The evolution of male genitalia: patterns of genetic variation and covariation in the genital sclerites of the dung beetle Onthophagus taurus

Authors


Clarissa M. House, Evolutionary Biology Research Group, School of Animal Biology (M092), University of Western Australia, Nedlands WA 6009, Australia.
Tel.: 61 08 9380 2221; fax: 61 02 9385 1558;
e-mail: clarissa@cyllene.uwa.edu.au

Abstract

Three main hypotheses, have been invoked to explain divergent genital evolution, the lock and key, pleiotropy, and sexual selection hypotheses, each of which make different predictions about how genital traits are inherited. Here we used a half-sib breeding design to examine the patterns of genetic variation and covariation between male genital sclerites, and their covariance with general body morphology in the dung beetle Onthophagus taurus. We found CVA's and CVP's were similar for both genital and general morphological traits and that CVR's were large for both trait types. We found that male genital sclerites were negatively genetically correlated with general morphological traits. Variation in male genital morphology has direct implications for a male's fertilization success and the resulting sexual selection acting on male genitalia is predicted to maintain high levels of additive genetic variance. Contrary to this prediction, we found that individual genital sclerites all had low levels of additive genetic variance and large maternal and environmental sources of variation. Our data suggest that the genital sclerites in O. taurus are not inherited independently but as a genetically integrated unit. More importantly, the way the different sclerites function to influence male fertilization success reflects this genetic integration. Even though levels of VA in individual genital sclerites may be low, there may still be sufficient VA in multivariate trait space for selection to generate evolutionary change in the overall morphology of male genitalia.

Introduction

In animals with internal fertilization and promiscuous mating, male genitalia tend to evolve rapidly and divergently (Eberhard, 1985). Studies of the evolutionary processes responsible for these patterns have focused on the relationship between male genital morphology and mating behaviour, insemination, fertilization success and female genital morphology (reviewed in Simmons, 2001). Surprisingly, little attention has been paid to the underlying genetic architecture of genital traits, despite its importance for how these traits will evolve when subject to selection (but see Arnqvist & Thornhill, 1998; Preziosi & Roff, 1998).

Three main hypotheses have been proposed to explain the evolution of male genitalia: the lock-and-key, the pleiotropy, and the sexual selection hypotheses. Under the lock-and-key hypothesis, selection for preinsemination reproductive isolation is predicted to favour male genitalia (the key) that provides an exact mechanical fit to female genitalia (the lock) so that male genitalia become highly canalized (Eberhard, 1985; Proctor et al., 1995; Arnqvist, 1997). The pleiotropy hypothesis proposes that variation in genitalic morphology is selectively neutral and that male genitalia evolve via pleiotropic effects of genes that code for both genital and general morphological characters (Mayr, 1963). Finally, the sexual selection hypothesis proposes that fertilization success is nonrandom with respect to genital morphology (Eberhard, 1985). This will occur if differences between males in their genital morphology are related to the ability to remove rival sperm (sperm competition), control fertilization (sexual conflict), and/or induce post-copulatory preferential sperm utilization (cryptic female choice) (Arnqvist, 1997).

In his seminal paper Arnqvist (1997) proposed that a more complete understanding of how male genitalia evolve would come through detailed single species studies. Moreover, he provided a tentative framework to test alternative hypotheses which generate a number of predictions about the nature of selection operating on, and the resulting patterns of morphological and genetic (co) variation in, male genitalia within species. According to the lock-and-key hypothesis, strong stabilizing selection for preinsemination reproductive isolation is expected to reduce phenotypic and residual coefficients of variation (CVP and CVR, respectively) so that the expression of genital traits is impervious to environmentally induced deviation (Arnqvist, 1997). Accordingly, genitalia that have evolved according to the lock-and-key hypothesis should show low levels of genetic and phenotypic variation (VA and VP, respectively) and genetic correlations between genitalia and general morphological traits should be weak or absent (Arnqvist, 1997).

Differentiating between the pleiotropy and sexual selection hypotheses is more difficult due to a number of predictions being common to both hypotheses. As selection operating on genitalia is assumed to be neutral in the pleiotropy hypothesis (Mayr, 1963), the only selection operating on genital morphology is indirect and is driven by the strength and the sign of genetic correlations between genital and general morphological traits (Arnqvist, 1997). Genitalia that have evolved according to the pleiotropy hypothesis are thus predicted to exhibit high levels of VA, VP and strong genetic correlations with general morphological traits (Arnqvist, 1997). In contrast, according to the sexual selection hypothesis, variation in male genitalia has direct implications for a male's fertilization success (Arnqvist, 1997). Therefore, male genitalia are expected to be subject to strong directional sexual selection (Arnqvist, 1997). Theoretically, traits that are subject to strong directional selection are expected to show reduced levels of VA (Kirkpatrick & Ryan, 1991). However, counter to this prediction, traits subject to sexual selection typically show greater levels of VA than do general morphological traits (Pomiankowski & Møller, 1995). Furthermore, both quantitative models (Haldane & Jayakar, 1963; Lande, 1975,1982; Felsenstein, 1976; Wolf et al., 1998) and empirical studies (Jia & Greenfield, 1997; Moore & Moore, 1999; Hunt & Simmons, 2002; Kotiaho et al., 2003) suggest that levels of VA can be maintained through a number of different evolutionary mechanisms. Thus, it is tempting to predict that relatively high levels of VA should be maintained in genital traits under strong and persistent sexual selection. However, the relative magnitude of variation to be expected in sexual traits is a point of debate (Merilä & Sheldon, 2000; Hunt et al., 2004; Tomkins et al., 2004; Blows & Hoffman, 2005).

While Arnqvist's (1997) review presents a coherent framework for constructing testable predictions for the evolution of male genitalia in single species studies, many of the predictions understate the complexity of genital morphology. Male genitalia are frequently comprised of sets of functionally related structures that collectively form the entire genital unit (see the numerous examples provided in Eberhard, 1985). That is, male genitalia are often morphologically integrated and are expected to be highly developmentally, functionally and genetically correlated (Cheverud, 1984,1988). Thus, an understanding of the processes that have shaped genital evolution often requires estimation of the genetic variability within sets of genital structures and the covariances among them. This gives information regarding the potential for genital structures to respond to selection, and indicates whether sets of traits are likely to evolve independently or as an integrated unit (sensuCheverud, 1984,1995).

