The evolution of polyandry: patterns of genotypic variation in female mating frequency, male fertilization success and a test of the sexy-sperm hypothesis

Authors


L. W. Simmons, Evolutionary Biology Research Group, Department of Zoology, The University of Western Australia, Nedlands, WA 6009, Australia.
Tel.: +8 9380 2221; fax: +8 9380 1029; e-mail: lsimmons@cyllene.uwa.edu.au

Abstract

Abstract The sexy-sperm hypothesis predicts that females obtain indirect benefits for their offspring via polyandy, in the form of increased fertilization success for their sons. I use a quantitative genetic approach to test the sexy-sperm hypothesis using the field cricket Teleogryllus oceanicus. Previous studies of this species have shown considerable phenotypic variation in fertilization success when two or more males compete. There were high broad-sense heritabilities for both paternity and polyandry. Patterns of genotypic variance were consistent with X-linked inheritance and/or maternal effects on these traits. The genetic architecture therefore precludes the evolution of polyandry via a sexy-sperm process. Thus the positive genetic correlation between paternity in sons and polyandry in daughters predicted by the sexy-sperm hypothesis was absent. There was significant heritable variation in the investment by females in ovaries and by males in the accessory gland. Surprisingly there was a very strong genetic correlation between these two traits. The significance of this genetic correlation for the coevolution of male seminal products and polyandry is discussed.

Introduction

Multiple mating by females with different partners (polyandry) is taxonomically widespread and an increasingly discussed evolutionary puzzle (Keller & Reeve, 1995; Zeh & Zeh, 1996, 1997; Yasui, 1998; Arnqvist & Nilsson, 2000; Jennions & Petrie, 2000). Where females gain immediate benefits from males, such as nuptial gifts, nest sites, or male protection, the occurrence of polyandry is understandable. The problem arises in species where females gain no immediate benefits from their behaviour and where mating incurs costs in the form of time, energy, or increased mortality (Daly, 1978; Arnqvist, 1989; Chapman et al., 1995).

Theoretically, polyandry could arise in the absence of immediate benefits to females, if they obtained some form of indirect genetic benefit for their offspring (Keller & Reeve, 1995; Zeh & Zeh, 1996, 1997; Yasui, 1997). Polyandry facilitates both sperm competition (Parker, 1970; Simmons, 2001b) and cryptic female choice (Thornhill, 1983; Eberhard, 1996) and, theoretically, these mechanisms of postcopulatory sexual selection could provide the necessary indirect benefits for its evolution and maintenance. By mating with several different males and inciting sperm competition, females could increase the probability that their eggs are fertilized by competitively superior sperm and thereby increase the probability that their sons will produce competitively superior sperm, the so-called ‘sexy-sperm’ hypothesis (Sivinski, 1984; Harvey & May, 1989; Curtsinger, 1991; Keller & Reeve, 1995). The precise mechanism underlying the fertilization advantage of superior males (e.g. sperm number or quality, seminal fluid composition, genital morphology) is unimportant, only that some males have a higher fertilization success than others and that fertilization success is heritable. The argument is parallel to that proposed for the evolution of secondary sexual traits in males and female mating preferences, where the female preference and male trait coevolve in a Fisherian runaway process; polyandry is analogous to the female preference because it provides the mechanism by which the male trait, fertilization success, is assessed (Curtsinger, 1991; Keller & Reeve, 1995).

If fertilization success were in some way correlated with male viability, polyandry could also provide indirect benefits to females via a classic good genes process. The so-called ‘good-sperm’ hypothesis predicts that by inciting sperm competition, females could again increase the probability that their eggs were fertilized by competitively superior sperm, but in this case the offspring they produce would have higher general viability because of the correlation between a male's fertilization success and his general viability (Madsen et al., 1992; Parker, 1992; Keller & Reeve, 1995; Yasui, 1997).

