Sperm competition and the evolution of testes size in terrestrial mammalian carnivores

Introduction

Sperm competition, the competition between male ejaculates to fertilize ova (Parker 1970, 1998), is recognized as a strong selective force driving the evolution of behavioural and morphological sexual traits, including testes size. In general terms, the degree of sperm competition will be related to a species’ pattern of social organization (Parker et al. 1997), as this will dictate the number of males that mate with each sexually receptive female and the timing of matings relative to the female's fertilization ‘window’. Data across many taxa indicate that males have larger testes relative to their body size where sperm competition is high (e.g. Harcourt et al. 1981; Gage 1994; Møller & Briskie 1995; Stockley et al. 1997; Parker 1998). In recent decades, the advent of molecular analysis techniques has indicated that social mating systems cannot be considered a reliable proxy for genetic mating patterns (Birkhead & Møller 1996; Clutton-Brock & Isvaran 2006). Consequently, relative testes size (RTS) has been assumed to be a more reliable index of the degree of sperm competition experienced by males within a species than overt patterns of mating (e.g. Gage & Freckleton 2003), on the grounds that sperm expenditure correlates with RTS and that RTS increases with the degree of competition (Parker & Ball 2005).

Underlying behavioural and ecological factors associated with interspecific variation in extra-pair paternity, which will affect the degree of sperm competition, have been analysed extensively in birds (Griffith, Owens & Thuman 2002; Pitcher, Dunn & Whittingham 2005) and to a lesser extent in mammals (Clutton-Brock & Isvaran 2006; Isvaran & Clutton-Brock 2007). In birds, most of the variation in the degree of sperm competition is associated with opportunities for promiscuity (local breeding density and breeding synchrony), the need for paternal care, life-history patterns (reproductive lifespan; review in Griffith et al. 2002) and social mating system (Pitcher et al. 2005). Moreover, RTS correlates positively with levels of extra-pair paternity regardless of social mating system in birds (Møller & Briskie 1995) and with the incidence of multiple-sire litters in rodents (Ramm, Parker & Stockley 2005). A preliminary comparative analysis of mammals that form breeding groups indicated that extra-group paternity was inversely related to the length of the mating season and to the number of females in breeding groups but not to the social mating system (Isvaran & Clutton-Brock 2007).

Terrestrial carnivores represent a diverse taxon found in a range of ecological conditions, with contrasting patterns of reproduction and modes of sociality and with the widest body size range of any mammalian order (Gittleman 1985). Carnivores also exhibit great variation in male-female association and dispersion and, hence, in inter- and intra-sexual contact rates. For example, in predominantly solitary families such as the Felidae and Mustelidae, social polygyny or promiscuity are the typical patterns of mating (Sandell 1989). Conversely, in social families such as the Canidae, social monogamy is considered the typical mating pattern (Kleiman 1977). The reproductive mechanisms of some species, such as embryonic diapause and delayed implantation, are thought to increase male–male competition (Mead 1989; Birkhead & Møller 1993), whilst superfetation may extend the length of the female reproductive cycle, increasing the likelihood of sperm competition. The degree of sperm competition may also vary between spontaneous and induced ovulators (Kenagy & Trombulak 1986), since the first male to copulate with a female in species where ovulation is induced is likely to have a competitive advantage (Gomendio, Harcourt & Roldán 1998). Moreover, comparative evidence from white blood cell counts suggests that sexually transmitted diseases connected to mating promiscuity have shaped carnivore immune systems (Nunn, Gittleman & Antonovics 2003). Thus, a broad range of observations and evidence suggests that several ecological and life-history traits might affect carnivore sperm competition. However, previous inter-taxonomic comparisons of RTS in the Carnivora have been limited to a few species (n = 7, Kenagy & Trombulak 1986; n = 10, Anderson, Nyholt & Dixson 2004; n = 10, Ramm 2007).

In this paper, we use phylogenetic comparative methods to investigate the relationship between several predicted correlates of the degree of sperm competition in terrestrial carnivores and RTS, using RTS as an index of the level of sperm competition. Specifically, we test the hypotheses that the degree of oestrous synchrony (i.e. length of the mating season) and the average number of breeding attempts within an individual's lifetime are inversely related to the ability of males to monopolize females (e.g. Mauck, Marschall & Parker 1999) leading to an increase in RTS. We also hypothesize that spontaneous ovulation, as opposed to induced ovulation, is directly related to sperm competition levels (Kenagy & Trombulak 1986) and hence RTS is expected to be greater in spontaneous ovulators. Finally, we test the hypothesis that patterns of social organization (i.e. social mating system) do not reliably reflect the degree of male–male competition for access to females, and hence are a poor predictor of RTS.

