SEARCH

SEARCH BY CITATION

Keywords:

  • cuticular hydrocarbons;
  • disruptive selection;
  • mate choice;
  • sexual selection;
  • sexual dimorphism;
  • stabilizing selection

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Sexual dimorphism is thought to result from directional sexual selection acting on male signal traits, with female signal traits given little, if any, attention. Here, we examine male mating preferences in the Australian field cricket, Teleogryllus oceanicus. Using a multivariate selection analysis approach, we found that male preferences have the potential to exert selection on female cuticular hydrocarbons, chemical compounds widely used as sexual signals in insects. In addition to finding both stabilizing and disruptive preference gradients, we also found weak negative directional preference for female cuticular hydrocarbons. We contrast our results with a recent study examining sexual selection via female choice on male T. oceanicus cuticular hydrocarbons and suggest that differences in the form and intensity of sexual selection between the genders may provide part of the net selection differential necessary for the evolution of sexual dimorphism in this species.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Sexual dimorphism is presumed to reflect adaptive divergence in response to selection favouring different optimal character states in the two sexes. There must, therefore, be a net selection differential between the sexes for sexual dimorphism to evolve (Lande, 1980; Cox & Calsbeek, 2009). Sexual selection could drive differentiation between the sexes, if the signal trait is used during mate choice. Mate choice is often assumed to be the prerogative of females, because females tend to have a putatively larger reproductive investment than males. As such, the net selection differential resulting in sexual dimorphism is often thought to be a product of directional sexual selection acting on male signal traits, such as elaborate plumage or courtship displays (see Kraaijeveld et al., 2007 for review). However, recent evidence suggests that spermatogenesis is far from limitless and males can also show a high selectivity towards their mates, thus maximizing their reproductive success (Bonduriansky, 2001; Wedell et al., 2002). Therefore, in some species, males may also be choosy, indicating the potential for sexual selection to act on the same signal trait in both sexes. The opportunity for sexual selection to act on female as well as male traits means that the net selection differential among the sexes cannot be fully understood without an analysis of sexual selection on both sexes (Kraaijeveld et al., 2007).

Cuticular hydrocarbons are chemical compounds found on the cuticle of most terrestrial arthropods. These chemical compounds are used widely as sexual signals in the selection of mates in insects, and a recent study has shown that these compounds are sexually dimorphic in a range of species, with either qualitative or quantitative differences in compounds between the genders (Thomas & Simmons, 2008b). Despite the large number of species that display sexual dimorphism for cuticular hydrocarbons, to date, research on mutual sexual selection acting on these compounds is limited to species in the genus Drosophila. Recent work on D. serrata has shown that males and females differ in the form and strength of sexual selection acting on cuticular hydrocarbons; female mate preference resulted in primarily linear sexual selection on male cuticular hydrocarbons, whereas male mate preference resulted in primarily nonlinear sexual selection on female cuticular hydrocarbons (Chenoweth & Blows, 2005). Aside from studies on Drosophila, characterization of the form and intensity of sexual selection acting on cuticular hydrocarbons in both males and females is lacking.

Most interest in Orthopteran sexual signals has focussed on auditory communication and therefore female mate choice. However, increasing evidence suggests that cuticular hydrocarbons can provide an additional sexual signal on which females and males can base their choice of mates. Cuticular hydrocarbons are known to be important sex recognition cues for male and female crickets (Otte & Cade, 1976; Rence & Loher, 1977). Moreover, recent research has shown that males can use these chemical signals to discriminate between females of different mating status (Thomas & Simmons, 2007, 2009a). In the current study, we use a multivariate approach to estimate the form and intensity of male preferences for cuticular hydrocarbons of female field crickets, Teleogryllus oceanicus. The cuticular hydrocarbons of this species are sexually dimorphic with both quantitative and qualitative differences being found between the genders (Thomas & Simmons, 2008b). We contrast our results with a recent study examining sexual selection on male T. oceanicus cuticular hydrocarbons (Thomas & Simmons, 2009b). In so doing, our study provides one of only a few studies that investigate how male mating preferences differ from female mating preferences for the same signal trait.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Animals for this experiment were the direct offspring from field collected females originating from a banana plantation in Carnarvon, Western Australia. All crickets were maintained on a 12 : 12 h light : dark cycle in a constant temperature room (25 °C). Crickets were supplied with water and fed cat chow ad libitum. Sexes were separated before the penultimate instar. Following the imaginal moult, experimental crickets were housed individually in boxes (7cm × 7cm × 5cm) and left to mature for 14 ± 3 days to ensure sexual receptivity, before being used in experiments.

