EJACULATE QUALITY AND CONSTRAINTS IN RELATION TO SPERM COMPETITION LEVELS AMONG EUTHERIAN MAMMALS

Authors


Abstract

The outcome of sperm competition is influenced by the relative quantity and quality of sperm among competing ejaculates. Whereas it is well established that individual ejaculate traits evolve rapidly under postcopulatory sexual selection, little is known about other factors that might influence the evolution of ejaculates. For example, the metabolic rate is likely to affect the sperm production rate and the cellular activity or metabolism of sperm, and it has recently been suggested to constrain the evolution of sperm length in large but not small mammals. I thus examined in eutherian mammals how ejaculate quality traits vary with one another and with testis mass, body size, and metabolism. I found all ejaculate traits to covary positively with one another and to increase with relative testis mass. When controlling for testis mass, small-bodied species showed superior sperm quality (but not sperm number). Furthermore, sperm motility and viability were positively associated with the mass-corrected metabolic rate, but the percentage of morphologically normal and acrosome-intact sperm were not. These results indicate that body size and the energy budget may also influence the evolution of ejaculate quality, although these influences appear to vary among traits.

When more than one male mates with the same female and their sperm compete for fertilization, the resulting paternity share is typically biased toward the male transferring the most numerous and highest-quality sperm (e.g., Snook 2005; Pizzari and Parker 2009; Simmons and Fitzpatrick 2009). Consequently, sperm competition exerts selection on males to maximize both sperm quantity and quality (Snook 2005; Pizzari and Parker 2009; Simmons and Fitzpatrick 2012). Sperm quality is determined by a combination of variables, including sperm velocity, the proportions of motile, viable, structurally normal and acrosome-intact sperm, the sperm capacitation ability, and absolute and relative dimensions of different sperm components. These traits are essential for (1) successful migration of sperm to the egg(s) (particularly through the selective environment of the female reproductive tract in internal fertilizers); (2) sperm storage; (3) sperm capacitation; and (4) interaction with, and penetration of, the ovum (reviewed in Florman and Ducibella 2006; Suarez 2008; Holt and Fazeli 2010).

Polyandrous species tend to invest comparatively more in sperm production than monogamous species as indicated by their relatively larger testes (reviewed by Vahed and Parker 2012), denser sperm-producing tissue (Lüpold et al. 2009b; Rowe and Pruett-Jones 2011) and increased testicular function (Ramm and Stockley 2010; Lüpold et al. 2011). However, there is growing evidence that the production of sperm and ejaculates is energetically costly (e.g., Dewsbury 1982; Thomsen et al. 2006). Because the limited energy resources for sperm production have to be allocated between sperm quantity and different sperm-quality traits (e.g., sperm length), there are likely to be constraints in the evolution of ejaculates (Parker et al. 2010). Consistent with this idea, some studies have reported a trade-off between sperm size and number (Pitnick 1996; Oppliger et al. 1998; Immler et al. 2011), or among sperm-quality traits themselves, such as between sperm viability/longevity and sperm velocity/length (e.g., Birkhead and Fletcher 1995; Levitan 2000; Gage et al. 2002). However, several other studies have found no such trade-offs but rather positive covariation between different ejaculate traits (e.g., sperm length and velocity: Gomendio and Roldan 2008; Fitzpatrick et al. 2009; Lüpold et al. 2009a; sperm velocity and longevity: Kortet et al. 2004; sperm viability, motility, and morphological normality: Malo et al. 2005). More recently, two comparative studies across Muroid rodents and Australian fairy-wrens (Maluridae), respectively, have reported positive covariation among multiple measures of ejaculate quality or between ejaculate traits and relative testis mass (Gómez Montoto et al. 2011; Rowe and Pruett-Jones 2011). These results suggest that ejaculate quality traits may evolve in concert under sexual selection, possibly because the competitive fertilization success of an ejaculate is determined by a suite of different parameters (Pizzari and Parker 2009; Fitzpatrick et al. 2012; Lüpold et al. 2012). However, positive covariation among ejaculate traits could be found across species even if they were traded off within species. For example, species may vary in their overall resource allocation to ejaculates, with species under higher levels of sperm competition allocating relatively more resources toward ejaculates (Reznick 1985; van Noordwijk and de Jong 1986; Parker 1998).

