Sexual selection across sensory modalities: female choice of male behavioral and gustatory displays

Sexual selection across sensory modalities: female choice of male behavioral and gustatory displays Abstract The role of cuticular hydrocarbons in sexual displays has received considerable interest over the last two decades. For example, multiple studies have documented significant directional and nonlinear sexual selection acting on the cuticular hydrocarbon profiles of both male and female insects. The majority of these studies have excluded other sensory modalities that may influence attractiveness and measured selection using laboratory raised individuals. Furthermore, much of this work has been conducted using drosophilid fruit flies and crickets, and investigations using different taxa are necessary to improve our understanding of broader taxonomic trends. Here, we extend our understanding of sexual selection on cuticular hydrocarbons by measuring selection imposed by female mate choice on male bull-horned dung beetles, Onthophagus taurus. Both male and female beetles used in our study were collected from the field, ensuring that our estimates of selection incorporated some degree of naturally occurring variation in both cuticular hydrocarbon profiles and female mate preferences. Consistent with previous studies on this species, we found significant directional selection on male courtship displays. We also found significant nonlinear selection on the male cuticular hydrocarbon profile acting independently of the influence of behavioral courtship. Our data are consistent with a role for cuticular hydrocarbons in the mating system of this species and suggest that female O. taurus use multiple sensory modalities to assess different aspects of male quality. Introduction Sexual selection through mate choice has resulted in the evolution of some of the most elaborate traits found in biology (Darwin 1871; Andersson 1994). A multitude of studies have documented how an individual’s mating success can be influenced by characteristics such as bright coloration (Hill 1991; Brooks and Endler 2001), loud vocalizations (Rebar et al. 2009; Tanner et al. 2017), and conspicuous behavioral displays (Byers et al. 2010; Callander et al. 2012). A recognition that mate choice can be influenced by multiple components within and across sensory modalities (Candolin 2003; Hebets and Papaj 2005; Partan and Marler 2005) has led to a multivariate approach to their analysis (e.g. Pryke et al. 2001; LeBas et al. 2003; Bentsen et al. 2006; Gerhardt and Brooks 2009; Cole and Endler 2015; Tanner et al. 2017). Multivariate selection analyses allow for the elucidation of the marginal effect of each measured trait on mating success and can help avoid potentially false conclusions of selection acting on a given trait that can arise from phenotypic correlations between that trait with other traits that are associated with fitness (Lande and Arnold 1983; Phillips and Arnold 1989; Blows 2007; Chenoweth et al. 2012). Although the role of chemical traits in mate choice has received relatively less attention (Johansson and Jones 2007; Coleman 2009), in the context of multivariate selection analyses, the cuticular hydrocarbons (CHCs) of insects have received considerable interest (Steiger and Stökl 2014). CHCs form a protective layer on the insect’s cuticle, primarily protecting it from desiccation, but also playing a role in close-range/contact sexual communication (Blomquist and Bagnères 2010). For example, multiple studies have found consistent and relatively strong directional sexual selection acting on the CHC profile of male Drosophila serrata (Blows et al. 2004; Hine et al. 2011; Gershman et al. 2014). These, and associated investigations, have contributed greatly towards our understanding of the multivariate nature of mate choice. Notably, by comparing the vector of directional selection gradients with the genetic (co)variance matrix (G), a lack of genetic variance in the direction of selection has been revealed, a perspective which contrasts with that gained from examining individual traits in isolation (Blows et al. 2004; Hine et al. 2004; Van Homrigh et al. 2007). However, a weakness of multivariate selection analyses is the potential influence of unmeasured characters (Lande and Arnold 1983; Mitchell-Olds and Shaw 1987). If a trait being examined is phenotypically correlated with an unmeasured character that is itself under selection, the resultant selection gradients may reflect this correlation rather than direct selection on the trait itself (Lande and Arnold 1983). Accounting for all traits influencing attractiveness is therefore an important goal in studies examining the multivariate nature of mate choice. Little attention, however, has been given to investigating how traits from other signal modalities act alongside CHCs in determining overall attractiveness. Where both CHCs and nonchemical traits have been incorporated into one study, the evidence to date suggests that the attractiveness of an individual’s CHC profile is independent of the attractiveness of trait(s) in other sensory modalities. For example, no association was found between the measures of male attractiveness for calling song and chemical cues (presumed to be CHCs) in the cricket Gryllus integer (Leonard and Hedrick 2010). Similarly, no correlation was found between an individual’s courtship song attractiveness and CHC attractiveness in the Australian field cricket, Teleogryllus oceanicus, (Simmons et al. 2013). Just as a correlated trait can influence estimates of selection gradients, a trait influencing attractiveness independently of the traits of interest should have little impact on the selection gradients when included in the analysis. For example, in an analysis that incorporated two wing-shape measures (McGuigan 2009), the estimated linear selection gradients on D. bunnanda CHCs were similar to selection gradients found when only CHCs were included (Van Homrigh et al. 2007), suggesting that these two sets of traits influence attractiveness independently. Regardless of whether the attractiveness of an individual’s CHC profile is correlated with the attractiveness of traits in other sensory modalities, overall attractiveness is still likely to be influenced by both CHCs and non-CHC traits. For example, Rybak et al. (2002) showed that both CHCs and song are important in male D. melanogaster courtship, with the removal/reduction of either modality decreasing the attractiveness of males. Incorporating non-CHC traits into multivariate selection analyses will likely provide a greater understanding of mate choice within the study system, as well as ensuring the estimated selection gradients are not biased by unmeasured characters. Estimates of selection gradients can also be influenced by the phenotypic (co)variation present in the study sample (Chenoweth et al. 2012). Consequently, our understanding of how mate choice influences the evolution of CHCs in nature will be enhanced by attempts to represent naturally occurring variation within the study sample. We are aware of only two species for which investigations of sexual selection on CHCs have sampled from natural populations. Hine et al. (2004) estimated selection gradients from separate binomial mate-choice tests (using lab-reared females) on both laboratory raised and field collected male D. serrata. Although the individual gradients differed, including four of the six being of opposite sign, the overall vectors of selection were not found to be significantly different. In the only study that has directly estimated selection on CHCs occurring in a natural population, Steiger et al. (2013) used a natural phenotypic marker of mating success to estimate linear and nonlinear selection acting on the CHCs of male sagebrush crickets, Cyphoderris strepitans. Importantly, along with natural variation in the CHC phenotype, Steiger et al. (2013) also included natural variation in female preference. As laboratory studies generally control the age and mating status of females used to assess male attractiveness, they implicitly assume that the average preference of the laboratory sample of females is representative of that of the larger natural population (and hence an unbiased estimate of selection gradients, see Gershman et al. (2014) for further discussion). Studies that sample from natural populations will therefore not only increase the likelihood of capturing the naturally occurring phenotypic (co)variance, but also the average population preference. The lack of multivariate selection studies