The developmental plasticity and functional significance of an additional sperm storage compartment in female yellow dung flies

Authors


Summary

  1. The mechanistic basis for and adaptive significance of variation in female sperm storage organs are important for a range of questions concerning sexual selection and speciation, as such variation influences the evolutionary trajectories of male fertilization related traits and may facilitate speciation through its effects on gamete recognition.
  2. Female yellow dung flies (Scathophaga stercoraria) usually develop three sperm storage compartments, and this subdivision may be an adaptation for sorting sperm during postcopulatory choice.
  3. Using lines artificially selected to express four spermathecae (4s), we explored the fitness consequences of the novel phenotype relative to the naturally prevalent three-spermatheca (3s) phenotype by manipulating the opportunity for postcopulatory sexual selection (females mated either with three or only one male prior to oviposition). In addition, we examined the developmental plasticity of spermathecal number in response to different larval food environments and estimated its genetic correlation with growth rate.
  4. Mating treatments with and without the opportunity for postcopulatory sexual selection revealed no significant fitness differences between alternative spermathecal phenotypes within selection lines despite overall benefits associated with multiple mating, and moderate egg-to-adult survival costs in response to artificial selection for 4s. Manipulations of the larval food environment revealed that the expression of 4s is highly plastic and tightly linked to environmental conditions promoting fast somatic growth and development. Likewise, siblings with fast intrinsic (genetic) growth were more likely to express 4s within and across food environments.
  5. The present results highlight a great potential for rapid evolutionary change in female sperm storage morphology through indirect selection on life-history traits, and further suggest genetic assimilation as a potential mechanism facilitating phylogenetic transitions in spermatheca number as frequently observed within the Dipterans.

Introduction

Traditionally, divergence in traits related to insemination and fertilization was considered to be caused by pleiotropy, or reinforcement following secondary contact (Dobzhansky 1937; Mayr 1963; Panhuis et al. 2001). More recently, it has become evident that these traits can be shaped by sexual selection (reviewed in Eberhard 1996; Birkhead & Møller 1998; Simmons 2001; Pitnick, Wolfner & Suarez 2009), which may lead to rapid trait divergence and even facilitate speciation by directly affecting mate or gamete recognition and competition (Panhuis et al. 2001; Simmons 2005; Howard et al. 2009; but see Morrow, Pitcher & Arnqvist 2003 for a comparative study in birds). In internally fertilizing species, the arena for sperm competition and choice is the female reproductive tract, which has been a focus of much recent comparative and experimental research on the processes of sperm storage and usage by the female. The emerging data not only reveal an astonishing diversity in female reproductive anatomy, reflected by frequent evolutionary gain, loss or functional reshuffling of species-specific sperm storage sites (e.g. Baminger & Haase 1999; Beese et al. 2009; Puniamoorthy, Kotrba & Meier 2010), but also that this diversity influences the evolutionary trajectories of male fertilization-related traits (e.g. Koene & Schulenburg 2005; Rönn, Katvala & Arnqvist 2007; Kuntner, Coddington & Schneider 2009; Higginson et al. 2012).

Several hypotheses have been proposed to explain divergence in female reproductive morphology. Sperm need to survive long enough to ensure fertilization (because in most taxa insemination and fertilization are temporally separated), so natural selection promoting sperm longevity and/or supply almost certainly will have shaped the female genital tract and associated sperm stores (Pitnick, Markov & Spicer 1999; Twig & Yuval 2005; Nakahara & Tsubaki 2007; Pitnick, Wolfner & Suarez 2009). Postcopulatory sexual selection through cryptic female choice or sexual conflict over paternity may play an important additional role in the evolutionary diversification of female reproductive anatomy. While experimental data relating variation in paternity outcomes to morphological variation of the female reproductive tract support this hypothesis (e.g. Miller & Pitnick 2002; Amitin & Pitnick 2007; Simmons & Kotiaho 2007), studies examining covariation between female reproductive anatomy and female fitness remain scarce. Thus, the relative importance of cryptic female choice and sexually antagonistic co-evolution for creating variation in female reproductive anatomy requires further investigation (Reinhardt et al. 2007; Pitnick, Wolfner & Suarez 2009; Birkhead 2010).