One way that physiology may influence the evolution of ejaculates is by variation in the basal metabolic rate (BMR), which is the daily energy expenditure for somatic maintenance (e.g., Glazier 2005). By reflecting maintenance costs, a relatively high BMR might indicate that little energy is available for other processes such as reproduction at a given energy budget (“compensation hypothesis”; Blackmer et al. 2005). However, because much of the BMR is explained by the metabolic activity of the digestive organs, a relatively high BMR might also indicate a higher level of food ingestion and processing, which would increase the potential total energy turnover and ultimately permit greater reproductive investments (“increased intake hypothesis”; Nilsson 2002).

The energetic costs of ejaculates can be considerable. For example, the caloric content alone of the 1–6 daily ejaculates produced by male Japanese macaques (Macaca fuscata) has been estimated to represent 0.8–6% of the BMR (Thomsen et al. 2006). The total energetic expenses are expectedly even higher when the production of sperm and seminal fluid, or simply the maintenance of testes and accessory glands, are included (Thomsen et al. 2006). The testicular metabolic rate has been studied in only a handful of species (Setchell and Waites 1964; Ewing 1967; Härkönen and Kormano 1971), but under the assumption that it is proportional to whole-body BMR, Kenagy and Trombulak (1986) estimated the maintenance costs of the testes alone to range between about 0.2% and 10% of the total BMR. The sum of all metabolic costs related to ejaculate production thus might not be trivial, and as outlined earlier, it seems plausible that individuals or species with a relatively high BMR may have a comparatively greater energy budget to allocate toward reproduction, thereby being better able to increase investments in ejaculates in response to postcopulatory sexual selection.

Furthermore, the cellular metabolism of various highly active and rapidly dividing cell types is approximately proportional to the whole-body metabolism (Savage et al. 2007), mediated by the relative mitochondrial membrane surface area or oxidative enzyme activity in their cells (e.g., Porter 2001; Glazier 2005). A direct link between sperm and whole-body metabolism remains to be established but seems plausible given that sperm are also highly active and go through rapid cell divisions during spermatogenesis. A recent comparative study in birds suggests that sperm energy reserves (i.e., ATP) increase with the sperm midpiece length (Rowe et al. 2013; also see Vladić et al. 2002), whereas Tourmente et al. (2011) reported an increase in midpiece length (and other sperm dimensions) with the mass-corrected BMR among marsupial mammals. Although Rowe et al. (2013) did not find a direct relationship between ATP content to in vitro sperm swimming speed (but see Froman and Feltmann 1998), the combination of these studies provides at least some indirect evidence that species with a relatively high BMR might, on average, be better able to maintain high sperm motility and viability than others. Consistent with this prediction, Gomendio et al. (2011) and Tourmente et al. (2011) have shown that small mammalian species produce longer sperm than larger species and hypothesized that, due to more efficient energy uptake and transport or energy use by cells (e.g., Glazier 2005), small species with their higher mass-specific metabolic rate may be energetically less constrained to evolve longer sperm in response to sexual selection than larger species that exhibit a lower BMR relative to body size. However, further studies are needed to understand how different ejaculate characteristics might (co)evolve and how possible metabolic constraints might influence the evolution of ejaculates.

Using data from the literature on a wide range of eutherian mammals, I tested the two hypotheses that (1) ejaculate quality traits covary positively with one another and with the degree of sperm competition and that (2) these traits are also positively associated with the metabolic rate controlled for body size.

Methods

DATA COLLECTION AND TREATMENT

For a total of 173 eutherian species, I obtained published data on body mass, combined testis mass (CTM), BMR, and five ejaculate parameters, including the total number of sperm per ejaculate and the respective proportions of motile, structurally normal, viable, and acrosome-intact sperm (see Table S1). For ejaculate traits, each species was represented by 9.6 ± 0.1 males (± SE; range 1–168), and although species belonged to 10 mammalian Orders, most were Artiodactyla (N = 45), Carnivora (46), Primates (34), or Rodentia (27; Table S1).

