Introduction

Female promiscuity is a feature of the breeding system of most passerines, often leading to extrapair paternity (e.g. [1]). Therefore, fitness in males that breed in social pairs but engage in extrapair copulations is the combined success of both extrapair and within-pair siring (e.g. [2]). Sperm competition theory proposes that males are selected to both protect from and overcome a rival male's paternity assurance mechanisms, and that traits involved in paternity defense and offense might be under antagonistic forces [3]–[5]. Several studies have looked at inter-male variation in a given trait and its concurrent and separate effect in within-pair and extrapair reproductive success in birds, yet these studies are largely limited to secondary sexual traits (i.e. plumage, song) or age (reviewed in Table S1, n = 20 species). In many cases, the trait has a directional effect on extrapair paternity but no effect on own-nest paternity, with two notable exceptions: crown ultraviolet hue in blue tits (Cyanistes caeruleus, [6]) and tail length in cape sugarbirds (Promerops cafer, [7]) have opposite effects. Focus on primary sexual traits and male copulation roles has largely been restricted to empirical studies of controlled matings in invertebrates (e.g. [8]–[10]), social status manipulation in domestic fowl (e.g. [11]), hormonal manipulation in a wild passerine [12] and alternative mating tactics in centrarchid fish (e.g. [13], [14]).

Most long-term studies of birds have difficulty measuring lifetime measures of fitness, due to low assignment of extra-pair sires and/or tracking fitness of dispersers (e.g. [1], [15]). Yet, accurate measurement of selection in the wild is best achieved when not restricted to a spatio-temporal snapshot of an otherwise well studied system [16], [17]. Lifetime paternity success provides a direct assessment of differential pre- and postcopulatory success [18]. However, using data from natural, unobserved matings has inherent unknown confounding effects, such as inter-male variation in sperm competition risk (likelihood of being cuckolded), relative timing and number of copulations for a given clutch [19]. The detailed knowledge of the unusual breeding biology of superb fairy-wrens (Malurus cyaneus) make this a good species to investigate the evolutionary consequences of sperm morphology variation on extrapair and within-pair siring success in a wild bird. Superb fairy-wrens are territorial, facultative cooperative breeders, with a permanent social pair bond [20], [21]. Males are philopatric and both sexes acquire their social mates passively rather than through selection on phenotype [21]. Males queue for dominance on their natal or adjacent territories [22], while females compete for rare vacancies through dispersal [21]. By contrast, females show strong precopulatory mate choice of extragroup sires, preferring males that acquire the nuptial plumage early, an honest signal of male quality [23]–[26]. Regardless of the quality of their own social partner (e.g. when it molted), all females make pre-dawn extragroup forays to mate with these preferred males, usually two or three days before egg-laying [27], leading to extragroup paternity in almost all broods [28]. Helpers (and/or neighbors) of attractive males gain some of these extragroup fertilizations [29], suggesting a hidden lek effect of dawn chorus displays [30] and/or ‘error-prone’ female choice [29]. Remarkably, unless socially paired to a son, females always copulate with their social partner within thirty minutes of returning from the extragroup foray, and/ or with unrelated helpers (Cockburn & Double, unpublished data). Thus sperm competition is pervasive, a view supported by the relatively large size of the testes, proportion of sperm producing tissue and cloacal protuberance [31]–[33].

Passerine sperm is morphologically complex, with helical shaped heads, a large, single fused mitochondrion wrapped around the flagellum ([34], [35], but see [36] for an exception), and move by rotation along the main axis [37]. Both passerine sperm form and function exhibit considerable additive genetic variance and generally show low condition-dependence ([38]–[40], but see [41] for an exception). Comparative work in birds has shown that, as sperm competition increases, mean sperm length increases, although not linearly [42], [43], and both inter-male and intra-male variation in sperm size is reduced [44]–[46]. Recent studies suggest that while there are significant associations between sperm morphology (absolute flagellum length and relative to head length) and sperm motility across species ([47], but see [48]), these can become uncoupled at the intraspecific level in taxa, especially in species under high sperm competition ([40], [49]–[51] but see [52]). It is therefore reasonable to assume that in taxa under strong sperm competition, one might find certain morphometric traits to be associated with cuckolding success and/or defense. We studied cuckolding success and defense in the superb fairy-wren.

Methods

This study was conducted on an intensively studied color-banded population at the Australian National Botanical Gardens, Canberra [21], [22], in which the breeding history, age and parentage are known for most birds [21], [23], [27]. Ethical approval for this research was given by the Australian National University Animal Experimentation Ethics Committee (permits: F.BTZ.63.04 and F.BTZ.06.07).