Alternatively, certain aspects of the female reproductive tract may evolve neutrally. Population genetic theory predicts that traits with sex-specific expression accumulate more deleterious mutations under selection-mutation balance and hence should show high genetic divergence and low phenotypic robustness compared to traits that are expressed in both sexes and consequently exposed to selection twice as often (De Visser et al. 2003; Mank & Ellegren 2009; Van Dyken & Wade 2010). Even if selectively neutral, morphological variation in the female reproductive tract can alter the evolutionary trajectories of male fertilization-related traits, as males are expected to adapt to physical conditions set by the sperm competitive environment inside the female (Clark, Begun & Prout 1999; Bjork et al. 2007; Tregenza, Attia & Buahaiba 2009). Thus, the interplay between selective and underlying developmental mechanisms, an important generator of morphological diversity in general (West-Eberhard 2003), may play a pertinent role in processes of divergence in reproductive traits in particular.

Intraspecific polymorphisms in the number of female sperm storage compartments (spermathecae) have been described in several insect species including acalyptrate flies, mosquitoes and earwigs (Sturtevant 1925; Jones & Ludlam 1965; Kamimura 2004). Such polymorphisms are interesting traits for studying the evolutionary and developmental processes underlying macro-evolutionary change in female reproductive anatomy. For example, within the Diptera, the Scatopsidae usually possess only one (Haenni 1997), the Drosophilidae two (Bächli 1998), the Scathophagidae three (McLean 1998) and the Chamaemyiidae four spermathecae (De Jong 2000), and reversible gains or losses appear to be common in several taxa (Pitnick, Markov & Spicer 1999; Presgraves, Baker & Wilkinson 1999; De Jong 2000).

In one of the best-studied species in the context of sexual selection, the yellow dung fly Scathophaga stercoraria (Diptera: Scathophagidae), females typically possess three spermathecae (hereafter denoted as 3s), but sometimes four spermathecae (4s) or even more (Simmons, Parker & Stockley 1999; Ward 2000; Ward, Wilson & Reim 2008). There are also intermediate phenotypes with bifurcated spermathecae and shared spermathecal ducts to the bursa copulatrix (where fertilization occurs) that illustrate the developmental transition between these alternative reproductive phenotypes (Ward, Wilson & Reim 2008). Sperm competition experiments have revealed reduced second male sperm precedence in 4s compared to 3s females (Ward 2000), in agreement with theoretical expectations that multiple spermathecae can be advantageous during postcopulatory sexual selection, if females are able to nonrandomly store and use sperm from different males (Hellriegel & Ward 1998; Snow & Andrade 2005; Ward 2007). Differential sperm storage and usage has been repeatedly demonstrated in S. stercoraria using phenotypic or molecular markers (Otronen, Reguera & Ward 1997; Hellriegel & Bernasconi 2000; Bussière et al. 2010; Demont, Martin & Bussière 2012). Since artificial selection for expressing 4s induced fecundity costs in female S. stercoraria, potential mate choice benefits could trade-off against costs that females incur from producing an additional sperm storage compartment (Ward 2007; Ward, Wilson & Reim 2008). Assessment of such costs is important because they can alter the strength and direction of selection acting on the focal and correlated traits within a given population (Jennions & Petrie 1997; Hunt, Brooks & Jennions 2005; Bussière et al. 2008).

The main objective of the present study was twofold. Recent population genetic data documented latitudinal variation for the 4s phenotype and striking phenotypic plasticity in spermatheca number in response to developmental temperature (Berger et al. 2011). These findings suggest that high metabolism or fast larval growth and development, which all systematically vary with latitude and temperature, might be important in driving these patterns. We therefore first aimed to better understand the nature of phenotypic variance in spermatheca number in S. stercoraria and its correlation with larval growth rates. We used lines of flies artificially selected for the 4s phenotype, which is rarely expressed in natural populations, primarily to generate sufficient 4s flies for our experiments. We manipulated larval resource acquisition in two different ways relative to the control diet: we restricted larval food uptake by diluting cow dung (the natural food source) with agar to impede food uptake, and we treated some of the larval food with the parasiticide ivermectin to reduce muscular efficiency during larval development (Römbke et al. 2009). Both treatments ultimately retard growth.

Secondly, the prevalence of phenotypic variants should be determined by their net selective advantage. Our other aim was therefore to obtain information on the relative fitness of both spermatheca phenotypes in a sexual selection framework. We addressed this question by providing females with the opportunity for postcopulatory sexual selection (when mated with three different males) or not (when mated with only one male) before oviposition. Complementary to earlier work with similar artificial selection lines that focused on paternity outcomes (Ward 2000; Ward, Wilson & Reim 2008), we assessed several components of offspring performance (egg-to-adult survival, development time, growth rate and size at metamorphosis) as a proxy for female fitness. We expected that, if having four spermathecae is beneficial for sperm storage and/or choosing high quality male sperm, this should positively affect offspring survival and performance when females have an opportunity to sort the sperm of rival males.