Although sperm-quality traits are typically measured in separate subsamples and assays, it is important to note that they are not necessarily independent. For example, the proportion of viable sperm, measured by the exclusion (live cells) or permeation (dead cells) of specific stains through the sperm membrane, sets a ceiling to the proportion of motile sperm, with motile sperm being all progressively and nonprogressively moving sperm as quantified manually or by computer-assisted sperm analysis. However, assay sensitivity or measurement error in the compiled studies may occasionally have resulted in a higher proportion of motile than viable sperm. Similarly, in the case of structural normality and acrosome integrity, both of which are typically assessed by visual inspection under a light microscope, the subset of abnormal sperm includes those with acrosome defects. Additional abnormalities include but are not limited to a bent midpiece or flagellum, two sperm heads or flagella, or a disproportionately large or small head.

All data of the four sperm-quality traits were restricted to fresh as opposed to cryopreserved samples. The methods for these measures varied slightly among sources (e.g., manual versus automated, or different types of stains), which can be an issue in comparative studies (Amann 1981; Holman 2009). It was not feasible to evaluate the reliability of measurements in each study, but with nearly every species assayed by a different laboratory and each measurement technique represented in all major taxa (and thus in a range of size classes), I specifically assumed that variation due to assay reliability was unlikely to cause a systematic error that would jeopardize the conclusions. In contrast, I statistically controlled for the method of sample collection, because samples of 32 species were collected by dissection of the epididymis rather than by natural or electro-ejaculation, such that the absence or presence of seminal fluid could confound the results (see Analyses). Because epididymal samples were unlikely to reflect the amount of ejaculates, I restricted all analyses involving total sperm number to ejaculated samples. In some cases, the total sperm number was not provided but it could be calculated as the product of the ejaculate volume and sperm concentration.

Overall, not all data were available for each species, but for each species I obtained data of at least two different ejaculate traits (for among-trait correlations) or at least one such trait along with data on CTM or BMR and body mass. If data were provided for breeding and nonbreeding males or for experimentally treated and control animals, I used those for the breeding or control individuals, respectively. Where applicable, data from multiple studies were combined by calculating a weighted mean.

ANALYSES

I conducted all analyses using the statistical package R version 2.12.2 (R Foundation for Statistical Computing 2011) and transformed all nonnormal data distributions by logarithmic or arcsine transformations to meet the parametric requirements of the statistical models. I accounted for statistical nonindependence of data points by shared ancestry of species using phylogenetic general linear models (PGLM; Pagel 1999; Freckleton et al. 2002; script kindly provided by R. Freckleton). The phylogeny was based on a phylogenetic supertree of extant mammals (Bininda-Emonds et al. 2007), with additional species included following specific phylogenies (for details, see Figure S1). This PGLM estimates the phylogenetic scaling parameter λ, with values of λ not significantly different from 0 indicating phylogenetic independence, and λ not significantly different from 1 indicating a complete phylogenetic association of the traits. Using likelihood ratio tests, I established whether the model with the maximum-likelihood estimate of λ differed from models with values of λ = 0 or 1, respectively, and present the P-values of these tests as superscripts following λ. I report these results together with the (partial) correlation coefficients r and 95% noncentral confidence intervals (95% CI) calculated from the t-values of the PGLM (Nakagawa and Cuthill 2007). Assessing the strength of relationships based on r is also preferable to Bonferroni corrections for multiple testing, as it avoids the increased probability of committing type II errors (Nakagawa 2004).

For analyses of sperm competition, I included both CTM and body mass as predictor variables. As a measure of male investments in sperm production, relative testis mass is a widely used proxy for the degree of sperm competition and often the best index available. Among mammals, relative testis mass is correlated with the level of multiple paternity (Soulsbury 2010), thus supporting the assumption that it generally reflects the level of sperm competition (although testes might occasionally be enlarged in response to a high mating frequency rather than sperm competition; Vahed and Parker 2012).