on CHCs that incorporate additional sexual traits provides an imperative for further work on the role of sexual selection in the evolution of CHCs. Additionally, expanding this work beyond drosophilid fruit flies and crickets, as recently done by Lane et al. (2016), will contribute to a taxonomically broader understanding of the evolution of insect CHCs. The dung beetle, Onthophagus taurus, has become a model species for studies of sexual selection and provides an ideal candidate for studies of selection on CHCs through female mate choice. Males vigorously drum the dorsal surface of females during courtship, and females prefer males that perform this courtship at a higher rate (Kotiaho et al. 2001; Kotiaho 2002; McCullough and Simmons 2016). Courtship rate is genetically correlated with a measure of male condition (Kotiaho et al. 2001), and females mated to males with a higher courtship rate have offspring of greater viability (Simmons and Holley 2011), suggesting the operation of good genes mate choice in this species (Garcia-Gonzalez and Simmons 2011). However, the influence of male traits from other sensory modalities on male attractiveness has not been determined. Mating occurs in narrow tunnels beneath dung pats, making a role for visual traits unlikely. We therefore hypothesized that olfactory traits, specifically CHCs, might play a role in the mating system of this species. Our aim in this study was to measure the strength and form of selection on O. taurus CHCs and, at the same time, assess how CHCs and courtship rate act together in influencing male attractiveness. METHODS Beetles were collected from Walpole, Western Australia, and transported to a controlled temperature room (set at 28 °C on a 12:12 h light:dark cycle) where they were maintained in single-sex groups of ~100 individuals in 10-L buckets that contained moist sand topped with cow dung. Mating trials took place two weeks after collection to ensure that all beetles were sexually mature. It is possible that time spent under controlled laboratory conditions altered the CHC profiles of the subject beetles and consequently the phenotypic (co)variance of our sample may differ from that found in the field. However, our sample should still capture more of the naturally occurring variation in CHC profiles when compared with the use of laboratory reared individuals, because of variation in age and mating history, for example. We therefore refer to our sample as “field collected” to contrast it to a sample derived from laboratory reared individuals, but highlight the need to interpret the results cautiously when referring to selection occurring in nature. Typically, females collected from the field are already mated, with estimates of the number of previous mates ranging between one and five (mean of 2.8 ± 1.1) (McCullough et al. 2017). We used no-choice trials where a single male and female were introduced into a chamber designed to mimic their natural breeding tunnels (Kotiaho et al. 2001). Pairs were checked every 2 min for 1 h and at each check a record was made if the male was displaying courtship activity (drumming the dorsal surface of the female with his forelegs). Courtship rate was calculated as the number of times a male was recorded displaying courtship at these 2-min intervals divided by time (60 min for unmated males or time elapsed before mating for mated males (Simmons and Holley 2011)). For data analysis, males that were observed courting were assigned either a 1 if mated or 0 if unmated, males that did not court were discarded. Beetles were frozen immediately following mating trials in preparation for CHC extraction and gas chromatography–mass spectrometry (GCMS) analysis. To extract CHCs, beetles were immersed in 1 mL of hexane for 5 min, after which the beetle was removed using clean forceps and the hexane allowed to evaporate. Prior to GCMS, the samples were re-suspended in 0.1 mL of hexane and transferred to a 0.1-mL glass insert within an autosampler vial. We used a Shimadzu QP2010 GCMS machine fitted with a 20 m × 150 µm × 0.15 µm VF1 column operating on the splitless mode. The initial oven temperature was set at 40 °C for 1 min, before increasing at 20 °C per minute until 200 °C where it was held for 1 min, and then increased at 10 °C per minute before reaching 300 °C, where it was held for 10 min. Peaks were identified from the NIST library by their mass spectra and retention indices. Retention indices were assigned by comparing compounds with a known sample of n-alkanes (C7–C40). Peaks were integrated using Shimadzu GCMSsolution Version 4.41. As is conventional for CHC analysis (Blows 1998), peak areas were transformed to logcontrasts by taking the log of their relative area divided by the relative area of a randomly chosen peak (in this case n-tricosane; we added 1 before taking the log to account for peak areas of 0). A principal component (PC) analysis was then performed on this set of logcontrasts as a data reduction tool, with only those components with an eigenvalue greater than 1 included in the analysis. We consider those peaks with loadings greater than or equal to 70% of the highest value as contributing significantly to a PC (Mardia et al. 1979). We used the multiple regression method outlined by Lande and Arnold (1983) to estimate linear and nonlinear selection gradients on our retained PCs and courtship rate. This approach first includes only the linear terms in a regression on relative fitness (in our case relative mating success) to estimate the strength of directional selection acting on each trait (β selection gradients). A second regression that includes all correlational and quadratic terms is then fit to estimate nonlinear selection (represented by the γ covariance matrix of correlational and quadratic selection) with nonzero quadratic terms representing curvature in the relationship between the trait and fitness. To allow comparison of selection across traits and studies, we standardized all traits to mean 0 and standard deviation 1 (PCs already have mean 0, so these were only variance standardized). All quadratic coefficients presented in our γ matrix were multiplied by 2 (Stinchcombe et al. 2008). As nonlinear selection is likely to be underestimated using this technique (Blows and Brooks 2003), we performed a canonical rotation of the γ matrix to estimate multivariate linear and nonlinear selection (Phillips and Arnold 1989). We visualized selection in multivariate space using thin plate splines by applying the canonical axes to the original phenotypic data. Overall significance of selection was tested by fitting a logistic regression (using absolute mating success as the response variable) to the model containing only the linear terms (for directional selection) and to the full second-order model using individual scores on the canonical axes (for nonlinear selection). As relative mating success was non-normally distributed, permutation testing was used for assessing the significance of individual terms, with our procedure for the canonical axes differing from that of Reynolds et al. (2010) in that we kept the canonical rotation constant (Chenoweth et al. 2012). As the detection of significant curvature in the fitness surface does not necessarily represent the presence of an intermediate fitness optima (or minima), we used the methods outlined by Mitchell-Olds and Shaw (1987) to test for significant stabilizing and disruptive selection on our canonical axes. This method involves two tests that constrain fitness to be at its maximum at one or the other extreme of the phenotypic range and tests whether this provides a significantly worse fit compared to the unconstrained model (see Chenoweth et al. (2007) for further details). Again, permutation testing was used to assess the significance of the relevant model term. To determine how the inclusion of the behavioral trait, courtship rate, influenced the selection gradient estimates for our CHC PCs, we repeated the first- and second-order multiple regressions, but this time excluding courtship rate. We then compared these linear and nonlinear selection gradients with those on our CHC PCs estimated from the model which included courtship rate. We calculated the angle between our two vectors of selection gradient estimates to compare how the overall estimated direction of selection differed between them (an angle of 0° would indicate perfect alignment, 90° that the overall direction of selection estimates were orthogonal, and 180° that selection was estimated to be in the opposite direction). Similarity between our two γ matrices