Materials and methods

Experimental Flies

We used five replicate artificial selection lines, of which three were up-selected for four spermathecae (4s lines). The remaining two lines were down-selected for three spermathecae (3s lines), serving as a control against which to evaluate correlated genetic responses to the 4s genetic background. The detailed selection procedure is described in Thüler (2009) and largely corresponded to that of Ward, Wilson & Reim (2008), so we provide only a brief summary here. Parental flies, originating from Fehraltorf near Zurich, Switzerland, were collected in 2006. Independent lines were established after the first generation of selection. The Zurich population contains about 15% females with 4s when flies are raised at 18 °C under unlimited dung supply (for other rearing temperatures see Berger et al. 2011), whereas the 4s frequency expressed in the field is below 2%. The seventh generation of selection was used for our experiments. This generation featured intermediate frequencies of the 4s phenotype in the up-selected lines (Fig. 1), which was optimal for our study in terms of statistical power to test for fitness differences between the two alternative spermatheca phenotypes and for exploring the nature of the phenotypic plasticity previously observed (Berger et al. 2011). Within the two 3s lines the frequency of females expressing 4s was very low (Fig. 1).

Figure 1.

Effects of food treatment on the mean (± 95% CI) percentage of females with four spermathecae (4s). Estimates are based on family means involving N = 160 families for the 4s and N = 57 families for the 3s selection regime, respectively.

Mating Experiments

In order to manipulate the diversity and quantity of sperm in storage, and thus the opportunity for postcopulatory sexual selection to occur, females from all selection lines were mated with either three different males or only a single male prior to oviposition. Because the focal 4s phenotype was largely absent in the lines selected for 3s (Fig. 1), sample sizes were chosen to be larger in the 4s lines (N = 170) than in the 3s lines (triple mated females: N = 57). Males and females were always mated within the same selection regime. If pairs did not copulate within 1 h, the male was replaced. All copulations took place in the absence of dung to avoid premature oviposition. One hour after the last mating (allowing sperm to move into storage: Sbilordo, Schäfer & Ward 2009; Bussière et al. 2010), females were provided with a smear of fresh dung on filter paper for oviposition. Before our experiments, all adult flies were maintained for 2–4 weeks postemergence on a diet of Drosophila, sugar and water at 15 °C, a light/dark cycle of 16:8 h, and 60% humidity, which are ideal conditions for adults (Blanckenhorn et al. 2010).

Common Garden Rearing of Offspring

Offspring were raised in common garden conditions in which we manipulated food acquisition in two different ways relative to the control with unmanipulated dung. The agar treatment was designed to reduce the efficiency of larval food uptake (equivalent to food limitation) and consisted of homogenized cow dung (stored at −80 °C prior to the experiment) diluted by adding an equal amount of a 2% agar solution. The ivermectin treatment consisted of cow dung with a concentration of 6·57 µg kg−1 (wet weight) ivermectin, a pharmaceutical used to treat livestock for arthropod and nematode parasites. Ivermectin disturbs the ion transport through the cell wall leading to a reduced muscular efficiency, ultimately retarding growth (Römbke et al. 2009). These treatments have qualitatively different effects on larval growth (see 'Results'). All flies were reared in a climate chamber at 15 °C with a light/dark cycle of 16L:8D and 60% humidity in transparent plastic vessels (4 × 4 × 4 cm), each filled with 30 g of the respective dung suspensions. We split individual clutches into batches of c. 14 eggs and randomly distributed them among the vessels with the three dung treatments. In the yellow dung fly, 2 g of fresh dung per larva is sufficient to prevent competition between larvae (Amano 1983). Any remaining eggs of the clutches, which typically consist of about 60 eggs, were discarded.

Measurements of Traits

As standard measures of offspring performance in this (Blanckenhorn et al. 2010) and other species, we scored egg-to-adult survival (number of emerging flies divided by the number eggs placed into the vials), body size (hind-tibia length), juvenile development time and juvenile growth rate (hind tibia length divided by development time). We checked daily for adult fly eclosion to obtain egg-to-adult development times. Spermatheca phenotypes of mothers and daughters were scored by dissecting flies using standard binocular microscopes as described in Ward, Wilson & Reim (2008). Individuals with a bifurcated, not fully divided third spermatheca were classified as 4s phenotypes (see Fig. 1 in Ward, Wilson & Reim 2008). Categorizing these intermediate phenotypes as 3s did not change the results qualitatively except that the overall 4s frequency was about 5% lower.