Like classical models of precopulatory sexual selection (Kirkpatrick & Ryan, 1991; Pomiankowski et al., 1991; Kokko et al., 2003), both the sexy-sperm and the good-sperm models of postcopulatory sexual selection predict that alleles coding for fertilization success will become associated with alleles coding for polyandry. Therefore, the hallmark of a process of indirect selection on the evolution of polyandry should be a positive genetic correlation between female mating frequency and male fertilization success (Keller & Reeve, 1995; Simmons, 2001b). Under the good-sperm scenario, there should also be a positive genetic correlation between fertilization success and viability (Kirkpatrick & Ryan, 1991; Yasui, 1997). However, as recently pointed out by Pizzari & Birkhead (2002), the mechanisms of genetic inheritance of sperm competition traits complicate predictions of the sexy-sperm and good-sperm hypotheses. An increasing number of studies are revealing that traits that might be important determinants of fertilization success, such as sperm morphology, testes size, or ejaculate features, exhibit patterns of sex-linked inheritance (reviewed in Pizzari & Birkhead, 2002; Simmons & Kotiaho, 2002). If male fertilization success is inherited through female-biased mechanisms, such as sex-linked loci and/or maternal effects, females cannot enhance the fertilization success of their sons through mating with males having high fertilization success. The sexy-sperm hypothesis is more applicable with male-biased mechanisms of inheritance, or where both parents contribute to their sons’ fertilization success, and then the predictions depend critically on whether there is sexual conflict (Parker, 1979; Rice & Holland, 1997; Rice, 2000) which seems inherent in sperm competition (Hosken et al., 2002; Pizzari & Birkhead, 2002). Thus, an examination of the genetic architecture of polyandry and fertilization success should be an integral part of any test of the sexy-sperm or good-sperm hypotheses.

Although there are some studies that have examined the genetics of traits thought to be important in sperm competition (reviewed in Keller & Reeve, 1995; Pizzari & Birkhead, 2002; Simmons & Kotiaho, 2002), there are few detailed quantitative genetic studies of male fertilization success itself. In their study of domestic fowl, Froman et al. (2002) reported a heritability of 0.30 for sperm motility, and motility strongly influences fertilization success. They also showed an exclusive maternal heritability of 0.15, suggesting that some form of maternally inherited element influenced this trait, and ultimately fertilization success. Edwards (1955) demonstrated a heritable component to the probability of fertilization by mouse sperm, and Radwan (1998) used sire on son regression to report a heritability of 0.28 for fertilization success in bulb mites, Rhizoglyphus robini. Hosken et al. (2001) showed that fertilization success in yellow dung flies evolved over 10 generations of polyandrous mating, implying significant additive genetic variance for this trait. Finally, genetic variation in fertilization success has been studied in some detail using different lines of Drosophila melanogaster (Clark et al., 1995). Variation in success in avoiding lost paternity to future males appears to be associated with variation at four different loci on the second chromosome that code for four different accessory gland proteins. However, much of this variation appears to be dominance variance, rather than additive variance (Hughes, 1997).

There is even less information on genetic variation in female mating frequency. Several studies of Drosophila have examined the genetics of remating speed (Pyle & Gromko, 1981; Gromko & Newport, 1988; Gromko, 1992; Singh & Singh, 2001) which may, or may not, correlate with female mating frequency. Solymar & Cade (1990) used mother on daughter regression, reporting a heritability of mating frequency of 0.69 for the field cricket Gryllus integer, and Torres-Vila et al. (2001) were successful in artificially selecting for high and low mating frequencies in the moth, Spodoptera exigua, reporting a realized heritability of 0.73. In a similar study using the moth Loesia botrana, a realized heritability for remating of 0.53 was obtained (Torres-Vila et al., 2002). Finally, Wedell (2001) reported upper limits to heritability for the propensity for female Pieris napi to store apyrene sperm (0.73) and the duration of the refractory period between matings (0.85) using a full sib breeding design. The duration of the refractory period is positively correlated with the numbers of apyrene sperm stored and negatively correlated with a female's lifetime mating frequency. Thus, the upper limit to heritability for lifetime mating frequency was estimated as 0.63 (Wedell et al., 2002).

