Do extra-group fertilizations increase the potential for sexual selection in male mammals?

Abstract

Fertilizations by males outside the social breeding group (extra-group paternity, EGP) are widespread in birds and mammals. EGP is generally proposed to increase male reproductive skew and thereby increase the potential for sexual selection, but the generality of this relationship is unclear. We extracted data from 27 mammals in seven orders and used phylogenetic comparative methods to investigate the influence of EGP and social mating system on measures of inequality in male fertilization success, which are indices of the potential for sexual selection. We find that EGP and social mating system can predict the potential for sexual selection in mammalian populations, but only when considered jointly and not individually. EGP appears to increase the potential for sexual selection but only when the degree of social polygyny is relatively low. When social polygyny is high, EGP appears to result in a more uniform distribution of reproduction and a decrease in the potential for sexual selection. A possible explanation to be investigated is that the phenotype of extra-group fathers differs systematically across social mating systems. Our findings have implications for the use of EGP and social mating system as indices of sexual selection in comparative analyses of trait evolution under sexual selection.

1. Introduction

Offspring fathered by males outside the social breeding pair or group (extra-pair paternity, or extra-group paternity EGP) are widespread in birds and mammals [1,2]. EGP can have large fitness consequences, for example, resulting in a strong skew in male reproductive success, even in socially monogamous populations [3,4]. A high skew creates the conditions for intense competition for fertilizations and a strong potential for sexual selection to act [5,6]. Accordingly, several conspicuous male traits in monogamous populations are thought to have evolved as a result of EGP generating high variance in male reproductive success [7,8]. EGP is generally thought to increase the variance in male reproductive success [4,9,10]. However, EGP can theoretically either decrease or increase variance depending on the correlation between within-group and extra-group paternity among males [6].

Tests of the generality of the cross-taxon relationship between the variance in fertilization success and EGP or even that between variance and social mating system are rare. In this paper, we examine these relationships using, as our measure, the variance in relative fitness, also called the opportunity for selection (I, [5]). Despite some disadvantages [11,12] this measure has an unambiguous meaning in selection theory [13]. The variance in relative fitness sets the upper limit to selection on a given trait and, therefore, can act as an index of the maximum potential for selection within a population. Using measures of reproductive success based on genetic markers, we focused on variation in reproductive success arising from variation in access to fertilizations rather than in survival to arrive as closely as possible to measuring the opportunity for sexual selection. We also used two other popular measures of skew to evaluate whether results were contingent on using I. We focused on measures of skew rather than on other measures, such as selection differentials and gradients, because we were not examining any particular trait but, more generally, characterizing the climate of sexual selection in a population.

We examined the relative strengths of the relationships between I and (a) extra-group paternity and (b) social mating system using phylogenetic comparative methods. We included social mating system because this is typically viewed as a major determinant of the distribution of reproductive success in a population; socially polygynous populations are expected to show higher male reproductive skew and sexual selection than do monogamous populations. We expected I (and the other measures of skew) to increase with EGP and with the degree of social polygyny.

2. Material and methods

We searched the published literature for annual reproductive success (the measure that approaches as close as the published literature currently allows to measuring variance in fertilization success) of males based on genetic measures of paternity (see electronic supplementary material for further details), and calculated (i) the opportunity for sexual selection (I, variance in male reproductive success divided by the square of mean male reproductive success), (ii) Ruzzante's Q [14] and (iii) Morisita's index [15]. We also extracted data on EGP, the proportion of offspring fathered by males outside the social breeding group, and the number of adult males and females in a breeding group.

We focused on species that form social breeding groups, because EGP is not defined for species where males and females associate very briefly. First, as a categorical measure of social mating system, based on overt associations between males and females, we classified the mating system of a population as: monogamy (one male–one female in breeding group), polygyny (one male–multiple females) and multi-male (multiple males–one or more females). Second, as a continuous measure of social mating system, we calculated breeding group sex ratio as the ratio of the mean number of adult females to that of adult males in a breeding group. This describes the mean number of females that a male associates with in monogamous and polygynous systems and the mean number of females per male for males in multi-male groups.

We used phylogenetic generalized least-squares modelling (PGLS, [16]) to account for potential non-independence among species due to common ancestry, and used the most recent mammal supertree [17] to construct phylogenetic relationships (figure S1). Because data on both EGP and I from the same species (N = 24) are fewer than data on I alone (N = 27), we first explored the relationship of I with social mating system and EGP separately. Next, we examined the relative relationships between I and social mating system and EGP by running a PGLS model with I (log-transformed to satisfy normality assumptions) as the response variable and breeding group sex ratio, EGP, and their interaction as predictors. We did not include categorical social mating system because it is moderately correlated with breeding group sex ratio and the latter is preferred because it provides finer detail. As a rough measure of model fit, we calculated the Pearson correlation between observed and predicted values of log I from the model. Additionally, we ran supplementary weighted PGLS models to check whether variation in sample size among studies might influence results and also examined for sources of potential bias. Finally, we relaxed the criterion of only using annual reproductive success measures to estimate I and carried out the analyses described above with the extended dataset (details in electronic supplementary material). All analyses were carried out in the statistical language R [18].

