Sexual selection theory makes clear predictions regarding male spermatogenic investment. To test these predictions we used experimental sexual selection in Drosophila pseudoobscura, a sperm heteromorphic species in which males produce both fertile and sterile sperm, the latter of which may function in postmating competition. Specifically, we determined whether the number and size of both sperm types, as well as relative testis mass and accessory gland size, increased with increased sperm competition risk and whether any fitness benefits could accrue from such changes. We found no effect of sexual selection history on either the number or size of either sperm morph, or on relative testis mass. However, males experiencing a greater opportunity for sexual selection evolved the largest accessory glands, had the greatest mating capacity, and sired the most progeny. These findings suggest that sterile sperm are not direct targets of sexual selection and that accessory gland size, rather than testis mass, appears to be an important determinant of male reproductive success. We briefly review the data from experimental sexual selection studies and find that testis mass may not be a frequent target of postcopulatory sexual selection and, even when it is, the resulting changes do not always improve fitness.

Female promiscuity occurs across a broad range of taxa and is virtually ubiquitous among insects (Ridley 1988; Andersson 1994; Arnqvist and Nilsson 2000). The conditions necessary for sperm competition to occur, that is contests among rival ejaculates to fertilize ova, frequently occur in insects as females typically mate multiply and store sperm from multiple males (Parker 1970a, 1984, 1988). Evolutionary responses to sperm competition take a myriad of forms, encompassing behavioral, physiological, and morphological adaptations, including alterations in mating capacity (Harcourt et al. 1981; Birkhead et al. 1987), increased ejaculate investment (e.g., Parker 1990; Simmons 2001), and interspecific morphological diversity in sperm (Smith 1984; Stockley 1997; Birkhead and Møller 1998; Simmons and Siva-Jothy 1998; Simmons 2001; Snook 2005).

Sperm morphological diversity also occurs within species (e.g., Swallow and Wilkinson 2002; Till-Bottraud et al. 2005) because selection can act on individual gamete shape and size (Parker 1970a). Intraspecific gamete diversity can be manifested as sperm heteromorphism, the simultaneous production of multiple sperm types by individual males (Silberglied et al. 1984; Swallow and Wilkinson 2002; Till-Bottraud et al. 2005). Sperm heteromorphism is neither a rare phenomenon, nor taxonomically limited in animals, being found in some vertebrates and in over half of the invertebrate taxonomic groups investigated (Till-Bottraud et al. 2005). One such sperm heteromorphic group is the Drosophila obscura group of fruitflies. Unlike the rest of its genus, males in this group produce two sperm types, short and long (Snook 1997). Although both are nucleated and motile, only long sperm are fertilization-competent (Snook et al. 1994; Snook and Karr 1998; Snook and Markow 2002) and yet short sperm comprise over half of the ejaculate (Holman et al. 2008). Given that spermatogenesis is costly and can limit male fitness (reviewed in Dewsbury 1982; Simmons 2001; Wedell et al. 2002), production of an ejaculate in which infertile gametes outnumber fertile ones likely entails a nontrivial energetic outlay from which corresponding fitness benefits are assumed to accrue (Till-Bottraud et al. 2005).

Over the last decade considerable empirical effort has been invested to identify the adaptive significance of the sterile caste of gametes in D. pseudoobsura. Studies show that sterile short sperm are not naturally selected (Silberglied et al. 1984; Sivinski 1984) to provision either the female, eggs, or developing embryos (Snook et al. 1994; Snook and Markow 1996; Snook and Karr 1998). Sterile sperm may have arisen through sexual selection (Silberglied et al. 1984; Sivinski 1984) if their presence, alongside fertile sperm, confers a competitive advantage on a male's ejaculate. Several potential sperm competitive functions have been proposed and tested (reviewed in Till-Bottraud et al. 2005; Holman and Snook 2006). However, there is no evidence that in D. pseudoobscura sterile sperm act defensively, either by blocking admission of rival ejaculates to female storage organs (Snook et al. 1994; Snook 1998) or by acting as “cheap filler” (Silberglied et al. 1984) to delay female remating (Snook 1998). Short sperm may act offensively to remove or destroy previously deposited rival sperm (Silberglied et al. 1984) although this hypothesis has not been tested. A recently proposed alternative hypothesis is that sterile sperm function to increase the probability that their “brother” fertile gametes will persist within the hostile spermicidal environment of the female reproductive tract (Till-Bottraud et al. 2005; Holman and Snook 2006). This hypothesis was supported when Holman and Snook (2008) found that the female reproductive tract of D. pseuddobscura is spermicidal and that males producing a larger proportion of short sperm in their ejaculate had a greater number of long sperm surviving in the female. This function could have arisen to increase a male's fitness during sperm competition given that the number of fertilizing sperm in the fertilizing set will increase. Alternatively, the evolution of a sterile antispermicidal sperm morph need not depend on the action of sexual selection if female spermicide is nonadaptive, for instance, if it occurs as a pleiotropic effect of immune activation in the reproductive tract following insemination (Holman and Snook 2006, 2008). Thus, while the putative function of short sperm may be resolved, the evolutionary route by which sperm heteromorphism has arisen in D. pseudoobscura remains a conundrum.