In the dung beetle, Onthophagus taurus, the patterns of fertilization success suggest that sexual selection acts on male genital morphology (House & Simmons, 2003). Two pairs of sclerites are important for a male's fertilization success (House & Simmons, 2003). Sclerites 4 and 5 serve a ‘defensive’ role and are associated with the first male's ability to defend his paternity against a second male. A male mating in the defensive role gains the highest fertilization success when sclerite 4 is small and sclerite 5 is large. In contrast, sclerites 1 and 2 serve an ‘offensive’ role and are associated with the second male's ability to preempt the first male's paternity. A male mating in the offensive role gains the highest fertilization success when sclerite 1 is small and sclerite 2 is large. These data show that sexual selection via sperm competition acts on male genital morphology. Moreover, they suggest that sperm competition may generate antagonistic selection on different aspects of male genital morphology (Parker, 1984). In this situation, ‘offensive’ and ‘defensive’ adaptations are opposed so that males may be successful in one role but not in the other. There are a number of empirical studies that provide support for this prediction (Clark et al., 1995; Civetta & Clark, 2000; Nilsson et al., 2003). However, when the same male O. taurus is mated in both the defensive and offensive role, his fertilization gains in alternative roles are positively correlated (House & Simmons, submitted). This finding has two important implications. First, it suggests that selection in O. taurus operates simultaneously for defensive and offensive adaptations rather than for specialization to a given role. Second, because the relationships between genital morphology and fertilization success shown in House & Simmons (2003) were also observed when individual males mated in the offensive and defensive role, it suggests that sclerites 1 and 2 and sclerites 4 and 5 are not functioning independently of each other. More specifically, for a given males fertilization success to be positively correlated in both roles, we predict that sclerites 1 and 4 and sclerites 2 and 5 must be positively genetically correlated whereas sclerites 2 and 4 and sclerites 1 and 5 must be negatively correlated.

Here we use a half-sib breeding design to estimate (1) the magnitude of additive genetic variance in the genital sclerites and in general morphological traits, (2) the patterns of genetic covariance within measures of the same genital sclerite and between genital sclerites and (3) the patterns of genetic covariance between genital sclerites and general morphological traits in the dung beetle O. taurus. Using our estimates of genetic (co) variation, we assess the relative importance of the lock-and-key, pleiotropy, and sexual selection hypotheses for explaining the evolution of male genitalia in this species. Moreover, we discuss the limitations of Arnqvist's (1997) within species predictions when male genitalia are structurally complex and are inherited as a morphologically integrated unit.

Methods

Parental stock

Beetles were collected from fresh cattle dung from a pasture in Denmark, in the southwest of Western Australia. A mixed sex population was kept in culture for 2 weeks to ensure all beetles were sexually mature and mated. Mated females were then placed in individual breeding chambers (25 × 6 cm PVC piping, three-quarters filled with moist sand and topped with 250 mL of cow dung) and left to construct brood masses. A brood is a mass of dung that provides the resources for the growth and development of a single offspring. Brood masses were sieved from the sand after 10 days, and incubated at 27 °C for 3 weeks. On emergence, adult beetles were sexed and maintained in single sex cultures with unlimited access to fresh dung for 2 weeks. These virgin beetles were used as the parents in our genetic breeding design.

Breeding design

We used a conventional half-sibling breeding design (Lynch & Walsh, 1998). Fifty sires were each housed with six dams for 7 days in containers (4.5 × 17.0 × 11.0 cm) that were half filled with moist sand and topped with fresh cow dung. Mated females were then placed in individual breeding chambers that were sieved after 7 days. Ten broods from each dam were scrapped of excess sand, weighed and placed in individual plastic containers (5 × 7 × 7 cm). Where dams produced greater than 10 broods, the excess broods were placed in individual containers but were not weighed. Emergent offspring were placed in 1.5-mL Eppendorf vials, frozen and then stored in 70% ethanol.

Morphometrics

Morphological variation in general and genital traits was examined by linear measurement for 548 males from 222 families. While simplistic, linear measurements have been shown to capture the same amount of variation in genital morphology as more complex multivariate shape analysis (see Arnqvist & Thornhill, 1998). The genitalia of O. taurus have several chitinous sclerites (House & Simmons, 2003). Tissue connected to the sclerites was macerated in 10% KOH (30 min) and cleared in 50% aqueous lactic acid solution (30 min) before the sclerites were mounted on slides with Eukitt fixative. The length of the sclerite outline, the distance between distinctive landmarks and the area bound by these same landmarks on five of the sclerites was extracted using the measurement explorer function of the Optimas Image Analysis package (see Fig. 1). Male body length, head length and width, pronotum width and elytra length was also measured using the Optimas Image Analysis package.

Figure 1.

Diagrammatic view of five of the genital sclerites removed from the endophallus. Sclerites 2 and 3 are closely associated in the endophallus as drawn, but were separated for measurement. Sclerite size was characterized by measuring the linear distance between fixed landmarks: a–b on sclerite 1; a–b and b–c on sclerites 2 and 3; a–c and b–d on sclerite 4; and a–b and a–c on sclerite 5. The relative position of the landmarks was characterized by measuring the area bound by the landmarks a–c on sclerite 2, 3 and 5 and a–d on sclerite 4. The complexity of the sclerite contour was characterized by measuring the length of its outline.

Repeated measures of the general body and genital traits were taken from a cohort of 12 and 23 beetles respectively. In each instance, morphological variation between males was significantly greater than within males and repeatability estimates were consistently high for all morphological measurements (range of repeatability estimates for genital traits: 0.73–0.97, F1,22 = 3.71–35.68, P = 0.0014–0.0001; and general traits: 0.81–0.98, F1,11 = 5.30–40.0, P = 0.004–0.0001).

