Quantitative genetics of sexual display, ejaculate quality and size in a lekking species
Correspondence author. E-mail: email@example.com
- The investment into extravagant sexual display and competitive sperm are two essential components of pre- and post-copulatory sexual selection. Even though the selective forces acting on sexual display and sperm characteristics have been extensively studied in recent years, the genetic architecture underlying the expression of these traits has been rarely explored.
- Here, we estimated the genetic variances and covariances of traits linked with ejaculate size and quality, and sexual display in the houbara bustard (Chlamydotis undulata undulata, Jacquin 1784).
- Using a very large pedigree-based data set, we show that sexual signalling and ejaculate size (but not ejaculate quality) are heritable and genetically positively correlated. The matrix of genetic covariances also provided support for some across-sex correlations: male and female gamete numbers are positively correlated, and more surprisingly, male display and female gamete numbers are also positively correlated.
- These results can have important implications for the understanding of the evolution of sperm traits and sexual display in animals.
Sexual selection is a complex phenomenon including processes occurring at different stages of the mating sequence (Andersson 1994), with male–male interactions and female choice being the major determinants of male mating success. In the last years, it has become well established that mate choice itself is a more complex process than previously thought, involving traits that are expressed before and after copulation (Birkhead & Pizzari 2002; Andersson & Simmons 2006). Pre-copulatory sexual selection refers to all processes occurring before mating. According to the indicator hypothesis, secondary sexual traits (ornaments, behaviours and chemical cues) are usually thought to have evolved to guide such pre-copulatory choice, because they may inform the choosy sex about the phenotypic/genetic quality of their bearers (Kokko, Jennions & Brooks 2006). It is, however, a common feature to observe that females mate with different partners within a single reproductive event (Simmons 2005). Multiple matings open the way to post-copulatory sexual selection where sperm from unrelated males compete for access to the eggs and where the specific environment experienced within the female reproductive tract can select sperm with a particular phenotype/genotype (Parker 1970; Eberhard 1996; Evans et al. 2003).
In the last decade, the study of pre- and post-copulatory sexual selection has comprised a very large number of theoretical, experimental and comparative work (Simmons 2001; Birkhead & Pizzari 2002; Kokko, Jennions & Brooks 2006; Andersson & Simmons 2006; Clutton-Brock 2007). However, how males allocate resources into traits involved in pre- vs. post-copulatory episodes of selection is a matter of debate. There are a few hypotheses that can help us to predict the sign of the covariation between traits involved in mate attractiveness and sperm quality. First, an economic view of sexual display and sperm quality predicts that the two traits should be negatively correlated (Parker 1998; Simmons & Emlen 2006). This resembles to the classical principle of resource allocation where energy allocated to one function is no longer available to other competing functions. In addition to this, males that allocate heavily in sexual display may engage in more copulatory events during a breeding bout compared to less preferred males, and this could produce a dilution/depletion of gametes per ejaculate (Preston et al. 2001). Finally, males can have alternative mating strategies. Individuals with a non-preferred phenotype have lower mating success, but by investing into traits determining sperm competition could, nevertheless, secure some paternity (Pizzari, Cornwallis & Froman 2007). All this should therefore result in a trade-off between the expression of ornaments targeted by female choice and ejaculate size/quality.
Alternatively, one could expect a positive covariation between traits involved in pre- and post-copulatory selection. It has been suggested that polyandrous females may gather indirect, genetic, benefits for the progeny by allowing sperm of different males to compete (Keller & Reeve 1995; Yasui 1997; Evans & Simmons 2008). The assumption of this so-called good-sperm hypothesis is that sperm competitiveness correlates with traits describing the genetic quality of males. By allowing sperm to compete, females could then select males with supposedly good genes and produce offspring with better viability (Hosken et al. 2003; Kekäläinen et al. 2010). The good-sperm hypothesis makes a clear parallel with the good-genes model of sexual selection, where sexual traits are supposed to reflect the genetic quality of their bearers (Simmons 2005). According to the good-sperm hypothesis, we might therefore expect a positive covariation between sexual display and ejaculate size/quality if these traits are linked to the underlying male genetic quality.