Here I present a quantitative genetic analysis of reproductive traits in the Australian field cricket, Teleogryllus oceanicus, in an attempt to test predictions of the sexy-sperm hypothesis. In a previous study, I used this species as a model to examine some of the predictions of the good-sperm hypothesis (Simmons, 2001a). Counter to the good-sperm hypothesis, the offspring of males who were successful in sperm competition did not have higher viability, and were not superior in other life-history traits such as adult body size or development speed. However, I did not examine the sperm competitive success of offspring sired by males who were themselves successful. There was considerable variation in sire fertilization success (the coefficient of variation was 53% when two males competed and 82% when four males competed). This variation in fertilization success cannot be explained by variation in the numbers of sperm transferred, or by the length of a male's sperm (Simmons et al., in press). Eberhard (1996, pp. 157–158) argued that field crickets may exercise cryptic female choice via a mechanism of selective sperm transport and utilization, thus biasing paternity towards certain males. Alternatively, males may vary in sperm mobility and/or fertilization efficiency, which generates variation in paternity as a result of sperm competition. Females could potentially gain indirect benefits through either mechanism depending on the genetics of fertilization success. Cryptic female choice is expected to generate the genetic correlation between fertilization success and polyandry predicted by the sexy-sperm hypothesis (Keller & Reeve, 1995; Simmons, 2001b). Here I document patterns of phenotypic and genotypic variation in two male traits that may be important in sperm competition, investment in testes and accessory glands, as well as fertilization success itself. I also examine patterns of phenotypic and genotypic variation in female investment in ovaries and female mating frequency. Finally, I examine genetic covariance between these male and female traits.

Methods

Breeding design

I used a standard half-sib breeding design (Falconer & Mackay, 1996; Lynch & Walsh, 1998). Parents were obtained by collecting 120 adult females from a banana plantation in Carnarvon, North Western Australia. These females were housed individually in small plastic containers (7 cm × 7 cm × 5 cm) supplied with cat chow and a Petri dish containing a pad of moist cotton wool for oviposition. Nymphs emerging from oviposition pads were reared in family groups of 100 individuals housed in 5-L containers with fly screen lids and supplied with cat chow and water ad libitum. The sexes were separated at the penultimate instar. From among the available 120 field derived families, 23 were randomly chosen to provide a single sire for the half sib breeding design. Each of the 23 sires were provided with four dams again from randomly chosen field derived families, subject to the criteria that the family had not provided a sire, and each family provided only one dam. The 23 sire and dam groups were housed in 5-L containers and left for 12 days to ensure all females had mated. Females were then housed individually and allowed to oviposit as described above. Two weeks later nymphs began to emerge from cotton wool pads. I reared 200 nymphs from each family, 100 in each of two 5-L containers, again separating the sexes at the penultimate instar. On adult emergence, all animals were housed individually. These offspring were assessed for the reproductive variables of interest. Because of the deaths of some dams, the final number of dams per sire varied between three and four with a mean of 3.21 ± 0.09. Throughout all rearing and experimental procedures, crickets were housed in a constant temperature room set at 25 °C with a 12 h : 12 h light : dark reversed cycle.

Assessment of polyandry

Five females from each family were placed individually into plastic containers (7 cm × 7 cm × 5 cm) provided with cat chow and a moist pad of cotton wool. Each day, at the onset of the dark cycle, food and cotton pad were removed and females were provided with a single sexually mature adult male taken from a randomly breeding stock culture. Pairs were observed closely to ensure that the male actively courted the female. Any male that did not begin courting within 15 min was replaced. Pairs were left for 1 h to mate. After 1 h each male was replaced with a fresh sexually active male. Females were thus provided with five opportunities to mate each day (males require a 70 min refractory period after mating so could only mate once in the 1 h period, Loher & Rence, 1978) and the process was repeated for four consecutive days, noting with how many of the males each female copulated. A total of 348 females were assessed for mating frequency. Females were frozen at the end of the 4-day period.

Assessment of fertilization success

To assess paternity I used a laboratory strain of T. oceanicus which has a double recessive genotypic marker that has a white-eyed phenotype (see Simmons et al., in press for details). As with polyandry, I attempted to assess the fertilization success of five males from each full sib family. However, as a result of limitations on the numbers of virgin white-eyed females available, the final sample sizes were reduced to an average of 2.0 ± 0.1 (range 1–5) males per dam family and 5.9 ± 0.5 (range 3–10) males per sire family. Nevertheless, all five males per dam family were kept for morphological measurement (see below).

Each male was allowed to mate once with a white-eyed female that had mated once with a white-eyed male 1–2 h previously. Thus, the fertilization success of males was assessed in the role of second male; sequence position has no influence on a males fertilizations success (Simmons, 2001a; Simmons et al., in press). Both matings were carefully observed to ensure successful spermatophore transfer, and that the spermatophore remained attached for longer than 30 min for sperm transfer (but note that variation in spermatophore attachment duration does not influence paternity (Simmons et al., in press)). A total of 136 males were assessed for fertilization success. Following matings, experimental males were frozen. Females were housed individually and allowed to oviposit for 14 days. When nymphs hatched they were scored for eye colour. The proportion of offspring sired by the second, experimental, male was calculated as the number of black-eyed offspring divided by the total number of offspring hatching.