Genetic analyses

Our genetic design was unbalanced because two sires failed to produce any progeny, several dams produced too few sons and several sclerites were damaged during dissection. To account for this unbalanced design, we fitted a standard nested model and used restricted maximum likelihood (REML) and ASREML (see http://www.vsn-intl.com/ASReml/) to estimate the variance and co-variances, respectively. This method is more reliable than conventional least squares anova when the breeding design is unbalanced (Lynch & Walsh, 1998). Due to the small size of the genital sclerites, the raw morphological measures were transformed to z-scores prior to genetic analysis. The weight of the brood mass that an offspring consumes during development represents a strong maternal effect (Hunt & Simmons, 2000) and is known to inflate estimates of genetic variance in this species (Hunt & Simmons, 2002; Kotiaho et al., 2003). We therefore controlled for brood mass weight in our genetic analyses by fitting it as a covariate. To compare levels of phenotypic and additive genetic variation between the genital sclerites and the general morphological traits, we calculated coefficients of variation (CV). CV's standardize measures of variation by the trait mean and therefore facilitate comparisons across traits that differ in size (Houle, 1992). Coefficients of variation (CV) were calculated as inline image where V is either the phenotypic (VP), additive genetic (VA) or residual variance (VPVA) and X is the mean trait size (Houle, 1992).

When a covariance is estimated between two traits with low VA, the resulting genetic correlation and standard error often have extreme values making biological interpretation difficult (Cheverud, 1988). Moreover, a genetic correlation cannot be calculated if either of the traits have a VA that is equal to zero (Lynch & Walsh, 1998). This was the case in O. taurus with many of the traits measured exhibiting zero VA (see Table 1). Consequently, our covariance matrices (genetic, maternal, environmental and phenotypic) only contain trait measures that have nonzero VA. Irrespective of the low confidence of covariance estimates, the sign of the covariances may still be of biological importance (e.g. Kingsolver & Wiernasz, 1991). To assist our interpretation in the overall patterns of covariance between traits we summarized the variation within each of the five sclerites and the general morphological traits using Principal Component Analysis (PCA). PC1 is generally considered as an absolute size vector (Bryant & Yarnold, 2003) and these scores were used to calculate our covariance estimates. Although two measures were taken for sclerite 1, only the outline measure had detectable VA. As a minimum of two variables is necessary for PCA, this measure was transformed to a z-score. As both z-scores and PC scores have a mean of zero and a standard deviation of one they are comparable units of measurement (Bryant & Yarnold, 2003).

Table 1.  Descriptive phenotypic and genetic statistics for male genital sclerites and general morphological traits in the dung beetle O. Taurus.
TraitMeasureMean ± SESire h2± SEDam h2± SECVPCVACVR
  1. The mean trait size is measured in mm. Heritability estimates provided in bold are significant at the P < 0.05 level using the G statistic in REML.

Sclerite 11 a–b0.37 ± 0.00200.53 ± 0.229.6609.66
1 outline0.97 ± 0.0020.17 ± 0.140.85 ± 0.254.660.994.55
Sclerite 22 a–b0.35 ± 0.00100.59 ± 0.233.7603.76
2 b–c0.41 ± 0.0010.33 ± 0.160.49 ± 0.233.851.113.68
2 area0.03 ± 0.00010.28 ± 0.150.46 ± 0.234.831.274.66
2 outline1.04 ± 0.0020.22 ± 0.150.60 ± 0.243.740.873.63
Sclerite 33 a–b0.32 ± 0.00100.58 ± 0.235.1905.19
3 b–c0.29 ± 0.0010.28 ± 0.140.18 ± 0.215.551.465.35
3 area0.01 ± 0.000100.57 ± 0.2212.07012.07
3 outline0.82 ± 0.0020.07 ± 0.120.64 ± 0.256.190.816.14
Sclerite 44 a–b0.23 ± 0.0010.13 ± 0.110.21 ± 0.227.531.367.41
4 b–d0.20 ± 0.00100.37 ± 0.247.6707.67
4 area0.02 ± 0.00010.20 ± 0.140.48 ± 0.235.541.255.40
4 outline0.66 ± 0.0010.10 ± 0.120.55 ± 0.245.100.815.03
Sclerite 55 a–b0.28 ± 0.00100.78 ± 0.234.4804.48
5 b–c0.18 ± 0.0010.02 ± 0.100.47 ± 0.2410.820.8210.79
5 area0.02 ± 0.00010.03 ± 0.100.23 ± 0.236.010.545.99
5 outline1.01 ± 0.00200.54 ± 0.224.1304.13
Body sizeBody length9.37 ± 0.0300.95 ± 0.247.6107.61
Head length2.31 ± 0.0100.67 ± 0.2511.46011.46
Head width2.86 ± 0.010.08 ± 0.131.18 ± 0.275.990.885.92
Pronotum width4.57 ± 0.0101.31 ± 0.266.8006.80
Elytra length5.01 ± 0.010.02 ± 0.131.36 ± 0.295.930.425.91

We used a Mantel's randomization test (Mantel, 1967) in Poptools (Microsoft Excel) to test for equality between the additive genetic, maternal, environmental and the phenotypic covariance matrices for the genital sclerites. Comparisons of matrices were performed using the pair-wise correlation values and 1000 randomizations were used to estimate a significance level for each pair-wise matrix comparison (Manly, 1997). This analysis performed on standardized covariance estimates yielded identical results.