Materials and methods

We collated data on testes size and body mass from the literature for species within all eight extant terrestrial carnivore families: Mustelidae, Procyonidae, Ursidae, Canidae, Hyaenidae, Felidae, Herpestidae and Viverridae (see Table S1 in Supplementary material). Unfortunately, there is a paucity of published information on testes mass per se. Therefore, to maximize sample sizes, we collated information on three measures of testes size: testes mass, testes and epididymes mass combined, and testes volume. However, it was evident that these measures were not recorded randomly across taxa. For example, testes mass was predominantly available for hunted species where carcasses were available for examination (e.g. Canidae, Mustelidae). Conversely, data for Felidae were mostly in the form of testes volume, having been derived from individuals in zoological collections involved in captive breeding programmes. Consequently, we (i) analysed each data set (i.e. testes mass, n = 40 species; testes and epididymes mass, n = 38 species; testes volume, n = 48 species) separately, and (ii) created a fourth data set, whereby testes mass for all species where these data were not directly available was estimated from regression analyses on subsets of data where two or more testes size measures had been recorded in the same study; the latter (n = 79 species) is referred to as the derived data set. As information on the pattern of ovulation was not available for all species, sample sizes vary between analyses. Only measures of combined testes size taken from healthy adult males at the peak of the reproductive season were included (Calhim & Birkhead 2007); any studies where it was not clear whether epididymes mass was recorded were excluded. In the testes and epididymes mass combined data set, two outliers were identified (Mustela erminea and Taxidea taxus); however, analyses run with and without these outliers were qualitatively equivalent so they are included in the results.

We gave preference to those studies where male body mass and testes size were both recorded. For studies reporting only testes size, body mass data were obtained for that species (or sub-species) from the same geographical region to account for possible intraspecific variation in body mass. To assess whether using body mass and testes size data from the same or different studies affected the results, we included a dichotomous variable in all preliminary analyses that coded whether these data had been recorded in the same study or had been amalgamated from a collection of studies. The results were, however, qualitatively unchanged by the inclusion or exclusion of this variable; for conciseness, we have not presented these results in this manuscript.

For each species, we compiled data on: length of mating season; reproductive lifespan in months (estimated as: residuals of the regression of [(longevity in captivity – male age at sexual maturity)/average male body mass] to correct for the allometric effect of body mass on longevity); social mating system (monogamous: one male, one female; polygynous: one male, multiple females; multi-male: multiple males, one or multiple females; after Isvaran & Clutton-Brock 2007); and type of ovulation (spontaneous, induced; Gittleman 1986; Nowak 2005; see also electronic supplementary material). Length of the mating season was taken as the number of months in which the testes of mature males showed active spermatogenesis. Where this information was not available, the mating season was estimated from known monthly patterns of birth rather than the number of months in which mating had been recorded, as the latter was often not known for lesser studied, that is, cryptic species. We calculated reproductive lifespan to assess whether an increased number of breeding attempts promotes sperm competition. For example, a small short-lived mammal may have multiple breeding attempts within a year, and hence experience higher sperm competition than a large long-lived mammal with fewer breeding attempts within a year and more across years. Longevity in captivity was used as a proxy for longevity in the wild as more complete data were available for the former. In addition, evidence from zoo data shows that age of reproductive cessation is positively related with species longevity in captivity (Ricklefs, Scheuerlein & Cohen 2003), and so provides a reliable indication of the maximum possible number of breeding attempts an individual can expect to have in a lifetime. To distinguish between polygynous and multi-male social mating systems in species that do not form breeding groups (e.g. roving strategies, scramble competition, etc.), we used the available data on intra-sexual territoriality.