Sexual selection also may act on other aspects of male reproductive physiology and morphology. Because sperm competitive success often depends on relative sperm numbers and the overall demand for sperm is a function of ejaculate size and mating rate (Short 1979; Parker 1982, 1984; Gomendio et al. 1998; Simmons 2001), males experiencing more intense sperm competition should invest more in testicular tissue, the machinery of spermatogenesis. Moreover, in Drosophila, gametes are accompanied by seminal fluid proteins arising from the accessory glands. These accessory gland proteins (Acps) mediate sperm competitive interactions and induce profound changes in female reproductive biology, potentially generating sexual conflict (Chen 1984; Wolfner 1997; Chapman 2001; Arnqvist and Rowe 2005). Males experiencing more intense sperm competition should therefore invest in larger accessory glands to benefit from the production of Acps associated with sperm competition and sexual conflict. This increase in ejaculatory investment is predicted to allow males to mate either with more females or more frequently (Dewsbury 1982; Simmons 2001; Wedell et al. 2002), thereby siring more progeny and garnering higher fitness than males selected under less intense or no sperm competition.

We took an experimental evolution approach to test these key predictions regarding the role of sexual selection as the driving force behind the evolution of sperm number and size, and testis and accessory gland size in male D. pseudoobscura. Experimental evolution has a proven record of generating unique data on how sexual selection shapes the reproductive biology of insects (Holland and Rice 1999; Hosken et al. 2001; Hosken and Ward 2001; Pitnick et al. 2001a,b; Wigby and Chapman 2004; Crudgington et al. 2005; Snook et al. 2005; Linklater et al. 2007; Reuter et al. 2008). With respect to ejaculate characteristics, for example, naturally promiscuous D. melanogaster males that were maintained monogamously evolved smaller testes and had fewer developing spermatocytes than control (promiscuous) males (Pitnick et al. 2001a). Similarly in the yellow dung-fly, Scatophaga stercoraria, enforced monogamy generated a rapid reduction in testis size relative to promiscuity (Hosken and Ward 2001) which, as predicted, resulted in poorer sperm competitiveness (Hosken et al. 2001). To date, studies that have used experimental evolution to manipulate insect mating systems have only considered sperm monomorphic species. The simultaneous production of multiple sperm types by individual male D. pseudoobscura represents an additional layer of ejaculatory complexity, which should afford us unique opportunities to test sexual selection theory regarding male postmating success.

We manipulated sexual selection in the naturally promiscuous D. pseudoobscura (Anderson 1974) by enforcing monogamy, maintaining predicted levels of female promiscuity, or elevating the potential for female promiscuity. In accordance with sexual selection theory, our previous research on these lines has shown divergence in critical aspects of male courtship and in the extent of intersexual conflict. Males evolving under female promiscuity initiated courtship song earlier and sang faster (Snook et al. 2005) than monogamy-selected males, whereas females housed with promiscuity-selected males produced fewer progeny (Crudgington et al. 2005) and suffered reduced survival (H. S. Crudgington and R. R. Snook, unpubl. data) relative to those housed with monogamy males. These findings clearly demonstrate the effectiveness of our sexual selection treatments. We therefore aimed to examine the consequences of alternating sexual selection via experimental evolution for male ejaculatory investment and mating behavior. Specifically, we quantified the (1) number and length of both the short sterile and the long fertile sperm morph contained within a male's first ejaculate, (2) relative testis mass, (3) size of the accessory glands that are responsible for Acp production (Chen 1984; Wolfner 1997; Chapman 2001), (4) number of copulations males achieved when presented with successive mating opportunities, (5) resulting progeny from the sequential matings as a surrogate for fitness, and (6) number of inseminations gained when males were confined en masse with multiple females. This study therefore allows us to experimentally test general sexual selection theory and to address the evolutionary route by which sperm heteromorphism may have arisen in this species.