Results

Descriptive statistics

Several measures of sclerite 2 and a single measure of sclerite 3 had significant, albeit low, sire heritabilities (Table 1). In contrast, 14 of the 18 genital measures demonstrated significant dam heritabilities (Table 1). Similar patterns were observed for the general morphological traits, except that none of the sire heritability estimates were significant while all of the dam heritability estimates were significant (Table 1). Dam variances include contributions from nongenetic maternal effects, common environment, dominance and epistasis (Lynch & Walsh, 1998). Consequently, any one of these effects could inflate our estimates of dam variance. We argue, however, that the large asymmetry between sire and dam variances presented here are largely due to nongenetic maternal effects because (1) where maternal effects have been quantified, they are typically much greater than effects due to dominance or epistasis (Cheverud & Moore, 1994) and (2) all offspring in this experiment were reared individually so that they experienced unique environmental conditions during their development. Thus, even after statistically controlling for the effects of brood mass weight, it appears that additional maternal effects have a large effect on offspring phenotype. The low sire heritabilities for genital and general morphological traits in O. taurus may be due to low additive genetic variance and/or high residual variance (Lynch & Walsh, 1998). Levels of CVA in genital sclerites were consistently low, averaging 0.63 ± 0.13 across the five sclerites (Table 1). This was of similar magnitude to the CVA exhibited in our five general morphological traits, averaging 0.22 ± 0.16 (Table 1). In contrast, CVR was over 10 times larger than CVA across the five genital sclerites (mean CVR across sclerites: 6.09 ± 0.59) and over 32 times larger across the five general morphological traits (mean CVR across general morphological traits: 6.97 ± 1.11) (Table 1). Thus, the low h2 estimates for both genital and general morphological traits in O. taurus appear to be the result of high residual variation in these traits. More importantly, though, our results demonstrate that genital and general morphological traits exhibit similar levels of genetic and phenotypic variation in this species (Table 1).

Covariance between measures of the same genital sclerite

Traits that are morphologically integrated are predicted to show similar patterns of genetic, phenotypic, maternal and environmental correlations (Cheverud, 1984,1995). We provide two lines of evidence to suggest that the different measures of the same sclerite are morphologically integrated. First, all of the genetic correlations within each of the five sclerites are positive (Table 2) suggesting that these different measures are not inherited independently but as an integrated unit. Second, the phenotypic (Table 2), maternal and environmental correlations (Table 3) are also positive suggesting that these nongenetic contributions also facilitate the morphological integration of measures of the same sclerite.

Table 2.  Estimates of additive genetic (above diagonal) and phenotypic (below diagonal) correlations for genital and general morphological traits in O. taurus.
 1 outline2 b–c2 area2 outline3 b–c3 outline4 a–c4 area4 outline5 a–c5 areaHWEL
  1. Co-variances were only calculated for those trait measures that exhibited nonzero additive genetic variances. Standard errors are provided in parentheses. All correlations control for brood mass weight. Significant correlations (± 2SE from zero) are presented in bold. HW, head width; and EL, elytra length.

  2. *Estimates converged on the boundary using ASREML.

1 outline −0.11 (0.40)−0.75 (0.42)−0.51 (0.50)0.07 (0.41)−0.22 (0.73)0.28 (0.59)−0.36 (0.47)−0.01 (0.65)−2.34 (27.49)−0.48 (1.54)−0.62 (0.97)−2.69 (11.62)
2 b–c0.26 (0.04)0.42 (0.26)0.72 (0.21)0.74 (0.20)0.48 (0.47)−0.43 (0.49)−0.68 (0.37)−0.65 (0.54)0.41 (1.81)−1.07 (2.26)−0.23 (0.64)−1.13 (2.64)
2 area0.24 (0.04)0.38 (0.03)0.99 (0.11)0.65 (0.25)0.78 (0.46)−0.99 (0.50)−0.66 (0.39)−1.23 (1.18)1.68 (9.65)−0.60 (1.17)0.27 (0.52)−0.001 (1.49)
2 outline0.28 (0.04)0.55 (0.03)0.73 (0.02)0.89 (0.26)1.26 (0.65)−0.68 (0.60)−0.71 (0.47)−1.23 (1.18)3.99 (39.76)−0.52 (1.34)0.61 (0.48)−0.22 (2.03)
3 b–c0.16 (0.04)0.40 (0.03)0.31 (0.04)0.38 (0.03)1.03 (0.46)−0.57 (0.47)−0.43 (0.39)−0.62 (0.56)3.74 (12.78)2.82 (21.27)−0.10 (0.58)−1.03 (3.18)
3 outline0.25 (0.04)0.33 (0.04)0.29 (0.04)0.39 (0.04)0.61 (0.02)−0.37 (0.92)−0.41 (0.69)−0.51 (0.98)4.08 (8.53)2.39 (10.77)0.57 (0.84)−0.89 (4.61)
4 a–c0.11 (0.04)0.12 (0.04)0.10 (0.04)0.09 (0.04)0.17 (0.04)0.11 (0.04)1.25 (0.31)1.34 (0.55)−0.80 (2.88)−1.00 (1.74)−0.30 (0.78)−1.06 (5.04)
4 area0.16 (0.04)0.24 (0.04)0.22 (0.04)0.24 (0.04)0.16 (0.04)0.19 (0.04)0.65 (0.02)0.97 (0.12)−0.90 (2.12)−2.45 (14.30)−0.23 (0.78)−0.15 (2.34)
4 outline0.16 (0.04)0.27 (0.04)0.27 (0.04)0.27 (0.04)0.17 (0.04)0.21 (0.04)0.59 (0.03)0.93 (0.01)−2.06 (15.83)−4.02 (20.83)−0.39 (1.16)0.14 (2.01)
5 a–c0.12 (0.04)0.10 (0.04)0.13 (0.04)0.13 (0.04)0.12 (0.04)0.09 (0.04)0.01 (0.04)−0.02 (0.04)0.01 (0.04)12.87 (37.8)3.00 (11.56)10.22 (53.93)
5 area0.17 (0.04)0.21 (0.04)0.22 (0.04)0.24 (0.04)0.19 (0.04)0.20 (0.04)0.06 (0.04)0.11 (0.04)0.10 (0.04)0.34 (0.04)−1.87 (5.46)0*
HW0.25 (0.04)0.43 (0.03)0.41 (0.03)0.51 (0.04)0.26 (0.04)0.30 (0.04)0.09 (0.04)0.28 (0.04)0.32 (0.04)0.07 (0.04)0.22 (0.04)0*
EL0.29 (0.04)0.43 (0.04)0.47 (0.03)0.55 (0.03)0.25 (0.04)0.33 (0.04)0.12 (0.04)0.30 (0.04)0.34 (0.04)0.06 (0.04)0.26 (0.04)0.86 (0.01)
Table 3.  Estimates of maternal (above diagonal) and environmental (below diagonal) correlations for genital and general morphological traits in O. taurus.
 1 outline2 b–c2 area2 outline3 b–c3 outline4 a–c4 area4 outline5 a–c5 areaHWEL
  1. Co-variances were only calculated for those trait measures that exhibited nonzero additive genetic variances. Standard errors are provided in parentheses. All correlations control for brood mass weight. Significant correlations (± 2SE from zero) are presented in bold.