was estimated using the matrix comparison method of Krzanowski (1979), which calculates the similarity of the matrix subspaces that describe the majority of the variation (see Rundle et al. (2008) for an example of how these point estimates can be used for the comparison of multiple γ matrices). In our case, we retained the first three eigenvectors of both matrices, with a score of 0 indicating that the matrices were completely unaligned, whereas a score of 3 would show perfect alignment. All analyses were performed in R (R Core Team 2017) using the FactoMineR package for the PC analysis (Lê et al. 2008) and the fields package for thin plate spline visualization (Nychka et al. 2015). RESULTS We analyzed hexane washes from a total of 119 male O. taurus, 21 (18%) of which were successful in securing a mating. Results from the GCMS analysis are presented in Table 1 and Figure 1 shows a typical chromatogram. The vast majority of the CHCs detected were methyl-branched compounds, with alkanes less common but representing some of the larger peaks. Our PC analysis returned 7 PCs with eigenvalues greater than 1 that cumulatively explained 83.13 % of the total variation (Table 1). PC1 was weighted positively by all but one compound and most strongly by longer chained compounds. We therefore interpret this axis as describing the relative abundance of heavier compounds. PC2 contrasts straight-chained alkanes with methyl-branched compounds that have a chain-length of 26 carbons or less. PC3 contrasts four methyl-branched compounds and two alkanes with the three longest-chained compounds. PC4 was positively loaded by two 2-methylalkanes, two alkanes, and two longer chained 15-methylalkanes. PC5 contrasts two 2-methylalkanes with an 8-methylalkane. PC6 contrasts a 2-methylalkane with a 7-methylalkane and PC7 is weighted by two short-chained alkanes. Though the patterns of variation explained by these PCs are not completely clear, and our description is necessarily a simplification, it is interesting to note that the 2-methylalkanes contribute to the phenotypic variation described by 5 of the 7 PCs. Figure 1 View largeDownload slide Chromatogram of a typical Onthophagus taurus CHC profile. The x-axis shows the retention time and the y-axis peak abundance. Numbers above peaks correspond to those in Table 1. Figure 1 View largeDownload slide Chromatogram of a typical Onthophagus taurus CHC profile. The x-axis shows the retention time and the y-axis peak abundance. Numbers above peaks correspond to those in Table 1. Table 1 Cuticular hydrocarbons of Onthophagus taurus, their mean relative amounts, and the results of the PC analysis PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 Only PCs with eigenvalues greater than 1 are shown. Loadings are shown in the body of the table and those in bold have values that are equal to or greater than 70% of the highest loading. View Large Table 1 Cuticular hydrocarbons of Onthophagus taurus, their mean relative amounts, and the results of the PC analysis PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 Only PCs with eigenvalues greater than 1 are shown. Loadings are shown in the body of the table and those in bold have values that are equal to or greater than 70% of the highest loading. View Large The multivariate selection analyses revealed that courtship rate was subject to directional selection, with nonlinear selection acting on the CHC profile. We detected overall significant linear selection (χ28 = 34.938, P < 0.001, r2(adjusted) = 0.258), with this effect driven by courtship rate as no significant linear selection was found for our CHC PCs (Table 2). We detected significant negative correlational selection between PC2 and PC5 (Table 2) and overall significant nonlinear selection on the canonical axes (χ216 = 83.140, P < 0.001). Significant concave selection was found on axes m1 and m2, along with significant convex selection on axes m7 and m8 (Table 3). Selection acting upon axes m1 and m2 was disruptive, with models constraining the highest fitness to be at only the minimum or maximum of the phenotypic variation providing a significantly worse fit (P <0.001 in both cases for m1, P = 0.014 and P < 0.001 for m2). We did not detect significant stabilizing selection on axis m8 (P = 0.164 and P = 0.422) nor on axis m7 (P = 0.078 and P = 0.066). Table 2 Results of the multiple regression analyses including courtship rate (CR) β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05, ***P < 0.001 View Large Table 2 Results of the multiple regression analyses including courtship rate (CR) β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05, ***P < 0.001 View Large Table 3 Results of the canonical analysis derived from the multiple regression including PCs 1–7 and courtship rate (CR) mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01, ***P < 0.001 View Large Table 3 Results of the canonical analysis derived from the multiple regression including PCs 1–7 and courtship rate (CR) mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01, ***P < 0.001 View Large Further inspection of the major canonical axes revealed that the disruptive selection was most strongly associated with the CHC profile. Visualizing selection acting on the major canonical axes, m1 and m8, revealed a fitness trough at intermediate values of m1 (as indicated above by the significant disruptive selection) and a peak at decreasing values of m1 and m8 (Figure 2). An examination of the eigenvectors of axis m1 revealed that the fitness trough was associated with intermediate values of PC2 and PC5, representing a lack of contrast between short-chained methyl-branched compounds with n-alkanes (PC2), as well as a lack of contrast between two 2-methyl-branched compounds and 8-methylhexacosane (PC5). Inspecting the eigenvectors for axes m1 and m8 together revealed that the fitness peak was associated with high values for PC2 and low values for PC5 (m1), along with high courtship rate and low values of PC6 (m8). An inspection of the loadings of the original CHC logcontrasts on these PCs revealed an association between mating success and investment in short-chained methyl-branched compounds (PC2) and 2-methyalkanes (PC5 and PC6). Figure 2 View largeDownload slide Thin plate spline visualization in 3 dimensions (a) and as a contour plot (b) of selection acting on the major canonical axes, m1 and m8. Relative fitness is represented by the vertical axis in (a) and through color in both (a) and (b) with red representing high fitness and blue representing low fitness. The points on (b) are the raw data points. Figure 2 View largeDownload slide Thin plate spline visualization in 3 dimensions (a) and as a contour plot (b) of selection acting on the major canonical axes, m1 and m8. Relative fitness is represented by the vertical axis in (a) and through color in both (a) and (b) with red representing high fitness and blue representing low fitness. The points on (b) are the raw data points. A comparison of analyses that either included or excluded courtship rate revealed that nonlinear selection on the CHC profile was largely independent of the effect of courtship rate on mating success. Removing courtship rate and repeating the multiple regression analyses resulted in the model of directional selection no longer explaining a significant amount of variation in male mating success (χ27 = 9.275, P = 0.234, r2(adjusted) = 0.017). Interestingly, of the small amount of variation that was explained, there was significant directional selection on PC6 (Table 4), which out of all the PCs was found to have the strongest correlational selection with courtship rate (Table 2), although this was not significant (P = 0.094). Comparing the vector of directional selection gradients on the CHC PCs from this analysis with that obtained from our analysis that included courtship rate revealed that the two vectors were oriented 48° from each other and that the individual β estimates were not significantly correlated (r = 0.689, P = 0.087). The exclusion of courtship rate had little effect on the nonlinear selection estimates, with a comparison of the γ matrices returning a Krzanowski value of 2.513 out of a maximum of 3 (or 84% of the score for complete similarity), and a significant correlation between the individual γ estimates (r = 0.869, P <0.001). This was reflected again in finding significant nonlinear selection on the canonical axes derived from the multiple regression that excluded courtship rate (χ214 = 59.431, P <0.001). Although the two sets of canonical axes are not comparable (as they were based on two