A few studies, both correlative and experimental, have explored the sign of the covariation between pre- and post-copulatory traits with some reporting positive covariation and others negative covariation (Matthews, Evans & Magurran 1997; Danielsson 2001; Malo et al. 2005; Locatello et al. 2006; Pitcher, Rodd & Rowe 2007; Hosken et al. 2008; Rogers et al. 2008; Rowe et al. 2010). However, most of these studies have looked at the phenotypic correlations among traits or have used fertilization success as a proxy for ejaculate quality. If pre- and post-copulatory traits exhibit standing genetic variation, then one might expect a genetic covariation between traits as well, especially if sperm competitiveness and sexual display reflect the genetic quality of males. Estimating genetic covariance among traits is also important because it helps us to predict the response to selection. A negative genetic correlation between ornaments and ejaculate quality would set a constraint to the evolution of both traits. Strong pre-copulatory selection due to female preference for ornamented traits would not produce an evolutionary change because of the antagonist genetic correlation with investment into sperm which is also under positive directional selection because of sperm competition. Positive genetic correlations would imply a positive feedback similar to the runaway process where the limit is set by the cost of producing more and more exaggerated ornaments (Lande 1980). Quantitative genetic studies of pre- and post-copulatory traits are understandably fewer than those that have focused on phenotypic correlations. We are aware of only three recent studies that have investigated the genetic covariance between ornaments and ejaculate size/quality in captive populations, and all of them have reported evidence in support of the trade-off hypothesis (negative genetic correlations between pre- and post-copulatory traits, Simmons, Tinghitella & Zuk 2010; Evans 2010; Engqvist 2011).
The houbara bustard (Chlamydotis undulata undulata) is a very useful model species to investigate the evolution of traits involved in pre- and post-copulatory sexual selection. Males have a very peculiar courtship display during the breeding season that involves behavioural traits, feather ornaments and acoustic signals. Previous work has shown that the expression of courtship display incurs costs. Males lose 10% of their body mass during the breeding season (Saint Jalme et al. 1996), and experimental evidence shows that immune-challenged males dramatically reduce their display activity (Chargé et al. 2010).
In addition to presumably strong pre-copulatory sexual selection, male houbara bustard also experience post-copulatory episodes of selection because, despite having a small clutch size (two to four eggs), 60% of broods have multiple sires (Lesobre et al. 2010), and, at least in captivity, the number of sperm tightly predicts the probability to fertilize the eggs in non-competitive fertilization trials (Saint Jalme, Gaucher & Paillat 1994). Moreover, in a previous study we showed that males exposed to an inflammatory insult produced lower number of spermatozoa than control males, resulting in a reduced hatching success (Chargé et al. 2010). Sperm characteristics (both ejaculate size and quality) could therefore determine male reproductive success in this species.
We explored the phenotypic and genetic covariation between pre- and post-copulatory traits in the houbara bustard by taking advantage of a large captive breeding programme at the Emirates Center for Wildlife Propagation (ECWP, Morocco), involving several hundreds of birds with known pedigree, over a period of 22 years, to estimate the genetic variances and covariances of ejaculate size, ejaculate quality, sexual display and body mass. In addition to this, we also estimated the genetic variances of female fecundity and body mass, and the covariances between male and female traits.
Material and methods
The houbara bustard is a middle-size bird inhabiting semi-arid areas in North Africa, Middle East and Central Asia, the subspecies Chlamydotis undulata undulata being restricted to North Africa. The species is sexually dimorphic in size with males larger than females, and in plumage with males harbouring ornamental feathers. Although life span is unknown for free-ranging birds, some captive individuals have reached the age of 24 years.
