Rapid diversification of male genitalia and mating strategies in Ohomopterus ground beetles

Authors


Yasuoki Takami, Department of Zoology, Graduate School of Science, Kyoto University, Kitashirakawa-oiwake, Sakyo, Kyoto 606-8502, Japan. Tel.: +81 75 753 4077; Fax: +81 75 753 4100; e-mail: takami@terra.zool.kyoto-u.ac.jp

Abstract

We analysed evolutionary diversification and covariation in male genitalia and four mating traits related to sexual selection, i.e. testis size, spermatophore size, copulation duration and post-copulatory guarding duration, in Ohomopterus ground beetles using phylogenetically independent contrasts. Male genital size and mating duration have evolved more rapidly than body size and the other traits studied. Male genital size was negatively correlated with copulation duration, suggesting that elongated male genitalia may enable decreased time investment in a single copulation because it is more effective at facilitating spermatophore deposition. Male genital size was positively correlated with spermatophore size, suggesting coevolution between offensive and defensive male mating tactics because the elongated male genitalia may be advantageous in displacement of rivals’ plug-like spermatophores, and decreased mating duration may intensify sperm competition. Thus, the remarkable diversity of male genitalia in Ohomopterus may have been facilitated by the interplay between inter- and intrasexual selection processes.

Introduction

In promiscuous species, sexual selection is a strong selective agent that may lead to morphological, physiological and behavioural adaptations of organisms (Parker, 1970; Simmons, 2001; Birkhead & Pizzari, 2002). Traits favoured by sexual selection may diversify rapidly, as observed in the interspecific variation in the bright plumage of birds and in the exaggerated horns of beetles (Andersson, 1994). Exaggerated male genitalia are also conspicuous examples of rapid diversification of traits subject to sexual selection (Eberhard, 1985; Arnqvist, 1998; Hosken & Stockley, 2004).

Three mechanisms of sexual selection, i.e. sperm competition, sexual conflict and cryptic female choice, are thought to affect the evolution of male genitalia (Eberhard, 1985, 1996; Alexander et al., 1997; Simmons, 2001; Hosken & Stockley, 2004; Arnqvist & Rowe, 2005). The importance of these mechanisms to genital evolution has been evaluated in many organisms using both experimental and comparative analyses. In damselflies, males use secondary genitalia to remove rival sperm stored in the female reproductive tract (Waage, 1979; Córdoba-Aguilar, 2002); thus, sperm competition may be the main factor in the diversification of male mating structures among damselfly families (Waage, 1986). In water striders, males grasp females (for coercive copulation) using genital spines located at the tip of the abdomen (Arnqvist, 1989; Bertin & Fairbairn, 2005), whereas females often resist these attempts by using their own genital spines to dislodge the males (Arnqvist & Rowe, 1995). Thus, sexual conflict may be a principal cause of the diversification of genital morphology and mating behaviour in water striders (Arnqvist & Rowe, 2002a, b; Rowe & Arnqvist, 2002). Citing numerous examples in insects, spiders and other animals, Eberhard (1985, 1996, 2004) has emphasized the importance of sexual selection via cryptic female choice to the evolution of mating systems. However, there is controversy regarding the specific mechanism responsible for the rapid diversification of male genital morphology (Eberhard, 2004; Hosken & Stockley, 2004), even though the possible mechanisms are not necessarily mutually exclusive.

A comparative analysis approach using multiple species has been used to examine whether the wide diversity of genital structures observed across taxa can be explained by hypothetical underlying mechanisms, such as sperm competition, cryptic female choice and sexual conflict (Eberhard, 1985, 1996, 2004; Arnqvist, 1998; Hosken et al., 2001; Arnqvist & Rowe, 2002a, b; Larivière & Ferguson, 2002; Stockley, 2002; Rodríguez et al., 2004). This method is designed to detect correlated evolution between genital morphology and a measure of selection intensity. In general, the degree of promiscuity may be the best measure of sexual selection intensity (Eberhard, 1985; Arnqvist, 1998), but it is difficult to measure. Testis size has been used as a good indicator of sexual selection intensity because it likely reflects the degree of polyandry within and among species (e.g. Gage et al., 2002; Gomendio et al., 2006). In the intersexual context of genital evolution, i.e. sexual conflict and cryptic female choice, female genital dimensions or morphological correspondence between the male and female genitalia have been recognized as a measure of sexual selection and have been used to predict male genital diversity (Arnqvist & Rowe, 2002a; Eberhard, 2004; Rodríguez et al., 2004). Because the mechanism that drives male genital diversity may vary among taxa, some specific measures of sexual selection intensity have been identified (e.g. physiological condition in mammals; Larivière & Ferguson, 2002; Stockley, 2002). Thus, including both general and specific measures of selection intensity may be critical for comparative analyses of male genital morphology to explore the underlying mechanisms of diversification.