different sets of traits), it is notable that the CHC PC loadings on the two m1 axes bear some resemblance (Tables 3 and 5) and that this axis displays the greatest curvature in our canonical rotation that excluded courtship rate (Table 5). Table 4 Results of the multiple regression analyses excluding courtship rate β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05 View Large Table 4 Results of the multiple regression analyses excluding courtship rate β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05 View Large Table 5 Results of the canonical analysis derived from the multiple regression of PCs 1–7 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01 View Large Table 5 Results of the canonical analysis derived from the multiple regression of PCs 1–7 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01 View Large Discussion Using multivariate selection analyses that incorporated traits from different sensory modalities, we have found significant nonlinear selection on the axes of major CHC phenotypic variation (CHC PCs) and significant directional selection on a behavioral sexual display (courtship rate) in the dung beetle, O. taurus. Our finding of significant directional selection on courtship rate is consistent with previous studies on mate choice using this species (Kotiaho et al. 2001; Kotiaho 2002; Simmons and Holley 2011; McCullough and Simmons 2016). What is novel about our results is the finding that CHCs also contribute to mating success in O. taurus. Furthermore, the high level of similarity between our two γ matrices (84% of the maximum similarity score) indicates that nonlinear selection on the CHC PCs does not depend on an individual’s courtship rate. Our study therefore provides an example of mate-choice mediated sexual selection acting on an insect’s CHC profile independently of the influence of a well-characterized courtship display. The fact that we have shown this in the laboratory using beetles collected from the field, that would have varied in a variety of life-history traits including age and previous mating history, gives support to the view that sexual selection is likely to be a persistent driver of the evolution of CHCs in natural populations of insects. Although nonlinear selection on the CHC PCs appeared to act independently of courtship rate, the estimates of directional selection gradients were influenced to some extent by whether courtship rate was incorporated into the analysis. This was most evident in the doubling of the selection gradient (and consequently altering its statistical significance) for PC6, the PC that showed the strongest correlational selection with courtship rate. Similarly, comparing the overall direction of selection acting on the CHC PCs estimated from the two analyses which included or excluded courtship rate resulted in β vector estimates that were angled moderately (48°) away from each other. Although total linear selection on the CHC PCs was not significant, the ability of a non-CHC trait to influence the estimate of the overall direction of selection has important implications. For example, such an influence will affect the estimated alignment of selection with genetic variation, potentially altering interpretations of the amount of genetic variation available for selection. The potential for unmeasured characters to influence selection gradients is well known (Lande and Arnold 1983; Mitchell-Olds and Shaw 1987), and we have shown that estimates of selection on CHCs are no exception to this rule. As CHCs are commonly used in studies of sexual selection on multivariate traits (Chenoweth and Blows 2003; Van Homrigh et al. 2007; Thomas and Simmons 2009b, 2010; Curtis et al. 2013; Steiger et al. 2013; Ingleby et al. 2014; Steiger et al. 2015; Lane et al. 2016), our results indicate that a fuller understanding of mate choice will be gained by incorporating non-CHC traits in future studies. This will of course likely necessitate much larger sample sizes than we have included here, an admittedly challenging logistical proposition. Here, we were interested in assessing selection imposed on male O. taurus CHC profiles through the action of female mate choice. We emphasize that this is highly unlikely to represent the total selection acting on CHCs in this species. The role CHCs play in desiccation resistance (Gibbs and Rajpurohit 2010), along with their production utilizing the shared resource of internal hydrocarbons (Schal et al. 1994; Wicker and Jallon 1995), are both likely to impose strong selection on CHCs. For example, experimental evolution studies have found that altering the environmental temperature (Sharma et al. 2012) and imposing fecundity selection (Blows 2002) lead to an evolutionary response in the CHC profile. The CHCs of some species are also important in male dominance displays (Roux et al. 2002; Kortet and Hedrick 2005; Thomas and Simmons 2009a, 2011b), and Lane et al. (2016) have recently shown that selection on CHCs imposed by female mate choice can differ to that imposed by male–male competition. In investigations of selection via female choice on male CHC profiles, the use of no-choice mating assays is a useful experimental technique that avoids the potential effect of male–male competition. Our study adds O. taurus to the limited number of species for which no-choice mating assays have been used to assess selection from mate choice on male CHCs (broad-horned flour beetle, Gnatocerus cornutus (Lane et al. 2016); Australian field cricket, T. oceanicus (Thomas and Simmons 2009b); decorated cricket, Gryllodes sigillatus (Steiger et al. 2015); and D. simulans (Ingleby et al. 2014)). The fact that male O. taurus CHCs are subject to selection from female choice is evidence that females of this species use CHCs in their assessment of potential mates. Although our study was not designed to test what benefits such assessment may provide, our results suggest at least one potential hypothesis. Chemical traits have been suggested as good candidates for genetic compatibility-based mate choice (Mays and Hill 2004), and there is some evidence that CHCs act in this manner. For example, female cucumber beetles, Diabrotica undecimpunctata, prefer males with more dissimilar CHC profiles and produce offspring with higher immunocompetence when mated to their preferred mates (Ali and Tallamy 2010). Mating is more likely to occur between T. oceanicus pairs that have more dissimilar CHC profiles, and genetic distance seems to be correlated with CHC dissimilarity in this species (Thomas and Simmons 2011a). This, and the finding that CHC attractiveness is uncorrelated with a potential signal of good genes (courtship song), suggests that compatibility-based mate choice is facilitated through CHCs in these crickets (Simmons et al. 2013). Compatibility and “good” genes benefits are theoretically uncorrelated (Puurtinen et al. 2009), such that individuals can use one trait to assess good genes benefits, and another trait to assess the genetic compatibility of a potential mate. Such a view of mate choice falls under the “multiple messages” hypothesis which states that multicomponent/multimodal signals evolve to communicate different mate characteristics (Candolin 2003; Hebets and Papaj 2005; Partan and Marler 2005). Our finding of disruptive selection on male O. taurus CHCs acting independently of courtship rate is consistent with this scenario. There is strong evidence that courtship rate signals good genes in this species (Kotiaho et al. 2001; Simmons and Holley 2011) and the disruptive selection on the CHC profile we detected is consistent with compatibility-based mate choice. Alternatively, variation in courtship rate and the CHC profile in our study sample may be associated with age or other life-history traits that would have varied among individuals collected from the field, and females may be using this variation to select mates based on these characteristics. Although our data are suggestive of a multiple-messages/good-genes/compatible-genes scenario, alternative hypotheses are plausible, and at this point further experimentation is required before firm conclusions can be drawn. Due to its correlative nature, multiple regression is essentially a hypothesis-generating tool that requires experimental manipulation of traits to test its predictions (Chenoweth et al. 2012). Here, we have provided initial evidence that CHCs play a role in the mating system of O. taurus and suggest hypotheses for future research. Results from multivariate selection studies on CHCs in other species have proven fruitful for providing