Morphological variables

All frozen crickets were later thawed, weighed to the nearest 0.01 mg and the width of their pronotums measured to the nearest 0.01 mm using digital callipers. Crickets were dissected and the weights of the testes and accessory glands determined for males, and the weight of the ovaries determined for females. For males, soma weight was calculated as body weight minus the sum of the weights of the testes and accessory glands. For females, soma weight was calculated as body weight minus ovary weight. For a subsample of 25 females the ovaries were dissected and the number of mature and developing eggs counted. The relationship between ovary weight and the number of eggs was strong and significant (r2 = 0.80; F1,23 = 93.67; P < 0.001) indicating that ovary weight is a good measure of female investment in oogenesis.

Genetic analyses

Mixed-model nested analyses of variance, with dams nested within sires as a random effect, were used to test statistically for sire and dam effects, using Satterthwaites's approximation of the error term to account for unequal sample sizes of offspring. Because of the unbalanced nature of the design, particularly for data on fertilization success, I used restricted maximum likelihood procedures to estimate observational components of variance. Narrow-sense (sire component) and broad-sense (dam component) heritabilities and their standard errors were calculated according to Becker (1984). Coefficients of variation (CVs) were calculated as inline image where V refers to phenotypic, sire, dam or residual variances (Houle, 1992). Within sex genetic correlations and their standard errors were calculated from standard multivariate nested analyses of covariance (Becker, 1984; Lynch & Walsh, 1998). Genetic correlations between the sexes were calculated following Via's (1984) method 3; within dams, each brother was arbitrarily paired with a sister to calculate covariances using multivariate nested analyses.

Results

Significant sire effects were found for accessory gland weight, ovary weight, and for female body size (Table 1). There were significant dam effects for all traits measured (Table 1). The narrow sense CVA's (sire component) were negligible for paternity (male fertilization success) and for polyandry (female mating frequency). Thus, the narrow-sense heritability estimates for these traits were also negligible (Table 2). In contrast, the broad-sense CVA's (dam component) were very high and yielded high estimates of broad-sense heritability, particularly in the case of paternity (Table 2). Accessory gland weight and ovary weight had quantitatively equivalent narrow and broad-sense CVA's and only slightly higher broad-sense than narrow-sense heritabilities.

Table 1.  Sibling analyses: mixed model nested analyses of variance for reproductive and morphological traits of Teleogryllus oceanicus.
TraitSourceSSd.f.MSFP
  1. To account for unequal sample sizes of offspring within sires, Satterthwaite's approximation of the error term was calculated as: *0.939 × dam[sire] + 0.061 × residual; †0.991 × dam[sire] + 0.009 × residual; ‡0.993 × dam[sire] + 0.007 × residual; §0.997 × dam[sire] + 0.003 × residual.

PaternitySire*2.668220.1230.6710.846
Dam[sire]8.745470.1861.8570.010
Residual6.611660.100  
PolyandrySire†291.752213.261.0090.473
Dam[sire]632.064813.171.8150.002
Residual2024.472797.26  
TestesSire‡225922102.691.1040.376
Dam[sire]43814793.231.4860.028
Residual1813028962.73  
Accessory glandSire‡413822188.102.0000.023
Dam[sire]44364794.392.096<0.001
Residual1301528945.04  
OvariesSire§1286042258452.2840.009
Dam[sire]1229924825621.7620.003
Residual4216192901454  
Soma (male)Sire‡59254222269341.3630.184
Dam[sire]93220747198341.980<0.001
Residual289510828910018  
Soma (female)Sire§41016622186441.6350.078
Dam[sire]54814848114191.907<0.001
Residual17364312905988  
Pronotum (male)Sire‡12.693220.5771.5220.113
Dam[sire]17.887470.3812.043<0.001
Residual54.0202900.186  
Pronotum (female)Sire§8.731220.3971.8530.037
Dam[sire]10.295480.2141.5720.014
Residual39.5652900.136  
Table 2.  Observational coefficients of variation* and narrow-sense (s) and broad-sense (d) heritabilities with their standard errors (approximated after Becker, 1984) for reproductive and morphological traits of Teleogryllus oceanicus.
TraitMeanSDCVPCVAsCVAdCVRhinline imageSEhinline imageSE
  1. *P, phenotypic; A, additive genetic; R, residual.