Ground beetles belonging to the subgenus Ohomopterus (genus Carabus) are endemic to the Japan archipelago; they consist of 15 species and many subspecies and exhibit diverse, elaborate genital morphologies (Ishikawa, 1987, 1991; Takami, 2000a). Male beetles have a sclerotized hook-like structure, called a copulatory piece, in the intromittent organ, and it shows remarkable variation among species (Fig. 1). Female beetles have a membranous pocket in the genitalia, called a vaginal appendix, which functions as a receptacle for the copulatory piece during genital coupling (Ishikawa, 1987).

Figure 1.

 Maximum likelihood (ML) tree with molecular clock assumption inferred from five nuclear DNA sequences of Ohomopterus taxa (left). Nodal support was given with bootstrap probabilities (> 50%). Asterisks indicate that the nodes were constrained to be monophyletic in ML analyses, or attached after the tree reconstruction based on taxonomic knowledge. Rectangles show the variations of five measured traits (middle): black, dark grey, light grey and white squares indicate 0–25%, 25–50%, 50–75% and 75–100% of ranked trait values respectively; CPL, copulatory piece length; TES, testis size; SPS, spermatophore size; CPD, copulation duration; PGD, post-copulatory guarding duration; the former three traits were standardized by body length. Male genitalia of representative species are also shown (right); CP, copulatory piece.

In Ohomopterus, the male genitalia plays an important role in inter- and intrasexual mating processes (Takami, 2000b, 2002, 2003, 2007). The copulatory piece is essential for the proper deposition of the spermatophore, as was shown in experiments using Carabus insulicola males with surgically shortened copulatory pieces (Takami, 2003). The length of the copulatory piece may be associated with copulation efficiency in terms of copulation duration because a comparison between species with short (Carabus albrechti) and long (C. insulicola) copulatory pieces showed that species with a short copulatory piece required a longer time to begin spermatophore deposition within the bursa copulatrix, resulting in longer copulation (Takami, 2000b). Such delayed ejaculation indicates the presence of female control (Eberhard, 1996), and highly muscular female genitalia of Ohomopterus support this possibility (Takami, 2002). Thus, a longer copulatory piece may be favoured to enhance spermatophore deposition ability over female control (Takami, 2002, 2003).

Further, the copulatory piece of Ohomopterus males may be an agent of intrasexual selection through sperm competition via spermatophore displacement. In double-mating experiments using C. insulicola, the second male had a high probability (80%) of displacing the first male's spermatophore, suggesting that spermatophore displacement is a highly effective offensive tactic in sperm competition (Takami, 2007). Because the spermatophore of this species strongly adheres to the bursa copulatrix (Takami, 2002), its removal is possible only using sclerotized and elongated genital parts, especially the copulatory piece. Second males who failed to displace the first males’ spermatophores (20%) discontinued copulation before insemination or could not deposit a spermatophore closer to the spermathecal opening. Therefore, the first male's spermatophore may act as a plug to prevent spermatophore deposition by a second male.

Generally, the intensity of sexual selection for genital morphology depends on the frequency and interval of female remating (Eberhard, 1985; Arnqvist, 1998; Hosken & Stockley, 2004); this may also be true in Ohomopterus. In double-mating experiments using C. insulicola, the P2 values (the fraction of offspring sired by the second male) were moderate (0.30–0.57) and were negatively correlated with the post-copulatory guarding duration of the first male because longer post-copulatory guarding by the first male allows more sperm to be transferred into the spermatheca (Takami, 2007). Male fertilization success also depends on the remating interval of females. When a female remated immediately after the first copulation, the second male had twice the paternity rate (P2 = 0.57) than when a female remated 24 h after the first copulation (P2 = 0.30). These results suggest that the effectiveness of spermatophore displacement for male fertilization success, as well as the selection intensity for genital morphology, is lowered by increasing the female mating interval. Several hours of prolonged copulation and guarding by males is enough to avoid immediate remating by females on a given day because these beetles are mostly nocturnal and mating occurs between nightfall and midnight (Sota, 1985).