The mating system of houbara bustard is based on an ‘exploded’ lek (Hingrat et al. 2008). During the breeding season (January–June), males engage in a complex behavioural display (Gaucher et al. 1996). The courtship is characterized by a circular running with the white feathers on the neck and the head fully erected. This gives a peculiar ‘snow ball’ shape that can be detected from long distances in the houbara semi-desert habitat. In addition to this, males perform a subsonic booming produced by the inflating of pectoral bags. Males devote several hours per day to courtship activity (Hingrat et al. 2008). Females visit displaying males and mate with several males within the same reproductive episode as suggested by the high proportion of broods with multipaternity (Lesobre et al. 2010). Only females provide parental care with the paternal contribution to reproduction being limited to the transfer of genetic material with the gametes.
Birds start reproducing when 1–4 years old from January to June. Females lay two to four eggs per clutch with an incubation period of about 23 days. Juveniles are semi-precocial with males exhibiting longer dispersal distances than females (Hardouin et al. 2012).
Housing of Birds
The houbara bustard is endangered across its entire distribution area mainly due to over-hunting and habitat degradation (Tourenq et al. 2005; Combreau, Launay & Lawrence 2001) leading to the creation of a captive breeding centre in Morocco to supplement wild North African populations (Lacroix 2003).
Breeding adults were housed outdoor in individual cages (2 × 4 m²) at the ECWP (Missour, Morocco). Food (dry granules) and water were provided daily ad libitum. The cages were arranged in rows. Males and females being isolated, reproduction occurred by artificial insemination. Males were regularly exposed to a dummy female by experienced ECWP staff members to collect semen (Saint Jalme, Gaucher & Paillat 1994). Males ejaculated in a Petri dish, and semen was immediately transferred into an Eppendorf tube and given to the adjacent laboratory where the assessment of ejaculate size and quality occurred. The average number of days between two consecutive collections was 3·4 days (SD = 3·63, min–max = 1−166 days). The number of sperm per ejaculate was assessed using a spectrophotometer at a wavelength of 600 nm, after dilution in Lake 7.1 diluent (Lake & Ravie 1984; Saint Jalme et al. 2003). Annual and seasonal variations in sperm count are detailed, respectively, by Preston et al. (2011) and Saint Jalme, Gaucher & Paillat (1994).
Proportion of live sperm (sperm viability) and morphologically normal sperm was assessed using an eosin–nigrosin method (Lindsay et al. 1999). Briefly, semen was stained with eosin–nigrosin, and a minimum of 100 sperm were screened under a light microscope (×1000). Aberrant sperm morphology includes double flagella, a swollen membrane or an extended nucleus. Whereas the number of sperm was assessed for each ejaculate, sperm viability and morphologically normal sperm were only measured once per year.
Pairs were formed by ECWP managers following a breeding programme based on their relatedness (mean kinship) assessed from pedigree analysis (Ballou & Lacy, 1995) and aiming at equalizing the representation of each founder line to maintain genetic diversity while avoiding inbreeding (Lesobre 2008).
Females were artificially inseminated with on average thirteen millions spermatozoa (mean ± SE = 13·2 × 106 ± 0·02 × 106). Eggs were collected daily to avoid brooding and placed in incubators (37·5–37·8 °C) with humidity set to reach a loss of 13–18% of initial egg mass, over the incubation period (23 days).
At hatching, chicks were transferred to a rearing facility and hand-fed by ECWP staff members according to a specific protocol depending on whether juveniles were integrated to the captive breeder flock or were released in the wild within the framework of the reinforcement programme.
The phenotypic data were collected on 2021 males and 2359 females during the period 1999–2008. The following traits were measured: body mass, courtship display, number of sperm per ejaculate (ejaculate size), proportion of live sperm per ejaculate, proportion of morphologically normal sperm per ejaculate (ejaculate quality) and number of eggs laid per year.
Body mass (±1 g) was regularly measured in both males and females at several instances during a year. We used body mass in fall when individuals have recovered from the breeding effort.
Sexual display could be assessed because this behaviour is well preserved in captive bred animals (Gaucher et al. 1996). Staff members of the ECWP recorded whether a male was displaying during at least one of three daily scans (dawn, early morning and afternoon), by moving around the caged area.
Sperm count was assessed as the number of spermatozoa per ejaculate, and sperm quality was evaluated following the protocol described above.
