finding mates (Taylor & Kekic, 1988). Both wings of each fly were removed using forceps, dipped in Histoclear and mounted under a cover slip on a glass slide using Aquamount. Once these slides were set, wings were photographed using a digital camera attached to a Leica DMLS compound microscope (Leica Microsystems, Wetzlar, Germany) at 40× magnification. The photographs were saved as black and white TIFF images and analysed using the computer program Object-Image (Vischer et al., 1994). Landmarks were placed on each wing in order to calculate wing area using methods described by Gilchrist & Partridge (1999). Wing size was measured using the right wing of each individual. In rare cases where the right wing was missing or damaged, the left wing was used instead.

Sex combs

In Drosophila, the sex comb is a structure consisting of a row of modified bristles or ‘teeth’ located on the anterior ventral surface of the first pair of legs in males and is highly variable across species (Kopp & True, 2002). The sex comb is thought to be involved in manipulating the female during copulation, and evidence has been found that the number of teeth making up the sex comb in D. melanogaster is under sexual selection (Ahuja & Singh, 2008). The number of teeth on the sex comb on each male’s right leg was counted and used as a measure of sex comb size. This was carried out by removing the right foreleg, mounting it on a slide with double-sided tape and observing the sex comb under a Leica DMLS compound microscope at 200× magnification. Where the right leg was missing or damaged, the left leg was measured in its place.

Genetic traits

Genetic traits were assessed using twenty-three polymorphic microsatellite loci, randomly distributed across the second and third chromosomes. Genotypes for these loci were scored from DNA extracted from each mating pair (96 males and 96 females) or randomly sampled individual (71 males and 70 females). DNA extraction and PCR protocols followed methods outlined in the study by Gockel et al. (2001). PCR products were analysed on an ABI 3730 Sequencer, and allele sizes were scored using Genescan-500 LIZ internal size standard and GeneMapper software version 3.7 (Applied Biosystems, Foster City, CA, USA). Information about the microsatellite loci, including primer sequences, is provided in the Table S1.

Genetic differentiation between flies collected on different days and mating status types was assessed by calculating Weir & Cockerham’s (1984) estimator of FST. Pairwise FST values and tests for differentiation among the groups were calculated using the fstat (version 2.9.3.2) software package (Goudet, 1995). Relationships between all wild-caught flies were visualized by constructing a phylogenetic tree of individuals, based on Cavalli-Sforza & Edwards (1967) distance and UPGMA algorithm, performed using the populations (version 1.2.31) software program (Langella 1999). For comparison, four D. melanogaster from a different site, 235 km away in the Swan Valley, Western Australia (31.85°S, 116.00°E), and four Drosophila simulans were included in the phylogenetic analysis.

Relatedness

Relatedness was calculated as Queller & Goodnight’s (1989) pairwise relatedness (R) using the software Kingroup (Konovalov et al., 2004). For individuals collected as part of a mating pair, we calculated three separate relatedness scores. All comparisons and averages were restricted to between individuals collected on the same day. First, R was calculated between the individual and the male or female it was collected in-copula with. Second, we calculated the average relatedness between the individual and all opposite sex individuals from the random collection. Third, we calculated the average relatedness of the individual to all opposite sex individuals collected as part of the mating sample (excluding comparisons to their actual mating partner).

For the multivariate analysis, the relatedness measure used for individuals from the mating sample was their relatedness to their copulation partner. For individuals in the randomly collected sample, who do not have a copulation partner to calculate relatedness with, each individual was randomly assigned an opposite sex individual collected as part of the mating sample on the same day. This relatedness score represents their relatedness to an individual who could potentially have mated with them, but was found mating with an alternative individual when collected.

Because there are known biases in relatedness estimators depending on individual heterozygosity (Roberts et al., 2006), some individuals are expected to have higher average pairwise relatedness estimates than others. To correct for this, we subtracted each individual’s average relatedness to all randomly sampled individuals of both sexes collected on the same day from their relatedness to their actual or potential mate. This new relatedness score then represents the difference between the relatedness they share with their mate and their average relatedness with the population. However, because performing this correction did not qualitatively alter any of our analyses, we present analyses based on the uncorrected scores only.

Statistical analysis

All statistical analyses were carried out using R (version 2.11.1 GUI 1.34 Leopard build) (R Development Core Team, 2010) unless otherwise stated, and means are presented ± 1 SE.