Materials and Methods


We manipulated opportunities for sexual selection by establishing three experimental evolution treatments: (1) enforced monogamy (M: one female and one male), (2) control promiscuity (C: one female and three males), and (3) elevated promiscuity (E: one female and six males). We established four replicate populations of each treatment, thus giving a total of 12 experimental populations (see Fig. 1). The establishment and maintenance of these lines has previously been described in full in Crudgington et al. (2005). Briefly, 4- to 7-day-old virgin males and females were combined in ratios appropriate to each treatment; hereafter referred to as “families.” We effectively controlled for Ne by increasing the number of families in the M treatment relative to C and E (see Crudgington et al. 2005; Bacigalupe et al. 2008; R. R. Snook and J. Slate, unpubl. data). Each family was housed in vials containing standard food media and live yeast for five days, after which they were transferred to fresh food vials for a further five days. All flies were then discarded and eggs laid in the second set of vials were allowed to develop. Parental flies therefore experienced a total interaction period of 10 days which is sufficient for female remating to occur and thus for sperm competition to potentially arise (Turner and Anderson 1984; Badcock 2006). On eclosion, for each treatment and replicate (handled separately), virgin progeny from all family vials were mixed, and males and females were randomly collected via CO2 anesthetization. Virgin males and females were housed separately in yeasted food vials for five to seven days after which they are randomly assigned to family vials to propagate the next generation. When collected for experiments, for each replicate and treatment, virgin males and females were collected via CO2 anesthetization and housed separately in food bottles with dried yeast for five days before use. For experiments, males and females were mouth-pootered into food vials.

Figure 1.

Schematic of experimental evolution design comprising four replicates (1, 2, 3, and 4) of three sexual selection treatments (M, C, and E). The original population derived from 50 wild-caught mated females collected in Tucson, Arizona USA. Following a minimum of three generations of laboratory rearing, we set up our replicate populations in a staggered fashion (see Crudgington et al. 2005) by randomly allocating virgin males and females from the ancestral population into one of the three treatments. To control for variation in Ne, we set up a variable number of families among the treatments: M = 90, C = 65, and E = 65.


The length of the wing vein IV (WVL) was measured for use as an index of body size (Robertson and Reeve 1952; Gilchrist et al. 2001) as follows: the right-hand wing, as viewed dorsally, was removed from each fly, mounted on a dissection slide with 25 μl PBS, and allowed to dry at room temperature for 24 h. Images of wings were captured and measured using Imagepro and Image-J software, respectively.


Individual 5-day-old virgin females were confined with two 5-day-old virgin males in vials containing standard media and live yeast. On initiation of copulation with one of the males, the nonmating male was discarded. The female reproductive tract was dissected within 4 h of copulation, before sperm begin migrating from the uterus to the sperm storage organs (Snook et al. 1994). Sperm were removed from the uterus as previously described (Snook et al. 1994) and mated flies were retained for subsequent body size measurement. Because the time needed to quantify all sperm within an ejaculate would be prohibitive, each sample was dispersed in 100 μl of phosphate buffer solution (PBS), 50 μl of this suspension was then added to a further 50 μl of PBS, and an estimate of total short and long sperm numbers was gained by extrapolating from the numbers in a 2.5 μl aliquot of the suspension. Two 2.5 μl aliquots were pipetted onto a microscope slide that was dried in an oven for 1 h at 65°C in preparation for sperm staining and subsequent counting. To gain measures of sperm length, the remaining 50 μl of the original ejaculate/PBS suspension was transferred to a fresh microscope slide and processed as described in Snook et al. (1994).

Sperm samples were fixed and stained with the DNA-specific epifluoresecent dye diaminophenylinidole (DAPI, Sigma-Aldrich, Gillingham, U.K.; 2.19 μM/PBS solution) and the numbers of short and long sperm in a 2.5 μl aliquot of each male's ejaculate were counted using epifluorescent microscopy. Digital images of six short and six long sperm from each male were captured and measured using Imagepro (Media Cybernetics, Bethesda, MD) and Image-J software (http://rsbweb.nih.gov/ij/) as Monte-Carlo computer simulations indicated that this number is sufficient to capture the extent of within-male variation equivalent to measuring 20 sperm of each type (Badcock 2006). Mean values for short and long sperm length for each male were then derived from the six measures of each type. The methods used for quantifying sperm numbers and lengths are highly repeatable in the Snook laboratory (Badcock 2006; Holman et al. 2008). Flies for this investigation were sourced from generation 43, 42, 42, and 42 of replicate 1, 2, 3, and 4, respectively.