Mating trials

Prior to their use in experiments, males were mated with a nonexperimental female, because (i) in some cases, a cricket’s choosiness can increase after their first mate (Bateman et al., 2001) and (ii) to purge males of their old spermatophores, forcing them to invest in the production of a fresh spermatophore when offered an experimental female. Spermatophores are discreet vessels containing sperm, which remain attached outside the female following mating. During mating trials, we recorded two measures of female attractiveness: (i) spermatophore production and (ii) courtship commencement. Spermatophore production was a measure of the time it took males to ejaculate and manufacture a spermatophore while in the presence of the experimental female; time costs involved in replenishing sperm reserves may cause males to be selective in their choice of mate (Härdling et al., 2008). Courtship commencement was a measure of the time it took males to commence their courtship song after extruding their fresh spermatophore. In most populations of T. oceanicus, courtship song appears to be essential to induce mounting by females (Libersat et al., 1994). To minimize observer disturbance during mating trials, we conducted trials in a room dimly lit by red incandescent lights. We also placed each pair of crickets in a small plastic box (7cm × 7cm × 5cm) that was then placed inside a larger plastic box (17cm × 12cm × 6cm), which was lined with packing material on three sides to minimize the mating pairs exposure to other mating pairs under observation.

Cuticular hydrocarbon analysis

To quantify differences in cuticular hydrocarbon profiles, we immersed freshly freeze-killed individual crickets in 5 mL of hexane for 5 min. We injected 1 μL of this sample into a gas chromatograph and mass spectrometer (Agilent GC-6890N; MS-5975 with inert Mass Selective Detector Agilent Technologies Australia, Phy Ltd, Forest Hill, Australia) operating in the splitless mode and fitted with a Stabilwax column of 30 m × 0.25 mm internal diameter using helium as a carrier gas. We optimized the separation of the extract by using a column temperature profile, in which the analysis began at a temperature of 50 °C for 1 min and rose to 250 °C (rate of increase 20°C min−1). The transfer line from the GC to the mass spectrometer was set at 250 °C. We analysed washes derived from 89 females. We also analysed hexane blanks to control for potential contamination of samples.

For data analysis, peaks were labelled by peak number, which corresponded to their retention times (Table 1). Cuticular hydrocarbon profiles of each female consisted of the relative abundances of 27 individual compounds. Note that a previous study using females of this species reported only 25 compounds (Thomas & Simmons, 2008b). In this study, we were able to differentiate 27 compounds because of the use of a more sensitive GCMS machine and different analysis software. We transformed these 27 peaks into log contrasts (using peak 6 as the divisor), as described previously for T. oceanicus cuticular hydrocarbon data (Thomas & Simmons, 2008a,b). We performed a principal components analysis (PCA) on these log contrasts. The PCA provided a new set of standardized uncorrelated variables, ideal for calculating multivariate selection gradients (Lande & Arnold, 1983).

Table 1.   The principal component analysis of trait loading shows the correlations between the relative concentrations of cuticular hydrocarbon peaks and the six components (PC1-6) extracted from the principal component analysis
PeakHydrocarbonMWRTPC1PC2PC3PC4PC5PC6
  1. Peaks that contribute significant amounts to the principal components are in bold. RT, retention time; MW, molecular weight. Peak numbers in parentheses correspond to those in Thomas & Simmons’s study (2008b).