Univariate analysis

For mating individuals, we carried out paired Student’s t-tests separately for males and females. First, we compared each individual’s relatedness to their actual mating partner with their average relatedness to all randomly collected opposite sex individuals. Second as a more conservative estimate ensuring comparison was made only to individuals known to have mated that morning, we compared each individual’s relatedness to their actual mate with their average relatedness to the opposite sex individuals collected as part of the mating sample (excluding their actual mate).

Next, we wanted to determine whether there was a difference between the average relatedness within the mating sample and the average relatedness within the random sample. We carried out a Student’s t-test comparing the mean relatedness of each individual collected as part of the mating sample to all other individuals from the mating sample (excluding their actual mate) with the mean relatedness of each individual collected as part of the random sample to all other individuals from the random sample. Data were checked to confirm equal variances prior to this test.

Multivariate analysis

We used generalized linear models (GLM) and response surface methodology (Phillips & Arnold, 1989; Blows & Brooks, 2003) to investigate the linear and nonlinear effects of relatedness (to actual mates for the mating sample and potential mates for random sample) on mating success. These methods are widely used to estimate selection coefficients operating on particular traits of individuals (e.g. Brooks et al., 2005). The linear component of these methods has also been used to investigate the effects of traits such as relatedness, which is a property of a pair of individuals (e.g. Wang & Lu, 2011). We constructed separate models to examine the male and female mating success. Before analysis, the effect of day of collection was tested for each trait using anova. For females, there was no effect of day of collection on any trait. For males, wing size was significantly affected by the day of collection. We therefore standardized these data by subtracting the collection day mean from the trait value and dividing this number by the standard deviation. All scores were converted to Z scores with a mean of 0 and a standard deviation of 1 prior to analysis.

Mating success

For mating success, we set the error distribution of the GLM to binomial because the response data were binary (0 or 1 for randomly sampled and mating individuals, respectively). The dependent variable was thus the binomial probability of mating, and the explanatory variables were relatedness, wing size and sex comb bristle number (males only). Response surfaces were fitted using the rsm package (version 1.40) (Lenth, 2009) in R. Before undertaking this analysis, individuals with missing data for one or more traits were excluded, and mating success was converted to a relative value by dividing each score by the mean. The response surfaces were used to derive the quadratic and correlational coefficients as well as to calculate eigenvalues and canonical weights for the multivariate analysis. Eigenvalues and quadratic regression coefficients were doubled as required when estimating nonlinear effects (Stinchcombe et al., 2008). Significance testing of eigenvalues was achieved using two methods: first, the classical double regression method was used (Phillips & Arnold, 1989) and second, because of the high false-positive rates associated with the double regression technique, the new permutation method recommended by Reynolds et al. (2010) was used. Double regression was carried out fitting a second-order polynomial linear model to the new multivariate variables. Permutation was carried out in R using code given in the Table S1 to Reynolds et al. (2010) using 10 000 permutations.

Offspring production

Offspring production was analysed only for females that were collected in copula. Day of collection was found to have a significant effect on offspring production. To correct for this, prior to analysis, the number of offspring produced by each individual was divided by the mean offspring production of all individuals collected on the same day. We used a GLM fitted with a quasi-Poisson error distribution to cope with overdispersion in the offspring count data. The explanatory variables were relatedness, male and female wing size and male sex comb number. Offspring production was then converted to relative offspring production, and response surface methodology was used to assess nonlinear effects using the same methods as outlined above for mating success.

Correlations between male and female traits

If size assortative mating is occurring, this could indirectly lead to assortative mating for relatedness, for example if more closely related individuals are more likely to be of similar sizes. We therefore used Pearson’s product–moment correlation to test for associations between morphological traits of males and females within mating pairs. If assortative mating for size causes assortative mating for relatedness, we would expect a relationship between the similarity in size of a pair, and their relatedness. To test for this possibility, we examined the correlation between pair relatedness and the difference in wing size between males and females of a pair.

Results

Genetic differentiation among samples

There was low but significant genetic differentiation between groups split by day of collection and mating status (FST = 0.004, = 0.001). Pairwise tests revealed significant divergences between different mating status types on each collection day, but not among collection days within the same mating status type (Table 1). All pairwise FST values were very low (range 0 to 0.008). The phylogenetic classification of individuals (Fig. 1) reveals no clear clustering of individuals with the same mating status. By contrast, individuals of D. melanogaster from a different site and D. simulans each cluster together.