Paternity0.600.3660.580.0054.2860.780.000.80 ± 0.46
Polyandry13.912.9020.870.0015.4814.000.000.55 ± 0.22
Testes (mg)21.678.3038.357.1821.3237.670.04 ± 0.180.31 ± 0.14
Accessory gland (mg)31.557.7924.7816.1119.5718.830.42 ± 0.180.62 ± 0.22
Ovaries (mg)89.9543.1948.1832.0532.6335.970.44 ± 0.010.46 ± 0.19
Soma (male) (mg)540.97110.3520.497.2415.7819.160.12 ± 0.150.59 ± 0.19
Soma (female) (mg)516.0286.4418.258.2512.6816.290.20 ± 0.190.48 ± 0.21
Pronotum (male) (mm)6.080.487.973.486.307.170.19 ± 0.220.62 ± 0.24
Pronotum (female) (mm)5.650.407.173.824.386.060.28 ± 0.170.37 ± 0.18

Across offspring, paternity was unrelated to any of the traits measured in this study. Polyandry was significantly correlated with ovary weight (r = 0.189; n = 348; P < 0.001). Phenotypic correlations between traits calculated from dam or sire family means are presented in Table 3. It has been argued that these sire-mean and dam-mean phenotypic correlations can provide conservative approximations of the underlying genetic correlations between traits (Simons & Roff, 1996; Roff, 1997; Lynch & Walsh, 1998). They are likely to underestimate the true genetic correlations because the variances of family means will be inflated by environmental variance (Lynch & Walsh, 1998). There was no significant correlation between paternity and polyandry, either across sires or dams. Polyandry was positively associated with ovary weight across dam family means but not sire family means. Neither of these correlations were significant after Bonferroni adjustment of probabilities (Table 3). The only traits that showed consistent strong and significant family mean correlations were measures of body size (pronotum width and soma weight) within the sexes, and accessory gland weight and ovary weight between the sexes. Narrow-sense genetic correlations calculated from nested analyses of covariance between pronotum width and soma weight were 0.921 ± 0.189 for males and 0.924 ± 0.101 for females, higher than the sire-mean correlations as expected. The narrow-sense genetic correlation between accessory gland weight and ovary weight calculated using method 3 in Via (1984) was close to 1 (1.125 ± 0.274) (see Fig. 1).

Table 3.  Phenotypic correlations across sires (above the diagonal) and across dams (below the diagonal) between reproductive and morphological traits of Teleogryllus oceanicus.
TraitPolyandryPaternityPronotum (male)Soma (male)TestesAccessory glandPronotum (female)Soma (female)Ovaries
  1. Across sires (N = 23) Pearson correlation coefficients >0.396 are individually significant at P = 0.05. Across dams (N = 69) the critical value for individual significance is 0.235. Coefficients in bold remain significant after Bonferroni adjustment for the number of correlations performed across sires or across dams.

Polyandry −0.1330.0490.1690.3000.1590.0180.0290.307
Paternity−0.100 −0.181−0.328−0.374−0.068−0.405−0.303−0.295
Pronotum (male)0.0450.107 0.8850.4870.2790.5770.5580.060
Soma (male)0.0810.0940.881 0.5830.3190.4790.4800.178
Testes0.148−0.1060.5840.656 −0.1660.1480.204−0.234
Accessory gland0.1270.1060.1950.145−0.218 0.2910.0930.724
Pronotum (female)0.058−0.1310.2960.2470.2140.161 0.8830.373
Soma (female)0.144−0.0130.3280.2870.1860.1450.870 0.101
Ovaries0.2870.1030.0340.113−0.0990.4340.2190.115 
Figure 1.

The sire family mean (±SE) accessory gland weight of male Teleogryllus oceanicus, plotted against the sire family mean (±SE) ovary weight of females. There is a strong genetic correlation between these two traits (see text for details).

Discussion

Patterns of genotypic variation

Quantitative genetic analyses revealed strong dam effects on paternity, polyandry and testes size. The higher broad-sense than narrow-sense heritabilities suggest some form of sex-biased inheritance for these traits.