a basis for further investigation. For example, artificial selection on the linear combination of CHCs found to be subject to sexual selection in D. serrata resulted in a correlated increase in male mating success (Hine et al. 2011). Similarly, a positive selection gradient on 2-methylhexacosane suggested an attractiveness function for this compound in D. serrata (Chenoweth and Blows 2005), with this role experimentally validated using a perfuming experiment (Chung et al. 2014). These findings indicate that multivariate selection analyses on CHCs can provide useful insights for further experimental work. We suggest that incorporating different sensory modalities will provide further insights into how CHCs influence attractiveness alongside other traits, leading to a fuller understanding of the multivariate nature of mate choice. Conducting these studies across a broad range of taxa will contribute to elucidating general patterns in the relative importance of CHCs in mate choice, as well as consistencies (or otherwise) in the information gained by mate assessment based on CHCs in different taxa. We hope that our findings encourage further research on CHC mediated mate choice across a broader range of species that incorporate traits from other sensory modalities. Funding This work was supported through an Australian Government Research Training Program Scholarship to J.D.B. and an ARC Discovery Project to L.W.S. Data accessibility Analyses reported in this article can be reproduced using the data provided by Berson and Simmons (2018). We thank Jessica Horn for performing the mating trials and assisting with CHC extraction, Anchal Gupta for integrating the CHC peak areas, and Rowan Lymbery for statistical advice. The quality of the manuscript was improved by comments from 2 anonymous referees. The authors acknowledge the facilities and the scientific and technical assistance of the Metabolomics Australia Facility at the Centre for Microscopy, Characterisation & Analysis, The University of Western Australia, a facility funded by the University, State, and Commonwealth Governments. References Ali JG , Tallamy DW . 2010 . Female spotted cucumber beetles use own cuticular hydrocarbon signature to choose immunocompatible mates . Anim Behav . 80 : 9 – 12 . Google Scholar CrossRef Search ADS Andersson M . 1994 . Sexual selection . Princeton (NJ) : Princeton University Press . Bentsen CL , Hunt J , Jennions MD , Brooks R . 2006 . Complex multivariate sexual selection on male acoustic signaling in a wild population of Teleogryllus commodus . Am Nat . 167 : E102 – E116 . Google Scholar CrossRef Search ADS PubMed Berson JD , Simmons LW . 2018 . Data from: sexual selection across sensory modalities: female choice of male behavioral and gustatory displays . Dryad Digital Repository . https://doi.org/10.5061/dryad.8p81b6h. Blomquist GJ , Bagnères AG . 2010 . Insect hydrocarbons: biology, biochemistry, and chemical ecology . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Blows MW . 1998 . Evolution of a mate recognition system after hybridization between two Drosophila species . Am Nat . 151 : 538 – 544 . Google Scholar CrossRef Search ADS PubMed Blows MW . 2002 . Interaction between natural and sexual selection during the evolution of mate recognition . Proc R Soc B . 269 : 1113 – 1118 . Google Scholar CrossRef Search ADS PubMed Blows MW . 2007 . A tale of two matrices: multivariate approaches in evolutionary biology . J Evol Biol . 20 : 1 – 8 . Google Scholar CrossRef Search ADS PubMed Blows MW , Brooks R . 2003 . Measuring nonlinear selection . Am Nat . 162 : 815 – 820 . Google Scholar CrossRef Search ADS PubMed Blows MW , Chenoweth SF , Hine E . 2004 . Orientation of the genetic variance-covariance matrix and the fitness surface for multiple male sexually selected traits . Am Nat . 163 : 329 – 340 . Google Scholar CrossRef Search ADS PubMed Brooks R , Endler JA . 2001 . Direct and indirect sexual selection and quantitative genetics of male traits in guppies (Poecilia reticulata) . Evolution . 55 : 1002 – 1015 . Google Scholar CrossRef Search ADS PubMed Byers J , Hebets E , Podos J . 2010 . Female mate choice based upon male motor performance . Anim Behav . 79 : 771 – 778 . Google Scholar CrossRef Search ADS Callander S , Jennions MD , Backwell PRY . 2012 . The effect of claw size and wave rate on female choice in a fiddler crab . J Ethol . 30 : 151 – 155 . Google Scholar CrossRef Search ADS Candolin U . 2003 . The use of multiple cues in mate choice . Biol Rev . 78 : 575 – 595 . Google Scholar CrossRef Search ADS PubMed Chenoweth SF , Blows MW . 2003 . Signal trait sexual dimorphism and mutual sexual selection in Drosophila serrata . Evolution . 57 : 2326 – 2334 . Google Scholar CrossRef Search ADS PubMed Chenoweth SF , Blows MW . 2005 . Contrasting mutual sexual selection on homologous signal traits in Drosophila serrata . Am Nat . 165 : 281 – 289 . Google Scholar CrossRef Search ADS PubMed Chenoweth SF , Hunt J , Rundle HD . 2012 . Analyzing and comparing the geometry of individual fitness surfaces . In: Svensson EI , Calsbeek R , editors. The adaptive landscape in evolutionary biology . UK : Oxford University Press . p. 126 – 149 . Google Scholar CrossRef Search ADS Chenoweth SF , Petfield D , Doughty P , Blows MW . 2007 . Male choice generates stabilizing sexual selection on a female fecundity correlate . J Evol Biol . 20 : 1745 – 1750 . Google Scholar CrossRef Search ADS PubMed Chung H , Loehlin DW , Dufour HD , Vaccarro K , Millar JG , Carroll SB . 2014 . A single gene affects both ecological divergence and mate choice in Drosophila . Science . 343 : 1148 – 1151 . Google Scholar CrossRef Search ADS PubMed Cole GL , Endler JA . 2015 . Variable environmental effects on a multicomponent sexually selected trait . Am Nat . 185 : 452 – 468 . Google Scholar CrossRef Search ADS PubMed Coleman SW . 2009 . Taxonomic and sensory biases in the mate-choice literature: there are far too few studies of chemical and multimodal communication . Acta Ethologica . 12 : 45 – 48 . Google Scholar CrossRef Search ADS Curtis S , Sztepanacz JL , White BE , Dyer KA , Rundle HD , Mayer P . 2013 . Epicuticular compounds of Drosophila subquinaria and D. recens: identification, quantification, and their role in female mate choice . J Chem Ecol . 39 : 579 – 590 . Google Scholar CrossRef Search ADS PubMed Darwin C . 1871 . The descent of man, and selection in relation to sex . London : J. Murray . Garcia-Gonzalez F , Simmons LW . 2011 . Good genes and sexual selection in dung beetles (Onthophagus taurus): genetic variance in egg-to-adult and adult viability . Plos One . 6 : e16233 . Google Scholar CrossRef Search ADS PubMed Gerhardt HC , Brooks R . 2009 . Experimental analysis of multivariate female choice in Gray treefrogs (Hyla versicolor): evidence for directional and stabilizing selection . Evolution . 63 : 2504 – 2512 . Google Scholar CrossRef Search ADS PubMed Gershman S , Delcourt M , Rundle HD . 2014 . Sexual selection on Drosophila serrata male pheromones does not vary with female age or mating status . J Evol Biol . 27 : 1279 – 1286 . Google Scholar CrossRef Search ADS PubMed Gibbs AG , Rajpurohit S . 2010 . Cuticular lipids and water balance . In: Blomquist GJ , Bagnéres AG , editors. Insect hydrocarbons: biology, biochemistry, and chemical ecology . Cambridge : Cambridge University Press . p. 100 – 120 . Google Scholar CrossRef Search ADS Hebets EA , Papaj DR . 2005 . Complex signal function: developing a framework of testable hypotheses . Behav Ecol Sociobiol . 57 : 197 – 214 . Google Scholar CrossRef Search ADS Hill GE . 1991 . Plumage coloration is a sexually selected indicator of male quality . Nature . 350 : 337 – 339 . Google Scholar CrossRef Search ADS Hine E , Chenoweth SF , Blows MW . 2004 . Multivariate quantitative genetics and the lek paradox: genetic variance in male sexually selected traits of Drosophila serrata under field conditions . Evolution . 58 : 2754 – 2762 . Google Scholar CrossRef Search ADS PubMed Hine E , McGuigan K , Blows MW . 2011 . Natural selection stops the evolution of male attractiveness . Proc Natl Acad Sci USA . 108 : 3659 – 3664 . Google Scholar CrossRef Search ADS PubMed Ingleby FC , Hosken DJ , Flowers K , Hawkes MF , Lane SM , Rapkin J , House CM , Sharma MD , Hunt J . 2014 . Environmental heterogeneity, multivariate sexual selection and genetic constraints on cuticular hydrocarbons in Drosophila simulans . J Evol Biol . 27 : 700 – 713 . Google Scholar CrossRef Search ADS PubMed Johansson BG , Jones TM . 2007 . The role of chemical communication in mate choice . Biol Rev . 