1 outline0.32 (0.23)0.15 (0.27)0.19 (0.24)−0.10 (0.39)0.23 (0.23)0.19 (0.35)0.61 (0.25)0.50 (0.25)0.14 (0.34)−0.20 (0.44)0.21 (0.19)0.42 (0.17)
2 b–c0.32 (0.11)0.77 (0.25)0.85 (0.17)0.80 (0.33)0.77 (0.21)0.11 (0.38)0.74 (0.24)0.58 (0.24)0.11 (0.38)0.52 (0.38)0.77 (0.15)0.72 (0.14)
2 area0.47 (0.10)0.23 (0.09)0.85 (0.13)0.63 (0.44)0.80 (0.27)0.38 (0.45)0.30 (0.14)0.64 (0.26)0.38 (0.45)0.60 (0.47)0.55 (0.20)0.84 (0.16)
2 outline0.45 (0.10)0.40 (0.07)0.66 (0.06)0.40 (0.38)0.49 (0.23)−0.11 (0.41)0.71 (0.26)0.64 (0.26)0.08 (0.40)0.77 (0.41)0.72 (0.15)0.84 (0.12)
3 b–c0.26 (0.11)0.23 (0.09)0.17 (0.09)0.31 (0.08)0.98 (0.28)0.22 (0.56)0.31 (0.43)0.53 (0.41)0.01 (0.63)0.25 (0.61)0.56 (0.31)0.29 (0.30)
3 outline0.31 (0.10)0.12 (0.10)0.08 (0.09)0.28 (0.09)0.52 (0.06)−0.13 (0.38)0.61 (0.26)0.49 (0.26)0.16 (0.40)0.99 (0.45)0.35 (0.20)0.41 (0.17)
4 a–c0.08 (0.10)0.20 (0.10)0.16 (0.10)0.22 (0.10)0.24 (0.09)0.20 (0.10)0.59 (0.21)0.85 (0.21)−0.21 (0.54)−0.49 (0.62)0.04 (0.32)0.39 (0.29)
4 area0.02 (0.11)0.23 (0.10)0.25 (0.10)0.20 (0.10)0.24 (0.10)0.08 (0.10)0.63 (0.05)0.56 (0.05)0.10 (0.40)0.46 (0.44)0.83 (0.19)0.86 (0.17)
4 outline0.02 (0.11)0.27 (0.10)0.26 (0.09)0.26 (0.09)0.17 (0.10)0.14 (0.10)0.49 (0.06)0.45 (0.01)−0.11 (0.40)0.46 (0.44)0.77 (0.18)0.74 (0.17)
5 a–c0.16 (0.11)0.10 (0.10)0.04 (0.10)0.08 (0.10)0.006 (0.09)−0.04 (0.10)0.06 (0.09)−0.03 (0.10)0.07 (0.10)0.52 (0.50)−0.63 (0.34)−0.42 (0.30)
5 area0.32 (0.10)0.22 (0.09)0.20 (0.09)0.16 (0.09)0.15 (0.09)0.01 (0.09)0.17 (0.09)0.10 (0.10)0.09 (0.09)0.28 (0.08)0.46 (0.34)0.58 (0.31)
HW0.38 (0.12)0.34 (0.10)0.40 (0.10)0.39 (0.09)0.24 (0.11)0.26 (0.11)0.15 (0.12)0.01 (0.12)0.11 (0.11)0.32 (0.12)0.23 (0.11)0.55 (0.04)
EL0.36 (0.12)0.43 (0.11)0.36 (0.09)0.45 (0.08)0.38 (0.11)0.34 (0.11)0.07 (0.12)−0.02 (0.12)0.11 (0.11)0.30 (0.12)0.26 (0.11)1.27 (0.13)

Covariance between the different genital sclerites

When a male is first to mate, his paternity is higher when sclerite 4 is small and sclerite 5 is large. Conversely, when he mates in the second male role his paternity is highest when sclerite 1 is small and sclerite 2 is large (House & Simmons, 2003). Moreover, because a male's success in the offensive and defensive roles are positively related, sclerites 1 and 4 and sclerites 2 and 5 are predicted to be positively related and sclerites 2 and 4 and sclerites 1 and 5 are predicted to be negatively related (Fig. 1). Interestingly, these functional associations were largely matched by the underlying patterns of additive genetic covariance. There were positive genetic correlations between the first principal components that summarized sclerites 2, 3 and 5 (Table 4, 5). In contrast, there were negative genetic correlations between sclerites 1 and 2, between sclerites 1 and 5, between sclerites 2 and 4, and between sclerites 4 and 5 (Table 5). Unlike the additive genetic covariances, the phenotypic (Table 5), maternal and environmental (Table 6) correlations were all positive and significant. Qualitatively similar patterns were observed in the raw covariance matrices presented in Table 2 and 3, although the low underlying VA in these traits, and the resulting extreme genetic correlation estimates and standard errors, make the interpretation more difficult.

Table 4.  Principle component analyses of the different measures of the same genital sclerite and for general morphological measures in O. taurus. PCA's are based on the 548 genital measures that have nonzero additive genetic variance.
 Sclerite 2Sclerite 3Sclerite 4Sclerite 5General Traits
2 b–c2 area2 outline3 b–c3 outline4 a–c4 area4 outline5 a–c5 areaHWEL
Factor loadings0.5110.5830.6310.7070.7070.5270.6080.5940.7070.7070.7070.707
Eigenvalue2.131.612.421.341.90
Variance (%)71.1980.3580.7867.0592.97
Table 5.  Estimates of genetic (above diagonal) and phenotypic (below diagonal) correlations between the PC 1 scores that capture the overall variation in each of the genital sclerites and the PC1 score for general morphology.
PC scoreSclerite 1Sclerite 2Sclerite 3Sclerite 4Sclerite 5General trait
  1. Standard errors are provided in parentheses and significant correlations (P < 0.05) are provided in bold. Measurement of the outline of sclerite 1 is presented as a z-score rather than as a PC1 score because this was the only genital measure that had nonzero additive genetic variance.