Data Analysis

We fitted general and generalized linear mixed models using maximum likelihood procedures implemented in the lme4 package (Bates, Maechler & Bolker 2011) for R version 2.14.1 (R Development Core Team 2011) for our analyses. We verified that these models conformed to appropriate parametric assumptions by inspecting diagnostic plots. Throughout, we plotted means and 95% confidence intervals based on the raw data rather than the parameter estimates obtained from the statistical models to provide complementary information on the underlying data structure.

We ran two sets of analyses. First, to assess the overall effects of food treatment and selection regime on the offspring performance variables (development rate, growth rate, body size and spermatheca number), we analysed only the female offspring of the whole data set because the focus was on the female spermathecae. In this first series of analyses, we included mating treatment as a main factor only (without interactions) to control for its effects, as our main aim was to assess phenotypic responses to environmental conditions. Quasi-genetic correlations among traits and across environments were calculated using Pearson's correlations based on family means.

Secondly, when investigating the effects of maternal spermathecal phenotype (3s or 4s), mating treatment (single vs. triple mated), and offspring food treatment on offspring performance variables, we analysed female and male offspring of mothers from the 4s selection lines only, because 4s expression was nearly nil in the 3s lines and thus uninformative.

Development time was square-root-transformed prior to analysis. In all cases, our analyses of survival and spermatheca phenotype featured binomial errors and logit transforms. In addition to the fixed factors, both line identity and cross identity (nested within line) of the offspring were included as random effects in all models. We checked whether crossing random effects with fixed factors improved the models, but it never did, so we report parameter estimates from models with uncrossed random effects. Some of our binomial models indicated over-dispersion of the residuals, and quasi-error families are no longer supported in generalized linear mixed models using lme4. Instead, for these models, we fitted each individual offspring's ID (the lowest and unreplicated level of observation) as an additional random effect. This correction partitions the over-dispersion to variation between individuals, and produces more conservative results compared to uncorrected over-dispersed models. We used likelihood ratio tests during model simplification to determine whether terms should be retained in the minimum adequate model, starting with the highest order interactions. We report anova tables including likelihood-ratio χ2 extracted using the car package for R (Fox & Weisberg 2011).

Results

In total, our common garden experiment resulted in 1978 female and 2386 male offspring. Since we were mainly interested in the nature of the developmental plasticity and the relative fitness of the alternative spermatheca phenotypes, we focussed on the phenotypic analyses of the mothers and daughters, but of course included the male offspring to calculate egg-to-adult survival.

Developmental Plasticity and Responses to the Selection Regime

Food treatment had a strong effect on all phenotypic traits studied (Figs 1 and 2, Tables 1 and 3). Importantly, the frequency of 4s females declined from 60·52% (N = 423) in the control to 33·10% in agar (N = 435) and 43·04% in the ivermectin treatment (N = 446) when averaging over the three 4s lines, indicating large phenotypic plasticity of the trait (Fig. 1). The frequency of females developing 4s in the 3s lines was very low under all food conditions, ranging from 0% (N= 188) in the agar, through 2·48% (= 171) in the ivermectin to 3·50% (N = 161) in the control treatment (Fig. 1).

Figure 2.

Effects of food treatment on mean (± 95% CI) egg-to-adult (a) development time, (b) body size and (c) growth rate. Note that no offspring with 4s emerged from the 3s lines raised on agar and only very few from the other two food treatments, hence the large confidence limits.

Food treatment also substantially affected development time, body size and consequently the growth rate of the offspring (Tables 1 and 3, Fig. 2). Across both selection regimes, flies reared on agar-treated dung enclosed on average 1·7 days later and on ivermectin-treated dung 3·5 days later, than flies in the control group (Fig. 2a). Agar instead caused a stronger reduction in body size compared to ivermectin: the mean female hind tibia length of flies on agar was reduced by about 9·9% while that of flies on ivermectin was only reduced by 1·8% relative to that of control flies (Fig. 2b). As a result, the reduction in growth rate was more similar for both food treatments relative to the control (14·6% in agar and 9·1% in ivermectin, respectively, Fig. 2c).

In addition to the expected artificially selected responses in spermatheca number, the parental selection regime significantly influenced the development time, body size and growth rate of the offspring (Table 1). In general, offspring of flies selected for 4s were slightly but significantly smaller, showed longer development times, and reduced growth rates compared to offspring from flies selected for 3s (Fig. 2). Interaction terms between food treatment and selection regime were never found to be significant (Table 1; Fig. 2). We also tested for effects of the selection regime and the three different food treatments on offspring survival (Table 1). Egg-to-adult survival was significantly lower in the 4s than in the 3s selection lines (Table 1, Fig. 3a). Flies reared on ivermectin-treated dung had poorer survival than those reared on the control diet, while agar-treated flies survived as well as those from the control dung (Table 1, Fig. 3a).