1 (1)Unresolved11.65−0.830.010.440.130.090.12
2 (2)C2940811.84−0.840.010.370.210.000.01
3 (3)C29:140611.95−0.690.130.560.140.120.14
4Unresolved12.18−0.710.120.390.270.150.05
5 (4)X-meC3143612.360.77−0.250.140.17−0.06−0.16
7 (6)Unresolved12.83−0.310.390.26−0.020.12−0.30
8 (7)C31:143412.860.46−0.040.06−0.350.50−0.06
9 (8)C31:143413.100.840.170.380.04−0.030.20
10 (9)C31:143413.140.76−0.360.210.250.110.09
11 (10)C31:243213.220.24−0.380.380.510.49−0.13
12 (11)C31:243213.300.79−0.460.250.02−0.100.09
13 (12)C31:243213.400.84−0.470.150.09−0.020.01
14 (13)C31:243213.490.760.530.070.040.110.01
15 (14)X-meC3346414.080.050.81−0.130.030.200.27
16 (15)Unresolved14.130.64−0.150.240.17−0.160.44
17 (16)C33:146214.280.400.530.000.51−0.070.06
18 (17)C33:146214.390.600.680.110.01−0.03−0.09
19 (18)C33:146214.660.660.430.13−0.16−0.050.43
20 (19)C33:146214.780.510.470.050.160.440.09
21C33:246014.890.25−0.050.480.720.080.08
22 (20)C33:246014.940.640.330.420.05−0.300.22
23 (21)C33:246015.060.800.430.270.18−0.100.03
24 (22)C33:246015.200.920.190.140.170.12−0.03
25 (23)C33:246015.300.860.300.030.180.24−0.09
26 (24)C35:248817.550.200.410.200.000.400.50
27 (25)C35:248817.710.320.770.090.120.07−0.17

Characterizing male mating preferences

We used standard selection analysis methodology to characterize the form and intensity of male preference functions (Lande & Arnold, 1983). Although, our response variables, spermatophore production and courtship commencement, capture variation in female attractiveness, they may or may not translate into variation in female fitness. We discuss how our measures of male preference might translate to variation in female fitness in the discussion. However, to avoid implying variation in fitness in our analysis, rather than use the term ‘selection’ we use the term ‘preference’.

To characterize the form of male preferences acting on female cuticular hydrocarbons, we used separate multiple regressions to estimate the vector of univariate linear preference gradients (β), and the matrix of quadratic and correlational preference gradients (γ) for both measures of female attractiveness (Lande & Arnold, 1983). Before analysis, we calculated a relative measure for each of our attractiveness measurements by dividing each datum by the mean. To ease interpretation, we then took the reciprocal of both attractiveness measures on the grounds that males should take longer to produce a spermatophore and to commence courtship when in the presence of an unattractive female.

To determine the significance of univariate linear and nonlinear preferences for both measures of male response, we assessed the fit of the respective models (see Thomas & Simmons, 2009b for methods). We used the overall significance of the regression model incorporating only the univariate linear (β) term, to evaluate whether linear preference was occurring. To evaluate the significance of nonlinear preference, we used response surface methodology (Phillips & Arnold, 1989), implemented in JMP®7. As we used the response surface methodology, there was no need to double the quadratic coefficients (Stinchcombe et al., 2008). This method results in a canonical analysis of the matrix of nonlinear preference gradients. The canonical analysis allows an interpretation of both concave and convex preference to be made on combinations of traits that describe the greatest amount of nonlinear variation on response surfaces (Phillips & Arnold, 1989; Blows & Brooks, 2003; Stinchcombe et al., 2008). The resulting eigenvectors (mi) from the canonical analysis denote the major axes that constitute the M matrix. These indicate how the original traits, or principal components, contribute to the major axes of the response surface. To evaluate the significance of multivariate linear and nonlinear preferences, new variables were created from the eigenvectors of γ, and a second quadratic regression was performed using these new variables (Blows & Brooks, 2003).