Due to shared ancestry, the phylogeny of species must be accounted for in comparative analyses (Felsenstein 1985; Harvey & Pagel 1991). We used a multiple regression analysis based on a phylogenetically corrected general linear model (PGLM). This method is equivalent to a generalized least-squares regression where phylogenetic dependence of data is incorporated into the error structure using a maximum likelihood framework (Pagel 1999; Freckleton, Harvey & Pagel 2002). In the course of these analyses, an index of phylogenetic dependence (λ) was calculated; λ varies from 0 (phylogenetic independence of traits) to 1 (phylogenetic dependence where the trait observed evolved under the Brownian model of evolution). In PGLM, λ is set to its maximum likelihood value rather than assuming clear-cut phylogenetic dependence/independence of data (Freckleton et al. 2002). We used the untransformed branch lengths in the phylogenetic tree published by Bininda-Emonds, Gittleman & Purvis (1999) to estimate phylogenetic distances. We assumed that the final divergence of M. sibirica itatsi, Felis catus, Canis familiaris dingo and Herpestes auropunctatus occurred at the same time as M. sibirica, F. silvestris, C. lupus and H. javanicus, respectively.

We selected variables to be included in the PGLM full models on the basis of their predicted a priori biological relevance, thereby avoiding the problems associated with stepwise procedures (Whittingham et al. 2006). The order in which the variables were entered into a PGLM reflected their relative influence on the dependent variable. Therefore, to remove the allometric effect of body size on testes size, we entered body mass first into the full models. We were also interested in assessing whether ovulation type would affect the relationship between testes size and length of the mating season (Gomendio et al. 1998). Therefore, an ovulation type × length of the mating season interaction term was included in each full model. Where data on specific variables are missing, the PGLM deletes species data in a case-wise fashion. Thus, in the results, model sample sizes reflect the maximum available data where all variables are present for each species.

To account for multiple statistical testing and to estimate the accuracy of our predictions, we estimated effect sizes (correlation coefficient r, sensu Cohen 1977) and non-central confidence intervals (CI) from t values obtained from PGLMs (Nakagawa & Cuthill 2007) using a function written by S. Nakagawa (available from the author). All variables were logarithmically (ln) transformed, except length of the mating season (inverse square-root transformation) to meet the underlying assumptions of the statistical tests. All analyses were run on the statistical package ‘r’ v2·5·0 (R Foundation for Statistical Computing 2007) using the APE package and an unpublished function written by R. Freckleton (available upon request).

Results

Testes mass was significantly highly correlated with both testes volume (F_1,6 = 617·7, P < 0·0001, R² adj. = 0·989, n = 7 studies: ln(testes mass) = 1·06 ln(testes volume) – 0·328) and testes mass and epididymes mass combined (F_1,23 = 7596·47, P < 0·0001, R² adj. = 0·997, n = 25 studies: ln(testes mass) = 0·995 ln(testes and epididymes mass) – 0·225). In the derived data set (n = 79 species), average testes mass, expressed as phylogenetically corrected residual testes mass/body mass, varied significantly across carnivore families (GLM: R² = 43·7%, F_1,7 = 7·86, P < 0·0001; Fig. 1), being highest in the Mustelidae (mean = 0·39, range = −0·02 to 0·76, n = 23 species) and Canidae (mean = 0·33, range = 0·18−0·58, n = 16 species) and lowest in the Felidae (mean = –0·68, range =–1·09 to –0·32, n = 22 species) and Hyaenidae, although the latter was not significantly different from the other families (mean =–0·86, range = –1·12 to –0·66, n = 3 species; Fig. 1).

Reproductive lifespan and mating system were not significantly related to RTS in any of the four data sets (Table 1a–d). For instance, in the largest derived data set (n = 79 species), there was no significant difference in average testes mass expressed as phylogenetically corrected residual testes mass/body mass among monogamous (mean = 0·11, range = –0·31 to 0·52, n = 18 species), polygynous (mean =–0·13, range = –0·54 to 0·09, n = 31 species) or multi-male species (mean = 0·08, range = –0·53 to 0·79, n = 28 species; GLM: F_1,2 = 0·89, P = 0·41; Fig. 2). In all four data sets, λ was significantly different from zero suggesting that, in contrast to extra-group paternity in mammals (Isvaran & Clutton-Brock 2007), phylogeny has played a significant role in the evolution of testes size in terrestrial carnivores (Table 2).