Table 1. Generalized nested linear model results of the effects of food treatment and selection regime on the egg-to-adult survival (both sexes), spermatheca phenotype (3s vs. 4s), development time, hind tibia length and growth rate of female offspring. We included the number of mates in all models to control for its potential effect (see Table 3 for the main analysis of spermathecal types). Terms denoted NS did not significantly reduce model deviance and hence were dropped from the minimum adequate model. The two-way interaction between food treatment and selection regime was never significant. Parameter estimates and corresponding standard errors are provided in italics
Sourced.f.SpermathecaeDevelopment timeHind tibia lengthGrowth rateSurvival
χ2 P χ2 P χ2 P χ2 P χ2 P
Food treatment270·41<0·001663·02<0·0011356·57<0·0011083·72<0·00112·290·002
Agar −1·179 ± 0·1420·131 ± 0·012−0·262 ± 0·008−0·012 ± 0·000−0·000 ± 0·000
Ivermectin −0·707 ± 0·1370·317 ± 0·012−0·046 ± 0·008−0·011 ± 0·000−0·196 ± 0·065
Selection regime126·04<0·0014·490·03432·55<0·0017·780·0054·370·037
4s 3·825 ± 0·7500·096 ± 0·045−0·075 ± 0·013−0·005 ± 0·002−0·586 ± 0·280
No. mates1 NS NS NS5·420·02010·420·001
3 mates    0·003 ± 0·0010·788 ± 0·244
Figure 3.

Interactive effects of maternal spermathecal phenotype and the number of mates on family mean (± 95% CI) offspring egg-to-adult survival rate across food treatments. (a) compares flies from the two selection regimes and (b) only mothers from the 4s selection lines that phenotypically expressed three vs. four spermathecae. Sample sizes of family means are given in brackets.

Correlations Among Traits within and between Food Treatments

Whereas the 4s selection regime resulted in longer development times, smaller body sizes and reduced growth rates compared to the 3s regime, within the 4s selection regime offspring that expressed 4s showed shorter development times, larger body sizes and faster growth rates than offspring that developed 3s (t-test: all P < 0·05, except for body size differences in the control treatment: t-test P = 0·063; see Fig. 2). Using Spearman's rank correlations of family means, we further explored the relationships between spermatheca expression and development time, body size and growth rate across and within the different food environments (Table 2). Once again, for this analysis only crosses involving 4s lines were used because of the low expression of the additional spermatheca in the 3s lines (Fig. 1). These Spearman's rank correlations may overestimate genetic correlations as they can be inflated by maternal and common environmental effects, and hence should be interpreted with caution. Mirroring the average responses across food treatments shown in Fig. 2, the development of 4s showed moderate but uniformly negative correlations with development time across and within food environments (range from −0·083 to −0·273), with six of the nine coefficients being statistically significant (P < 0·05). Correlations between 4s development and body size were generally positive in sign (range from 0·043 to 0·383), though only two coefficients were significant. The corresponding nine correlation coefficients for growth rate were all positive in sign (range from 0·159 to 0·275), with seven being significant (Table 2).

Table 2. Spearman's rank correlations of the 4s frequency with development time, body size and growth rate across and within food treatments (A = agar, C = control, I = ivermectin). All correlations are based on family means, thus being quasi-genetic. Only crosses involving 4s lines were included in the analysis
 Dev time (A)Dev time (C)Dev time (I)Tibia length (A)Tibia length (C)Tibia length (I)Growth rate (A)Growth rate (C)Growth rate (I)
  1. Correlations within food environments denoted by an might be inflated by common environment effects (*P < 0·05, **P < 0·01, ***P < 0·001).

4s Freq (A)0·221*,† (129)0·083 (118)0·099 (118)0·226*,† (129)0·383*** (117)0·058 (117)0·275**, (129)0·229* (117)0·163 (117)
4s Freq (C)0·273** (117)0·210*,† (126)0·237* (117)0·086 (117)0·163 (126)0·138 (115)0·264** (117)0·260**,† (126)0·279** (115)
4s Freq (I)0·182* (118)0·166 (118)0·164 (128)0·043 (118)0·160 (117)0·184*,† (126)0·159 (118)0·199* (117)0·183*,† (126)

Functional Significance and Sexual Benefits

The second main objective of our study was to explore the performance of offspring produced by both maternal spermatheca phenotypes. We analysed the effects of maternal phenotype and mating treatment (one vs. three mates) across food environments on egg-to-adult survival and other offspring life-history traits (Table 3). As above, only mothers from the 4s regime were included in this analysis.