Given these genital functions and mechanisms of sperm competition in Ohomopterus, we propose the hypothesis that the diversification of male genital morphology has been driven by sexual selection in Ohomopterus ground beetles. We expect several patterns of evolutionary covariation between genital morphology and other mating traits. First, if a longer copulatory piece increases the efficiency of spermatophore deposition, males should reduce the duration of a single copulation to increase their mating opportunities with other females or to avoid the costs of prolonged copulation (e.g. energetic expenditure and predation risk). Thus, negative evolutionary covariation between male genital morphology and copulation duration is expected. An increased efficiency and decreased duration of a single mating implies an increase in the risk of sperm competition. Second, because male fertilization success under sperm competition depends on whether the male can displace a rival's spermatophore, the plug-like spermatophore may select for displacement devices on the male genitalia. Larger spermatophores would be more effective plugs and more elaborate genitalia may be favoured to displace rivals’ spermatophores. Thus, positive evolutionary covariation between male genital morphology and spermatophore size is expected. Concomitantly, testis size, a good indicator of the degree of sperm competition, may show positive evolutionary covariation with male genital morphology. Finally, under higher risk of sperm competition, longer post-copulatory guarding may be favoured to avoid sperm displacement (Parker, 1974; Yamamura, 1986; Yamamura & Tsuji, 1989). However, as the guarding duration increases, the risk of sperm competition decreases (Parker, 1984). Therefore, the patterns of covariation between post-copulatory guarding duration and other traits could depend on sperm competition risk and cannot be predicted simply.

Our goal was to evaluate the hypothesis that the diversification of male genital morphology has been driven by sexual selection in Ohomopterus ground beetles. To examine the expected evolutionary covariation between genital morphology and other mating traits, we used comparative analyses of phylogenetically independent contrasts based on molecular phylogenetic trees.

Materials and methods

Taxon sampling, mating behaviour and genital morphology

Comparative analyses with nonrandom taxon sampling may suffer biased estimates of correlation, high type I error rates and low statistical power (Ackerly, 2000). Ohomopterus consists of 15 species, and many groups that vary geographically are classified as subspecies (Ishikawa, 1991). To avoid the above-mentioned statistical problems and include the entire range of diversity within this group, we sampled 22 taxa, including all species and several subspecies, which showed marked variation in genital morphology and body size (Table 1).

Table 1.   Male body length, copulatory piece length, testis weight, spermatophore weight, copula duration and post-copulatory guarding duration in 22 taxa of Ohomopterus ground beetles.
TaxaLocalityMale body length (mm)Copulatory piece length (mm)Testis weight (mg)Spermatophore weight (mg)Copulation duration (min)Post-copulatory guarding duration (min)
  1. Mean ± SD. (n) is shown.