The length of mating season was significantly negatively related to three of the testes size measures (Table 1a,c,d) but not to testes and epididymes mass combined (Table 1b). The relationship between testes size and length of mating season was further significantly affected by an interaction with ovulation type (Table 1a–d). Species with shorter mating seasons were found to have larger RTS than species with longer mating seasons (Table 1a,c,d) and the length of the mating season had a greater effect on RTS in spontaneous ovulators (Fig. 3a–c), suggesting that sperm competition is more pronounced in spontaneous versus induced ovulators. For the testes and epididymes mass combined data set, these relationships were not significant (Table 1b). Similarly, the pattern of ovulation was also significantly negatively related to testes volume and derived testes mass but not to testes mass or testes and epididymes mass combined (Table 1a–d). However, interpreting the main effects of an interaction term is not straightforward where the interaction term significantly affects the dependent variable, as in the case of all data sets except the testes and epididymes mass combined.

Discussion

Length of the mating season and ovulation type individually and in interaction explained most of the variation in RTS in terrestrial carnivores. Kenagy & Trombulak (1986) predicted that in species with shorter mating seasons, males should invest relatively more in spermatogenic tissue because of a higher demand for sperm production. As predicted, the length of the mating season significantly explained variation in RTS in all data sets except testes and epididymes mass combined. However, the interaction between length of the mating season and ovulation type also explained variation in RTS; for both induced and spontaneous ovulators within these three data sets, RTS increased as the length of the mating season decreased. It is difficult to tease apart main effects from their effects in interaction, and in all three data sets both length of the mating season and ovulation type affected variation in RTS. These results are, therefore, consistent with the hypothesis that increased female reproductive synchrony (shorter mating season) promotes male intra-sexual competition for receptive females (Emlen & Oring 1977) and male investment in spermatogenic tissue (Kenagy & Trombulak 1986). This is further supported by comparative evidence that extra-group paternity is inversely related, in part, to the length of the mating season (Isvaran & Clutton-Brock 2007) but also to male mate guarding behaviour (Clutton-Brock & Isvaran 2006).

In three of the four data sets analysed, RTS was more significantly affected by changes in the length of the mating system in spontaneous than in induced ovulators. Whilst previously predicted (Kenagy & Trombulak 1986), to our knowledge this is the first study to show this within mammalian taxa. As copulation triggers ovulation in induced ovulators, the first male to copulate with a female will fertilize most of her ova (Birkhead 2000). Males encountering a female that has already been mated may elect not to mate with her as chances are high that she will already be pregnant (Schwagmeyer & Parker 1990). Moreover, induced ovulators may enhance their likelihood of assuring paternity by mate guarding after copulation (Gomendio et al. 1998). In contrast, in spontaneous ovulators the male copulating closest to ovulation generally sires most of the offspring (Birkhead 2000). Consequently, the likelihood of multiple copulations is greater in spontaneous ovulators, leading to greater sperm competition and larger RTS. Our findings in the context of this interaction with ovulation type need, however, to be viewed as preliminary as data on the pattern of ovulation were not available for several carnivore species. Nonetheless, ovulation type appeared as a significant explanatory variable in three of the four data sets, strongly indicating that it is an important parameter contributing to the observed variation in RTS of mammalian carnivores. The implication for studies on other mammalian taxa is that ovulation type needs to be incorporated more often in comparative analyses to elucidate its effect on sperm competition.

After correcting for phylogeny and for the allometric effect of body mass, residual testes mass differed significantly among carnivore families, being highest in the Canidae and Mustelidae, and lowest in the Felidae and, albeit not significantly, Hyaenidae. As hypothesized, however, this variation in RTS was not strongly correlated with social mating patterns based on behavioural observations in these taxa, for example, both the Mustelidae and Felidae are broadly polygynous and solitary (Sandell 1989; Johnson, Macdonald & Dickman 2000), whilst the Hyaenidae vary in their degree of sociality and mating systems (Gittleman 1989). The relatively large testes size of the Canidae suggest that, despite being regarded as monogamous with paternal care (Kleiman 1977), male intra-sexual competition is likely to be higher than estimated by observational studies on behavioural mating patterns and subsequent parental care. Most canids are in fact spontaneous ovulators and monoestrous with short mating seasons, leading to high sperm competition levels. Indeed, recent genetic studies within this taxon have revealed high levels of extra-pair paternity (Roemer et al. 2001; Baker et al. 2004). This confirms the findings of previous comparative studies in a range of taxa (voles, Heske & Ostfeld 1990; birds, Birkhead & Møller 1996; mammals, Clutton-Brock & Isvaran 2006; Isvaran & Clutton-Brock 2007; but see primates Harcourt et al. 1981), indicating that inferring levels of sperm competition from social mating patterns can be misleading (Griffith et al. 2002).