Table 3. Generalized nested linear model results of the effects of food treatment, number of mates (1 vs 3) and maternal spermatheca phenotype (3s vs. 4s) on the egg-to-adult survival (both sexes), spermatheca phenotype, development time, hind tibia length and growth rate of the offspring. Only crosses involving females derived from the 4s lines are considered in the analyses. Terms for which no value is listed (NS) did not significantly decrease model deviance and hence were dropped from the minimal adequate model. The three-way interaction and the two two-way interactions between maternal spermatheca phenotype and the number of mates as well as food treatment were never significant. Parameter estimates and corresponding standard errors are given in italics
Sourced.f.SpermathecaeDevelopment timeHind tibia lengthGrowth rateSurvival
χ2 P χ2 P χ2 P χ2 P χ2 P
  1. a

    Parameter estimate of a nonsignificant factor, which nevertheless improved the overall fit of the model according to a likelihood-ratio test.

Maternal Spermatheca Phenotype29·57<0·002 NS NS NS NS
4s 0·398 ± 0·129   −0·420 ± 0·297a
No. Mates1 NS NS32·640·0016·110·0136·510·011
3 Mates   −0·051 ± 0·016−0·003 ± 0·0010·713 ± 0·729
Food Treatment161·06<0·001724·20<0·0011316·03<0·0017·780·005 NS
Agar −1·136 ± 0·1470·125 ± 0·012−0·214 ± 0·012−0·011 ± 0·001 
Ivermectin −0·680 ± 0·1420·314 ± 0·012−0·056 ± 0·012−0·012 ± 0·001 
No. Mates X Food1 NS NS44·11<0·0017·780·005 NS
3 Mates X Agar   −0·081 ± 0·016−0·003 ± 0·001 
3 Mates X Ivermectin   −0·016 ± 0·0160·002 ± 0·001 

As expected, maternal spermatheca phenotype was correlated with the daughter's spermatheca phenotype (Table 3), demonstrating significant standing genetic variation for the trait within our 4s lines. In addition, multiple mating enhanced offspring egg-to-adult survival (Table 3; Fig. 3a,b), which is consistent with the paternity assurance hypothesis predicting that multiple mating should increase the chances of obtaining viable or compatible sperm (see Tregenza et al. 2003 for similar results in S. stercoraria). If having four spermathecae permits more efficient sperm storage or partitioning of ejaculates to improve sperm choice, 4s females in the multiple mating group should have the highest fitness. Contrary to this prediction, the 4s phenotype did not result in higher offspring survival (either as a main effect or in interactions with number of mates; Table 3, Fig. 3b). In fact, the nonsignificant parameter estimates involving maternal spermatheca phenotype suggested nonsignificant but negative effects of 4s on survival (before they were excluded from the model in the course of model simplification: 4s × 3 mates interaction deviation = −0·101 ± 0·060; χ2 = 0·029; P = 0·866; 4S main effect deviation = −0·420 ± 0·297 SE; χ2 = 1·99; P = 0·158).

We further tested whether maternal spermatheca phenotype (as a main effect or in an interaction with mating treatment) significantly influenced body size, development time or growth rate of the offspring, but found no evidence for this (Table 3). However, the effect of multiple mating on body size and growth rate depended on food treatment, such that offspring from triple-mated females were smaller and had reduced growth rates when raised on agar-treated dung (Table 3).

Discussion

Our study of the developmental plasticity and fitness consequences of a discrete sperm storage polymorphism in the yellow dung fly Scathophaga stercoraria leads to two main conclusions. First, there is significant standing genetic variance for spermatheca number, and its expression is both phenotypically and genetically correlated with juvenile growth and development. Secondly, fitness benefits through sexual selection are unlikely to favour alleles for the rare 4s phenotype within natural dung fly populations, as offspring of mothers artificially selected for the 4s phenotype indicated reduced rather than increased egg-to-adult survival relative to mothers selected for three spermathecae independent of the mating treatment. In the following sections, we first discuss the developmental plasticity of spermatheca expression in light of adaptive and nonadaptive scenarios that might explain the polymorphism, and then more generally examine the implications of our findings for the evolutionary dynamics and phylogenetic transitions in female sperm storage organs and co-evolving male fertilization related traits.