albrechtiTsukui, Kanagawa20.88 ± 0.54 (24)0.813 ± 0.048 (9)3.39 ± 0.97 (7)1.40 ± 0.30 (5)92.64 ± 45.76 (14)14.07 ± 23.62 (14)
lewisianusKagosaka-pass, Yamanashi20.03 ± 0.53 (19)0.705 ± 0.041 (12)2.18 ± 0.54 (6)1.39 ± 0.34 (8)172.43 ± 60.08 (7)20.29 ± 13.92 (7)
lewisianus awakazusanusMt Tomisan, Chiba20.36 ± 0.45 (14)0.659 ± 0.028 (12)1.37 ± 0.26 (3)1.17 ± 0.21 (3)197.00 ± 59.98 (17)11.38 ± 15.36 (16)
kimuraiMt Ryusozan, Shizuoka18.53 ± 0.55 (4)0.659 ± 0.023 (4)1.86 ± 0.91 (3)0.66 ± 0.05 (3)139.33 ± 41.41 (3)0.33 ± 0.58 (3)
yamatoSuzuka, Mie19.54 ± 0.56 (15)0.597 ± 0.028 (15)1.27 ± 0.22 (6)0.86 ± 0.23 (14)229.20 ± 47.63 (15)3.67 ± 4.32 (15)
japonicusOhnogahara, Ehime20.08 ± 0.64 (18)0.602 ± 0.029 (18)2.78 ± 0.52 (6)1.68 ± 0.58 (15)214.00 ± 83.53 (16)17.93 ± 32.86 (15)
japonicus tsushimaeIs. Tsushima, Nagasaki23.29 ± 0.64 (7)0.799 ± 0.035 (7)2.66 ± 0.57 (5)1.34 ± 0.21 (7)109.71 ± 59.68 (7)52.00 ± 42.11 (7)
daisenMt Daisen, Shimane21.57 ± 0.76 (18)0.852 ± 0.044 (18)3.09 ± 0.94 (4)1.87 ± 0.33 (19)93.21 ± 38.77 (19)8.63 ± 15.23 (19)
dehaaniiSuzuka, Mie29.56 ± 1.02 (13)1.064 ± 0.054 (10)6.47 ± 1.04 (5)2.71 ± 0.87 (11)94.77 ± 33.09 (13)34.15 ± 54.84 (13)
tosanusOhnogahara, Ehime29.33 ± 0.70 (7)1.203 ± 0.044 (7)5.38 ± 1.05 (5)2.83 ± 0.54 (7)172.14 ± 35.18 (7)69.85 ± 45.67 (7)
tosanus ishizuchianusMt Ishizuchisan, Ehime24.43 ± 1.07 (13)1.056 ± 0.057 (13)4.62 ± 0.68 (5)1.73 ± 0.58 (12)76.42 ± 35.17 (12)49.67 ± 56.24 (12)
yaconinusSakyo, Kyoto27.21 ± 0.79 (15)1.233 ± 0.053 (15)5.88 ± 1.40 (6)3.91 ± 0.80 (14)50.07 ± 21.82 (14)11.38 ± 20.93 (13)
iwawakianusSuzuka, Mie23.53 ± 0.55 (17)1.143 ± 0.055 (13)3.36 ± 0.25 (6)3.78 ± 0.67 (10)149.23 ± 98.82 (13)28.46 ± 51.38 (13)
iwawakianus kiiensisMt Koyasan, Wakayama21.42 ± 0.41 (4)1.035 ± 0.022 (4)2.58 ± 0.59 (4)3.08 ± 0.31 (4)275.25 ± 72.89 (4)52.00 ± 73.54 (4)
esakiiFuji, Shizuoka22.67 ± 0.93 (11)1.098 ± 0.089 (11)3.63 ± 0.51 (5)2.59 ± 0.47 (9)69.00 ± 23.18 (13)14.46 ± 17.75 (13)
insulicolaTsukui, Kanagawa28.38 ± 0.98 (12)3.310 ± 0.030 (6)7.94 ± 1.63 (5)4.27 ± 0.87 (7)78.29 ± 40.74 (24)40.13 ± 41.01 (24)
arrowianusTokoname, Aichi26.74 ± 1.15 (19)2.818 ± 0.123 (19)5.51 ± 1.08 (4)5.20 ± 0.60 (6)29.0 ± 5.42 (7)0.57 ± 1.13 (7)
arrowianus komiyaiSakuma, Shizuoka21.66 ± 0.82 (13)1.491 ± 0.040 (13)4.25 ± 0.17 (2)2.67 ± 0.45 (13)44.54 ± 18.65 (13)4.15 ± 4.83 (13)
maiyasanusSakyo, Kyoto23.03 ± 0.63 (11)2.422 ± 0.138 (11)6.03 ± 0.88 (5)3.67 ± 0.50 (11)194.27 ± 44.03 (11)13.73 ± 20.50 (11)
maiyasanus suzukanusSuzuka, Mie24.84 ± 0.74 (8)2.784 ± 0.245 (7)5.37 ± 0.73 (8)3.80 ± 0.81 (17)129.53 ± 48.18 (17)27.12 ± 34.24 (17)
maiyasanus shigarakiAyama, Mie23.48 ± 0.56 (4)2.661 ± 0.073 (4)3.54 ± 0.93 (4)4.07 ± 1.16 (3)149.17 ± 31.11 (6)15.00 ± 20.36 (6)
uenoiMt Kongosan, Osaka26.17 ± 0.58 (21)8.598 ± 0.231 (13)6.02 ± 0.49 (7)5.89 ± 1.07 (14)32.42 ± 6.86 (19)16.53 ± 33.44 (19)

Sexually active beetles were caught using pitfall traps during the reproductive season (May to June). Males and females were housed separately in plastic boxes (12.5 × 12.5 × 9 cm3) with moistened moss in an incubator under long-day conditions (16L : 8D) at 20 °C to maintain sexual maturity. Beetles were maintained in these conditions for at least 48 h before mating trials to ameliorate the potential effect of previous mating in the wild. Beetles were fed minced beef and chopped apples every other day.

In the evening, under interior light at 23 ± 2 °C, we introduced a male and a female into a plastic box (12.5 × 12.5 × 9 cm3) and observed their mating behaviour. In this group of beetles, males initiate mating by mounting the female's back. For pairs that mated, we measured copulation duration and post-copulatory guarding duration, rounded to the nearest 1 min. Copulation duration was defined as the time from insertion of the aedeagus to its withdrawal, and post-copulatory guarding duration was defined as the time from withdrawal of the aedeagus to separation of the beetles. Beetles that did not commence mating within 10 min were returned to the incubator for later trials. After mating, beetles were immediately frozen.