The relationship between reproductive lifespan, that is, the number of reproductive attempts a male may experience in his lifetime, and sperm competition, is complex. According to theoretical models, short-lived species should tolerate (and thus experience) higher levels of extra-pair fertilizations (i.e. sperm competition), than long-lived species since it pays short-lived males to achieve as many fertilizations as possible (Mauck et al. 1999). However, recent experimental evidence from a short-lived passerine shows that there are reproductive advantages to long-term monogamy (Adkins-Regan & Tomaszycki 2007). In the current study, we found no evidence of an association between reproductive lifespan and RTS, suggesting that RTS in terrestrial carnivores has been shaped by male intra-sexual competition within a reproductive bout (one mating season) rather than over a reproductive lifespan.

Three of the four data sets showed consistent patterns regarding RTS, mating season and ovulation type. However, one data set, testes mass and epididymes mass combined, did not. In mammals the epididymis serves three main functions: maturation, storage and transport of spermatozoa (Jones 1999). It has also been proposed that the epididymis plays a major role in sperm competition. The amount of sperm stored in the epididymis varies considerably between species, as does the number of ejaculates that may be delivered daily (Jones 1999). Recent studies have shown that in monotremes sperm form bundles during the passage through the epididymis, and that bundle swimming velocity is threefold that of individual sperm, strongly suggesting a sperm competition role (Jones et al. 2007). In our study, epididymes mass varied between 10·7% and 46·2% of carnivore testes mass; unfortunately, the number of species for which we had epididymes data (excluding testes mass) was limited and we could not examine this trait alone. The inclusion of the epididymis appears to have a confounding effect on the analysis of variation in RTS, perhaps because of different selective pressures on the epididymis as opposed to the testis (e.g. see Anderson et al. 2004). Alternatively, some of the data contained in the data set were wrong, but exclusion of two outliers (M. erminea and T. taxus) did not change the results, so we think this is unlikely. It is clear, therefore, that more studies on the competitive function of the epididymis are needed to shed light on its role in mammalian sperm competition and that data are required from more species.

In contrast to the findings of a preliminary study on levels of extra-group paternity in mammals that form social breeding groups (Isvaran & Clutton-Brock 2007), we found a significant phylogenetic signal in all four data sets (0·54 < λ < 0·92), indicating that phylogeny has played a significant role in the evolution of carnivore testes size, and hence in shaping sperm competition in carnivores. These contrasting results may have arisen, however, from the fact that the two studies have considered the degree of sperm competition at two disparate levels. Analyses based on measures of extra-group paternity are more dependent on local conditions (i.e. breeding density, oestrous synchrony) and less on phylogeny (Griffith et al. 2002; Isvaran & Clutton-Brock 2007). Conversely, the assessment of sperm competition by measuring testes size (current study) analyses the ultimate evolutionary outcomes of that competition, and is more likely to identify a phylogenetic signal.

In summary, we found widespread variation in the RTS between terrestrial carnivores which was not associated with individual species’ social mating systems or reproductive lifespan. In contrast, we found that the length of the mating season was negatively associated to RTS, with carnivores with the shortest mating seasons exhibiting the largest RTS. This is consistent with the hypothesis that male intra-sexual competition, and hence sperm competition, is highest in these species. Moreover, this relationship was stronger in spontaneous than in induced ovulators. Additional data on ovulation type and reproductive behaviour of carnivores with contrasting patterns of ovulation need to be gathered to further our understanding of the interaction of this parameter and the length of the mating season. In particular, data from the complete spectrum of carnivore species are required to avoid potential biases that may arise from the over-representation of closely related species. All the above results were not significant when testes and epididymes mass combined were analysed, suggesting possible contrasting roles of the epididymis and the testis in sperm competition in terrestrial carnivores: at present our knowledge of the epididymis role in sperm competition is too limited to draw any conclusions. The analyses of all four data sets support the hypothesis that phylogeny has played an important role in shaping carnivore testes size.