Phenotypic plasticity may evolve as an adaptive response to a variable but predictable environment, allowing the production of an optimal phenotype in each of the alternative environments. Alternatively, it may reflect neutral or even maladaptive phenotypic responses due to low genetic robustness against extrinsic (i.e. environmental) and/or intrinsic (i.e. genetic) stressors (Gotthard & Nylin 1995; West-Eberhard 2003; Berrigan & Scheiner 2004). The suppression of the 4s phenotype in response to food limitation (via dilution with agar) and pharmaceutical residues could be interpreted as condition-dependent variation directly linked to resource acquisition, as expected for a costly mate choice trait that influences paternity (Hunt, Brooks & Jennions 2005; Ward 2007; Amitin & Pitnick 2007). Previous interpretations of the spermathecal polymorphism of S. stercoraria as a mate choice adaptation largely rested on the observation that double-mated 4s females in good body condition exhibit lower second male sperm precedence than 3s females, suggesting that the 4s phenotype confers increased sperm sorting capabilities (Ward 2000, 2007; Ward, Wilson & Reim 2008). In our study, triple-mated females with 4s, however, showed no evidence for enhanced egg-to-adult survival, suggesting that a greater variance in paternity does not confer fitness benefits. It is nevertheless possible that other fitness components besides those that we measured do improve – this possibility requires further investigations. Also, we found no significant interactions between maternal 4s phenotype and the mating treatment. Such interactions could have been expected under postcopulatory mate choice or conflict, and a range of adaptive scenarios related to sperm storage, particularly in light of the fact that multiple mating enhanced egg-to-adult survival in the present study. On the contrary, our data indicated negative effects on offspring growth and survival that appeared to represent correlated responses to artificial selection for 4s, in addition to the fecundity costs reported earlier (Ward, Wilson & Reim 2008). Together with the observation that genetic variation encoding the polymorphism remains largely hidden from direct selection in nature (Berger et al. 2011), our results question that the expression of the 4s phenotype maybe adaptive and trade-off against costs that females incur from sperm discrimination (Ward 2000, 2007; Ward, Wilson & Reim 2008).

Alternatively, the plasticity in spermatheca development could arise because of developmental instability, which might increase in magnitude as juvenile growth and development hasten (Dmitriew 2011). Since food limitation (including the agar dilution used here) and pharmaceutical residues reduce levels of 4s phenotype expression as well as juvenile growth and development (this study), whereas warm rearing temperatures, even when detrimentally high, augment all these traits (Berger et al. 2011), the combined results of both studies implicate a shared metabolic pathway regulating somatic growth and the development of the female reproductive tract in S. stercoraria. This interpretation is supported by the positive family-mean correlations between the traits scored in this study (Table 2), indicating that siblings with fast intrinsic (genetic) growth are more likely to develop the 4s phenotype (provided that allelic variation for the phenotype is present). Positive genetic correlations between 4s expression and juvenile growth and development have also been identified in a natural population of S. stercoraria (R. Walters, D. Berger, M. A. Schäfer & W. U. Blanckenhorn unpublished data), implying that these correlations are not merely a spurious correlated response to the enforced artificial selection. Also, northern Sstercoraria populations, which have evolved fast compensatory growth in response to seasonal time constraints (Demont et al. 2008; Scharf et al. 2010), exhibit higher 4s frequencies than southern populations when reared under standardized laboratory conditions, but not in the field (Berger et al. 2011).

Interestingly, a seemingly opposite pattern is found when comparing growth and development rates across the 3s and 4s selection regimes, with slower growth within 4s lines. This result, although only based on a comparison between five selection lines in total, would imply that aspects of the genetic architecture related to developmental instability were selected in parallel to the expression of 4s. Thus, it is perhaps not surprising that allelic variation causing both increased 4s expression and seemingly lower fitness is less pronounced within natural populations as compared to variation arising from the artificial selection programme. Substantial developmental costs associated with intraspecific variation in female reproductive morphology have also been implicated from a bidirectional artificial selection programme on the length of seminal receptacle, the primary organ of sperm storage in Drosophila melanogaster. Fruit flies selected for a long receptacle not only showed a substantial delay in egg-to-adult development time but further indicated a correlated reduction in longevity, which might counterbalance the evolution of longer receptacles in nature (Miller & Pitnick 2003).