To visualize the nonlinear male preference functions, we compared the response surfaces comprising the major axes of the separate canonical rotations using a nonparametric approximation of the response surfaces (Brodie et al., 1995). The response surface was generated using the Tsp function of the fields package in R, which generates a thin-plate spline by minimizing the GCV score (ver 2.4.1, R Development Core Team, http://www.r-project.org). Visualization of these splines suggested that male preferences may be generating both stabilizing and disruptive selection. We tested this observation statistically using methods originally outlined by Mitchell-Olds & Shaw (1987). Briefly, this method tests for a nonintermediate optimum by constraining the attractiveness optimum at either the maximum or minimum (stabilizing preference) or median (disruptive preference) of the phenotypic distribution. We then test whether this new constrained phenotypic distribution produces a significantly worse fit than the unconstrained (original) model, using standard techniques for testing a general linear hypothesis. Further details outlining the implementation of this procedure can be found in Chenoweth et al.’s study (2007).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Principal component (PC) analysis returned 26 components. We used only those components where the eigenvalue was >1 (Norman & Streiner, 1994) in our selection analyses. Six components had an eigenvalue >1 and collectively explained 81% of the variance in cuticular hydrocarbon blend. The percentage of variance explained was 42.43, 16.39, 7.55, 6.19, 4.59 and 4.03 for components 1–6 respectively. The principal component analysis matrix of trait loading (Table 1) shows how the original peaks contribute to each PC. Correlations >0.7 times the largest correlation were considered to have contributed significantly to the PC (Mardia et al., 1979). PC1 was most strongly loaded by peaks corresponding to C31 and C33 alkenes, whereas the remaining PCs can be interpreted as representing relative abundances of a range of hydrocarbons (Table 1).

Female cuticular hydrocarbon blends explained a significant proportion of the variation in both measures of male preference. The percentage of variance explained by these models was 43% and 39% for the measures spermatophore production and courtship commencement, respectively. The standardized univariate linear, quadratic and correlational preference gradients are presented in Tables 2 and 3. We found significant univariate linear preference on PC2 for both measures of male response (Table 2 and 3). We also found correlational preference between PC4 and PC2 and quadratic preference on PC5 when spermatophore production was considered the response variable (Table 2). When courtship commencement was considered the response variable, we found significant quadratic preference on PC4 (Table 3).

Table 2.   The vector of standardized linear selection gradients (β) and the matrix (γ) of standardized quadratic and correlational selection gradients for the six principal components with eigenvalues greater than one. The response variable is inverse relative spermatophore production
Inverse relative time to produce spermatophore
 βPC1PC2PC3PC4PC5PC6
  1. *P < 0.01, **P < 0.05.

PC1−0.166−0.006−0.010−0.0200.076−0.0240.021
PC2−0.423*0.082−0.0680.272**−0.208−0.194
PC3−0.2290.056−0.287−0.316−0.090
PC4−0.0140.224−0.280−0.400
PC5−0.325−0.589**0.112
PC6−0.056−0.025
Table 3.   The vector of standardized linear selection gradients (β) and the matrix (γ) of standardized quadratic and correlational selection gradients for the six principal components with eigenvalues greater than one. The response variable is inverse relative courtship commencement
Inverse relative courtship commencement
 βPC1PC2PC3PC4PC5PC6
  1. *P < 0.01, **P < 0.05.