I further examined whether ejaculate quality was explained by the BMR. Rather than using the mass-specific metabolic rate (i.e., BMR divided by body mass) as the predictor (see Gomendio et al. 2011; Tourmente et al. 2011), I conducted multiple regressions with BMR and body mass as two independent variables, for two reasons. First, if a trait is allometrically related with body mass (i.e., trait does not vary as a fixed proportion of body mass), expressing it as a ratio can introduce spurious correlations that are explained solely by variation in body mass (e.g., Tanner 1949; Nevill and Holder 1995; Packard and Boardman 1999). Basal metabolic rate exhibits a negatively allometric relationship with body mass (allometric coefficient approximately 0.75; Savage et al. 2004), so the use of the mass-specific metabolic rate as an independent variable can be statistically problematic and is therefore discouraged (e.g., Packard and Boardman 1999). Second, using BMR and body mass separately can reveal additional body size effects that are independent of metabolism.

I further controlled all models on sperm quality for potential influences by the method of sample collection (dissection vs. ejaculate). In most cases, however, sampling method had no significant effect, and removing this factor resulted in a lower score of Akaike's information criterion (AICc; controlled for small sample sizes). In these cases, I report the results of simplified models.

Results

Controlling for phylogeny and the method of sample collection (dissection vs. ejaculate), all ejaculate quality traits were positively correlated with one another (Table 1) and with relative testis mass (Table 2A).

Table 1. Phylogenetically controlled associations among different ejaculate traits. Numbers above the diagonal represent effect sizes r with 95% noncentral confidence intervals in parentheses, those below the diagonal indicate P-values with sample sizes in parentheses. The method of sample collection had no significant effect on any relationship (all P > 0.28), so results of bivariate analyses are reported
TraitTSNNORMMOTVIABACI
Total sperm number (TSN) 0.340.330.400.27
  (0.17–0.49)(0.16–0.47)(0.19–0.56)(0.04–0.46)
Sperm normality (NORM)0.0003 0.350.340.40
 (113) (0.18–0.47)(0.15–0.49)(0.22–0.55)
Sperm motility (MOT)0.007<0.0001 0.530.30
 (124)(135) (0.37–0.64)(0.10–0.46)
Sperm viability (VIAB)0.00020.0003<0.0001 0.30
 (78)(100)(101) (0.08–0.49)
Acrosome integrity (ACI)0.021<0.00010.0090.008 
 (73)(94)(92)(74) 
Table 2. Minimum adequate models of the associations of ejaculate quality with (A) combined testis mass and (B) basal metabolic rate, all corrected for phylogeny and body mass. The partial correlation coefficients r are presented with the lower (LCL) and upper 95% confidence limits (UCL). Unless the sampling method is presented, its effect was not significant (all P > 0.17) and the simplified model had a lower Akaike's information criterion (AICc) score (ΔAICc = 1.06–3.22)
TraitPredictordfr (LCL, UCL)tPλ1
  1. a

    Superscripts following the phylogenetic scaling parameter λ estimates denote significance levels of likelihood ratio tests (first superscript: against λ = 0; second superscript: against λ = 1).

  2. b

    Due to serious collinearity between BMR and body mass (all VIF > 30), data of sequential regression models (Graham 2003) are shown, with BMR as original variable and body mass as residuals of its regression on BMR (res. BM). The result of BMR should be interpreted as the BMR effect in addition to the contribution it already made through its relationship with body mass. For original models, see Table S2.

