Dr Jacek Radwan, Institute of Environmental Sciences, Jagiellonian University, ul. Ingardena 6, 30-060 Cracow, Poland. Tel.: +48 12 633 63 77; fax: +48 12 634 19 78; e-mail: email@example.com
Under the ‘good genes’ mechanism of sexual selection (SS), females benefit from mate choice indirectly: their offspring inherit genes of the preferred, high quality fathers. Recent models assume that the genetic variance for male quality is maintained by deleterious mutations. Consequently, SS can be predicted to remove deleterious mutations from populations. We tested this prediction by relaxing selection in populations of the bulb mite, thus increasing their rate of accumulation of deleterious mutation. SS, allowed to operate in half of these populations, did not prevent the fitness decline observed in the other half of the relaxed selection lines. After 11 generations of relaxed selection, female fecundity in lines in which males were allowed to compete for females declined compared with control populations by similar amount as in monogamous lines (17.5 and 14.5%, respectively), whereas other fitness components (viability, longevity, male reproductive success) did not differ significantly between both types of lines and control populations.
Benefits of mate choice by females of species in which males provide nothing but gametes are thought to arise indirectly, via offspring inheriting well adapted genes from their fathers who signal their genetic quality with elaborated sexually selected traits (reviewed in Andersson, 1994). This requires that substantial genetic variation exists in populations. The variation is currently thought to be maintained mainly by deleterious mutations (Charlesworth & Hughes, 1999; Lynch et al., 1999). Such mutations should affect the condition of individuals, and those in the best condition should win in the competition for mates, either via direct contest or because they can afford more elaborated traits attractive to the opposite sex (Houle & Kondrashov, 2002). On this assumption, it has been also hypothesized that by purging populations of deleterious mutations, sexual selection (SS) can contribute to the maintenance of sexual reproduction (Agrawal, 2001; Siller, 2001). Although several studies have demonstrated increased fitness of progeny as a result of mate choice (e.g. Partridge, 1980; Moore, 1994; Petrie, 1994; Promislow et al., 1998), these effects tend to be rather weak (Møller & Alatalo, 1999), and their genetic basis unknown. Recently, it has been demonstrated that visible mutations that decrease female productivity also tend to decrease male mating success in Drosophila melanogaster (Whitlock & Bourguet, 2000). However, the mutations studied had much stronger effects than average mutations appearing in populations (Whitlock & Bourguet, 2000). Our goal was to study the effectiveness of SS in removing mutations occurring in bulb mite populations at the natural rate.
The good genes process of SS is usually discussed in the context of precopulatory mate choice, but cryptic choice, i.e. some form of paternity manipulation occurring after copulation (Eberhard, 1996), can have the same function. Similarly, male success in competition with other males, either directly or through sperm competition (Parker, 1970), can be expected to correlate with male genetic quality, so that females mating with the winners can increase fitness of their progeny (Cox & Le Boeuf, 1977; Berglund et al., 1996; Yasui, 1997). SS in the bulb mite, Rhizoglyphus robini (Acari: Astigmata: Acaridae) involves contests over access to females and competition between sperm from different males over access to ova (Radwan, 1997; Radwan & Klimas, 2001). Males of this species are dimorphic, with ‘fighter’ males that possess a thick, sharp, third pair of legs used to stab other males, and benign ‘scrambler’ males that have unmodified legs. Fighters achieve higher reproductive success than scramblers in mixed populations (Radwan & Klimas, 2001). Given significant heritability of male morph (Radwan, 1995, 2003b), this suggests that they may be of better genetic quality. Indeed, in another male-dimorphic acarid mite, Sancassania berlesei, fighter males develop from nymphs in better condition than those giving rise to scramblers (Radwan et al., 2002). Although precopulatory discrimination between potential mates cannot be excluded, females are highly promiscuous (Radwan & Siva-Jothy, 1996), making mate choice more likely post-insemination. Konior et al. (2001) have found that females mated to several males produce daughters of higher fecundity than monogamous females, which suggest that polyandry yields genetic benefits.