82 : 265 – 289 . Google Scholar CrossRef Search ADS PubMed Kortet R , Hedrick A . 2005 . The scent of dominance: female field crickets use odour to predict the outcome of male competition . Behav Ecol Sociobiol . 59 : 77 – 83 . Google Scholar CrossRef Search ADS Kotiaho JS . 2002 . Sexual selection and condition dependence of courtship display in three species of horned dung beetles . Behav Ecol . 13 : 791 – 799 . Google Scholar CrossRef Search ADS Kotiaho JS , Simmons LW , Tomkins JL . 2001 . Towards a resolution of the lek paradox . Nature . 410 : 684 – 686 . Google Scholar CrossRef Search ADS PubMed Krzanowski WJ . 1979 . Between-groups comparison of principal components . J Am Stat Assoc . 74 : 703 – 707 . Google Scholar CrossRef Search ADS Lande R , Arnold SJ . 1983 . The measurement of selection on correlated characters . Evolution . 37 : 1210 – 1226 . Google Scholar CrossRef Search ADS PubMed Lane SM , Dickinson AW , Tregenza T , House CM . 2016 . Sexual selection on male cuticular hydrocarbons via male-male competition and female choice . J Evol Biol . 29 : 1346 – 1355 . Google Scholar CrossRef Search ADS PubMed Lê S , Josse J , Husson F . 2008 . FactoMineR: an R package for multivariate analysis . J Stat Softw . 25 : 1 – 18 . Google Scholar CrossRef Search ADS LeBas NR , Hockham LR , Ritchie MG . 2003 . Nonlinear and correlational sexual selection on ‘honest’ female ornamentation . Proc R Soc B . 270 : 2159 – 2165 . Google Scholar CrossRef Search ADS PubMed Leonard AS , Hedrick AV . 2010 . Long-distance signals influence assessment of close range mating displays in the field cricket, Gryllus integer . Biol J Linn Soc . 100 : 856 – 865 . Google Scholar CrossRef Search ADS Mardia KV , Kent JT , Bibby JM . 1979 . Multivariate analysis . London : Academic Press . Mays HL , Hill GE . 2004 . Choosing mates: good genes versus genes that are a good fit . Trends Ecol Evol . 19 : 554 – 559 . Google Scholar CrossRef Search ADS PubMed McCullough EL , Buzatto BA , Simmons LW . 2017 . Benefits of polyandry: molecular evidence from field-caught dung beetles . Mol Ecol . 26 : 3546 – 3555 . Google Scholar CrossRef Search ADS PubMed McCullough EL , Simmons LW . 2016 . Selection on male physical performance during male-male competition and female choice . Behav Ecol . 27 : 1288 – 1295 . Google Scholar CrossRef Search ADS McGuigan K . 2009 . Condition dependence varies with mating success in male Drosophila bunnanda . J Evol Biol . 22 : 1813 – 1825 . Google Scholar CrossRef Search ADS PubMed Mitchell-Olds T , Shaw RG . 1987 . Regression analysis of natural selection: statistical inference and biological interpretation . Evolution . 41 : 1149 – 1161 . Google Scholar CrossRef Search ADS PubMed Nychka D , Furrer R , Paige J , Sain S . 2017 . fields: tools for spatial data. Version 9.6. Boulder (CO): University Corporation for Atmospheric Research. doi : Partan SR , Marler P . 2005 . Issues in the classification of multimodal communication signals . Am Nat . 166 : 231 – 245 . Google Scholar CrossRef Search ADS PubMed Phillips PC , Arnold SJ . 1989 . Visualizing multivariate selection . Evolution . 43 : 1209 – 1222 . Google Scholar CrossRef Search ADS PubMed Pryke SR , Andersson S , Lawes MJ . 2001 . Sexual selection of multiple handicaps in the red-collared widowbird: female choice of tail length but not carotenoid display . Evolution . 55 : 1452 – 1463 . Google Scholar CrossRef Search ADS PubMed Puurtinen M , Ketola T , Kotiaho JS . 2009 . The good-genes and compatible-genes benefits of mate choice . Am Nat . 174 : 741 – 752 . Google Scholar CrossRef Search ADS PubMed R Core Team . 2017 . R: a language and environment for statistical computing. Version 3.3.3 . Vienna (Austria) : R Foundation for Statistical Computing . Rebar D , Bailey NW , Zuk M . 2009 . Courtship song’s role during female mate choice in the field cricket Teleogryllus oceanicus . Behav Ecol . 20 : 1307 – 1314 . Google Scholar CrossRef Search ADS Reynolds RJ , Childers DK , Pajewski NM . 2010 . The distribution and hypothesis testing of eigenvalues from the canonical analysis of the gamma matrix of quadratic and correlational selection gradients . Evolution . 64 : 1076 – 1085 . Google Scholar CrossRef Search ADS PubMed Roux E , Sreng L , Provost E , Roux M , Clement JL . 2002 . Cuticular hydrocarbon profiles of dominant versus subordinate male Nauphoeta cinerea cockroaches . J Chem Ecol . 28 : 1221 – 1235 . Google Scholar CrossRef Search ADS PubMed Rundle HD , Chenoweth SF , Blows MW . 2008 . Comparing complex fitness surfaces: among-population variation in mutual sexual selection in Drosophila serrata . Am Nat . 171 : 443 – 454 . Google Scholar CrossRef Search ADS PubMed Rybak F , Sureau G , Aubin T . 2002 . Functional coupling of acoustic and chemical signals in the courtship behaviour of the male Drosophila melanogaster . Proc R Soc B . 269 : 695 – 701 . Google Scholar CrossRef Search ADS PubMed Schal C , Gu X , Burns EL , Blomquist GJ . 1994 . Patterns of biosynthesis and accumulation of hydrocarbons and contact sex pheromone in the female German cockroach, Blattella germanica . Arch Insect Biochem Physiol . 25 : 375 – 391 . Google Scholar CrossRef Search ADS PubMed Sharma MD , Hunt J , Hosken DJ . 2012 . Antagonistic responses to natural and sexual selection and the sex-specific evolution of cuticular hydrocarbons in Drosophila simulans . Evolution . 66 : 665 – 677 . Google Scholar CrossRef Search ADS PubMed Simmons LW , Holley R . 2011 . Offspring viability benefits but no apparent costs of mating with high quality males . Biol Lett . 7 : 419 – 421 . Google Scholar CrossRef Search ADS PubMed Simmons LW , Thomas ML , Simmons FW , Zuk M . 2013 . Female preferences for acoustic and olfactory signals during courtship: male crickets send multiple messages . Behav Ecol . 24 : 1099 – 1107 . Google Scholar CrossRef Search ADS Steiger S , Capodeanu-Nagler A , Gershman SN , Weddle CB , Rapkin J , Sakaluk SK , Hunt J . 2015 . Female choice for male cuticular hydrocarbon profile in decorated crickets is not based on similarity to their own profile . J Evol Biol . 28 : 2175 – 2186 . Google Scholar CrossRef Search ADS PubMed Steiger S , Ower GD , Stökl J , Mitchell C , Hunt J , Sakaluk SK . 2013 . Sexual selection on cuticular hydrocarbons of male sagebrush crickets in the wild . Proc R Soc B . 280 : 20132353 . Google Scholar CrossRef Search ADS PubMed Steiger S , Stökl J . 2014 . The role of sexual selection in the evolution of chemical signals in insects . Insects . 5 : 423 – 438 . Google Scholar CrossRef Search ADS PubMed Stinchcombe JR , Agrawal AF , Hohenlohe PA , Arnold SJ , Blows MW . 2008 . Estimating nonlinear selection gradients using quadratic regression coefficients: double or nothing ? Evolution . 62 : 2435 – 2440 . Google Scholar CrossRef Search ADS PubMed Tanner JC , Ward JL , Shaw RG , Bee MA . 2017 . Multivariate phenotypic selection on a complex sexual signal . Evolution . 71 : 1742 – 1754 . Google Scholar CrossRef Search ADS PubMed Thomas ML , Simmons LW . 2009a . Male dominance influences pheromone expression, ejaculate quality, and fertilization success in the Australian field cricket, Teleogryllus oceanicus . Behav Ecol . 20 : 1118 – 1124 . Google Scholar CrossRef Search ADS Thomas ML , Simmons LW . 2009b . Sexual selection on cuticular hydrocarbons in the Australian field cricket, Teleogryllus oceanicus . BMC Evol Biol . 9 : 12 . Google Scholar CrossRef Search ADS Thomas ML , Simmons LW . 2010 . Cuticular hydrocarbons influence female attractiveness to males in the Australian field cricket, Teleogryllus oceanicus . J Evol Biol . 23 : 707 – 714 . Google Scholar CrossRef Search ADS PubMed Thomas ML , Simmons LW . 2011a . Crickets detect the genetic similarity of mating partners via cuticular hydrocarbons . J Evol Biol . 24 : 1793 – 1800 . Google Scholar CrossRef Search ADS Thomas ML , Simmons LW . 2011b . Short-term phenotypic plasticity in long-chain cuticular hydrocarbons . Proc R Soc B . 278 : 3123 – 3128 . Google Scholar CrossRef Search ADS Van Homrigh A , Higgie M , McGuigan K , Blows MW . 2007 . The depletion of genetic variance by sexual selection . Curr Biol . 17 : 528 – 532 . Google Scholar CrossRef Search ADS PubMed Wicker C , Jallon JM . 1995 . Influence of ovary and ecdysteroids on pheromone biosynthesis in Drosophila melanogaster (Diptera: Drosophilidae) . Eur J Entomol . 92 : 197 – 202 . © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral Ecology Oxford University Press

Sexual selection across sensory modalities: female choice of male behavioral and gustatory displays

Loading next page...