(A) Relative testis mass      
Total sperm numberTestis mass1200.31 (0.15, 0.46)3.640.00040.70<0.001, 0.004
 Body mass1200.12 (−0.05, 0.29)1.380.170 
Sperm normalityTestis mass1290.37 (0.22, 0.50)4.55<0.00010.36<0.001, <0.001
 Body mass129−0.31 (−0.45, −0.15)−3.710.0003 
Sperm motilityTestis mass1370.27 (0.10, 0.41)3.220.0020.140.01, <0.001
 Body mass137−0.21 (−0.36, −0.05)−2.530.013 
 Method137−0.18 (−0.33, −0.01)−2.530.033 
Sperm viabilityTestis mass930.30 (0.10, 0.46)2.990.0040.140.11, <0.001
 Body mass93−0.22 (−0.39, −0.02)−2.140.035 
Acrosome integrityTestis mass850.38 (0.18, 0.53)3.750.00030.350.008, <0.001
 Body mass85−0.30 (−0.47, −0.10)−2.900.005 
(B) Relative basal metabolic rate2
Sperm normalityBMR660.03 (−0.21, 0.26)0.250.8030.73<0.001, 0.19
 res. BM66−0.14 (–0.36, 0.10)−1.160.251 
Sperm motilityBMR690.18 (–0.06, 0.39)1.500.1380.140.31, <0.001
 res. BM69−0.13 (–0.35, 0.11)−1.100.275 
 Method69−0.31 (−0.50, −0.08)−2.720.008 
Sperm motilityBMR520.28 (0.01, 0.49)2.070.0440.620.04, 0.11
(ejaculates only)res. BM520.03 (−0.23, 0.29)0.250.806 
Sperm viabilityBMR470.31 (0.03, 0.52)2.220.031<0.0011.0, <0.001
 res. BM470.25 (−0.04, 0.48)1.740.088 
Acrosome integrityBMR420.08 (−0.22, 0.36)0.510.6160.040.90, 0.12
 res. BM420.08 (−0.22, 0.37)0.650.520 

In comparison to the positive relationships with relative testis mass, the above models also yielded negative body mass effects on all sperm-quality measures but not on sperm number (Table 2A). Thus, controlling for testis mass, relatively small species appear to produce higher-quality sperm than large species. Although CTM was strongly correlated with body mass (r = 0.72 [95% CI = 0.65–0.78], df = 154, t = 13.0, P < 0.0001, λ = 0.95<0.001, 0.37), there was no serious collinearity in the above multiple regressions, as suggested by variance inflation factors (VIF) of < 9.5 (i.e., below the threshold of 10; Kleinbaum et al. 1998).

To further explore the body size effects on the four sperm-quality traits, I tested whether they were explained by the BMR. The basal metabolic rate increased with body mass (r = 0.98 [0.97– 0.99], df = 80, t = 51.1, P < 0.0001, λ = 0.270.04, <0.001), with a negatively allometric slope of 0.74 (0.71–0.76). This slope did not deviate significantly (r = −0.13, df = 80, P = 0.24) from the theoretically predicted allometric slope of 0.75 (e.g., Savage et al. 2004) and indicates that large species exhibit a lower metabolic rate per unit mass than small species. Controlling for body mass, species with relatively large testes exhibited a relatively high BMR (CTM: r = 0.24 [0.02–0.43], df = 76; t = 2.20, P = 0.03; body mass: r = 0.91 [0.87–0.93], df = 76; t = 19.11, P < 0.0001; λ = 0.120.49, <0.001). Again, there was no serious collinearity among the predictor variables (VIF = 8.0).

Controlling for body mass, sperm motility and viability both increased with BMR, whereas the proportions of structurally normal and acrosome-intact sperm did not (Table S2). To remedy the serious collinearity (VIF > 30), I performed sequential regressions (Graham 2003), with BMR entered as the original variable and body mass as the residuals from a regression on BMR (i.e., after removing the variation of BMR). By regressing predictors, this approach differs from the correctly criticized regression of residuals (e.g., Freckleton 2002), in which the residuals of the predictor of interest are used in the second-step regression. In the sequential regression, the result of BMR as the focal variable is not interpreted as a direct effect but rather as its effect in addition to the contribution it already made through its relationship with body mass, and information on body size per se is lost (Dormann et al. 2013). This approach removed collinearity among predictors (VIF < 1.08) and largely confirmed the earlier results (Table 2B; Fig. 1). However, the relationship with sperm motility was statistically significant only when the samples collected by dissection were excluded after a highly significant effect of sample collection (Table 2B).

Figure 1.

Associations of residual metabolic rate with (A) sperm motility (N = 55; r = 0.28, P = 0.04) and (B) sperm viability (N = 50; r = 0.0.31, P = 0.03), with all axes controlled for body mass in sequential regression analyses (see main text). Relationship (A) is restricted to ejaculated samples (see Table 2B). Relationship (B) remains significant after excluding the two extreme values (Mus spretus top left and Elephas maximus top right; r = 0.42, P = 0.003).