Table 1.   Matrix of FST for comparisons between samples grouped by mating status and day of collection.
 DayMatingBystanders
123123
  1. P-values were obtained after 1500 permutations. Significance level for multiple comparisons is 0.003. Significant values are highlighted in boldface.

Mating1
20.002
30.001−0.001
Bystanders10.0060.0040.006
20.0080.0080.0070.002
30.0040.0070.006> 0.001−0.001
Figure 1.

 UPGMA tree of individuals. A total of 333 wild-caught individuals from the study site, as well as four D. melanogaster from a different site (sv) and four D. simulans (sim) are represented. Mating individuals are depicted by red coloured branch tips and bystanders with blue coloured branch tips.

Univariate analysis

Paired t-tests revealed that males were more closely related to the female they were collected copulating with (mean R = 0.016 ± 0.014) than they were on average to a randomly selected female from the population (mean R = −0.016 ± 0.004) (t95 = 2.272, = 0.025). However, using the more conservative approach, their average relatedness to all opposite sex individuals collected as part of the mating sample (mean R = 0.012 ± 0.005) was not found to be significantly different from their relatedness to their actual mate (t95 = 0.285, = 0.777).

For females, there was a similar, but statistically nonsignificant trend for greater relatedness to their actual mate (mean R = 0.016 ± 0.014) than their average relatedness to opposite sex individuals from the random sample (mean R = −0.011 ± 0.004) (t95 = 1.928, = 0.056). Again, when the more conservative comparison was made to only opposite sex individuals that were collected as part of the mating sample (mean R = 0.012 ± 0.005), there was no significant difference (t95 = 0.282, = 0.778).

A two-sample t-test revealed that opposite sex individuals collected as part of the mating sample were more closely related to each other (R = 0.0124 ± 0.005) than were the randomly collected opposite sex individuals to each other (R = −0.0124 ± 0.006) (t164 = −3.206, = 0.002).

Multivariate analysis

Linear, correlational and quadratic effects for male mating success are given in Table 2. The analysis detected a significant, positive linear effect between relatedness and mating success. There was one significant correlational (interaction) effect, relatedness by wing size (Table 2). None of the other linear, quadratic or correlational effects were significantly different from zero.

Table 2.   The three male linear coefficients and matrix of three quadratic and three correlational nonlinear coefficients for mating success in males.
 Linear (β)Nonlinear (γ)
RelatednessWing areaSex comb
  1. ± standard errors.

  2. Significant values are highlighted in boldface, *< 0.05.

Relatedness0.355 ± 0.166*0.058 ± 0.070.098 ± 0.039*−0.012 ± 0.044
Wing area−0.282 ± 0.1850.016 ± 0.0720.019 ± 0.052
Sex comb0.214 ± 0.1850.032 ± 0.074

The results of the canonical analysis investigating nonlinear effects on male mating success are given in Table 3. Significance testing of the associated eigenvalues using classical double regression indicated that the eigenvalue associated with the new multivariate m1 axis was significantly different from zero. However, P-values calculated by the more conservative permutation method, as advocated by Reynolds et al. (2010), indicated that this effect was likely to have been caused by random factors.

Table 3.   Eigenanalysis for nonlinear effects of predictor variables on mating success in males.
 m1m2m3
  1. Significant values are highlighted in boldface, *< 0.05.

Eigenvalue0.2360.065−0.117
Double regression P-value0.016*0.3890.287
Permutation P-value0.0840.3990.689
Canonical weights
 Relatedness0.776−0.1570.611
 Wing area0.6310.154−0.761
 Sex comb0.0250.9750.219

None of the linear, quadratic or correlational effects for female mating success were significantly different from zero (Table 4). However, the effect of relatedness on mating success in females was in the same direction and of similar magnitude to the effect in males (Tables 2 and 4).

Table 4.   The two female linear coefficients and matrix of two quadratic and one correlational nonlinear coefficents for mating success in females.
 Linear (β)Nonlinear (γ)
RelatednessWing area
  1. ± standard errors.

Relatedness0.237 ± 0.1640.054 ± 0.106−0.099 ± 0.070
Wing area0.008 ± 0.1620.026 ± 0.108

No significant nonlinear effects were detected for female mating success using either the double regression or permutation methods (Table 5).