Measures of dry testis and somatic mass for 10 5-day-old virgin males from each selection line population were gained using frozen flies. After being defrosted at room temperature, each male was dissected and processed as previously described for Drosophila (Pitnick et al. 2001a; Holman et al. 2008). In this assay, our body size measure was somatic mass rather than WVL. Flies for this investigation were sourced from generation 67, 66, 65, and 63 of replicate 1, 2, 3, and 4, respectively.


Mean accessory gland area was determined for 20 5-day-old virgin males using frozen flies. For each male, the abdomen was separated from the rest of the body using dissection pins and the abdominal contents were gently transferred to PBS. The accessory glands were then isolated and placed in 8 μl of mineral oil on a microscope slide. An image of the paired glands was captured using Motic Images software (Motic, Ipswich, U.K.) and each gland was measured using Image-J software. Mean accessory gland area for each male was then derived from these two measures. Males were also processed for wing vein length as described above. Flies for this investigation were sourced from generation 75, 74, 78, and 71 of replicate 1, 2, 3, and 4, respectively.


We performed two experiments to determine male mating capacity, one in which males were presented with sequential mating opportunities with virgin females and another in which males were confined simultaneously with multiple virgin females. In the first of these experiments 20 5-day-old virgin males were confined individually with two 5-day-old virgin females in vials containing standard media and live yeast. Males and females were from coevolved lines; to accurately assess male mating capacity we needed to ensure in this experiment that male signals and female receptors matched (see Pizzari and Snook 2003 and Rowe et al. 2003 for discussion about interpopulation crosses). Two females were added as we were surveying male mating capacity and did not want to confound exhaustion of male mating capacity with female disinterest. On initiation of copulation with one of the females, the unmated female was unobtrusively removed. Following termination of copulation the male was immediately transferred to a fresh vial, again containing two virgin females. This sequence of mating, removal of the unmated female, and transfer of the male to a vial containing two fresh virgin females, was repeated until the male failed to achieve further copulations (defined as the lapsing of 30 min since the male was confined with fresh females). Following mating, each female was housed individually in a food vial for four days before being transferred to a fresh vial for a further four days. Thus, each female was permitted to oviposit for a total of eight days after which she was discarded and her emerging progeny was collected. The progeny resulting from each of the males' successive copulations were counted and summed to provide an additive index of male fitness. Males were retained for quantification of wing vein length as described above. Males excluded from analyses included those that failed to mate, those that sired fewer than 10 progeny, and those that escaped during the experiment and thus for which data were incomplete. Flies for this experiment were sourced from generation 60, 59, and 57 of replicate 2, 3, and 4, respectively.

For the second male mating capacity experiment we housed 10 5-day-old virgin males individually with 20 5-day-old virgin females in a food bottle for 8 h, after which each male was removed. Females were then frozen for subsequent dissection (as described above for the sperm trait assay) to determine whether they had been inseminated. Flies for this experiment were sourced from generation 62, 61, 60, and 58 of replicate 1, 2, 3, and 4, respectively.


All data, with the unit of replication as individual flies, were analyzed using a linear mixed effects model in which replicate (random factor) was nested within sexual selection treatment (fixed factor). For all of the sperm trait and testis mass data, which had normally distributed errors, we used JMP (data on short sperm number were log transformed). In the first mating capacity experiment, data on the number of sequential mates had Poisson-distributed errors (count data) and so were analyzed using the lme4 package for R, which allowed for analysis as a generalized linear mixed effects model, again with replicate nested within sexual selection treatment. We analyzed the progeny data from the first mating capacity experiment and data on the number of females inseminated from the second mating capacity experiment (both count data) using the same model but specifying a quasi-Poisson error distribution because the data were overdispersed. The P values for these analyses were derived through Markov chain Monte Carlo sampling (Baayen 2008), whereas the effect of replicate was evaluated by comparing the standard deviation for replicate with the residual deviation (Crawley 2007). Male (or female) body size data were initially included in models as covariates but were only retained if they explained a significant amount of variation in the response variable. However, retention of body size as a covariate did not qualitatively alter model outcomes. Analyses were performed using JMP ver. 7.0 and R ver. 2.6.l (The R Foundation for Statistical Computing).