PC10.0380.006−0.002−0.0110.0200.0140.022
PC2−0.085*0.0120.0090.039−0.0070.069
PC3−0.0100.005−0.009−0.0610.010
PC4−0.092−0.051**−0.0840.008
PC50.0340.0070.105
PC6−0.0310.025

Both of our measures of male response generated significant multivariate nonlinear preference on female cuticular hydrocarbons (Tables 4 and 5). To visualize the different nonlinear preference functions, we compared female attractiveness surfaces comprising the major axes of the canonical rotation. For the measurement of spermatophore production, we found two axes of nonlinear preferences, as indicated by the significant positive (m1) and negative (m6) eigenvalues (Table 4). We also found two axes of multivariate linear preferences (m4 and m6). The m6 axis is associated with high abundances of PC5, whereas the m1 axis contrasts negative values of PC3 & 6 with positive values of PC4. When plotting the response surface for the two major canonicals that represent the eigenvectors with the strongest positive (m1) and negative (m6) eigenvalues, the surface is a saddle with peaks at the extremes of the m1 axis, suggesting disruptive selection along this axis (Fig. 1). However, using the statistical procedures recommended by Mitchell-Olds & Shaw (1987), we found that a model constrained to have an attractiveness peak at the median value for m1 did not give a significantly worse fit (t87 = 1.19, P = 0.239). Along the m6 axis, the surface is suggestive of directional nonlinear selection (Fig. 1).

Table 4.   M matrix of eigenvectors from the canonical analysis of γ for the six principal components. θ is the strength of directional selection along each canonical axis as estimated from multiple linear regression. Significance of eigenvalues assessed using a second quadratic regression on the m variables. The response variable is inverse relative spermatophore production
Inverse relative spermatophore production
miθλiPC1PC2PC3PC4PC5PC6
  1. The eigenvalue (λi) of each eigenvector (mi) is given in the third column.

  2. θ, multivariate linear selection; λ, multivariate non-linear selection.

  3. *P < 0.05, ***P < 0.001.

m1−0.0860.467*0.0560.426−0.2120.773−0.134−0.393
m2−0.1940.128−0.1150.2680.852−0.123−0.228−0.350
m30.1100.0090.676−0.6130.2570.283−0.054−0.133
m4−0.367*−0.0200.7110.597−0.029−0.260−0.0260.263
m5−0.057−0.176−0.1440.0280.3200.458−0.1720.798
m6−0.416***−0.668*0.0120.1120.2470.1720.9470.005
Table 5.   M matrix of eigenvectors from the canonical analysis of γ for the six principal components. θ is the strength of directional selection along each canonical axis as estimated from multiple linear regression. Significance of eigenvalues assessed using a second quadratic regression on the m variables. The response variable is inverse relative courtship commencement
Inverse relative courtship commencement
miθλiPC1PC2PC3PC4PC5PC6
  1. The eigenvalue (λi) of each eigenvector (mi) is given in the third column.

  2. θ, multivariate linear selection; λ, multivariate non-linear selection.

  3. *P < 0.05, **P < 0.01, ***P < 0.001.

m1−0.0040.084**0.1460.258−0.183−0.1150.6130.700
m2−0.114***0.041**0.0130.6710.4340.323−0.4190.284
m30.0030.0090.8450.047−0.4000.293−0.161−0.109
m40.059*−0.0010.460−0.4440.725−0.1710.0480.186
m50.032−0.042**0.2170.5320.199−0.4630.330−0.554
m6−0.042−0.088***−0.076−0.0410.2320.7430.558−0.275
image

Figure 1.  The attractiveness surface of the major canonical axes m1 and m6 of the attractiveness measure, inverse relative spermatophore production. The two axes represent the eigenvectors with both the strongest nonlinear selection (highest eigenvalues) and the strongest positive (m1) and negative (m6) eigenvalues.

Download figure to PowerPoint

The preference surface obtained using the measure of relative courtship commencement was broadly similar (Table 5, and Figs 2 and 3). Both m1 and m6 were again the two major canonicals that represent the eigenvectors with the strongest positive (m1) and negative (m6) eigenvalues. However, the m2 and m5 axes were also found to be significant (Table 5). We plotted two attractiveness surfaces; one for the two major canonicals that represent the eigenvectors with the strongest positive (m1) and negative (m6) eigenvalues, and one for the two remaining significant canonicals (m2 and m5). We found that the surface is a peak along both the m1 and m2 axes, suggestive of stabilizing preferences (Figs 2 and 3). When we tested these peaks statistically, we found that a model constrained to have an attractiveness peak at the minimum or maximum did not produce a significantly worse fit along the m1 axis (minimum, t85 = −0.85, P = 0.397; maximum, t85 = −0.94, P = 0.352; Fig. 2), whereas the trend towards stabilizing preference along the m2 axis did give a significantly worse fit along the m2 axis (minimum, t85 = 2.50, P = 0.015; maximum, t85 = 3.98, P < 0.001; Fig. 2). Along both the m5 and m6 axes, there appears to be strong directional nonlinear preference (Figs 2 and 3). We also found significant multivariate linear preference along the m2 axis (Table 5).