We used the cumulative number of eggs laid per year to measure female fecundity.
Pedigree of the Captive Population
The data set used here is based on a pedigree including all the 19 915 birds born in captivity from 1986 to 2008. Pedigree was available for all captive birds because pairs were formed artificially and eggs individually incubated. Genetic programme was hinged on the maintenance of genetic diversity and avoidance of inbreeding. Therefore, inbreeding was negligible (see Fig. S1, Supporting Information). In general, the same male was used to inseminate a given female during a breeding season; however, when semen collection failed, the ejaculate of another male was used, incurring potential doubts about paternity. Reliability of pedigree structure was thus reinforced by performing microsatellite analyses to identify the sire (Lesobre 2008). The pedigree used in the present analyses included eight generations (see detailed information on pedigree in the Fig. S2, Table S1, Supporting Information).
For most of the individuals (67%), paternity was assigned with certainty either because females were inseminated with a single male during the entire breeding season, or because microsatellite analyses made it possible to identify the sire (Lesobre et al. 2010). To make sure that our estimates were not biased because of paternal misassignment, we re-computed all the additive genetic variances and covariances using this restricted sample size. The results were similar to those obtained on the entire data set, showing that paternal misassignment has a negligible effect on the estimates.
Quantitative Genetic Analyses
To estimate variance components of courtship display, ejaculate size and quality, body mass and female fecundity, we fitted an individual ‘animal’ model (Henderson 1973; Lynch & Walsh 1998; Kruuk 2004), using information from pedigree and phenotypic values of captive birds. Animal models are mixed models allowing decomposing the phenotypic variance of a trait into its additive genetic and other components of variance, using restricted maximum likelihood. The analyses were run with the software ASReml, release 2.0 (Gilmour et al. 2002).
Univariate models were used to estimate heritability for each trait. Age of birds was included as a fixed factor to take into account any effect of immaturity and/or senescence on reproductive traits (Preston et al. 2011). Birds were housed in outdoor cages and thus were undergoing climatic variability between years. We therefore controlled for inter-annual environmental variation including year (n = 10) into the model as a fixed factor. Year of birth was included as a random effect. Individual identity was fitted twice in the model: once as a factor linked to the pedigree to estimate additive genetic variance and second time as an individual identity unlinked to pedigree to take into account multiple measurements on the same individual (also called permanent environment effect). We included in the model year of sampling and age of birds as fixed effects as they explained a statistically significant fraction of variation in all traits (all P's < 0·001, see details in the Table S3, Supporting Information). For the number of sperm per ejaculate, we ran an additional model based on a detailed data set in which the number of days between two consecutive sperm collections was also included as a fixed factor. This was not done for sperm viability and morphologically normal sperm because these traits were assessed only once per year.
Partitioning of variance was performed according to eqn (eqn 1).
where Vp is the total phenotypic variance, Va is the additive genetic variance, Vpe is the variance of permanent environment, Vb is the variance of year of birth, VmA is the variance due to maternal genetic effects, Vm is the variance due to environmental maternal effects, and Vr is the residual variance.
To assess whether the additive genetic variance was significant, the likelihood of the model including the additive genetic effect was compared to the likelihood of a model in which the additive genetic effect was not included. The difference in likelihood of both models (ΔL) was compared to a χ² distribution (χ² = 2*ΔL) with one degree of freedom as models differed by one parameter only. We then calculated heritability in the narrow sense from the estimated additive genetic variances (Falconer & Mackay 1996). Standard errors around (co)variances and heritability estimates were computed directly in ASReml that uses the delta method.
We proceeded in the same way to assess maternal effects. We compared univariate models including or not environmental maternal effects. If maternal effects were significant, we disentangled environmental maternal effects from genetic dam effects by including dam identity linked to the pedigree in the models. Dam effects were not significant; therefore, they were not included in the final models for the main results but are reported in Table S4 (Supporting Information). Hence, for all traits presented here, eqn (eqn 1) is reduced to
We computed coefficients of variation (CV) of the effects included in the model by dividing the standard deviation by the mean phenotypic value (Houle 1992). Standard errors of CV were calculated following the formula given by Garcia-Gonzalez et al. (2012) (eqn (eqn 2)):
where σ[V] denotes the standard error of V and σ[μ] denotes the standard error of the mean.