Relaxation of natural selection via enforcing random, monogamous mating and equal contribution of each parent to the next generation leads to accumulation of deleterious mutations and a consequent decline in fitness, which may reach 2% per generation (reviewed in Burt, 1995; Lynch et al., 1999). Indeed, the possibility of a similar decline may be a reason for health concern in modern human societies, where natural selection is much relaxed (Crow, 1997). If variation in traits associated with male success in competition over access to females and their gametes reflects variation in mutation load, possibly mediated via male condition, then SS should effectively remove deleterious mutations from populations (Agrawal, 2001; Siller, 2001). We tested this prediction by allowing mutations to accumulate in three replicate lines with relaxed natural selection under the ‘middle-class-neighbourhood’ (MCN) design (random, monogamous mating, each pair contributing a pair of progeny to the next generation; Shabalina et al., 1997). We then compared the consequent decline in fitness to that observed in another three ‘sexual selection’ lines that were treated identically except for allowing competition between males over access to females and over fertilization of their eggs. Recent models have shown that by removing deleterious mutations from populations, SS can fully compensate for the two-fold cost of sex (Agrawal, 2001; Siller, 2001). As asexual populations, by sparing costs of producing males, can in theory grow twice as fast as sexual populations (Maynard Smith, 1978), full compensation requires that at least a half of the total selection against deleterious mutations occurs via SS (Agrawal, 2001; Siller, 2001). If this was fulfilled in our study populations, than SS should have slowed down accumulation of deleterious mutations in SS lines by at least a half compared with MCN lines, where SS was absent.
Maintenance of experimental populations and controls
Bulb mites are diploid, sexually reproducing pests of subterranean structures of plants and foods in storage (reviewed in Diaz et al., 2000). It occurs in colonies of variable size, which may include a few to hundreds of adults (Ascerno et al., 1983; Nakao, 1991, personal observations). Sex ratio at emergence of adults is near unity (Gerson et al., 1983; personal observations). The mites used in this study came from a stock culture derived from a colony of about 200 individuals found on onions in a garden near Cracow, Poland, in 1998 and kept in the laboratory as a large population (>1000 individuals, subdivided into six subpopulations mixed once a month) for about 100 generations before commencement of this research. The subpopulations were kept in 2.5 cm diameter, 2 cm high jars, maintained at 22–26 °C, >90% humidity, and fed once a week, a 3 : 1 mixture of powdered yeast and wheat germ ad libitum. Once a month about 1/4 of the food and debris containing several hundred mites at different stages of development were transferred to fresh jars.
The same feeding, humidity and temperature conditions were also maintained throughout all the experiments described below. Individually isolated mites, pairs and small groups of mites were kept in 0.8 cm diameter glass tubes (2 cm height) with Plaster of Paris bases soaked with water, and were provided with food ad libitum.
In, 2001, MCN and SS lines were started in three replicates each. In MCN lines each of the 60 females was randomly assigned an unrelated male and placed in 0.8 diameter cell for 7 days. Each SS line also comprised 60 males and 60 females, but they interacted in groups consisting of five females and five males, each group placed in a 0.8 cm diameter cell for 7 days. Under similar conditions, Radwan & Klimas (2001) observed high fight-related mortality, reaching 50% within 12 days, with a single fighter eventually monopolizing females in 15% of cases. Following 7 days of interactions, females from both MCN and SS lines were placed individually in separate tubes to oviposit. After their progeny reached maturity (about 15 days), one progeny of each sex from each female was used to start the next generation. This procedure was carried out for 11 generations, with simultaneous maintenance of three unmanipulated control populations (C). Each control consisted of several hundred mites kept in a 2.5 cm jar under the same conditions as the stock culture (fed once a week, transferred to a fresh jar once a month), except that they were not subdivided into periodically mixed subpopulations. Another control population (Ccold) was maintained in refrigerator at 10 °C, which by prolonging generation time to about 3 months, should reduce the possibility of its evolution.