image

Figure 2.  The attractiveness surface of the major canonical axes m1 and m6 of the attractiveness measure, inverse relative courtship commence. The two axes represent the eigenvectors with the strongest positive (m1) and negative (m6) eigenvalues.

Download figure to PowerPoint

image

Figure 3.  The attractiveness surface of the major canonical axes m2 and m5 of the attractiveness measure, inverse relative courtship commence. The two axes represent significant eigenvectors with positive (m1) and negative (m5) eigenvalues.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Our study provides evidence of a number of significant axes indicative of stabilizing and nonlinear directional male mating preferences acting on female cuticular hydrocarbons. We recognize that our sample size is relatively small (N = 89) for a formal selection analysis and it is possible that with a larger sample size, we may have found even more significant axes. Nevertheless, for both of our male preference measures, we found significant nonlinear directional preferences along the m6 axes. PC5 and its constituent hydrocarbons were major contributors to variation along the m6 axes for both measures of male preference. Four peaks contribute significantly to PC5, three of which display sexual dimorphism: peak 26 is found only in females, and peaks 8 and 20 are found in relatively larger quantities in males than females (Thomas & Simmons, 2008b). When spermatophore production time is considered the response variable, our selection analysis suggests that males display a preference for females with cuticular blends containing high amounts of the female specific compound 26 (C35:2). That males find this compound attractive in females is indicated by their readiness to ejaculate and manufacture a spermatophore when this compound is in high abundance (extreme negative scores on the m6 axes in Fig. 1). In contrast, when courtship commencement is considered the response variable, we find that males display a rapid onset of courtship towards females with cuticular blends containing low amounts of compound 26 (high scores on the m6 axis in Fig. 2). This discrepancy can be explained by two factors. First, in addition to PC5, PC4 is also a major contributor along the m6 axis for the measurement courtship commencement. This PC accounts for three hydrocarbons, one of which (Peak 17) is more abundant in females than males. Males commence courtship with females faster if they display a blend of hydrocarbons consisting of low levels of compounds 17 and 26. It seems to be a balance in the amount of these two female compounds that makes females relatively attractive to males. Secondly, the intensity of nonlinear directional preferences acting on female cuticular hydrocarbons is of an order of magnitude higher for the measurement spermatophore attachment (−0.668; Table 4) than courtship commencement (−0.088; Table 5).

For the measure courtship commencement, we found peaks along the m1 and m2 axes (Figs 2 and 3), indicative of stabilizing preferences. However, when tested quantitatively, only stabilizing preference along the m2 axis was found to be significant. Stabilizing selection is convex nonlinear selection, in which the optimum phenotype is at an intermediate point in the range of phenotypes in the population (Lande & Arnold, 1983). Our data suggest that male preferences may impose stabilizing selection on female CHCs. For the attractiveness measure courtship commencement, PC2 and its constituent hydrocarbons were major contributors to variation along the m2 axis (Table 5), whereas the m5 axis contrasts PC2 with PC6 (Table 5). These PCs account for a range of hydrocarbons (Table 1), four of which are found in relatively higher abundance in females than males (Thomas & Simmons, 2008b). Our selection analysis suggests that the attractiveness peak at the extreme negative score of the m5 axis is associated with an imbalance between the relative amounts of these female-abundant compounds (high levels of compound 15, 26 and 27 and low levels of compound 19; Fig. 3, Table 5).