Cuckolding success was measured as the lifetime number of illegitimate offspring that survived to four weeks after fledgling (when census provides the most reliable fitness measure; A. Cockburn, pers. comm.). Cuckolding success was computed as a categorical variable using the total number of extrapair offspring a male sired (see Fig. S1). Since the frequency distribution can be interpreted as bimodal, with modes at zero and 3–4 young (which correspond functionally to siring no young or a single brood of extrapair chicks), and a lowest point at 2 young (Fig. S1). Failing to acquire an extrapair copulation (EPC) or to convert an EPC into an extrapair fertilization, which are impossible to differentiate in our data, reflect low competitiveness. Also, since EPC forays are so time-restricted [27], one can assume that there is little between individual variation in the number or EPCs per bout. Therefore, males producing a single extrapair young can be considered to have low success as a sperm competitor. We therefore managed our data distribution by analyzing cuckolding success as a binomial response, where the two classes were 0–1 and 2+ young. Using this cut-off rather than 0–2 versus 3+ young is further supported, as (i) one of the two males producing two EPO gained 100% success in the brood in which it obtained extrapair paternity and (ii) the other was a subordinate individual that successfully cuckolded within its social group. Therefore, both males can be considered successful sperm competitors. Because male age is a strong predictor of extrapair fertilization success, through its positive effect on the nuptial plumage molt date [23], [26], male breeding experience (i.e. number of breeding seasons it was alive) was included as a covariate in the Generalized Linear Model (glm function with logit link and (quasi)binomial error distribution; n = 59 males). In order to assess the robustness of this analysis, it was repeated using GLM with negative binomial errors (glm.nb function), an alternative interpretation of Fig. S1 distribution (see Supplementary Material for more details). We present the results of the latter in the Supplementary Material). Cuckolding defense success refers to the proportion of sired fledglings in all broods sampled while a given male was dominant. Broods assigned to males while in an incestuous pair with their own mothers were excluded (n = 4 broods), since inbreeding avoidance by females may bias within-nest paternity success [21]. We used Generalized Linear Mixed Models, GLMMs (lmer function with logit link and binomial error distribution), to estimate the fixed effects of sperm morphometric traits, with male identity incorporated as a random factor. Helper number was included as a covariate, since dominant males without helpers are cuckolded less, possibly since paternity assurance increases care [21], [23], [28]. One measure of sperm defense per male (average number of sired young across broods) was not appropriate since the total number of fledglings, female identity and helper number differed across broods. A total of n = 255 broods belonging to 47 males was used.

Sperm samples were collected non-invasively by collecting the liquid part of the faeces (see [53]) from n = 59 adult males in December 2005. This method [53] has been shown to provide reliable sperm morphometry data. Sperm morphometry was measured using digital imaging software (Leica IM50) and photographs taken using light microscopy. Three independent sperm traits were directly measured (flagellum, head and straight midpiece lengths) and three composite traits were calculated (total length, flagellum∶head and midpiece∶flagellum length ratios) to the nearest 0.1 µm (for more details see [38] and [47]). In order to minimize autocorrelation between predictor variables, independent and composite sperm traits were tested separately (note that independent traits were not correlated with each other, r<|0.17| and p>0.2). Although five sperm per male are generally used to describe sperm morphometry in passerine birds ([38]), we included all sampled males in the analysis irrespective of number of sperm measured each (mean = 7 sperm per male, range = 1 to 10; see Fig. S2) since this species has low intra-male variation (Table S2), reasonable intra-male repeatability ([45]; Table S2), and a single sperm captures c. 70% of this intra-male variation (see [54]; Fig. S3). In fact, the superb fairy-wren shows a two fold difference between inter- and intra-male coefficient of variation for sperm length, one of the highest for which comparable variation indices are available (inter∶intra CV ratio = 1.9; range in 26 species = 0.8 to 2.1; [44]–[46]). Nonetheless, we conservatively weighted all analyses by sampling effort category (low weight given to males with fewer than five sperm measured). Although we cannot report across-year repeatability in sperm traits, it is unlikely that sperm sampling restriction to 2005 would significantly confound our results for the following three reasons. First, in contrast to the only study that has shown environmentally-induced plasticity in sperm morphology in a passerine [41], there is no evidence that primary sexual traits and hormonal profiles at the time of sperm production differ across males of different social status or social group in this species (e.g. [25], [31]). Second, we found no variation in any sperm morphometric trait with respect to age, status or group composition at the time of sampling (MANOVA, p>0.2, Table S3). Third, evidence from a two passerine species suggests considerable within-individual across-year repeatability (Agelaius phoeniceus, S Lüpold, pers. comm.; Troglodytes aedon, E. Cramer, pers. comm.). Nonetheless, all analyses were repeated with the subsample of males with at least 5 sperm measured (see Table S4).

Finally, the inclusion of males that are potentially still reproductively active past the 2008–9 breeding season, the last year we have completed paternity assignment data (n = 20/59 and n = 17/47 males for extrapair and within-pair success, respectively) create a further and potentially confounding factor in our sample. We highlight the possibly biased data points in the Figures and provide the results using a restricted dataset excluding those males in Table S4. Model simplification was achieved using stepwise removal of terms and comparison of alternative models' fit using likelihood ratio tests [55]. Quasibinomial error structure was used in cases where overdispersion needed to be accounted for [55]. All analyses were conducted in R version 2.10.1 (R Development Core Team). Effect sizes and their 95% confidence intervals [56] were calculated for each alternative data subset's minimal adequate final models, based on the standardized variable methods proposed by [57] (see Table S5).