Within- and between-Sex Genetic Correlations
Genetic correlations and correlations between the other random effects (permanent environment, birth year and residual) were estimated using bivariate animal models (eqn (eqn 3)). We included year of sampling and age of birds as fixed effects, in the same way that fitted in the univariate models.
where 1 and 2 represent, respectively, trait 1 and trait 2, Vp is the total phenotypic variance, Va is the additive genetic variance, Vpe is the variance of permanent environment, Vb is the variance of year of birth, Vr is the residual variance, and Cov represents the covariance between traits.
In the case of between-sex genetic correlations (e.g. model investigating correlation between male display and female fecundity), both traits cannot be measured on the same individual, and phenotypic covariance is undefined. However, the animal model can estimate the additive genetic covariance across sexes as a function of the covariance between opposite-sex relatives. In these models of sex-limited traits, covariances between residuals and permanent environment effects cannot be estimated and were set to zero (Brommer et al. 2007; Wilson et al. 2010).
As described previously, the significance of the covariance terms was estimated by comparing the likelihood of the full model to the likelihood of a reduced model where the genetic covariance between traits was set to zero.
We partitioned the phenotypic variance of seven traits measured in male and female houbara bustards (Table 1, see Table S3 (Supporting Information) for significance of random effects and estimates of fixed effects). Heritabilities ranged from 0·02 for the proportion of live sperm and the proportion of morphologically normal sperm to 0·36 for female body mass. Heritabilities were statistically significant for all traits except for the two variables describing ejaculate quality. Controlling for maternal effects or restricting the analyses to individuals with certain paternity did not change the results (Table 1). The heritability of ejaculate size was also significant when correcting for the number of days between subsequent sperm collections (h2 ± SE = 0·16 ± 0·03, P < 0·001).
Table 1. Phenotypic means, variance and coefficients of variation of random effects (±standard errors) for seven traits measured in male and female houbara bustards. Heritabilities and standard errors were calculated with the REML method. Values in bold indicate statistically significant heritabilities (P < 0·001). Fixed effects included in the animal models are reported in Tables S2 and S3-B (Supporting Information)
|Male body mass||204·22||1915·3 ± 5·4||13990·0 ± 3256·0||2962 ± 1739||16010 ± 2514||9243 ± 606|| 0·33 ± 0·07 ||0·06 ± 0·01||0·05 ± 2·10−2|| 0·34 ± 0·07 |
|Female body mass||143·61||1313·3 ± 3·1||7074·0 ± 1248·0||1167 ± 632·9||6915 ± 923·2||4758 ± 266·9|| 0·36 ± 0·06 ||0·06 ± 0·01||0·05 ± 1·10−2|| 0·39 ± 0·06 |
|Courtship display||41·98||49·3 ± 0·5||213·20 ± 34·04||61·19 ± 29·57||212·6 ± 27·52||753·6 ± 15·88|| 0·17 ± 0·03 ||0·30 ± 0·02||0·56 ± 0·01|| 0·16 ± 0·03 |
|Ejaculate size||18·92||18·8 ± 0·3||51·68 ± 11·25||12·64 ± 6·97||72·84 ± 9·61||173·76 ± 4·75|| 0·17 ± 0·03 ||0·38 ± 0·04||0·7 ± 0·01|| 0·14 ± 0·03 |
|Sperm viability||0·06||0·9 ± 1·10−3||7·89·10−05 ± 6·27·10−05||2·04·10−04 ± 1·18·10−04||1·25·10−04 ± 8·89·10−05||3·09·10−03 ± 1·20·10−04||0·02 ± 0·02||0·01 ± 3·92·10−3||0·06 ± 0·12||0·02 ± 0·02|
|% normal sperm||0·12||0·8 ± 2·10−3||1·97·10−04 ± 2·78·10−04||1·25·10−05 ± 4·78·10−05||2·94·10−03 ± 4·12·10−04||5·93·10−03 ± 2·56·10−04||0·02 ± 0·03||0·02 ± 0·01||0·10 ± 0·21||0·01 ± 0·03|
|Nb eggs laid||5·16||6·3 ± 0·1||5·61 ± 0·77||0·51 ± 0·28||5·03 ± 0·55||13·58 ± 0·26|| 0·23 ± 0·03 ||0·38 ± 0·03||0·58 ± 0·01|| 0·23 ± 0·03 |
To compare the evolvability of the different traits, we also computed the coefficient of additive genetic variation (CVA) and the associated standard errors. CVA of body mass, courtship display, ejaculate size and the number of eggs laid were low to moderate (from 0·06 for body mass to 0·38 for ejaculate size and number of eggs laid) with very low standard errors. Conversely, ejaculate quality had the largest CVA with the largest associated SE. This was paralleled by the coefficient of residual variation (CVR), with both variables describing ejaculate quality having the largest CVR.