Male genitalia were then dissected and the length of the copulatory piece was measured to the nearest 0.01 mm under a binocular microscope using an ocular micrometer. A pair of testes was removed and stored in absolute ethanol for at least 2 days; they were then dried in a desiccator, and the dry weight was measured to the nearest 0.01 mg using an electronic balance. Female genitalia were also dissected to remove the spermatophore from the bursa copulatrix; the spermatophore was then weighed to the nearest 0.01 mg. The weights of testes and spermatophores were converted to the cubic-root equivalent for comparison with other measurements (testis size and spermatophore size respectively). We pinned and dried the beetle carcasses and then measured male body length to the nearest 0.01 mm using electronic calipers; male body length was defined as the distance from the anterior margin of the labrum to the apex of the elytra and was used to control for interspecific variation in body size. Measurements were log-transformed before analyses and averaged within taxa. Interspecific differences among measurements were tested using anova.

Phylogenetic analysis

Phylogenetic trees of Ohomopterus have been proposed based on morphology (Takami, 2000a), mitochondrial DNA (mtDNA) sequences (Su et al., 1996) and nuclear DNA sequences (Sota & Vogler, 2001, 2003). However, there are large discrepancies among these phylogenetic trees, mostly between data derived from mtDNA-based analyses and other data sets, because of frequent hybridization and introgression of mitochondria among species (Sota & Vogler, 2001; Sota, 2002). Sota & Vogler (2003) proposed a reliable species phylogeny for the group based on five nuclear DNA loci, but their parsimony analyses provided little information on branch lengths.

Including branch lengths in comparative analyses has the potential to improve estimates of evolutionary correlations among traits (Martins & Garland, 1991). We performed maximum likelihood (ML) analyses on partial data from Sota & Vogler (2003) and Sota & Ishikawa (2004). For 20 taxa within Ohomopterus, five nuclear DNA loci were available (wingless, phosphoenolpyruvate carboxykinase, elongation factor-1α, cytochrome c and carab1; 2457 bp in total; Sota & Vogler, 2003). As outgroups, we used Carabus fiduciarius (subgenus Isiocarabus, the sister group of Ohomopterus), Carabus vanvolxemi (subgenus Carabus s. str.), Leptocarabus procerulus and Apotomopterus porrecticollis. Because data were not available for two subspecies of C. maiyasanus (maiyasanus and suzukanus), these were attached to the branch of the conspecific subspecies shigaraki with equally segregated branch lengths and the topology based on current taxonomic understanding (Fig. 1; Ishikawa & Kubota, 1994).

Appropriate substitution models for the nuclear gene sequence data were determined using Modeltest version 3.06 (Posada & Crandall, 1998). An ML tree was constructed under the GTR + G + I substitution model with parameters obtained from Modeltest using PAUP* version 4.0b10 (Swofford, 2002), using a heuristic search of 100 random-addition analyses with TBR branch swapping and the MulTrees option activated. To ensure the monophyly of each species, as well as among related species, topological constraints were induced to four nodes (Fig. 1; Sota & Vogler, 2003). This was necessary because the nuclear data failed to resolve the monophyly of three species that we examined. Statistical support for the branches was calculated from bootstrap probabilities based on 100 pseudoreplications.

An ML tree with a molecular clock assumption was obtained separately using the same tree search method in PAUP* (Fig. 1). The likelihood-ratio test for ML trees with and without molecular clock assumptions was, however, significant, and the hypothesis of a molecular clock was rejected [−2Δ = −2(7910.5 − 7890.9) = 39.3, d.f. = 23, P < 0.025]. Thus, we used two types of trees: ultrametric tree with the molecular clock assumption and nonultrametric tree without the assumption.

Phylogenetic signal

Before conducting comparative analyses of the traits, we assessed the presence of phylogenetic signals in the traits using random taxon reshuffling (Maddison & Slatkin, 1991; Blomberg et al., 2003; Laurin, 2004). First, we calculated the sum of squared changes of a continuous trait on the ML tree. Second, we generated 10 000 random trees based on the uniform speciation model (Yule model), keeping the depth of each tree in accordance with the molecular clock assumption and constructed a null distribution of the sum of squared changes of the trait. We then tested whether the observed sum of squared changes was less than 95% of the null distribution. These procedures were based on the ultrametric tree with molecular clock assumption and were performed using Mesquite version 1.05 (Maddison & Maddison, 2004).