In addition to nonlinear preferences, we also found significant multivariate linear preferences acting on both of our attractiveness measures (Tables 4 and 5). The intensity of the multivariate linear selection acting on female cuticular hydrocarbons appears to be of an order of magnitude greater than that previously reported for female preferences acting on males [largest significant θ for females −0.416 (table 4), and males −0.061 (Thomas & Simmons, 2009b)]. This result is in stark contrast to that found in Drosophila, where males experience stronger multivariate linear sexual selection on cuticular hydrocarbons than females (Chenoweth & Blows, 2005). This same pattern is also true for univariate directional preferences (β), which appear to be relatively greater in intensity in females than males [largest significant value for females −0.423 (Table 2) and males 0.018 (Thomas & Simmons, 2009b)]. In fact, the intensity of directional preferences found acting on females is considerably stronger than the median magnitude of directional sexual selection gradients (β = 0.16) found across all taxa by a recent comprehensive review (Kingsolver et al., 2001).

In our previous study of sexual selection acting on male cuticular hydrocarbons, we measured the number of mates obtained as a direct measure of male fitness (Thomas & Simmons, 2009b). As such, in that study, we were able to conclude that selection acted on male cuticular hydrocarbons. Here, we have measured male preference functions for female cuticular hydrocarbons, and it is unclear how these preferences might impact female fitness. Even if female attractiveness leads to variation in female mating frequency, typically female fitness is traditionally thought not to depend on the number of mates obtained (Bateman, 1948; Shuster & Wade, 2003). However, previous studies on T. oceanicus have shown that the survival of embryos produced by females is increased when they obtain seminal fluid proteins from different males (Simmons, 2001; García-González & Simmons, 2005, 2007). Moreover, male T. oceanicus have been shown to adjust the quality of their ejaculates in relation to female quality (Thomas & Simmons, 2007). As such, the number of mates obtained and the quality of their ejaculates are expected to impact directly on the numbers of offspring females produce, and thus their fitness. If male preferences based on female cuticular hydrocarbons do indeed impact female mating success and subsequent fitness, we would expect the patterns of selection on females to reflect qualitatively at least, the patterns of preferences we have observed in this study.

Interestingly, two compounds that were found to be subject to stabilizing female preference in males (Thomas & Simmons, 2009b) were found to be subject to nonlinear directional male preference in females (Peaks 8 & 20) (Tables 1, 4 and 5), suggesting that males and females can exercise preference on the same signal trait in fundamentally different ways. Male preferences for the same cuticular hydrocarbons as females could arise initially as a correlated response to the evolution of female mating preferences for the same signal trait in males (Amundsen, 2000). However, a genetic correlation between the sexes will hamper the evolution towards sexual dimorphism when selection on males and females is in opposite directions (Lande, 1980; Kraaijeveld et al., 2007). Only one quantitative genetic study, to date, has estimated the genetic correlation between the sexes for cuticular hydrocarbon traits and found that intersex genetic correlations are low for many of the hydrocarbons, due partly to sex-limited expression of X-linked genetic factors (Chenoweth & Blows, 2003). Unfortunately, the generality of this pattern is difficult to assess without further quantitative genetic studies. Moreover, we are unable to make a formal quantitative comparison of our attractiveness surfaces between males and females, because the cuticular hydrocarbon profiles of males and females are not the same; there are quantitative differences in some peaks, whereas others are unique to females. This means that the PC scores for the female profile are qualitatively different and cannot be contrasted directly with those published previously for males (Thomas & Simmons, 2009b). Regardless, the difference in the form and strength of preferences between the genders reported here may provide part of the net selection differential between the sexes necessary for the evolution of sexual dimorphism in this species.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

This work was supported by funding from the Australian Research Council, the University of Western Australia and the West Australian Centre of Excellence in Science and Innovation Program. Thanks to M. Blows for helpful suggestions on the analysis, and M. Beveridge and A. Denholm for assistance with animal husbandry.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References