Sclerite 1−0.46 (0.44)−0.004 (0.47)0.03 (0.50)−0.61 (0.89)−0.82 (0.78)
Sclerite 20.32 (0.04)0.87 (0.22)−0.81 (0.44)0.13 (0.70)−0.001 (0.46)
Sclerite 30.23 (0.04)0.46 (0.03)−0.40 (0.49)2.15 (1.96)−0.18 (0.68)
Sclerite 40.17 (0.04)0.27 (0.04)0.19 (0.04)−0.92 (0.90)−0.28 (0.71)
Sclerite 50.18 (0.04)0.25 (0.04)0.20 (0.04)0.06 (0.04)−0.28 (1.35)
General trait0.28 (0.04)0.58 (0.03)0.33 (0.04)0.29 (0.04)0.19 (0.04)
Table 6.  Estimates of maternal (above diagonal) and environmental (below diagonal) correlations between the PC 1 scores that capture the overall variation in each of the genital sclerites and the PC1 score for general morphology.
PC scoresSclerite 1Sclerite 2Sclerite 3Sclerite 4Sclerite 5General trait
  1. Standard errors are provided in parentheses and significant correlations (P < 0.05) are provided in bold.

Sclerite 10.22 (0.21)0.11 (0.26)0.51 (0.25)0.02 (0.33)0.32 (0.18)
Sclerite 20.54 (0.10)0.69 (0.20)0.71 (0.23)0.47 (0.33)0.81 (0.10)
Sclerite 30.33 (0.10)0.29 (0.08)0.35 (0.30)0.45 (0.39)0.41 (0.20)
Sclerite 40.03 (0.11)0.27 (0.10)0.22 (0.10)0.12 (0.39)0.73 (0.19)
Sclerite 50.30 (0.10)0.21 (0.10)0.03 (0.10)0.10 (0.10)−0.06 (0.29)
General trait0.38 (0.12)0.55 (0.08)0.36 (0.11)0.08 (0.11)0.35 (0.11)

Not surprisingly, the maternal, environmental and phenotypic matrices for the genital sclerites were statistically similar, whereas the additive genetic matrix was dissimilar to all others (Table 7). Thus, while maternal and environmental sources of variation have common effects on the phenotype of the genital sclerites, the additive genetic sources appear to have opposing effects.

Table 7.  Pairwise comparison of the genetic, maternal, environmental and phenotypic matrices for the genital sclerites in O. taurus.
MatricesGeneticMaternalEnvironmentalPhenotypic
  1. Comparisons were made using a Mantel's test and correlations provided in bold are significant at P < 0.05. Significance tests were attained using 1000 randomizations. Significant correlations mean that the two matrices are significantly equal (i.e. not dissimilar).

Genetic0.120.040.23
Maternal0.260.61
Environmental0.78
Phenotypic

Covariance between the genital sclerites and general morphology

There was an overall trend for the genetic correlation between the PC's describing variation in each of the genital sclerites and general morphology to be negative (Table 5). This pattern should be interpreted with caution, however, because even though all of the genetic correlation estimates were negative, only the correlation estimate between sclerite 1 and general morphology was larger than the associated standard error. In contrast, the phenotypic (Table 5), maternal and environmental (Table 6) correlations between the PC's describing the genital sclerites and general morphology were positive and in the majority of cases significant.

Discussion

Variation in male genital morphology is ubiquitous in animals with internal fertilization and promiscuous mating systems. The mechanism(s) that are responsible for this diversity is the topic of ongoing debate (Eberhard, 1985; Arnqvist, 1997; Hosken & Stockley, 2004). Three major hypotheses currently dominate the literature on genital evolution: the lock-and-key, the pleiotropy and the sexual selection hypotheses (reviewed in Arnqvist, 1997). Arnqvist (1997) advocated the need for more thorough within species studies to compliment the more traditional comparative approaches that have dominated this field of research. More importantly though, he provided a number of within species predictions to assist in distinguishing between the major hypotheses of genitalic evolution, based on (1) the nature of selection operating on male genitalia and (2) the resulting patterns of phenotypic and genetic (co) variation in male genitalia morphology. While there has been a general increase in the number of within species studies on genitalic evolution, most have focussed on levels of phenotypic variation in genital vs. general morphological traits (Arnqvist & Thornhill, 1998; Eberhard et al., 1998; Danielsson & Askenmo, 1999; Kelly et al., 2000; Jennions & Kelly, 2002; House & Simmons, 2003) and how this variation in genitalia influences male fertilization success (reviewed in Hosken & Stockley, 2004). Far fewer studies have examined the underlying genetic architecture of genital traits or tested the various quantitative genetic predictions associated with the major hypotheses of genitalic evolution (but see Arnqvist & Thornhill, 1998; Preziosi & Roff, 1998).

Our quantitative genetic data on the inheritance of male genitalia in O. taurus, coupled with our previous knowledge of how selection operates on male genitalia in this species (House & Simmons, 2003), allows us to assess the relative support for the main hypotheses of genitalic evolution. According to the lock-and-key hypothesis, strong stabilizing selection for preinsemination reproductive isolation is expected to reduce phenotypic and residual coefficients of variation so that the expression of genitalia is largely impervious to environmentally induced deviation (Arnqvist, 1997). Consequently, genitalia that have evolved under the lock-and-key hypothesis should demonstrate lower levels of VA and VP relative to general morphological traits and genetic correlations between genitalia and general morphological traits should be weak or absent (Arnqvist, 1997). In O. taurus, the high CVR's and large nongenetic maternal effects on the expression of genital traits are inconsistent with the view that genitalia are highly canalized. Furthermore, like the study of Arnqvist & Thornhill (1998) on the water strider Gerris lateralis, we have found that male genitalia in O. taurus are as genetically and phenotypically variable as general morphological traits. Therefore our results are largely inconsistent with the predictions of the lock-and-key hypothesis, a result that is largely supported by the findings of numerous comparative (Eberhard, 1985; Arnqvist, 1998; Eberhard et al., 1998) and within species studies (Goulson, 1993; Arnqvist & Thornhill, 1998).