Comparative analysis

Statistical tests of correlated evolution were based on phylogenetically independent contrasts (Felsenstein, 1985; Harvey & Pagel, 1991; Garland et al., 1992). We used 21 independent contrasts within Ohomopterus. Because our phylogenetic tree contained branch lengths as well as topology, raw contrasts were standardized by the standard deviations of the contrasts (Garland et al., 1992). The standardization of contrasts extracts the evolutionary rate of the traits per unit branch length (i.e. time) and makes it possible to compare contrasts. Because we found nonsignificant correlations between the standard deviations of contrasts and the absolute values of standardized contrasts (P > 0.05 in all cases), we concluded that the logarithmic transformation of the branch lengths was unnecessary (Garland et al., 1992).

Interspecific variation in the traits was evaluated using the absolute values of standardized independent contrasts, which show the relative evolutionary rates of a trait at any given node (Garland et al., 1992). A larger absolute value of contrast indicates more rapid evolution. The relative rates of trait evolution were compared using a Friedman test and pairwises sign test, in which proportional changes were compared because all measurements were log transformed.

Correlated evolution between the traits was tested using linear regression through the origin (Garland et al., 1992). To assess the effect of body size on the other traits, as well as to evaluate the relative variation among them, standardized independent contrasts of copulatory piece length, testis size, spermatophore size, copulation duration and post-copulatory guarding duration were regressed on those of male body length. The signs of these independent contrasts were altered in relation to the positivization of male body length contrasts (Garland et al., 1992). To examine whether interspecific variation in copulatory piece length was explained by the four other variables, we constructed univariate and multivariate regression models. Multiple regression models were constructed using a backward stepwise manner in which nonsignificant terms at P > 0.10 were removed from the final models. Additionally, we checked correlations between the four independent variables and confirmed no strong correlations (r21 = −0.32 to 0.24, P > 0.15), except between copulation duration and post-copulatory guarding duration (r21 = 0.56, P = 0.0078). The calculation of contrasts was performed using the PDAP module for Mesquite (Midford et al., 2003). Other statistical analyses were performed using JMP version 5 (SAS Institute Inc., 2004).

Results

In all six traits (Table 1), there were highly significant differences among taxa (anova, P < 0.0001 for all traits). The sequence of mating behaviour for all species examined was similar to that of C. insulicola reported by Takami (2002). Random taxon reshuffling of the data revealed the occurrence of significant phylogenetic signals in all six traits (P < 0.0001 in all traits). Thus, the use of comparative analyses with the phylogenetic tree was justified.

Evolutionary rates

The absolute values of standardized independent contrasts differed significantly between the traits (Friedman test, inline image = 65.95, P < 0.0001), indicating that there was a significant difference in evolutionary rate among traits. Pairwise comparisons between traits revealed that copulatory piece length, copulation duration and post-copulatory guarding duration evolved significantly more rapidly than male body length, and that post-copulatory guarding duration showed the most rapid evolution (sign test, P < 0.05 after sequential Bonferroni correction, Table 2). However, the evolutionary rates of testis size and spermatophore size were almost equal to that of male body length (Table 2).

Table 2.   Tests for variation and allometric slopes in six traits of Ohomopterus ground beetles based on 21 standardized independent contrasts.
 Absolute value of standardized independent contrastAllometric slope on male body length
RMAMAOLS
MeanSESlopeSlopeSlope (95% CI)SEtd.f.P
  1. Absolute values of standardized independent contrasts are compared by sign test, in which different letters indicate significant difference (P < 0.05 after sequential Bonferroni correction). Allometric slopes were evaluated by linear regression through the origin. RMA, reduced major axis; MA, major axis; OLS, ordinary least squares for which 95% confidence intervals and statistical tests are shown.

Male body length0.993a0.133       
Copulatory piece length3.520b0.8794.5606.9242.920 (1.385, 4.455)0.7833.73200.0013
Testis size1.188ab0.1591.2971.2700.898 (0.529, 1.267)0.1775.0720< 0.0001
Spermatophore size0.951a0.1891.1141.1520.840 (0.508, 1.172)0.1595.2920< 0.0001
Copulation duration5.039c0.777−5.292−14.287−1.931 (−4.228, 0.366)1.101−1.75200.095
Post-copulatory guarding duration8.484c0.9677.92437.6331.929 (−1.803, 5.652)1.7891.08200.29

A significant effect of male body length on copulatory piece length, testis size and spermatophore size was detected (Table 2). The slope of the allometry, representing evolutionary allometry (Klingenberg, 1996), was evaluated using reduced major axis (RMA), major axis (MA) and ordinary least squares (OLS) regressions. The slope of copulatory piece length was significantly steeper than one, even in the OLS regression, which was the most conservative model. These results indicate positive allometry. In contrast, all three regression slopes for testis size and spermatophore size were almost equal to unity, indicating that these traits have evolved isometrically. To remove the effect of male body length on copulatory piece length, testis size, spermatophore size and copulation duration (including a weak negative effect), residuals from OLS regression lines were computed and used in the following analyses (Garland et al., 1992).