Dam effects include additive genetic variance, common environmental variance, and maternal effects (Falconer & Mackay, 1996; Roff, 1997; Lynch & Walsh, 1998). Common environmental effects are unlikely to explain the patterns observed here because all families were raised under identical controlled conditions as pre-adults. Within families individuals were randomly distributed across two separate cages, and as adults they were housed individually during experimental trials to determine paternity and mating frequency. Thus, family members did not experience consistent common environmental conditions throughout their development. Maternal effects on the other hand could contribute to the patterns of variation observed and these typically inflate estimates of heritability (Falconer & Mackay, 1996; Roff, 1997; Lynch & Walsh, 1998). Maternal effects can arise due to differences in, for example, maternal provisioning of young. The types of characters that are influenced by maternal effects are predominantly aspects of body size. The influences are generally strong during early development but rarely persist into adulthood (Mousseau & Dingle, 1991; Mousseau & Fox, 1998). Although there is no maternal provisioning after oviposition, differences in the nutrient content of eggs could influence the size of offspring at hatching, and in crickets, differences in hatchling size can become amplified during development (Simmons, 1987). Broad-sense heritabilities for pronotum width, soma weight, accessory gland weight and ovary weight were on average 2.4 times higher than narrow-sense heritabilities, as might be expected from maternal effects on offspring development.

In contrast, the patterns of genetic variance in paternity and polyandry, and perhaps also testes weight, were qualitatively different. For these traits narrow-sense heritabilities were negligible so that the differences between broad and narrow-sense heritabilities were very much greater than they were for morphological traits. These patterns implicate some form of maternally biased inheritance other than, or in addition to, nongenetic maternal effects. Solymar & Cade's (1990) study of mating frequency in the cricket G. integer reported an estimate of heritability remarkably close to that reported here for T. oceanicus (0.69 vs. 0.55, respectively). Their study used a method of mother on daughter regression so that the estimate cannot be influenced by common environmental effects. However, because of the method used it is not possible to say whether there is a similar maternally biased inheritance in G. integer.

Maternally biased inheritance could arise due to loci coding for polyandry and paternity being on the X-chromosome (in crickets males are XO and females XX, Hewitt, 1979). There is increasing evidence to suggest that traits important in sexual selection are disproportionately influenced by alleles at X-linked loci (Reinhold, 1998; Hurst & Randerson, 1999; Saifi & Chandra, 1999; Ritchie, 2000). Indeed, in field crickets a number of traits important in sex and reproduction are already established as being X-linked. For example, temporal parameters of the males’ calling song and female preferences for male calls in Teleogryllus are both X-linked, as revealed by crosses between the congeneric species T. oceanicus and T. commodus (Bentley & Hoy, 1972; Hoy & Paul, 1973; Hoy, 1974; Hoy et al., 1977). In the confamilial Gryllus bimaculatus, X-linked inheritance is implicated for sperm length (Morrow & Gage, 2001). If sperm length and testis size were positively correlated, as they seem to be in some taxa (Pitnick, 1996; Morrow & Gage, 2000), this latter finding would be concordant with the X-linked inheritance of testis size suggested here for T. oceanicus. In general, a surprising number of traits potentially involved in sperm competition have been shown to be linked to the X, and sometimes to the Y-chromosome (Ward, 2000; Wang et al., 2001; Froman et al., 2002; Pizzari & Birkhead, 2002; Simmons & Kotiaho, 2002). Theory suggests that alleles involved in reproduction should become linked to the sex chromosomes, and particularly the X-chromosome when there is antagonistic coevolution between males and females (Rice, 1984; Rice & Holland, 1997; Gibson et al., 2002). Male and female interests regarding polyandry are unlikely to coincide (Simmons & Gwynne, 1991), and indeed there is evidence to suggest that male Teleogryllus transfer substances in their ejaculate that can inhibit the female's phonotactic response (Loher, 1981). Sexual antagonism between loci coding for male and female influences over remating could thereby underlie patterns of maternally biased inheritance of polyandry found here.

An alternative form of maternally biased inheritance could arise from cytoplasmic factors. It is difficult to see how such factors could influence testis weight or polyandry, but they could be important determinants of paternity, and explain the notably higher broad-sense heritability for this trait over all others. Mitochondrial DNA is inherited through the maternal line, and the mitochondria that provide the energy source for sperm, and thus contribute to its motility, may impact substantially on a male's fertilization success (Cummins, 1998; Kao et al., 1998; Ruiz-Pesini et al., 2000; Froman et al., 2002). Cross-generational studies of inheritance will be necessary to distinguish between alternative mechanism for the patterns of maternally biased inheritance observed in this study.