During two generations before fitness assays (i.e. generations 12–13), the treatment of both SS lines and all controls was changed to identical to that of MCN lines, i.e. females were mated to a single, randomly selected male and contributed a pair of progeny to the next generation. The purpose of this change was to reduce effects of factors other than deleterious mutations, on fitness components measured subsequently. These factors include maternal effects, which can affect fitness of progeny, e.g. in response to the number of males a female mates with (Kozielska et al., in press), but possibly also in response to other factors that differed between the treatments, such as presence or absence of other females. Additionally, it allowed to reduce linkage disequilibria that could have arisen because of selection for beneficial gene combinations in SS and C lines. For unlinked loci, random mating and relaxed selection cause linkage disequilibria to decay by a half each generation (Lynch & Walsh, 1998).
Fitness essays started from isolation of 10 eggs to 0.8 diameter cells from each of 30 to 50 females from each line of each treatment, and their survival to adulthood was recorded. One female and one male was then randomly chosen from each family and used for further essays. Females were each assigned a 2–4-day-old male randomly selected from the base population, and the number of eggs they laid was recorded for 2 weeks, which is representative of their lifetime oviposition lasting on average about 3 weeks (Konior et al., 2001). Any missing males were replaced. Female survival was also recorded. Male reproductive success was measured in competition with one rival male over two females for 5 days. A rival male aged 2–4 days was randomly selected from the base population and than irradiated with 20 krad from Co60 source. This dose prevents hatching of all eggs fertilized by irradiated males, but does not significantly affect their sperm-competitive abilities (Radwan, 1997). As almost all eggs normally hatch (Radwan & Siva-Jothy, 1996), male reproductive success was measured as the number of hatched eggs laid during the 5 days of the test and 3 days following the test.
Variance in male fitness in sexual selection lines
An estimate of variance in male reproductive success in the SS treatment was obtained from an independent experiment performed on mites from the source population. To mimic the situation in SS lines, one focal male was placed in a cell with four sterilized males (sterilization procedure as above) and five females for 7 days. Two eggs were then collected from each female, and the reproductive success of the focal male was measured as the proportion of eggs hatched.
No significant differences between C and Ccold reference lines were found in any of the fitness components assayed (Table 1), so there was no indication of evolution in any of them. In subsequent analyses, Ccold was thus included as a fourth control to increase the statistical power of comparisons with MCN and SS lines, but it can be noted that excluding it did not change any of our conclusions.
Table 1. Mean values and standard deviations (in brackets) for each fitness component measured.
MCN, middle-class-neighbourhood lines; SS, sexual selection lines; C, unmanipulated controls; Cc, cold control. The differences between controls were tested either with anova (F reported) or Kruskal–Wallis (H) when normality assumptions were not met. anova testing for fitness differences between treatments was performed on population mean values (see e.g. Holland & Rice, 1999 for justification); normality of the distribution of the mean values can be assumed based on the central limit theorem (Mendenhall, 1983). Bartlett's test did not reveal significant departures from homogeneity of variances for any fitness component analysed. As there were four components measured, the Bonferroni-corrected significance level was taken as 0.05/4 = 0.0125.
H3 = 3.96, n = 161
H3 = 0.75, n = 136
F3 = 0.89, n = 139
H3 = 2.41, n = 100
The only significant difference between the three treatments was in female fecundity (Table 1), which was the highest in the control treatment. As Ccold showed the highest fecundity of all controls (Fig. 1), we carried out a more conservative test including only unmanipulated controls, but the conclusion did not change (F2,6 = 11.16, P = 0.009). Post hoc Neumann–Keuls test showed that the significant differences were between MCN and C (P = 0.01) and between SS and C (P = 0.01), but not between MCN and SS (ns; Fig. 1). This shows that the fitness of R. robini females declines under the MCN design, but SS does not seem to counteract the decline.
We did not find any change in the fighter morph proportions between our lines (mean values: MCN, 0.906; SS, 0.914; C, 0.922; anova, F2,7 = 0.047, ns). Thus, there was no indication that relaxed natural selection affects genes determining male morphs.