Table 5.   Eigenanalysis for nonlinear effects of predictor variables on mating success in females.
 m1m2
Eigenvalue0.140−0.060
Double regression P-value0.2110.528
Permutation P-value0.3210.638
Canonical weights
 Relatedness−0.757−0.653
 Wing area0.653−0.757

Offspring production

The linear coefficients for offspring production revealed a significant positive effect of female wing size and a significant negative effect of increasing relatedness within mating pairs (Table 6). There was also a significant correlational effect between female wing size and relatedness (Table 6). Permutation revealed two significant axes of nonlinear effects from the response surface analysis, m1 and m3 (Table 7). Canonical weights show that m1 was weighted most heavily by male sex comb tooth number and relatedness, whereas m3 was weighted most heavily by female wing size and relatedness (Table 7). Visualization of these two multivariate variables using a fitness surface (Fig. 2) shows peak offspring production for low values of m1. This represents pairs with low relatedness and males with smaller sex combs. There is also a lower fitness peak at high values of m1 and high values of m2. This indicates that mating pairs with large female wings, large male sex combs and high relatedness had low offspring production.

Table 6.   The four linear coefficients and matrix of four quadratic and six correlational nonlinear coefficients for offspring production.
 Linear (β)Nonlinear (γ)
RelatednessFemale wing areaMale wing areaSex comb
  1. ± standard errors.

  2. Significant values are highlighted in boldface, *< 0.05.

Relatedness−0.331 ± 0.138*0.236 ± 0.264−0.368 ± 0.168*0.059 ± 0.1870.499 ± 0.193*
Female wing area0.328 ± 0.153*−0.134 ± 0.25−0.048 ± 0.249−0.101 ± 0.215
Male wing area0.059 ± 0.175−0.478 ± 0.320−0.034 ± 0.224
Sex comb−0.142 ± 0.1640.486 ± 0.334
Table 7.   Eigenanalysis for nonlinear effects of predictor variables on offspring production.
 m1m2m3m4
  1. Significant values are highlighted in boldface, *< 0.05, **< 0.01.

Eigenvalue0.9650.046−0.408−0.492
Double regression P-value< 0.001**0.8000.8720.630
Permutation P-value0.050*0.8360.024*0.426
Canonical weights
 Relatedness0.635−0.451−0.607−0.157
 Female wing area−0.2800.606−0.7450.002
 Male wing area0.018−0.148−0.1250.981
 Sex comb0.7200.6380.2490.115
Figure 2.

 Relative offspring production along the two significant multivariate axes, m1 and m3. Blue areas of the graph represent low offspring production, red areas represent high offspring production. The black dots indicate individual data points.

Correlations between male and female traits

There was no significant correlation between male and female traits within mating pairs for either wing size (r = 0.167, t85 = 1.558, = 0.123) or between female wing size and male sex comb tooth number (R2 = 0.032, t88 = 0.296, = 0.768).

Discussion

Our results provide clear evidence that individuals are not mating at random with respect to relatedness in this population. Both the multivariate and univariate analyses indicate that males were more likely to be mating with a related female. Increased relatedness within pairs is consistent with the hypothesis of optimal inbreeding, under which individuals are expected to often choose related mates in order to maximize their inclusive fitness. However, further analysis suggests that elevated relatedness between all individuals collected mating (not just between actual pairs) may be responsible for the observed higher relatedness within copulating pairs, without involving any direct mate preference based on the relatedness. The genetic analysis indicates that the higher relatedness between all mating individuals was unlikely to be the result of sampling across reproductively isolated populations or species. Instead, the data suggest that the mating individuals may have represented a subset of the total population. One possible scenario that could lead to this result would be if mating individuals were more likely to be of a particular age, and that individuals of a similar age are more likely to be related. Mating ability and courtship gene expression have been found to decline with age in laboratory experiments on D. melanogaster (Mack et al., 2000; Ruedi & Hughes, 2009). Such an effect may result in younger males being more frequently collected in our mating sample. However, for such an effect to produce the results we detected, females would also need to be of a similar age. Studies on the effect of female age on remating rate in D. melanogaster find no effect of age on female remating, other than very young females having lower rates of remating than older females (Fuerst et al., 1973; Vanvianen & Bijlsma, 1993), a factor that would not be expected to lead to increased relatedness between males and females.