Correlated evolution

Copulatory piece length was not associated with testis size in any of the univariate or multivariate regressions based on ultrametric and nonultrametric trees (Table 3). Instead, variation in copulatory piece length was significantly explained by that of spermatophore size and copulation duration in univariate and multivariate regressions based on the ultrametric tree. Spermatophore size and copulation duration were positively and negatively associated with copulatory piece length respectively (Table 3, Fig. 2). A similar trend was found in the models based on the nonultrametric tree, but the significance of copulation duration became marginal (Table 3). Post-copulatory guarding duration was never associated with copulatory piece length.

Table 3.   Effects of testis size, spermatophore size, copulation duration and post-copulatory guarding duration on copulatory piece length.
 UnivariateMultivariate
β ± SEtd.f.Pβ ± SEtd.f.P
  1. Univariate and multivariate linear regression models are constructed based on 21 independent contrasts, and are forced through the origin. Optimal models selected by backward stepwise procedure were shown in multivariate regressions.

Ultrametric tree with molecular clock assumption    Selected model (r2 = 0.35)F = 5.172, 190.016
 Testis size1.39 ± 0.941.48200.155    
 Spermatophore size2.22 ± 0.992.25200.0362.04 ± 0.922.23190.038
 Copulation duration−0.30 ± 0.14−2.12200.047−0.28 ± 0.13−2.11190.049
 Post-copulatory guarding duration−0.11 ± 0.09−1.21200.239    
Nonultrametric tree without molecular clock assumption    Selected model (r2 = 0.32)F = 4.472, 190.026
 Testis size1.39 ± 0.931.49200.151    
 Spermatophore size2.15 ± 1.022.10200.0482.02 ± 0.962.11190.048
 Copulation duration−0.28 ± 0.14−1.95200.065−0.26 ± 0.13−1.97190.064
 Post-copulatory guarding duration−0.12 ± 0.09−1.29200.211    
Figure 2.

 Partial regression plots: (a) between residual copulatory piece length contrasts and residual spermatophore size contrasts; and (b) between residual copulatory piece length contrasts and residual copulation duration contrasts, based on the multiple regression model with the ultrametric tree (Table 3).

Discussion

Comparative analyses of empirical data from 22 Ohomopterus taxa using a molecular phylogenetic tree show two significant results. First, male genital morphology and duration of both copulation and post-copulatory guarding have evolved more rapidly than the sizes of the male body, testis and spermatophore. Second, the rapid diversification of male genital morphology is associated with variation in spermatophore size and copulation duration. The former result provides quantitative confirmation for the general trend in male genital morphology, i.e. rapid diversification among species, and suggests that the male genital morphology and mating duration of Ohomopterus are subject to strong selective forces. The latter result supports the hypothesis that the diversification of male genital morphology has been driven by sexual selection in Ohomopterus ground beetles, as discussed below.

Multiple indirect measures of selection may covary together in response to actual selection intensity, potentially making the interpretation of patterns of correlated evolution between traits difficult because true causal relationships are obscured by correlated evolutionary responses. Here, however, covariation among the four measures of sexual selection was mostly weak and not significant, indicating that they have evolved independently from each other. The only significant correlation was between copulation duration and post-copulatory guarding duration. Although post-copulatory guarding duration was not significantly associated with the trait of interest (copulatory piece length), the negative association between post-copulatory guarding duration and copulatory piece length indicates similar relationship between copulation duration and copulatory piece length (Table 3). Therefore, these durations should be interpreted as a unit. In the following section, we attempt to interpret the observed patterns of correlated evolution as hypothesized causal relationships, keeping in mind that there may be covariation in mating duration and traits not examined.

Evolution of male genitalia

Rapid diversification of male genitalia in Ohomopterus ground beetles may have resulted from two different sexual selection processes: intersexual selection (sexual conflict or cryptic female choice) and intrasexual selection (sperm competition). In the intersexual process, a longer copulatory piece may have improved spermatophore deposition efficiency and allowed a decrease in the duration of a single copulation. This could then increase the frequency of the second process, i.e. immediate remating by once-mated females and subsequent spermatophore displacement. More elongated male genitalia may be advantageous in displacing a rival's plug-like spermatophore. Consequently, the male genitalia and the spermatophore may coevolve, with the elaboration of one trait leading to further elaboration of the other. Elongated copulatory pieces could also be beneficial in the intersexual process. Thus, the remarkable diversification of the copulatory piece of Ohomopterus ground beetles may have been facilitated by the interplay between inter- and intrasexual selection processes.