We found no evidence for a genetic maternal effect (see Supporting Information). Environmental maternal effect was statistically significant only for the percentage of morphologically normal sperm (see Supporting Information).
At the phenotypic level, courtship display and ejaculate size were correlated with body mass (RP = 0·38, P < 0·001; RP = 0·14, P = 0·006, respectively). Similarly, in females, body mass and fecundity were also correlated (RP = 0·31, P < 0·001), with heavier females laying more eggs. Ejaculate size was positively correlated with courtship display (RP = 0·22, P < 0·001). Sperm viability and mo RP hologically normal sperm were not correlated with courtship display, ejaculate size or male body mass (all RP's < 0·10, all P's > 0·2). Sperm viability and mo RP hologically normal sperm tended to be positively correlated even though the P value did not reach the significance threshold (RP = 0·27, P = 0·063).
All the following results are presented in Table 2 for models using the entire pedigree, while results based on pedigree with paternity certainty are presented below between brackets (see also Table S5, Supporting Information for results on other random effects and fixed effects).
Table 2. Genetic correlations between five traits measured in male and female houbara bustards. Only the five traits with statistically significant heritabilities have been considered for the estimation of genetic correlations. Values are followed by standard errors with P values in brackets. Statistically significant correlations are in bold. Fixed effects are similar to those in univariate model (Table S5B). Covariances for other random effects are reported in Table S5A
|Male body mass||0·97 ± 0·05 (< 0·001)||0·27 ± 0·13 (0·036)||0·44 ± 0·16 (0·007)||0·33 ± 0·13 (0·012)|
|Female body mass|| ||0·13 ± 0·12 (0·294)||0·18 ± 0·15 (0·221)||0·16 ± 0·11 (0·138)|
|Courtship display|| || ||0·34 ± 0·13 (0·005)||0·56 ± 0·10 (< 0·001)|
|Ejaculate size|| || || ||0·47 ± 0·12 (< 0·001)|
Courtship display was positively genetically correlated with ejaculate size (Table 2, RG ± SE = 0·27 ± 0·14, P = 0·046). Male body mass was positively genetically correlated with ejaculate size and courtship display (Table 2, RG ± SE = 0·48 ± 0·17, P = 0·005, and RG ± SE = 0·23 ± 0·13, P = 0·080). Female body mass and the number of eggs laid were, however, not genetically correlated (Table 2, RG ± SE = 0·18 ± 0·10, P = 0·099).
Positive genetic correlations between ejaculate size and courtship display might arise because of shared correlation with a third trait (body mass). However, when including body mass in the model, the genetic correlation between the number of sperm per ejaculate and courtship display remained unchanged (RG ± SE = 0·34 ± 0·13, P = 0·006 vs. RG ± SE = 0·34 ± 0·13 in Table 2).
We also investigated the covariance between traits expressed in the two sexes. Gamete production (number of spermatozoa in the ejaculate and number of eggs laid) was positively correlated across sexes (Table 2, RG ± SE = 0·42 ± 0·13, P = 0·001). Interestingly, courtship display in males was positively genetically correlated with female fecundity (Table 2, RG ± SE = 0·61 ± 0·10, P < 0·001).