The pleiotropy hypothesis assumes that selection operating on male genitalia is neutral and that genitalic evolution proceeds as a result of indirect selection acting on other genetically correlated morphological traits (Mayr, 1963). Male genitalia that have evolved under the pleiotropy hypothesis are therefore predicted to have high levels of VA and VP and strong genetic correlations with general morphological traits (Arnqvist, 1997). Previously, we have shown that male genitalia in O. taurus are subject to sexual selection, with four of five genital sclerites directly relating to a males fertilization success (House & Simmons, 2003). Coupled with our current finding of low levels of VA in male genitalia, this provides little support for the pleiotropy hypothesis in its strictest sense. However, we did find consistent negative genetic correlations between the PCs describing the variation in each of the genital sclerites and general body size, suggesting that pleiotropy may still play a role in the evolution of male genitalia in this species. Likewise, genetic correlations between genital morphology and body size have been demonstrated in the water striders, Aquarius remigis (Preziosi & Roff, 1998) and G. incognitus (Arnqvist & Thornhill, 1998). In A. remigis, the sign of genetic correlation between body size and genital morphology is positive (Preziosi & Roff, 1998) and selection for increased body size is well documented in this species (Fairbairn, 1988; Fairbairn & Preziosi, 1994). Thus, even though levels of VA in genital morphology are low in A. remigis (h2 = 0.11), strong selection on body size and the pleiotropic effects of genes governing body size and genital morphology may facilitate genitalic evolution in this species (Preziosi & Roff, 1998). Unlike the study of Preziosi & Roff (1998), however, the negative genetic correlations between male genitalia and body size in O. taurus suggests that there may be a trade-off between investment in genital and general body traits (Roff, 2002). Indeed, a recent experimental manipulation on O. taurus using surgical ablation of the genital precursor cells suggests that genitalia and general morphology traits may actually compete for a common resource pool during development (Moczek & Nijhout, 2004). Therefore, strong selection for body size in this species (Hunt & Simmons, 2001) is likely to constrain the evolution of male genitalia. A similar argument was made for the evolution of genital shape in G. incognitus (Arnqvist & Thornhill, 1998).

The sexual selection hypothesis predicts that variation in male genitalia has direct implications for a male's fertilization success and is therefore expected to be the subject of strong directional selection (Arnqvist, 1997). Although traits that are subject to strong directional selection are often predicted to have low levels of VA (Kirkpatrick & Ryan, 1991), a review of the published literature suggests that CVA's are actually significantly higher in sexual than nonsexual traits (Pomiankowski & Møller, 1995). Moreover, both Pomiankowski & Møller's (1995) nonlinear sexual selection model and Rowe & Houle's (1996) genic capture model provide theoretical reasons to account for the higher CVA in sexual traits. Although the proximate mechanisms differ, both models predict higher CVA's for sexual traits as a larger number of loci contribute to trait expression providing greater potential for mutation, environmental deviation and/or genotype-environment interactions to occur (Pomiankowski & Møller, 1995; Rowe & Houle, 1996). Although male genitalia in O. taurus are subject to sexual selection (House & Simmons, 2003), the observed levels of VA for genitalia were relatively low (mean h2 = 0.10). Thus, despite the growing number of empirical studies demonstrating the importance of sexual selection to the evolution of male genitalia (reviewed by Hosken & Stockley, 2004), our quantitative genetic data is in disagreement with the general prediction of the sexual selection hypothesis, that genitalia should exhibit high levels of VA (Arnqvist, 1997). However, until sufficient studies of the genetic architecture of genital morphology are available we are unable to conclude that genital traits generally have lower VA than other sexual traits or morphological traits in general. Indeed, our findings suggest that VA is equally low for general morphological traits in O. taurus, a result that is consistent with a number of other genetic studies in this species (Hunt & Simmons, 2002; Kotiaho et al., 2003).

The argument that low VA in male genitalia is evidence against the sexual selection hypothesis should be interpreted with caution for two major reasons. Firstly, there is still much controversy over how much VA to expect in traits, such as male genitalia, that are closely related to fitness. In a survey of heritability estimates, Mousseau & Roff (1987) found that life-history traits have lower heritability estimates than morphological or behavioural traits, and suggested this was a result of their closer proximity to fitness. Moreover, a number of single species studies on natural populations have also demonstrated a negative association between the heritability of a trait and its correlation with total fitness (Gustafsson, 1986; Kruuk et al., 2000). However, these findings may simply reflect greater levels of environmental and/or nonadditive genetic variance in fitness related traits (Houle, 1992; Merilä & Sheldon, 2000; Blows & Hoffman, 2005). The above inconsistencies suggest that, in general, it may be premature to make firm conclusions regarding the relative magnitude of VA to expect in sexual traits that are closely related to fitness (Merilä & Sheldon, 2000). Moreover, given the large number of mechanisms that can influence levels of VA, it is unlikely that broad generalizations concerning VA across different traits will ever be possible (Blows & Hoffman, 2005).