Discussion

Across eutherian mammals, I found all measures of ejaculate quality to covary positively with one another and with relative testis mass. These results extend previous findings for Muroid rodents (Gómez Montoto et al. 2011) and Malurid birds (Rowe and Pruett-Jones 2011) across a wide range of mammalian taxa, suggesting that the coevolution among different ejaculate quality traits may be a more general pattern. That selection is likely to favor a suite of ejaculate traits collectively is consistent with Gómez Montoto et al.'s (2011) study indicating that a composite measure of sperm quality can be more tightly associated with relative testis mass than any parameter alone, but also with intraspecific studies showing that the adaptive significance of ejaculate quality is best explained in a multivariate framework (e.g., Fitzpatrick et al. 2012; Lüpold et al. 2012).

The factors driving the positive phenotypic covariation among ejaculate traits remain unclear. One possibility, however, might be genetic covariance (e.g., pleiotropy or linkage disequilibrium; e.g., Simmons and Moore 2011). Second, ejaculate traits may be functionally or developmentally linked, including cases where they depend on each other simply by definition (e.g., dead sperm cannot be motile). For example, although structurally abnormal sperm can show some level of motility, only sperm with normal morphology and a normal interaction with the female tract may be functionally motile to traverse the uterotubal junction, the primary barrier to the oviduct (e.g., Krzanowska 1974; Scott 2000). Third, some traits may evolutionarily constrain others, as suggested by Fitzpatrick et al.'s (2009) study on the relationships between sperm morphology and velocity in cichlid fish, in which the initial target of postcopulatory sexual selection may have been sperm energetics (e.g., ATP production) to increase sperm velocity, and a longer flagellum may then have evolved secondarily. Overall, the evolution of ejaculates involves numerous “moving parts,” including complex interactions between different ejaculate traits as also between sperm and their functional environment (e.g., Pitnick et al. 2009; Pizzari and Parker 2009).

Consistent with previous comparative studies (Gómez Montoto et al. 2011; Rowe and Pruett-Jones 2011), I found no evidence for an evolutionary trade-off among different ejaculate traits. However, this does not necessarily contradict Parker et al.'s (2010) models that predict trade-offs in resource allocation between ejaculate traits. Even in the presence of intraspecific trade-offs, one could find positive interspecific covariation in at least five different scenarios. First, sperm competition levels as measured by relative testis size covaried with the mass-corrected BMR, whether this was a direct link or mediated through another variable. A greater energy production capacity (or relaxed energetic constraints) might provide physiological conditions that facilitate the evolution of multiple ejaculate traits simultaneously if these are favored by postcopulatory sexual selection. Second, different ejaculate traits may (independently) coevolve with the female reproductive tract, for example in response to increased challenges imposed by highly polyandrous females to select for the most competitive ejaculate or to avoid the increased risk of polyspermy (reviewed in Pitnick et al. 2009). Third, the quality of sperm may be controlled during spermatogenesis, with apoptotic elimination of developing spermatids that exhibit abnormal or suboptimal characteristics (e.g., Aziz et al. 2007). If this filter becomes more rigorous due to stronger selection on overall ejaculate quality as the level of sperm competition increases (Lüpold et al. 2011), we could find positive phenotypic covariation between different ejaculate characteristics as well as between each trait and the sperm competition index. Fourth, the biochemistry of the sperm membrane, particularly the mixture of different phospholipids and fatty acids, is essential for sperm maturation, viability, and function (e.g., Sebastian et al. 1987; Connor et al. 1998). Any interspecific variation in this composition, particularly in response to postcopulatory sexual selection, might simultaneously have effects on several aspects of sperm quality. Finally, similar to the sperm membrane, the composition of the seminal fluid is important for various parameters of sperm function and plays a pivotal role in sperm competition (e.g., Poiani 2006). Therefore, the properties of the seminal fluid may vary among species in a way that increases the overall sperm functionality in response to sperm competition even in the absence of direct genetic or functional links between different sperm traits (Poiani 2006). Taken together, species under higher levels of sperm competition appear to have evolved superior overall ejaculate quality in a concerted manner, presumably by optimizing the various stages of manufacturing and maintaining competitive ejaculates.