Although we only studied a single female reproductive trait here, there is good reason to expect similar levels of phenotypic plasticity for other aspects of reproductive morphology as well, since the spermathecal polymorphism correlates with a number of traits including sperm storage capacity and the number and length of the spermathecal ducts (Ward, Wilson & Reim 2008; Thüler 2009). Even if effectively neutral or maladaptive, the phenotypic plasticity found here almost certainly alters the physical environment where sperm compete and hence could influence the evolutionary trajectories of male fertilization related traits via male–female interactions affecting paternity (Clark, Begun & Prout 1999; Tregenza et al. 2003; Tregenza, Attia & Buahaiba 2009). For example, Pitnick and co-workers describe how phenotypic plasticity and genetic variation in the female seminal receptacle influences the evolution of sperm length and other traits related to paternity assurance across and within species of Drosophila (Miller & Pitnick 2002; Amitin & Pitnick 2007; Bjork et al. 2007). Similarly, loss of spermathecae in certain lineages of stalk-eyed flies was found to correlate with phylogenetically independent gains in sperm length (Presgraves, Baker & Wilkinson 1999), and a comparative analysis of the Scathophagidae (including S. stercoraria) showed that the sperm storage capacity of the spermathecae co-evolves with testes size (Minder, Hosken & Ward 2005).

Although we addressed the relative fitness of the artificially selected 4s phenotype in different mating environments, we cannot clarify the conditions that produced appreciable, naturally occurring allelic variation for the phenotype in the first place. Accumulation of deleterious mutations due to the sex-specific trait expression (De Visser et al. 2003; Mank & Ellegren 2009; Van Dyken & Wade 2010) could play a part, but may not explain the prevalence of the polymorphism in its entirety. However, by showing that the development of the 4s phenotype, which is largely absent in nature, correlates with extrinsic (environmental) and intrinsic (genetic) larval growth and development, and that the phenotype responds readily to selection with only moderate reductions in egg-to-adult survival, our study clearly demonstrates a great potential for evolutionary change. Because larval growth trajectories are shaped by a broad range of environmental factors such as food availability, temperature or season length (Blanckenhorn 1998; Gotthard 2008; Dmitriew 2011), environmentally mediated changes in growth and development could facilitate phylogenetic transitions in spermatheca number, which are frequently observed within the Diptera (e.g. Pitnick, Markov & Spicer 1999; Presgraves, Baker & Wilkinson 1999; De Jong 2000), by exposing hidden genetic variation to selection. Such initial stages may then be followed by persistent natural or sexual selection decoupling developmental instability from the morphological phenotypes and lead to the genetic assimilation of a novel trait (Waddington 1953; West-Eberhard 2003; Flatt 2005).

Similar results and conclusions were obtained in early studies of Drosophila with remarkably similar spermathecal phenotypes to those seen in S. stercoraria. Based on their study of D. melanogaster lines selected for three spermathecae (wild type females typically possess only two), Mather & Harrison (1949) noticed that variation in spermatheca number correlates with larval growth and development, and they discussed ‘the capacity of the reservoir of hidden genetic variability within a species’ for evolutionary change. Interestingly, Wexelson (1928) was able to recover the fertility of his Drosophila 3s mutant lines (which originated from strains originally selected for cell degeneration) after backcrossing them into wild-type strains while maintaining the 3s phenotype. Furthermore, Hadorn & Garber (1944) found that recessive mutants for spermatheca, which is likely a tissue-specific allele of the gene engrailed involved in the sex-determining pathway (Chase & Baker 1995), are highly temperature sensitive, like in S. stercoraria (Berger et al. 2011). These similarities might not have been necessarily expected a priori in light of the species' phylogenetic distance, but may be consistent with the growing evidence that repeated evolutionary change of homologous morphological structures is often governed by genetic modifications of the same developmental pathway (Abouheif & Wray 2002; Shapiro et al. 2004; Hoekstra et al. 2006). The comparative study of the molecular and developmental pathways regulating spermathecal polymorphisms across species should therefore be a promising avenue for future research on the diversity and evolvability of female sperm storage organs as well as the corresponding co-adapting male traits related to insemination and fertilization.

Acknowledgements

We are grateful to Karin Thüler for kindly providing us with the flies used in the experiment. We further like to thank Ursula Briegel, Chris Shirley and Samuel Tanner for their help in collecting the data and Yves Choffat for logistic support. The project is part of an SNF grant to W. U. Blanckenhorn. M. A. Schäfer was supported by a scholarship (SCHA 1502/2-1) of the Deutsche Forschungsgemeinschaft (DFG), D. Berger by a grant from the Swedish Research Council (VR) and L. F. Bussière by the University of Stirling during the preparation of the manuscript. None of the authors declare any sources of conflict of interest.

Data accessibility

Data are deposited in the Dryad repository: http://dx.doi.org/10.5061/dryad.31072

Ancillary