 
/lp/ou_press/sexual-selection-across-sensory-modalities-female-choice-of-male-mIAKoSPOtk
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1045-2249
eISSN
1465-7279
D.O.I.
10.1093/beheco/ary085
Publisher site
See Article on Publisher Site

Abstract

Abstract The role of cuticular hydrocarbons in sexual displays has received considerable interest over the last two decades. For example, multiple studies have documented significant directional and nonlinear sexual selection acting on the cuticular hydrocarbon profiles of both male and female insects. The majority of these studies have excluded other sensory modalities that may influence attractiveness and measured selection using laboratory raised individuals. Furthermore, much of this work has been conducted using drosophilid fruit flies and crickets, and investigations using different taxa are necessary to improve our understanding of broader taxonomic trends. Here, we extend our understanding of sexual selection on cuticular hydrocarbons by measuring selection imposed by female mate choice on male bull-horned dung beetles, Onthophagus taurus. Both male and female beetles used in our study were collected from the field, ensuring that our estimates of selection incorporated some degree of naturally occurring variation in both cuticular hydrocarbon profiles and female mate preferences. Consistent with previous studies on this species, we found significant directional selection on male courtship displays. We also found significant nonlinear selection on the male cuticular hydrocarbon profile acting independently of the influence of behavioral courtship. Our data are consistent with a role for cuticular hydrocarbons in the mating system of this species and suggest that female O. taurus use multiple sensory modalities to assess different aspects of male quality. Introduction Sexual selection through mate choice has resulted in the evolution of some of the most elaborate traits found in biology (Darwin 1871; Andersson 1994). A multitude of studies have documented how an individual’s mating success can be influenced by characteristics such as bright coloration (Hill 1991; Brooks and Endler 2001), loud vocalizations (Rebar et al. 2009; Tanner et al. 2017), and conspicuous behavioral displays (Byers et al. 2010; Callander et al. 2012). A recognition that mate choice can be influenced by multiple components within and across sensory modalities (Candolin 2003; Hebets and Papaj 2005; Partan and Marler 2005) has led to a multivariate approach to their analysis (e.g. Pryke et al. 2001; LeBas et al. 2003; Bentsen et al. 2006; Gerhardt and Brooks 2009; Cole and Endler 2015; Tanner et al. 2017). Multivariate selection analyses allow for the elucidation of the marginal effect of each measured trait on mating success and can help avoid potentially false conclusions of selection acting on a given trait that can arise from phenotypic correlations between that trait with other traits that are associated with fitness (Lande and Arnold 1983; Phillips and Arnold 1989; Blows 2007; Chenoweth et al. 2012). Although the role of chemical traits in mate choice has received relatively less attention (Johansson and Jones 2007; Coleman 2009), in the context of multivariate selection analyses, the cuticular hydrocarbons (CHCs) of insects have received considerable interest (Steiger and Stökl 2014). CHCs form a protective layer on the insect’s cuticle, primarily protecting it from desiccation, but also playing a role in close-range/contact sexual communication (Blomquist and Bagnères 2010). For example, multiple studies have found consistent and relatively strong directional sexual selection acting on the CHC profile of male Drosophila serrata (Blows et al. 2004; Hine et al. 2011; Gershman et al. 2014). These, and associated investigations, have contributed greatly towards our understanding of the multivariate nature of mate choice. Notably, by comparing the vector of directional selection gradients with the genetic (co)variance matrix (G), a lack of genetic variance in the direction of selection has been revealed, a perspective which contrasts with that gained from examining individual traits in isolation (Blows et al. 2004; Hine et al. 2004; Van Homrigh et al. 2007). However, a weakness of multivariate selection analyses is the potential influence of unmeasured characters (Lande and Arnold 1983; Mitchell-Olds and Shaw 1987). If a trait being examined is phenotypically correlated with an unmeasured character that is itself under selection, the resultant selection gradients may reflect this correlation rather than direct selection on the trait itself (Lande and Arnold 1983). Accounting for all traits influencing attractiveness is therefore an important goal in studies examining the multivariate nature of mate choice. Little attention, however, has been given to investigating how traits from other signal modalities act alongside CHCs in determining overall attractiveness. Where both CHCs and nonchemical traits have been incorporated into one study, the evidence to date suggests that the attractiveness of an individual’s CHC profile is independent of the attractiveness of trait(s) in other sensory modalities. For example, no association was found between the measures of male attractiveness for calling song and chemical cues (presumed to be CHCs) in the cricket Gryllus integer (Leonard and Hedrick 2010). Similarly, no correlation was found between an individual’s courtship song attractiveness and CHC attractiveness in the Australian field cricket, Teleogryllus oceanicus, (Simmons et al. 2013). Just as a correlated trait can influence estimates of selection gradients, a trait influencing attractiveness independently of the traits of interest should have little impact on the selection gradients when included in the analysis. For example, in an analysis that incorporated two wing-shape measures (McGuigan 2009), the estimated linear selection gradients on D. bunnanda CHCs were similar to selection gradients found when only CHCs were included (Van Homrigh et al. 2007), suggesting that these two sets of traits influence attractiveness independently. Regardless of whether the attractiveness of an individual’s CHC profile is correlated with the attractiveness of traits in other sensory modalities, overall attractiveness is still likely to be influenced by both CHCs and non-CHC traits. For example, Rybak et al. (2002) showed that both CHCs and song are important in male D. melanogaster courtship, with the removal/reduction of either modality decreasing the attractiveness of males. Incorporating non-CHC traits into multivariate selection analyses will likely provide a greater understanding of mate choice within the study system, as well as ensuring the estimated selection gradients are not biased by unmeasured characters. Estimates of selection gradients can also be influenced by the phenotypic (co)variation present in the study sample (Chenoweth et al. 2012). Consequently, our understanding of how mate choice influences the evolution of CHCs in nature will be enhanced by attempts to represent naturally occurring variation within the study sample. We are aware of only two species for which investigations of sexual selection on CHCs have sampled from natural populations. Hine et al. (2004) estimated selection gradients from separate binomial mate-choice tests (using lab-reared females) on both laboratory raised and field collected male D. serrata. Although the individual gradients differed, including four of the six being of opposite sign, the overall vectors of selection were not found to be significantly different. In the only study that has directly estimated selection on CHCs occurring in a natural population, Steiger et al. (2013) used a natural phenotypic marker of mating success to estimate linear and nonlinear selection acting on the CHCs of male sagebrush crickets, Cyphoderris