3. Results

I varied widely (0.4–7.45) among the sampled species (electronic supplementary material, table S1), which were widely distributed across the mammalian phylogeny and included 17 families in seven orders. In exploratory univariate analyses, I showed no large differences between categories of social mating system (PGLS, likelihood-ratio test χ² = 2.431, N = 27, d.f. = 2, p = 0.297, r (predicted − observed) = 0.29). I was also not consistently related to breeding group sex ratio (χ² = 0.345, N = 27, d.f. = 1, p = 0.557, r = 0.11, figure 1c) or to EGP (χ² = 0.744, N = 24, d.f. = 1, p = 0.388, r = 0.17, figure 1a) in univariate analyses.

Figure 1.

The relationship between the opportunity for sexual selection I and (a) EGP without considering breeding group sex ratio (BGSR); (b) EGP when BGSR is uniform (less than 1.1) (open circles, dashed line), i.e. weak polygyny, versus when BGSR is high (≥1.1) (closed circles, solid line); (c) BGSR without considering EGP; and (d) BGSR when EGP is low (≤20%) (open circles, dashed line) versus when EGP is high (greater than 20%) (closed circles, solid line). Lines in (b) and (d) are drawn using coefficients from PGLS models (table 1). EGP and BGSR were treated as continuous variables in PGLS models; they have been categorized here for visualizing the interaction. Lines were drawn using the median value of each category.

In analysis of their joint influence, EGP and breeding group sex ratio together explained considerable variation in I (table 1). The relationship between EGP and I was strongly modulated by breeding group sex ratio (note interaction term in table 1). I was positively related to EGP when breeding group sex ratio was low (weak polygyny), and switched to a declining relationship when breeding group sex ratio was high (strong polygyny) (figure 1b). Similarly, the relationship between I and breeding group sex ratio varied with EGP (figure 1d). Analyses of Ruzzante's Q and Morisita's index yielded similar findings (electronic supplementary material, tables S3 and S4), as did analyses that incorporated variation among studies in sample size and those with the extended dataset (electronic supplementary material, tables S2 and S5, and figure S2).

Table 1.

Phylogenetic comparative analysis of the relationship between I (log_e-transformed) and extra-group paternity and breeding group sex ratio. N = 24 species. Since the interaction was statistically significant, statistical hypothesis tests for main effects are not shown. r (predicted − observed) = 0.50.

	coefficient	95% CI	likelihood-ratio χ2	d.f.	p
intercept	0.177	−0.246 to 0.6
breeding group sex ratio (loge-transformed)	0.843	0.133–1.554
%EGP	0.015	0.0005–0.029
breeding group sex ratio (loge-transformed) × %EGP	−0.026	−0.048 to −0.004	5.631	1	0.018

4. Discussion

EGP was related to the potential for sexual selection, but contrary to expectations, the relationship was not consistent and, instead, varied dramatically with the degree of social polygyny. When breeding group sex ratios were relatively even (e.g. monogamous systems, multi-male systems with even sex ratios in breeding groups), EGP appeared to increase the potential for sexual selection. This positive association weakened and even switched when examined across species with an increasing level of social polygyny. That is, at the other extreme, when breeding group sex ratios were strongly female-biased (highly polygynous systems, multi-male systems with female-biased groups), EGP appeared to decrease male reproductive skew. Our findings thus imply an unsuspected relationship between overt (social mating system) and covert (EGP) mating tactics and the potential for sexual selection in male mammals.

A possible explanation for these results is that the dynamics of male and female covert mating tactics that result in EGP vary substantially, but predictably, in different animal societies. In societies where breeding group sex ratios are even (e.g. socially monogamous systems), females may show stronger preferences when choosing genetic mates than when choosing social mates [8]. For example, because of constraints, such as resource-limitation or biparental care, females may choose particular male phenotypes only through extra-pair copulations. This could lead to a skew in fertilization success favouring males of a particular phenotype, and an increase in this skew with increasing opportunities for extra-pair copulations. By contrast, in societies with strong social polygyny, male reproductive alternatives to the overt defence of females may be favoured. For example, in white-lined bats and in elephant seals, species with high levels of EGP, a common alternative male tactic is to remain on the periphery of harem groups and attempt to sneak-matings [19,20]. Furthermore, males typically adopt either the breeding group defence or the covert tactic and not both, either because of time and energy constraints or because of morphological/physiological/behavioural specializations that maximize the efficiency of alternative tactics. This would result in a re-distribution of fertilizations from the socially successful males to males pursuing covert tactics, and a decrease in variance with an increase in extra-group copulation opportunities. Information on the phenotypes of extra-pair fathers in different societies, needed to test this explanation, is still relatively scarce.

Although systematic phylogenetic comparative tests of the effect of EGP on I have not been conducted previously, a study of a fish species reports reduced I under high levels of sneak-mating [21], while studies from birds report that extra-group paternity typically increases the variance in reproductive success (although the magnitude of increase is debated, e.g. [4,10,22]). This result may be simply because most birds are socially monogamous, and analysing polygynous species may reveal patterns more similar to mammals.

Overall, our findings imply that social mating system and EGP, separately considered, are not effective indices of the overall climate of sexual selection experienced by a population. We may need to reconsider the manner in which they are used as indices of sexual selection in comparative analyses of trait evolution (typically done because measuring fertilization success is much more challenging than characterizing social mating system or even extra-group paternity, [10]).

Data accessibility

Data are available as electronic supplementary material.

Authors' contributions

K.I. conceived of the study; K.I. and S.S. conducted data extraction and analysis; K.I. wrote and S.S. made critical contributions to the manuscript, and both authors are accountable for the content and approved the final version of the manuscript.

Funding

We thank the Indian Institute of Science for supporting this research.