Spermatophore size is significantly associated with male genital morphology, although testis size is not (Table 3). This may be caused by spermatophore structure; the spermatophore of Ohomopterus ground beetles consists of a large gelatinous mass secreted from the accessory glands and contains only a small amount of sperm (Takami, 2002). Although relative testis size may be a measure of sperm mass in an ejaculate, it may not be a good predictor of spermatophore size, which may be the focus of selection. Because we did not consider the size of accessory glands, it remains to be tested whether their size is associated with spermatophore size and male genital morphology among species.

We found a weakly negative, but nonsignificant association between male genital morphology and post-copulatory guarding duration (Table 3). Post-copulatory guarding duration is the most labile of the measured traits and it may show large environmental variance, possibly owing to a behavioural response to individual mating conditions (e.g. physiological conditions or perceived risk of sperm competition).

Variation in the male mating strategy

Adaptations for paternity assurance in mating systems under sperm competition are thought to be driven by two opposing selective forces: selection favouring offensive tactics for pre-emption of stored rival sperm and counterselection favouring defensive anti-pre-emption tactics (Parker, 1970, 1984). The offensive tactics involve takeover and subsequent sperm displacement, whereas the defensive tactics consist of mate guarding by a mating plug, prolonged copulation, pre- and post-copulatory guarding and takeover avoidance behaviour (Parker, 1984; examples summarized in Birkhead & Møller, 1998 and Simmons, 2001). Selection for these two tactics is conflicting, and theory predicts that the only evolutionary stable strategy (ESS) is to invest in both tactics (Parker, 1984).

Prolonged copulation and guarding can benefit males by hindering female remating (Alcock, 1994) and benefit females by providing an opportunity to exercise cryptic female choice (Eberhard, 1996), but it may be costly to both sexes because of increases in predation risk and energetic expenditures (Dickinson, 1997). Optimal copulation and guarding duration could be determined by the balance between these costs and benefits according to the physiology and environment of each species (Shuster & Wade, 2003). Some factors, such as operational sex ratio or population density, could affect the frequency of mating and adaptive investment into male mating strategies (Yamamura, 1986; Simmons, 2001).

In Ohomopterus, copulation and post-copulatory guarding durations had the largest variation among species and their evolution was positively correlated, suggesting that they are strongly selected together. The total duration ranged from 30 min to longer than 5 h (Table 1), the latter was likely long enough to prevent female remating on that day. Prolonged copulation and guarding could be efficient defensive tactics when they are sufficiently long, whereas shorter durations could increase the spermatophore deposition efficiency and number of matings. Thus, these contrasting aspects of copulation and guarding duration may be associated with variation along a continuum of relative investment in offensive and defensive male mating tactics among species of Ohomopterus. The remarkable diversity of male genital morphology (involved in offensive tactics) may also be seen as an evolutionary consequence of the same continuum, on which a more elongated copulatory piece represents a more offensive male mating strategy.

In conclusion, we found that the rapid diversification of male genital morphology is associated with the size of the plug-like spermatophore and copulation duration in Ohomopterus ground beetles. Elongated male genitalia may have been favoured to improve spermatophore deposition efficiency by intersexual selection and to displace rival spermatophores by intrasexual selection, and mating durations may have been associated with the degree of sperm competition. The observed diversity in male genital morphology and other mating traits can be seen as variation in relative male investment into offensive (spermatophore displacement) and defensive (prolonged mating and plug-like spermatophore) tactics. Our results indicate that the rapid diversification of male genital morphology may be driven by the interplay between multiple sexual selection processes.

Acknowledgments

We thank I. Dohzono, H. Fukuda and N. Nagata for their help in collecting beetles. YT thanks T. Yamasaki and R. Ishikawa for their supervision. Supported by the Sasakawa Scientific Research Grant from the Japan Science Society (1998–1999, nos 10–239 and 11–225k), Research Fellowships for Young Scientists from Japan Society for the Promotion of Science (1999–2000, no. 07249 and 2003–2005, no. 04747), a Grant-in-aid from JSPS (no. 15207004), and a Grant for the Biodiversity Research of the 21st Century COE (A14) from the Ministry of Education, Culture, Sports, Science and Technology.

Ancillary