We provide here evidence for a positive genetic covariance between traits involved in pre- (courtship display) and post-copulatory sexual selection (ejaculate size) in males of a bird species with a lekking-based mating system. This shows that males that are favoured by female choice will also produce ejaculates that are favoured in the process of egg fertilization. As suggested by Birkhead & Pizzari (2002), when traits are under positive selection for copulation and fertilization success, they might become phenotypically and genetically integrated, as found in the present study.
All traits measured in this study, with the exception of sperm viability and the proportion of morphologically normal sperm, showed moderate, but statistically significant heritabilities. Traits closely related to fitness, under strong directional selection, should have depleted additive genetic variance and lower heritability than morphometric or physiological traits (Fisher 1930). However, Price & Schluter (1991) suggested that low heritability of fitness-related traits can be due to high level of residual variance rather than reduced additive genetic variance. Indeed, empirical evidence gathered in the wild generally supports the idea that low heritability of fitness-linked traits arises because of large residual variance (Houle 1992; Merilä & Sheldon 2000; Kruuk et al. 2000; Coltman et al. 2005). In this study, body mass, assumed to be less related to fitness than ejaculate quality or the number of eggs laid, had the highest heritability, whereas the two traits describing ejaculate quality had the smallest values. This pattern was mostly due to larger residual variance of ejaculate quality traits. The analysis of coefficient of variation suggests that ejaculate quality traits have both the largest CVA (with large SE) and the largest CVR. This is a similar pattern to the one reported by Merilä & Sheldon (2000), who found that traits more tightly linked with fitness had lower heritability and larger CVA and CVR than traits less important for fitness. It should be noted that the estimates obtained here for ejaculate size and quality lie somehow below the mean of heritabilities of production and performance traits recently reviewed by Simmons & Moore (2009). The comparison of CVA seems more uncertain because recent work has highlighted a high proportion of misreported CVA values in the literature (Garcia-Gonzalez et al. 2012).
For a microevolutionary change to occur, a trait needs to be heritable and under selection (Endler 1986). In its simpler version, the response to selection is the product of the heritability and the strength of selection (the selection differential) (Arnold & Wade 1984; Falconer & Mackay 1996). Our finding of low (zero) heritability for traits describing ejaculate quality therefore suggests that they should not respond to the selection due to sperm competition and/or cryptic female choice. On the contrary, ejaculate size and courtship display seem to have the potential for a microevolutionary change. Nevertheless, it should be noted that the response to selection will also depend on the selection on correlated traits and on the sign of these correlations (Lande & Arnold 1983). Investigating the matrix of phenotypic and genetic correlations among traits should therefore help us to predict how selection will indeed produce a microevolutionary shift or whether negative genetic correlation will constrain the evolution of a trait in spite of positive directional selection.
The analysis of the correlation matrices revealed an overall consistent pattern of phenotypic and genetic covariation among traits. Traits that were significantly correlated at the phenotypic level were also correlated at the genetic level, the exception being female body mass and the number of eggs laid that were poorly correlated at the phenotypic level and had a highly statistically significant genetic correlation. All the statistically significant correlations were positive. There have been a handful of studies that investigated the matrix of genetic correlations between sexual advertisements and sperm traits (Simmons, Tinghitella & Zuk 2010; Engqvist 2011). All of them have reported evidence for negative correlations in captive populations. For instance, Evans (2010) investigated the matrix of genetic correlations between courtship, mating strategy (sneaking) and ejaculate quality in guppies (Poecilia reticulata). The findings show a strong negative genetic correlation between courtship and sperm velocity (whereas courtship was not genetically correlated with sperm viability). Interestingly, sneaking was positively genetically correlated with sperm velocity but negatively correlated with sperm viability, suggesting a complex pattern of association among mating strategy and sperm traits (Evans 2010). We measured three traits describing ejaculate size and quality in male houbara bustards. However, only one of the three traits was heritable (sperm count). The lack of additive genetic variance for sperm viability and normal sperm morphology prevented us to estimate the genetic correlations between sperm traits.