Second, we argue that Arnqvist (1997) prediction that VA in male genitalia should be high if subject to persistent sexual selection is simplistic and underestimates the complexity present in most genital structures. In general, selection is unlikely to act on single traits in isolation (Lande & Arnold, 1983) and there is growing evidence that correlational selection represents a common and important agent of evolution in natural populations (Brodie, 1992; Schluter & Nychka, 1994; Sinervo & Svensson, 2002; Blows & Brooks, 2003). Furthermore, individual traits forming complex structures are often genetically correlated and are therefore inherited as an integrated unit, rather than as discrete entities (Lande & Arnold, 1983). As a result, there is a critical distinction between the level of VA in individual traits and VA in multiple traits that together form complex structures (Blows & Hoffman, 2005). Even though levels of VA in individual traits are low, there may be sufficient VA in multivariate trait space for selection to operate on (Schluter, 1996). Thus, when considering how much VA is present in a particular trait and how it will respond to selection, it is necessary to view it within the context of a wider set of morphologically integrated traits (Blows & Hoffman, 2005). In this sense, the key parameters describing the evolution of morphologically integrated traits are the additive genetic variance–covariance matrix (G) and how this is aligned with the major axis of selection (Lande, 1979; Lande & Arnold, 1983; Schluter, 1996), rather than levels of VA in each of the individual traits. Indeed, comparing levels of VA across individual traits has been suggested to have little biological relevance (Lynch & Walsh, 1998, p. 561).

Our study suggests that male genitalia in O. taurus should be viewed as a complex structure that is morphologically integrated at two discrete levels. First, we demonstrate that the different morphological measures of the same genital sclerite were positively genetically correlated, suggesting that the same underlying genes govern these traits. Moreover, the maternal, environmental and phenotypic correlations were also positive, suggesting that these causal factors also facilitate the integration of morphological measures within sclerites. Second, and more importantly, we demonstrate that the functional relationships between the different sclerites are largely matched by the structure of the G matrix. When two males compete for fertilizations the paternity of the first male is greatest when sclerite 4 is small and sclerite 5 is large, whereas the paternity of the second male is greatest when sclerite 2 is large and sclerite 1 is small (House & Simmons, 2003). Males, however, do not possess genitalia that are specialized to either the defensive (first male) or offensive (second male) role (House & Simmons, submitted). Rather, there is a positive correlation between a given males fertilization success when mated in both the defensive and offensive role, which suggests that males are simultaneously selected for competition in both roles (House & Simmons, submitted). These functional relationships are reflected in the G matrix with negative genetic correlations between sclerites 2 and 4, 4 and 5, 1 and 5 and 1 and 2, and positive genetic correlations between sclerites 2 and 5 and 1and 4 (see Table 5). Thus, sexual selection for high fertilization success through the appropriate scaling of the different genital sclerites seems to have lead to the integration of these traits at the genetic level (Cheverud, 1984,1995).

Hypotheses of morphological integration predict that functionally and developmentally related traits are inherited as genetically, phenotypically and environmentally integrated units (Cheverud, 1984,1995). While this was the case for the different morphological measures of the same genital sclerite, we found that the genetic, maternal, environmental and phenotypic correlation matrices across the different sclerites were dissimilar in structure. More specifically, although the maternal, environmental and phenotypic matrices were similar in structure to each other, each was dissimilar to the G matrix. In O. taurus, females provide the nutritional environment for the developing larvae (i.e. the brood mass) and these maternal effects are a major determinant of offspring phenotype (Hunt & Simmons, 2000,2002; Kotiaho et al., 2003). Our findings suggest that maternal effects also have large effects on male genital morphology in this species and that these effects are mediated by factors other than the weight of the brood mass, which was statistically controlled for in our analysis. It is not surprising, therefore, that the maternal, environmental and phenotypic correlation matrices are similar in structure. It is of considerable interest, however, that the maternal, environmental and phenotypic matrices are dissimilar to the G matrix. Previous research on O. taurus has shown that females differentially provision their offspring based on their partners phenotype and that these maternal effects can amplify the genetic sire contribution to offspring phenotype (Kotiaho et al., 2003). Contrary to this finding, we demonstrate that that the expression of additive genetic (co) variation for male genitalia may be masked or even opposed by maternal and environmental (co)variation. This is particularly important because at least one maternal effect, the weight of the brood mass, has a genetic basis and therefore can influence the evolution of offspring phenotype via indirect genetic effects (Hunt & Simmons, 2002). Further studies examining how the variation present in the additive genetic, maternal and environmental matrices are aligned with the major axes of selection on male genitalia will be valuable in determining the relative importance of these causal factors to genitalic evolution in O. taurus.

Concluding remarks

Animal genitalia often show rapid morphological differentiation between species. Our understanding of the processes that have shaped genital evolution may be greatly improved by the use of quantitative genetic analyses. An important first step in this process is the estimation of genetic parameters, namely patterns of (co) variation, as they describe the potential for genital traits to evolve in response to selection. However, there are a number of limitations to what can be inferred from these estimates. We argue that many of the quantitative genetic predictions of the major hypotheses of genital evolution (Arnqvist, 1997) may considerably underestimate the complexity of male genital morphology. Moreover, the findings of single species studies are not always representative of widespread patterns. The latter point is often addressed using comparative studies and this approach has proved useful in testing phenotypic predictions (Arnqvist, 1998). For example, Arnqvist (1998) compared monandrous and polyandrous clades of water striders to show that male genital morphology was phenotypically more divergent in polyandrous species, thus providing comparative support for the sexual selection hypothesis. We believe that it would be extremely valuable to apply a similar logic and compare the genetic architecture of genital morphology in species subject to different selective regimes. While the across species comparison of genetic architecture has been used to address other aspects of evolutionary biology, such as the evolution of the G matrix (Begin & Roff, 2001,2003,2004; Roff, 2002), this approach has not yet been applied to the question of male genital evolution. However, before such an approach can be taken, more single species studies that estimate the genetic (co)variance structure of male genital morphology are required. We therefore encourage further detailed single species studies on the genetic architecture of genital morphology and a synthesis of this information within a comparative framework.

Acknowledgments

We thank Rob Brooks for assistance in the use of REML and ASREML, John Hunt for lengthy discussions about what can be inferred from estimates of genetic variance and covariance and Wolf Blackenhorn and an anonymous referee for valuable comments on an earlier version of this manuscript. We also thank Ian Dadour for the use of facilities and Stephanie Maumelat for technical assistance. This research was funded by an Australian Postgraduate Award to C.M.H. and grants from the Australian Research Council to L.W.S.

Ancillary