Although sexual selection replicate had a significant effect on all the sperm traits, there was no significant effect of sexual selection treatment on any sperm trait (see Table 1 for test statistics and Table 2 for trait means and standard errors). With the exception of the proportion of short sperm, neither male nor female WVL was a significant covariate with any of the sperm traits examined. Additionally, there were no differences in either male or female body size among treatments (nested ANOVA: replicate nested within treatment and individuals as the unit of replication; males, P= 0.242; females, P= 0.44).

Table 1.  Test statistics from nested analysis of variance (ANOVA) or analysis of covariance (ANCOVA) for a suite of sperm traits following virgin copulations. An index of male or female body size (WVL) was retained in models (ANCOVA) only when it significantly covaried with the sperm trait. M, monogamy; C, control promiscuity; E, elevated promiscuity.
Sperm traitSexual selection treatment (M, C, E)Sexual selection replicate (1–4)WVL covariate
Log number of short spermF2,9=1.697, P>0.2F9,11=4.128, P<0.0001 
Number of long spermF2,9=0.27, P>0.7F9,11=2.874, P=0.0032 
Proportion of short spermF2,9=1.247, P>0.3F9,11=7.754, P<0.0001F1,211=5.883, P=0.016
Short sperm lengthF2,9=0.264, P>0.7F9,11=3.819, P=0.0002 
Long sperm lengthF2,9=1.357, P>0.3F9,11=5.076, P<0.0001 
Short sperm head lengthF2,9=0.319, P>0.7F9,11=5.395, P<0.0001 
Short sperm tail lengthF2,9=0.268, P>0.7F9,11=3.58, P=0.0004 
Long sperm head lengthF2,9=1.002, P>0.4F9,11=5.794, P<0.0001 
Long sperm tail lengthF2,9=1.277, P>0.3F9,11=4.476, P<0.0001 
Table 2.  Sperm trait and testis mass measures for virgin males for each replicate population (e.g., R1=Replicate 1) of each sexual selection treatment; values are nontransformed means±standard errors. M, monogamy; C, control promiscuity; E, elevated promiscuity.
Sperm-related traitR1 MR1 CR1 ER2 MR2 CR2 ER3 MR3 CR3 ER4 MR4 CR4 E
Short length95.7490.2787.4579.8791.0585.8386.9091.6182.4988.6687.4796.02
Long length319.26305.45305.51301.63312.15299.05311.47304.81302.12306.85312.48308.65
Short head14.7114.3613.4512.1514.6614.0413.5214.2013.3014.2213.3115.31
Short tail81.0675.974.067.7276.4071.7973.3877.4269.1874.4374.1580.94
Long head60.4459.5657.7857.1161.5557.9660.3958.0659.0660.1160.3959.59
Long tail259.06245.89247.73244.53250.61241.09251.08246.75243.06246.74252.59249.10
Testis mass31.3129.9831.6426.5728.6628.9929.0629.4929.8626.7330.0131.63


Replicate, but not sexual selection history, had a significant impact on male testis mass (treatment effect: F2,9= 1.401, P > 0.2; replicate effect: F9,11= 2.223, P= 0.02580). Somatic mass significantly positively covaried with testis mass (F1,107= 11.453, P= 0.001) (see Table 2 for trait means and standard errors). Additionally, there were no differences in male total body mass among treatments (nested ANOVA: replicate nested within treatment and individuals as the unit of replication; P= 0.69).


Both sexual selection treatment and replicate had a significant effect on mean male accessory gland size (treatment effect: F2,9= 16.25, P= 0.001; replicate effect: F9,11= 6.156, P < 0.0001), with a Tukey HSD test showing that E males had a significantly larger mean accessory gland area than either M or C males (P < 0.05); M and C males did not differ (P > 0.05)(Fig. 2). Male WVL was not a significant covariate with male accessory gland size. Additionally, there were no differences in male body size among treatments (nested ANOVA: replicate nested within treatment and individuals as the unit of replication; P= 0.82).

Figure 2.

Sexual selection treatment had a significant effect on the mean area of the paired accessory glands of virgin males; values are means ± standard errors. M, monogamy; C, control promiscuity; E, elevated promiscuity; Rep, replicate population.