The results of this study showed that when natural selection is relaxed through elimination of variance in male and female reproductive success, fitness of bulb mite populations declined. That this decline was not an artefact of evolution in control lines can be argued on two grounds. First, these lines were maintained under the same conditions to which the population had adapted in the laboratory for about 100 generations before the beginning of this study. Secondly, we found no evidence that they diverged from cold control (Ccold). Although it had passed through only three to four generations, Ccold could nevertheless evolve, especially if selection caused by the new temperature was strong. However, it is unlikely that Ccold and unmanipulated controls evolved in the same direction at similar rate, so the lack of significant difference between them indicates that neither evolved at a substantial rate.
In the bulb mite, fighters were shown to achieve higher reproductive success than scramblers in populations containing both morphs (Radwan & Klimas, 2001). Moreover, reproductive success of morphs was not frequency-dependent (Radwan & Klimas, 2001), so the reasons for the maintenance of genetic variation for male morph (Radwan, 1995, 2003b) remain to be explained. One possibility is that genetic variance for male condition is maintained by mutation-selection balance, with scrambler morph being expressed by males in poor condition (Gross & Repka, 1998; Radwan & Klimas, 2001; see also Radwan et al., 2002). This hypothesis predicts a decline in proportion of fighters in mutation-accumulation MCN lines compared with controls, but this prediction was not supported by our results. Thus, there was no evidence for better genetic quality of males expressing the fighter morph. However, this does not exclude the possibility that among fighters, those bearing fewer deleterious mutations are more successful competitors. Males with lower mutation loads could also be favoured by females at pre- or post-copulatory stage, and their sperm might be more competitive. If male success in these elements of sexual competition decreases with the load of deleterious mutations, then SS can be expected to decrease the mutation load of populations (Agrawal, 2001; Siller, 2001; Houle & Kondrashov, 2002). However, we have found no evidence in support of this expectation: retention of the variance in male reproductive success in SS lines did not prevent the decline in female fecundity, which was not significantly different from that observed in MCN lines. Similarly, in D. melanogaster SS has been recently shown to fail to promote the spread of alleles favourable in a new environment (Holland, 2002).
For SS to fully compensate for the two-fold cost of sex, it should account for at least a half of the total selection against deleterious mutations (Agrawal, 2001; Siller, 2001). To meet this requirement, the difference in female fecundity between our MCN and SS lines should be higher than a half of the difference between MCN lines (mean = 112.66) and controls (mean = 131.20, excluding Ccold which had the highest mean). This difference should thus be at least 9.54, whereas the observed difference was between MCN and SS lines was −4.01. As the 95% confidence interval on this difference spanned from −13.08 to 5.07, our results indicate that the effectiveness of SS in the bulb mite is lower than would be required to fully compensate for the two-fold cost of sex. We are currently exploring induced mutagenesis as a method to test for even smaller effects.
A decrease in fecundity of the SS females could result, in part, from changes in frequencies of genes involved in intersexual ontogenic conflict (Chippindale et al., 2001; Rice & Chippindale, 2001). Female evolution in these lines was minimized by enforcing equal contribution of their offspring to the next generation, so that any alleles that decrease female fitness, but simultaneously increase male reproductive success could sharply rise in frequency. However, this argument is undermined by the lack of significant differences between male reproductive success in SS lines and that of MCN or C males, so accumulation of deleterious mutations seems the most likely reason for the decline in fitness under relaxed natural selection.