Overall, these results seem to provide little support to the hypothesis that investment into pre-copulatory traits comes at the expense of post-copulatory traits and rather indicate that these traits (courtship display and ejaculate size) are phenotypically and genetically integrated (Birkhead & Pizzari 2002). These results fit within the general framework provided by the good-sperm hypothesis. According to this hypothesis, females that engage in multiple matings can gather indirect genetic benefits if males that win the sperm competition have superior genetic quality (Yasui 1997). This draws a clear parallel with the good-genes hypothesis where females choose a mate because sexual displays reflect the underlying genetic quality of males (Andersson & Simmons 2006). If secondary sexual traits and traits linked to sperm competitiveness are genetically variable and under positive selection, they might become positively correlated.
Genetic correlations can arise because of genes with pleiotropic effect or because of linkage disequilibrium between loci (Falconer & Mackay 1996). Female preference for exuberant males might have promoted the evolution of genetic covariance between secondary sexual traits and traits linked to sperm competitiveness, through linkage disequilibrium, if these traits are heritable (as also shown here) and phenotypically linked because of common condition dependence. We have previously shown that both courtship display and ejaculate quality are condition dependent because an experimental manipulation of male health status produced a simultaneous decrease in the expression of both traits (Chargé et al. 2010). Once the genetic covariance has been established, female preference should tend to reinforce it and reinforce the choice as well. Under this scenario, female preference can act as a major selective force driving the correlated evolution of the number of gametes released in the ejaculate and the expression of secondary sexual traits.
Because of genetic covariation between sexes, the selection pressure exerted by females on male gamete number can produce a correlated response on female gamete number. In the houbara bustard, the number of sperm released in the ejaculate and the number of eggs laid genetically covary, as do courtship display and female fecundity, setting the stage for correlated responses in females. Such correlated responses have obvious consequences in terms of reinforcement of female preference, because choosy females should produce attractive sons and fertile daughters. Interestingly, the positive covariation between male secondary sexual traits and female fecundity goes against the idea of antagonistic sexual selection in this species.
A major concern for theory on the indirect benefits of mate choice has been to envisage how variability is maintained in spite of directional selection exerted by females. The genetic matrix estimated here is based on temporally constant and benign environmental conditions as experienced by captive individuals. How the signs of genetic covariances are affected when traits are expressed in more constrained, natural environment is unknown. However, other studies have reported extensive effects of environmental conditions on the G-matrix of life-history traits, potentially contributing to the maintenance of variation of fitness-linked traits (Kruuk, Slate & Wilson 2008).
Most studies that have previously investigated the quantitative genetics of sperm traits have been conducted under laboratory conditions (Simmons & Moore 2009), as was the case for the present study. Measuring ejaculate size and quality for a large number of individuals (n = 2021 males in this study) across several generations in a natural setting is a daunting task and obviously a major limitation, explaining why most studies have been conducted in the laboratory. Nevertheless, it should be fully acknowledged that the adaptation to novel environmental conditions as those experienced in the laboratory might produce a shift towards positive genetic correlations between traits (Service & Rose 1985). A great deal of effort should therefore be devoted to replicate these results in the wild.
To conclude, our results highlight the importance of considering the genetic architecture of traits to fully appraise the evolutionary forces acting on primary and secondary sexual traits as well as the benefits of female preference.
The Emirates Centre for Wildlife Propagation (ECWP, www.ecwp.org) provided the data and funding for this project under the leadership of the International Fund for Houbara Conservation (IFHC). We are grateful to H.H. Sheikh Mohamed bin Zayed Al Nahyan, Crown Prince of Abu Dhabi and Chairman of the International Fund for Houbara Conservation (IFHC), and H.E. Mohammed Al Bowardi, Deputy Chairman of IFHC, for their support. A final thanks to Gwénaëlle Leveque who coordinated data collection and to all ECWP staff involved in behavioural observations and assessments of ejaculate quality.