For the first male mating capacity experiment in which males were provided with sequential mating opportunities, sexual selection treatment caused significant divergence in male mating capacity with E males mating with more females sequentially than either M or C males; M and C males did not differ (Fig. 3; see Table 3 for model description and test statistics). Male WVL was not a significant covariate with mate number. Similarly, there was a significant impact of sexual selection history on the total number of progeny sired by males through their sequential matings with the pattern of among-treatment differences in total progeny number mirroring that of mating capacity: E males fathered more progeny than either M or C males; M and C males did not differ (Fig. 4; see Table 3 for model description and test statistics). Male WVL did not significantly covary with total progeny number. Additionally, there were no differences in male body size among treatments (nested ANOVA: replicate nested within treatment and individuals as the unit of replication; P= 0.806).

Figure 3.

Sexual selection treatment had a significant effect on the number of mates gained by males when sequentially presented with virgin females; values are means ± standard errors. M, monogamy; C, control promiscuity; E, elevated promiscuity; Rep, replicate population.

Table 3.  Statistical model statements, model error structures, and test statistics from the male mating capacity experiment in which virgin males were sequentially presented with mating opportunities with virgin females. The response variables were “number of sequential matings” and “total number of progeny sired.” M, monogamy; C, control promiscuity; E, elevated promiscuity; SD, standard deviation.
Response variableGeneralized linear mixed effects model using lme4 in RModel error structureComparisons among sexual selection treatmentsReplicate SD and residual SD respectively
Number of sequential matingsresponse∼treatment +(1|replicate)PoissonE>M, z=4.539, P<0.00012.236×10−5, na
   E>C, z=4.89, P<0.0001 
   M=C, z=0.306, P>0.7 
Total number of progeny siredresponse∼treatment +(1|replicate)quasi-Poisson (data overdispersed)E>M, t=4.733, P<0.00011.407, 12.182
   E>C, t=5.91, P<0.0001 
   M=C, t=0.441, P>0.6 
Figure 4.

Sexual selection treatment had a significant effect on the total number of progeny that males sired through their sequential matings; values are means ± standard errors. M, monogamy; C, control promiscuity; E, elevated promiscuity; Rep, replicate population.

In the second male mating capacity experiment in which males were confined with multiple females for 8 h, sexual selection treatment was a significant determinant of the number of females inseminated by males: E males inseminated more females than either M or C males; M and C males did not differ (Fig. 5; see Table 4 for model description and test statistics).

Figure 5.

Sexual selection treatment had a significant effect on the number of virgin females inseminated by males following 8-h. M, monogamy; C, control promiscuity; E, elevated promiscuity; Rep, replicate population.

Table 4.  Statistical model statements, model error structures, and test statistics for the male mating capacity experiment in which virgin males were housed individually with 20 virgin females. The response variable was “number of inseminated females.” M, monogamy; C, control promiscuity; E, elevated promiscuity; SD, standard deviation.
Response variableGeneralized linear mixed effects model using lme4 in RModel error structureComparisons among sexual selection treatmentsReplicate SD and residual SD, respectively
Number of inseminated femalesresponse∼treatment+ (1|replicate)quasi-Poisson (data overdispersed)E>M, t=6.5, P<0.00018.963 × 10−6, 0.401
   E>C, t=5.92, P<0.0001 
   M=C, t=0.57, P>0.5 


Our experimental evolution study found that the evolution of a sterile caste of sperm does not appear to be driven by sexual selection in D. pseudoobscura. We found no consistent divergence in sterile short sperm number and length among our sexual selection treatments. Additionally, we found no evidence of consistent divergence in long sperm number and length. Recently, Holman and Snook (2008) found that short sperm function to protect long sperm from female spermicide in this species but whether this function was maintained through natural or sexual selection was untested. The results presented here offer no support that the protective anti-spermicidal property of sterile sperm (Holman and Snook 2008) is maintained by sexual selection.

A caveat to these conclusions, however, is that we examined gamete characteristics using the first ejaculate of sexually mature males when abundant sperm reserves had probably been amassed. Such conditions may not reflect the natural circumstances of competitive matings given that costs of multiple ejaculate production can render males sperm depleted (Dewsbury 1982; Wedell et al. 2002). Relative testis mass has been typically used to estimate spermatogenic investment and is predicted to increase under promiscuity relative to monogamy (e.g., Short 1979; Birkhead and Møller 1998; Gomendio et al. 1998). However, relative testis mass did not differ among sexual selection treatments, providing no evidence that selection under female promiscuity promotes greater investment by males in spermatogenesis than monogamy.