Performance of relaxed selection lines compared with controls could also be affected by inbreeding. Because there was no variance in family size in MCN lines, the effective population size was approximately Ne = 2N = 240, giving 1/2Ne = 0.0021 increase in inbreeding per generation. In SS males, each female contributed a pair of progeny to the next generation, but there was variance in male reproductive success. This variance was estimated at 5.3 (n = 21), so Ne = 8N/(5.3 + 4) = 103.2 (Falconer, 1989). Thus, after 11 generations, the inbreeding coefficient was 0.023 in MCN lines and 0.053 in SS lines. With inbreeding depression for fecundity of 0.87 (Radwan, 2003a), this should result in fertility decline of 2% for MCN lines and 4.6% for SS lines, i.e. much less than the observed 14.4 and 17.5% (with Ccold conservatively excluded). Inbreeding depression in SS lines could in fact be smaller than 4.6%, because purging of inbreeding depression might have occurred (e.g. Fowler & Whitlock, 1999). Another argument against the role of inbreeding in causing the decline in fecundity of MCN and SS females is provided by the lack of a simultaneous decline in male reproductive fitness. Such a decline could be expected if inbreeding was significant, because at least one of the components of male fitness, sperm competitiveness, shows a very high inbreeding depression (52% at inbreeding coefficient F = 0.25; M. Konior and J. Radwan, unpublished data, compared with 22% for fecundity, Radwan, 2003a).
Fitness decline in MCN and SS lines could also be partly due to breaking down epistatic combinations of alleles that could be maintained in linkage disequilibrium by selection acting in controls (Lynch et al., 1999). However, before fitness was assayed, controls as well as SS and MCN lines, underwent two generations of random, monogamous mating under relaxed selection, which should decrease linkage disequilibria by 75% (Lynch & Walsh, 1998), so this effect must have been small compared with the effect of deleterious mutations. Moreover, if epistatic interactions are important in determining individual quality, and if sexually selected traits reflect this quality as assumed by the indicator mechanism (Kotiaho et al., 2001 ; Andersson, 1994), SS should also maintain favourable epistatic combinations of genes. Thus, even if the decline in fecundity in MCN and SS lines was, in addition to accumulation of deleterious mutations, as a result of breaking down epistatic interactions, the lack of significant difference between them indicates that SS was not effective in maintaining favourable linkage disequilibria.
The reasons why fitness decline was significant in only one of the four components measured, that is female fecundity, are not obvious. Not all traits, however, seem to accumulate mutations at the same rate (see e.g. Shabalina et al., 1997; Lynch et al., 1999 and references therein), and those controlled by higher numbers of loci constitute larger ‘targets’ for deleterious mutations (Houle, 1998). Although direct estimates are not available, fecundity may indeed be expected to constitute a large mutational target, as it is likely to be controlled not only by a large number of genes involved in oogenesis, but also by all genes involved in accruing and allocating resources for egg production, including genes that affect development (Houle, 1998). In the bulb mite, probability of extinction of inbred lines was found to be negatively related to fecundity of the females from whom the lines originated, which suggests the mutational nature of a substantial part of genetic variation for this trait (Radwan, 2003a). If other traits measured did not decline under relaxed selection because they are relatively smaller mutational targets, then SS could not be expected to effectively compensate for the cost of sexual reproduction. This is because for such compensation to work, deleterious mutations that are removed through differential male reproductive success must also substantially affect female fitness, so that in equilibrium sexual females, bearing fewer mutations, could outcompete asexual females (Agrawal, 2001; Siller, 2001). If the number of genes affecting male reproductive success is small compared with that affecting female fitness, as the results of this study may suggest, or if the deleterious effects of mutations have sex-limited expression, SS cannot effectively remove mutations that are detrimental to females.
Male success in intrasexual contests has often been argued to be an indication of high genetic quality (reviewed in Berglund et al., 1996). However, recent evidence shows that this may not necessarily be the case and female mating preferences may serve to reverse reproductive hierarchy based on male dominance relationships (Moore & Moore, 1999). Hence, future research should attempt to discriminate between the effects of intra- and inter-SS to see if the former obscures the adaptive effects of the latter.
We thank A. Łomnicki and M. Jasieński for inspiration and discussions, J. Tomkins, N. LeBas and M. Kozielska and anonymous referees for their comments on previous versions of the manuscript. This project was supported by the State Committee for Scientific Research KBN 0408/P04/2001/21.