Results

After controlling for the positive effect of male breeding experience (i.e. breeding season number), success at siring extrapair offspring declined with both flagellum length (Fig. 1; GLM with quasibinomial errors; season number: estimate (± s.e.) = 0.98±0.27, z = 3.64, p = 0.0006; flagellum length: estimate (± s.e.) = −0.67–44.83±16.56, z = −2.71, p = 0.009; n = 59 males) and flagellum∶head length ratio (Fig. 1; GLM with quasibinomial errors; season number: estimate (± s.e.) = 0.90±0.26, z = 3.43, p = 0.001; flagellum∶head ratio: estimate (± s.e.) = −18.59±9.23, z = −2.01, p = 0.049, n = 59 males). Therefore, cuckolding success was associated with sperm with a shorter flagellum and relatively larger heads. The alternative negative binomial GLM method shows these results are robust (Table S4). Removal of males that could be reproductively active past the end date for paternity data (n = 12 males who could change from low to high extrapair success late in their lives), did not qualitatively change the results (flagellum length p = 0.018; flagellum∶head ratio p = 0.048; Table S4 and Table S5). The same was true for the restriction of the dataset to males with at least five measured (Table S4 and Table S5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Negative associations between extrapair reproductive success and sperm morphology.

(A) flagellum length and (B) relative length of flagellum to head section. Extrapair paternity success was transformed into a categorical variable based on the observed binomial frequency distribution peaks of fledged extrapair sired young (low = one or fewer, high = two or more). Males that are potential active breeders past the date of the current paternity assessment (2008/9 season) are represented by the open circles (n = 20 of the total n = 59 males). Fitted curves were calculated using the regression estimates from fitted models (male breeding experience included as a covariate).

https://doi.org/10.1371/journal.pone.0028809.g001

In contrast, males with sperm with a longer flagellum and relatively shorter heads were more successful at preventing cuckoldry (Fig. 2; binomial GLMMs with male identity as a random factor; flagellum length: estimate (± s.e.) = 20.08±8.31, z = 2.42, p = 0.016; flagellum∶head ratio: estimate (± s.e.) = 11.50±4.31, z = 2.67, p = 0.008; n = 255 broods assigned to 47 males). The number of helpers did not affect within-brood paternity success in this sample (p>0.2). Although our measure of cuckolding avoidance (proportion of fledgling sired in all assigned nests) can be biased either way by future potential breeding attempts (n = 17 dominant males alive past the last available paternity analysis), removal of these data points did not change any of the previous results (flagellum length p = 0.03; flagellum∶head ratio p = 0.008). Note, however, that restriction of the dataset to males with at least five measured sperm considerably decreased effect size estimates (c. by one third) and rendered the results not significant for either sperm trait (Table S4 and Table S5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Positive associations between within-pair reproductive success and sperm morphology.

(A) flagellum length and (B) relative length of flagellum to head section. Within-pair reproductive success is measured as the proportion of young at a social dominant male's nest(s). Analyses were conducted using male identity as a random factor (i.e. mixed models). Plots refer to mean paternity for all broods sampled. Males that are potential active breeders past the date of the current paternity assessment (2008/9 season) are represented by the open circles (n = 17 of the total n = 47 males). Fitted curves were calculated using the regression estimates from non-mixed effects fitted models.

https://doi.org/10.1371/journal.pone.0028809.g002

Midpiece size (absolute or relative to flagellum length) was not associated with either extra- or within-pair reproductive success (all models, p>0.4; see Table S4). In summary, the same sperm morphometric traits have opposite effects for cuckoldry success and cuckoldry avoidance.

Discussion

We found evidence of opposite selection on male sperm traits in superb fairy-wrens, a species under intense sperm competition. Male fairy wrens with shorter flagella and relatively larger heads sired more extrapair offspring, but were less likely to secure paternity at their own nest than males with the opposite sperm phenotype. To our knowledge, this is the first evidence for a naturally occurring selective trade-off for sperm morphology in a wild population.

Sperm size, design and numbers are known to influence the outcome of sperm competition (e.g. reviewed in [18], [58], [59]). Fertilization efficiency of sperm is a complex trait that is further influenced by female (cryptic) choice processes, often mediated by ejaculate storage [60], [61]. Although we currently lack data on female fairy-wren sperm storage morphology, we can speculate on possible mechanisms that explain the association between sperm design and reproductive success based on what we know about (i) the likely sperm competition processes operating in this species, (ii) the differences in the timing and number of copulations between the pair and the extrapair mate, and (iii) the unique features of passerine sperm morphology.

Supporting Information

Acknowledgments

We thank all members of the A. Cockburn research group that helped with life history data collection, L. Fromhage and A. Webster for valuable comments on earlier versions of the manuscript, and S. Nakagawa for statistical advice.