Several explanations may be forwarded for the lack of divergence in spermatogenic-related traits. First, heritable variation for these traits may be low. However, although we lack specific information on the underlying genetic variance of these traits for our populations, we have shown in this and other work that a number of other male reproductive traits have been targeted by sexual selection (Crudgington et al. 2005; Snook et al. 2005). Second, sperm competition may not be a potent force in this species. This seems highly improbable, however, given that female multiple mating (Anderson 1974) and sperm storage occurs, and that second male sperm precedence approaches 0.8 (Badcock 2006). As sperm precedence values exceeding 0.5 are likely to indicate an important role for sperm competition (Simmons and Siva-Jothy 1998), selection for adaptations conferring sperm competitive success/avoidance seems inevitable.

Another potential reason for not observing the predicted responses in spermatogenic traits relates to the issue of evolutionary trade-offs. Relaxation of sexual selection under monogamy is predicted to drive a reduction in ejaculate production given that ejaculates are expensive (Dewsbury 1982; Wedell et al. 2002), with released resources being diverted to traits of greater priority. However, it is feasible that relaxation of sexual selection under monogamy failed to drive the predicted reduction in male gametic traits because an absence of sexual selection and relatively benign culturing conditions rendered monogamy males relatively well resourced and therefore able to maintain spermatogenic investment. Alternatively, given that the promiscuity regimes have favored males investing in premating traits (Snook et al. 2005), trade-offs between investment in pre- and postcopulatory traits may have prevented the predicted upregulation of spermatogenesis in these treatments relative to monogamy.

Despite the lack of divergence in sperm-related traits, we found that E males had a greater mating capacity than either M or C males, both when presented with sequential and simultaneous mating opportunities with coevolved females. A consequence of the enhanced sequential mating success of E males was a correspondingly greater number of progeny sired. This finding cannot be attributed to differences in female productivity among the three sexual selection treatments as there was no consistent among-treatment pattern in the number of offspring produced across the sequential matings (H. S. Crudgington and R. R. Snook, unpubl. data). Nor can this response be attributed to variation in testis size as this trait did not differ among treatments. We argue that differences in male mating capacity may be related to E males having larger accessory glands. The importance of large accessory glands to male reproductive success has been demonstrated previously in D. melanogaster (Bangham et al. 2002) and Crytodiopsis dalmanni (Baker et al. 2003; Rogers et al. 2005a,b). In both these species male mating capacity was greatest in males with larger accessory glands, but not larger testes. Sexual selection is an obvious candidate process for driving this relationship; our findings support this proposition as males with a history of elevated sexual selection evolved the largest accessory glands and had higher mating capacities, both when offered sequential mating opportunities and when confined with multiple females. Intriguingly, direct selection for rapid multiple mating ability in E males cannot account for the evolution of larger accessory glands because this regime entailed six males being housed with a single female. Thus, the probability of males encountering multiple receptive females is zero. Moreover, the probability of repeated rapid copulation with the single available female is very low given that females tend to be resistant to remating for several days after an initial copulation (N. S. Badcock and R. R. Snook, unpubl. data).

Elevated sexual selection could favor larger accessory glands if they (1) allow males to partition limited sperm numbers across more matings (Wedell et al. 2002), consistent with the idea that male mating capacity is more constrained by supplies of Acps than those of sperm (Hihara 1981), or (2) allow more Acps to be transferred per copulation, which may improve male fitness by modifying aspects of female biology (Wolfner 1997; Chapman 2001). Unlike the evolutionary significance of Acps in D. melanogaster (Chapman 2001; Wolfner 2002), the functions of Acps in D. pseudoobscura have not been studied. However, cross-species genetic comparisons show that over half the Acp genes found in D. melanogaster are detectable in D. pseudoobscura (Mueller et al. 2005) and similar patterns of positive Darwinian selection have been observed in these congeners regarding several Acp melanogaster subgroup orthologs (Dixon Scully and Hellburg 2006). It is therefore entirely possible that at least some functional overlap exists between D. melanogaster and D. pseudoobscura Acps and therefore that Acp production in D. pseudoobscura is sexually selected. It would be useful to examine whether, in addition to accessory gland size, Acp profiles or efficacies have also diverged according to variation in sexual selection intensity in our lines, and if so, what impact individual proteins have on the fate of competing ejaculates, as well as on female productivity and ultimately on male fitness.