Whatever the cause of the increased relatedness within pairs, we detected no sign that mating between close relatives was avoided in this population. Our analysis of offspring production also indicates that the high level of relatedness observed within pairs is likely to come at a direct cost to the production of offspring. There are several factors that could explain reduced offspring production between closely related pairs. It is possible that individuals invested less in offspring production when mating with a close relative, a form of cryptic choice for inbreeding avoidance. Another possibility is that offspring produced by more closely related pairs suffered lower egg-to-adult survival due to inbreeding depression. In a previous laboratory study investigating similar levels of inbreeding in pedigrees derived from this population we found significant linear inbreeding depression for egg-to-adult viability (Robinson et al., 2009), making the later scenario probable. Moderate levels of inbreeding have also been found to reduce both female fecundity (Latter & Robinson, 1962) and male mating success (Sharp, 1984). Given these costs, we would have expected inbreeding avoidance through kin recognition to have evolved.

A possible explanation for the lack of inbreeding avoidance in this population is that the inclusive fitness benefits from inbreeding may have lead to inbreeding tolerance despite the costs from inbreeding depression, as predicted by recent theoretical models (Kokko & Ots (2006). It is therefore possible that the level of inbreeding in this population does not represent a net cost to fitness. It has also been suggested that there may be direct genetic benefits for the offspring produced by intermediate levels of inbreeding due to a balance between factors such as inbreeding depression and outbreeding depression (Bateson, 1983; Shields, 1983). However, a previous laboratory experiment on this same population of D. melanogaster found no evidence of a benefit to intermediate levels of inbreeding and instead found significant linear inbreeding depression for egg-to-adult viability (Robinson et al., 2009), making direct genetic benefits unlikely. Another factor that may be at play is post-copulatory inbreeding avoidance, which may act to reduce the cost of mating with a close relative (Tregenza & Wedell, 2000). There have been mixed results from studies on post-copulatory inbreeding avoidance in D. melanogaster. Using pedigrees derived from a natural population, Mack et al. (2002) found that sperm competitive ability was lower when males mated with related females. As multiple mating is common in this species (Jones & Clark, 2003), this may reduce the costs of inbreeding depression as a greater number of offspring may be sired by more distantly related mates. A recent study on a laboratory strain of D. melanogaster found no evidence of post-copulatory inbreeding avoidance via sperm competitive ability or mating latency, copulation duration, egg laying rate or remating interval (Ala-Honkola et al., 2011). However, it is unclear how long-term laboratory adaptation might impact the operation of sexual selection. It may be, for example, that the loss from laboratory strains of deleterious alleles that underlie inbreeding depression results in relaxed selection via mate choice.

In contrast to previous work on D. melanogaster, we found no linear effects of wing size or sex comb tooth number on mating success. Larger size in D. melanogaster males has been found to be associated with increased mating success in several laboratory studies (Partridge et al., 1987a; Wilkinson, 1987; Pitnick, 1991), and under field conditions (Partridge et al., 1987b; Taylor & Kekic, 1988). Although the finding of no effect of wing size in our study may reflect a lack of statistical power, there are also differences between our study and previous field studies that could account for the different findings. Taylor & Kekic (1988) sampled at hourly intervals throughout the day, whereas we restricted our collections to the period of the day where the majority of mating activity occurs. Sexual selection may be stronger at times outside the peak mating activity, leading to the different results. Further, Markow (1988) found that although larger males were more likely to be present at breeding sites, within the breeding sites there was no effect of body size on mating success, suggesting that smaller individuals may be excluded from the mating sites. It is therefore possible that larger individuals in our population of D. melanogaster did indeed have higher mating success, but that the contrasting smaller individuals were not present at the breeding site and were therefore not represented in our random sample. Partridge et al. (1987b) investigated laboratory-reared individuals of the same age marked and released into the field. Smaller individuals may have also been absent from the mating site in their study while still allowing sexual selection to be detected.

In conclusion, we detected evidence of elevated relatedness between mating individuals in this natural population, with no evidence of precopulatory inbreeding avoidance despite likely costs to offspring fitness. There are a number of factors including mate preference based on relatedness and/or some form of cohort effect that could potentially explain these results. In order to investigate the role of mate preferences in generating the nonrandom patterns of mating we have observed in this field population, an experimental approach is needed in which factors such as age differences among individuals can be tightly controlled.

Acknowledgments

This study was funded by the Australian Research Council and the University of Western Australia. We would like to thank Howard Park and Houghton’s wineries for allowing the collection of D. melanogaster from their properties.

Ancillary