Keeping testis mass constant, all measures of sperm quality (but not sperm quantity) decreased with body size. Although part of this size effect may be explained by the associated metabolic rate, body size itself may influence the evolution of ejaculates. A direct effect of body size could be expected if sperm tend to become increasingly diluted in a relatively large compared to small female reproductive tract (Harcourt et al. 1981; Short 1981). If so, large animals may be under selection for a compensating increase in ejaculate size. Theoretical models predict that if the spatial constraints around the site of fertilization are negligible, sperm size (as a sperm-quality measure) has only a small additional effect on ejaculate competitiveness with any increase in sperm density, such that selection should favor sperm number over sperm size (Parker et al. 2010). Conversely, in situations of strong spatial constraints (e.g., many insects), selection should be biased toward increased sperm size due to greater net gains compared to increasing sperm numbers (Parker et al. 2010). Recent empirical findings in passerine birds and drosophilid flies are consistent with both these predictions (Immler et al. 2011). If Parker et al.'s (2010) models also apply to other sperm-quality traits than sperm size, the negative covariation of body size with the four sperm-quality traits would lend at least partial support for these models, although a corresponding positive relationship with sperm number would be predicted but was not found. Alternatively, there could be differential selection on ejaculate traits among small or large species for other reasons, including potential differences in the patterns of postcopulatory sexual selection or different functional or evolutionary constraints that are yet to be determined. Further investigation is clearly needed.

Another possible link between body size and ejaculate quality derives from small animals having relatively higher cellular activity (e.g., Porter 2001), which might allow them to produce more competitive ejaculates. A study across shrews (Soricidae) suggests that a relatively high BMR accelerates sperm production and may thus increase the daily sperm output for a given testis size (Parapanov et al. 2008; also see Ramm and Stockley 2010 for a similar relationship based on body mass across a wider range of mammals). In addition, Gomendio et al. (2011) and Tourmente et al. (2011) suggested that small species might be energetically less constrained to evolve longer sperm in response to sperm competition. A similar pattern may also apply to other ejaculate traits, with small species being better able to respond to selection by evolving not just longer but also functionally better sperm.

In my comparative dataset, I found both sperm motility and viability to increase with BMR and to decrease with body mass when CTM was kept constant. These results are consistent with the prediction that species with small body size and species with a higher mass-corrected metabolic rate should be able to produce functionally superior ejaculates. However, the proportions of morphologically normal and acrosome-intact sperm were not associated with the mass-corrected BMR. I cannot exclude the possibility that this difference between ejaculate traits may at least partly be explained by the use of different species combinations among analyses due to data availability, but it may also reflect biological patterns. It seems plausible that sperm motility and viability, two traits that are directly linked to cellular metabolism and energetics, may be influenced more by variation in whole-body metabolism through its relationship with cellular metabolism (Savage et al. 2007). It now remains to be seen whether these associations of sperm motility and viability with BMR may be linked to the evolution of the mammalian midpiece, which increases with both sperm competition level and size-corrected metabolism (e.g., Tourmente et al. 2011).

In conclusion, a number of ejaculate traits were positively associated among one another and with the level of sperm competition as measured by relative testis mass. In addition to selection on sperm-quality traits themselves, it seems likely that these traits may also coevolve with both the seminal fluid and female reproductive tract. With testis size kept constant, all sperm-quality traits but not sperm number covaried negatively with body mass. Finally, sperm motility and viability increased with the relative metabolic rate, whereas the structural normality and acrosome integrity did not, suggesting a BMR effect primarily on ejaculate traits that may be more tightly linked to cellular activity and energetics. All these results combined add to the increasing number of studies highlighting the complex and multifarious nature of ejaculates, including various factors imposing selection or evolutionary constraints on them. Given the number of different parameters determining the overall quality of ejaculates, it is critical to understand the ejaculate evolution in a multivariate framework.

Associate Editor: A. Córdoba-Aguilar

ACKNOWLEDGMENTS

I thank W. T. Starmer and M. Ritchie for insightful discussions; R. Montgomerie, S. Pitnick, A. Córdoba-Aguilar, and three anonymous reviewers for valuable comments on the manuscript; and the Swiss National Science Foundation (fellowship PA00P3_134191) for funding. The author declares that he has no conflict of interest.

Ancillary