Sexual selection theory is explicit regarding the action of postcopulatory sexual selection on the evolution of spermatogenic traits: investment in testicular tissue and spermatogenesis is expected to increase in the presence of sperm competition and decrease in its absence (Parker 1970a, 1984; Simmons 2001). Recently, a number of experimental evolution studies have addressed this question in insects and, given the considerable body of data amassed, it is now appropriate to ask how theoretical expectations fare in light of these studies. The short answer is, surprisingly, not particularly well. In line with predictions, the imposition of obligate monogamy in the yellow dungfly, S. stercoraria, drove the evolution of both smaller testes and poorer sperm competitiveness relative to promiscuity (Hosken and Ward 2001; Hosken et al. 2001). Interestingly, however, monogamous and wild-type males did not differ in testis size (Hosken et al. 2001) despite the ubiquity of sperm competition in wild populations (Parker 1970b). In D. melanogaster, enforced monogamous males evolved smaller testes and reduced sperm production as expected (Pitnick et al. 2001a) but there was no reduction in the sperm competitiveness of monogamous males relative to promiscuous controls (Pitnick et al. 2001a). In another D. melanogaster experimental evolution study in which males were maintained at varying sex ratios (and thus, varying sexual selection intensities), the predicted increase in testis size for males subject to greater sexual selection was not observed (Wigby and Chapman 2004). In contrast, another sex ratio manipulation in D. melanogaster found that when the population had an extremely female-biased sex ratio, males evolved larger testes (Reuter et al. 2008). However, a male-biased treatment was not imposed (Reuter et al. 2008) and thus the results may not be directly comparable to the other experimental evolution studies in which the manipulated sex ratio was primarily male-biased. In D. pseudoobscura, we show that experimental male-biased manipulation does not generate the predicted patterns of sperm and testicular investment.

These valuable datasets suggest that spermatogenic-related characters are not automatic targets of postcopulatory sexual selection, indicating a more variable, or even less prominent, role for these traits in determining male reproductive success. However, inconsistencies between theoretical predictions and empirical outcomes might simply be due to idiosyncratic differences in experimental design or insufficiently sensitive or naturalistic assays. Alternatively, they may arise when traits of interest fail to be directly targeted because selection regimes inadvertently drive trade-offs between, for example, pre- and postmating male reproductive traits. Testes size evolution also may be driven by male mating rate rather than sperm competition per se, such that in male-biased treatments, which increase sperm competition but correspondingly decrease male mating rate, testes size may not be targeted (Reuter et al. 2008).

As with testes size, experimental evolution studies report different responses in accessory gland size to sexual selection manipulation. In two D. melanogaster experimental evolution lines, male accessory gland size did not diverge according to male evolutionary histories of varying sex ratios (and therefore varying levels of sexual selection) (Wigby and Chapman 2004; Linklater et al. 2007; Reuter et al. 2008). Moreover, and unlike our data suggest for D. pseudoobscura, D. melanogaster males subject to a male-biased sex ratio treatment became sperm and accessory gland depleted after fewer matings than the equal- or female-biased sex ratio treatment and thus those males had poorer fitness (Linklater et al. 2007). In D. pseudoobscura, we found that more male-biased lines evolved larger accessory glands than less male-biased or equal sex ratio lines and that males experiencing greater sexual selection intensity could mate with more females, consequently gaining greater fitness. Whether variation in the responses of these two species represents a functional difference in the role of Acps awaits further dissection. In general, exploring inconsistent and unexpected responses to mating system manipulation, both within and among experimental evolution studies, in addition to examining the precise consequences of manipulating sexual selection in different ways, will undoubtedly improve our understanding of exactly how suites of traits interact to determine competitive outcomes among males and which selective pressures are responsible.

Associate Editor: T. Chapman


We thank L. Holman, A. Turner, K. Hutchence, L. Bacigalupe, L. Brüstle, K. Green, J. Hope, K. Melville, H. Ramm, and J. Edwards for all their help in the laboratory. Many thanks also to J. Rolff, A. Beckerman, R. Freckleton, S. Nakagawa, and D. Gillespie for their invaluable statistical advice. The manuscript was greatly improved by the comments of two anonymous reviewers and the associate editor. This research was supported by the US National Science Foundation grant and NERC grants to RRS.