The sexy-sperm hypothesis

Consistent with previous studies of T. oceanicus, males varied in their ability to gain fertilizations when in sperm competition with other males (Simmons, 2001a; Simmons et al., in press). Testis mass did not influence fertilization success, consistent with Simmons et al.'s (in press) finding that the relative numbers of sperm transferred to the female have little impact on paternity. Nevertheless, there were intrinsic differences in fertilization success between families of males that were inherited via their mother. Dam effects explained over 60% of the variance in paternity. Maternally biased inheritance of fertilization success has also been documented in domestic fowl (Froman et al., 2002). When fertilization success is transmitted via female-biased mechanisms, females are less likely to acquire indirect benefits for their sons via mating with males with superior sperm competitiveness. Selection via sperm competition through the paternal line will not reduce variance in the male trait, which would account for the high broad-sense heritability of paternity. Moreover, the conditions for the evolution of polyandry are restrictive (Pizzari & Birkhead, 2002). Polyandry and cryptic female choice can only be favoured when male fertilization success provides direct benefits for females, and/or female expression of the alleles that code for fertilization success in males is beneficial to females. Neither of these conditions apply to the sexy-sperm hypothesis (Pizzari & Birkhead, 2002). If the mechanism of female-biased inheritance of fertilization success was due to X-linkage, then females could, theoretically, enhance the average fertilization success of their grandsons. Although such a process would only generate half the selection pressure for the evolution of polyandry via a sexy-sperm process compared with autosomal inheritance, a genetic correlation between polyandry and fertilization success would still be expected. However, there was no evidence of a genetic correlation between these traits. Contrary to the good-sperm hypothesis, Simmons (2001a) found no correlation between fertilization success and offspring viability. Thus, the sexy-sperm hypothesis for the evolution of polyandry in T. oceanicus can be rejected.

Sexual selection and the male accessory gland

A surprising result to emerge from this analysis was the strong genetic correlation between accessory gland weight in sons and ovary weight in daughters. Insect accessory gland products (Acps) stimulate oogenesis and oviposition, suppress female sexual receptivity and can influence sperm storage and utilization (see reviews in Eberhard, 1996; Simmons, 2001b). In some cases, notably in Diptera, the action of Acps have been found to be costly for females (Chapman et al., 1995, 1998). Acps are generally thought to serve male interests and there is good evidence in Drosophila that selection via sexual conflict has been critical in their evolution (Rice, 1996, 1998).

Acps that stimulate oogenesis and oviposition have been identified in Teleogryllus. The accessory glands consist of several hundred blind-ended tubules of varying size and structure (Kaulenas et al., 1975). Collectively the tubules of the accessory gland secrete over 30 protein fractions ranging from 10 000 to 100 000 in molecular weight (Kaulenas et al., 1975). Some of these proteins are undoubtedly responsible for construction of the spermatophore as males with surgically removed glands are unable to produce a spermatophore (Loher & Edson, 1973). However, many of the proteins are incorporated into the ejaculate where they are converted by prostaglandin synthetase that is secreted by the testes (Destephano & Brady, 1977) into prostaglandins that stimulate increased oogenesis and oviposition by the mated female (Stanley-Samuelson & Loher, 1983, 1986; Stanley-Samuelson et al., 1987). In general, polyandry increases lifetime egg production by female crickets (Simmons, 1988; Wagner et al., 2001) so that the action of Acps in this group of insects may be beneficial for females, rather than costly. In this study there was no association between accessory gland weight and paternity, suggesting that Acps do not influence sperm competition. But females who mated more frequently had heavier ovaries and were developing more eggs.

An alternative scenario to sexual conflict for the evolution of Acps was suggested by Cordero (1995, 1998) and Eberhard & Cordero (1995). They argued that stimulatory Acps could arise in males through a Fisherian sexual selection process. According to their hypothesis, females bias reproduction towards males able to transfer quantitatively and/or qualitatively superior Acps by variation in their patterns of receptivity, oogenesis and oviposition following insemination. Such cryptic female choice of males based on Acps would favour the coevolution of polyandry, allowing females to sample the ejaculates of multiple males. Even if Acps were costly for females, as they seem to be in Dipterans, females could obtain indirect benefits for their offspring through male manipulation of their reproduction, if males who were successful in manipulating females produced sons with more manipulative ejaculates. Thus, sexual selection for indirect benefits and sexual conflict need not be independent evolutionary processes (Kokko et al., 2003). Like other models of indirect selection, these arguments predict the evolution of a genetic correlation between the mechanism of preference, increased oogenesis, and the male trait, Acp production; alleles that code for the oogenesis response in females should become positively associated with alleles that code for the production of Acps by males.

The data for Teleogryllus are at least consistent with the hypothesis that Acps arise under indirect selection from cryptic female choice. Ovary weight was directly and strongly correlated with the number of eggs produced and it seems reasonable to assume that accessory gland weight should be positively associated with Acp production. There was a phenotypic correlation between the number of times females mated and the weight of their ovaries; this correlation was individually significant across dam family means, indicative of a positive broad-sense genetic correlation between polyandry and oogenesis. More importantly however, the genetic correlation between ovary weight (and thus egg production) and accessory gland weight, predicted by an indirect process of sexual selection, was strong and highly significant. Thus, although males with large accessory glands do not sire a greater proportion of a female's offspring, females that elevate their investment in oviposition following stimulation by male seminal fluids should produce a greater total number of daughters who likewise respond to male Acps and sons who invest more heavily in Acp production.

A purely Fisherian model of selection assumes that the genetic correlation between ovary weight and accessory gland weight reflects the establishment of linkage disequilibrium between genes that influence accessory gland weight in males and those that influence ovary weight in females. However, Lande (1984) argues that selection alone may be unable to maintain levels of linkage disequilibrium as high as that suggested by the genetic correlation between ovary weight and accessory gland weight found in this study. The alternative to linkage disequilibrium is that a single gene or set of genes have pleiotropic effects on ovary and accessory gland investment by males and females. Pleiotropy might arise if during development the same stem cells give rise to ovarian and accessory gland tissue. Thus male and female fecundity could potentially be under common genetic control. Recent research has highlighted antagonistic coevolution between Acps that enhance male fitness and decrease female longevity (Chapman et al., 1995; Rice, 1996; Holland & Rice, 1998). Antagonistic pleiotropy can lead to intersexual ontogenetic conflict evolution (Rice & Chippindale, 2001) and much research is currently focused on the assumption that Acps are costly for females and drive sexual selection via sexual conflict. However, synergistic pleiotropy between male fitness enhancing Acps and female fecundity, such as that suggested here, should promote rather than hinder the evolution of Acps and female responses to them. Indeed, pleiotropic effects of genes influencing Acp production and oogenesis should favour the evolution of polyandry. The variance in quality and/or quantity of Acps received by polyandrous females should be lower, and on average their immediate fecundity higher than that of monandrous females. Moreover, on average, the probability that a female will produce some offspring sired by males carrying genes for high fecundity should be greater for polyandrous females than for monandrous females. Consistent with this scenario, a recent study of bulb mites, Rhizoglyphus robini, found that polyandrous females produced daughters of higher average fecundity than did monandrous females (Konior et al., 2001). In general, fitness benefits for females associated with male Acps are probably more widespread than is currently recognized (Simmons, 2001b).

Concluding remarks

This study adds to a growing body of evidence that traits important in reproduction exhibit strong sex-biased inheritance. Although the genetic architecture of polyandry and paternity precludes evolution of polyandry under the sexy-sperm and good-sperm hypotheses, the results suggest a role of indirect sexual selection in the evolution of male investment in the accessory gland. Mechanisms of cryptic female choice are notoriously difficult to identify from phenotypic studies because biases towards certain males may be equally interpreted as the result of sperm competition (Simmons, 2001b). Quantitative genetic approaches such as the one described here offer a powerful means with which to unravel even the most cryptic mechanism of female choice; genetic covariation between trait and preference will be the evolutionary footprint of cryptic female choice, but not a product of post-copulatory male contest competition. Thus, the discovery of a genetic correlation between female investment in oogenesis and male investment in accessory glands raises the possibility that female response to Acps could represent a mechanism of cryptic female choice in crickets. A thorough exploration of the influence of male investment in accessory glands on Acp production and male and female fitness in T. oceanicus seems warranted.

Acknowledgments

My thanks to Maxine Beveridge, Kylie Shau-Gaull, and Sharne Bentley, without whose assistance the volume of work involved in screening offspring could not have been achieved. Mark Blows and David Hosken provided valuable feedback on the manuscript. This research was supported by the Australian Research Council.

Ancillary