strepitans. Importantly, along with natural variation in the CHC phenotype, Steiger et al. (2013) also included natural variation in female preference. As laboratory studies generally control the age and mating status of females used to assess male attractiveness, they implicitly assume that the average preference of the laboratory sample of females is representative of that of the larger natural population (and hence an unbiased estimate of selection gradients, see Gershman et al. (2014) for further discussion). Studies that sample from natural populations will therefore not only increase the likelihood of capturing the naturally occurring phenotypic (co)variance, but also the average population preference. The lack of multivariate selection studies on CHCs that incorporate additional sexual traits provides an imperative for further work on the role of sexual selection in the evolution of CHCs. Additionally, expanding this work beyond drosophilid fruit flies and crickets, as recently done by Lane et al. (2016), will contribute to a taxonomically broader understanding of the evolution of insect CHCs. The dung beetle, Onthophagus taurus, has become a model species for studies of sexual selection and provides an ideal candidate for studies of selection on CHCs through female mate choice. Males vigorously drum the dorsal surface of females during courtship, and females prefer males that perform this courtship at a higher rate (Kotiaho et al. 2001; Kotiaho 2002; McCullough and Simmons 2016). Courtship rate is genetically correlated with a measure of male condition (Kotiaho et al. 2001), and females mated to males with a higher courtship rate have offspring of greater viability (Simmons and Holley 2011), suggesting the operation of good genes mate choice in this species (Garcia-Gonzalez and Simmons 2011). However, the influence of male traits from other sensory modalities on male attractiveness has not been determined. Mating occurs in narrow tunnels beneath dung pats, making a role for visual traits unlikely. We therefore hypothesized that olfactory traits, specifically CHCs, might play a role in the mating system of this species. Our aim in this study was to measure the strength and form of selection on O. taurus CHCs and, at the same time, assess how CHCs and courtship rate act together in influencing male attractiveness. METHODS Beetles were collected from Walpole, Western Australia, and transported to a controlled temperature room (set at 28 °C on a 12:12 h light:dark cycle) where they were maintained in single-sex groups of ~100 individuals in 10-L buckets that contained moist sand topped with cow dung. Mating trials took place two weeks after collection to ensure that all beetles were sexually mature. It is possible that time spent under controlled laboratory conditions altered the CHC profiles of the subject beetles and consequently the phenotypic (co)variance of our sample may differ from that found in the field. However, our sample should still capture more of the naturally occurring variation in CHC profiles when compared with the use of laboratory reared individuals, because of variation in age and mating history, for example. We therefore refer to our sample as “field collected” to contrast it to a sample derived from laboratory reared individuals, but highlight the need to interpret the results cautiously when referring to selection occurring in nature. Typically, females collected from the field are already mated, with estimates of the number of previous mates ranging between one and five (mean of 2.8 ± 1.1) (McCullough et al. 2017). We used no-choice trials where a single male and female were introduced into a chamber designed to mimic their natural breeding tunnels (Kotiaho et al. 2001). Pairs were checked every 2 min for 1 h and at each check a record was made if the male was displaying courtship activity (drumming the dorsal surface of the female with his forelegs). Courtship rate was calculated as the number of times a male was recorded displaying courtship at these 2-min intervals divided by time (60 min for unmated males or time elapsed before mating for mated males (Simmons and Holley 2011)). For data analysis, males that were observed courting were assigned either a 1 if mated or 0 if unmated, males that did not court were discarded. Beetles were frozen immediately following mating trials in preparation for CHC extraction and gas chromatography–mass spectrometry (GCMS) analysis. To extract CHCs, beetles were immersed in 1 mL of hexane for 5 min, after which the beetle was removed using clean forceps and the hexane allowed to evaporate. Prior to GCMS, the samples were re-suspended in 0.1 mL of hexane and transferred to a 0.1-mL glass insert within an autosampler vial. We used a Shimadzu QP2010 GCMS machine fitted with a 20 m × 150 µm × 0.15 µm VF1 column operating on the splitless mode. The initial oven temperature was set at 40 °C for 1 min, before increasing at 20 °C per minute until 200 °C where it was held for 1 min, and then increased at 10 °C per minute before reaching 300 °C, where it was held for 10 min. Peaks were identified from the NIST library by their mass spectra and retention indices. Retention indices were assigned by comparing compounds with a known sample of n-alkanes (C7–C40). Peaks were integrated using Shimadzu GCMSsolution Version 4.41. As is conventional for CHC analysis (Blows 1998), peak areas were transformed to logcontrasts by taking the log of their relative area divided by the relative area of a randomly chosen peak (in this case n-tricosane; we added 1 before taking the log to account for peak areas of 0). A principal component (PC) analysis was then performed on this set of logcontrasts as a data reduction tool, with only those components with an eigenvalue greater than 1 included in the analysis. We consider those peaks with loadings greater than or equal to 70% of the highest value as contributing significantly to a PC (Mardia et al. 1979). We used the multiple regression method outlined by Lande and Arnold (1983) to estimate linear and nonlinear selection gradients on our retained PCs and courtship rate. This approach first includes only the linear terms in a regression on relative fitness (in our case relative mating success) to estimate the strength of directional selection acting on each trait (β selection gradients). A second regression that includes all correlational and quadratic terms is then fit to estimate nonlinear selection (represented by the γ covariance matrix of correlational and quadratic selection) with nonzero quadratic terms representing curvature in the relationship between the trait and fitness. To allow comparison of selection across traits and studies, we standardized all traits to mean 0 and standard deviation 1 (PCs already have mean 0, so these were only variance standardized). All quadratic coefficients presented in our γ matrix were multiplied by 2 (Stinchcombe et al. 2008). As nonlinear selection is likely to be underestimated using this technique (Blows and Brooks 2003), we performed a canonical rotation of the γ matrix to estimate multivariate linear and nonlinear selection (Phillips and Arnold 1989). We visualized selection in multivariate space using thin plate splines by applying the canonical axes to the original phenotypic data. Overall significance of selection was tested by fitting a logistic regression (using absolute mating success as the response variable) to the model containing only the linear terms (for directional selection) and to the full second-order model using individual scores on the canonical axes (for nonlinear selection). As relative mating success was non-normally distributed, permutation testing was used for assessing the significance of individual terms, with our procedure for the canonical axes differing from that of Reynolds et al. (2010) in that we kept the canonical rotation constant (Chenoweth et al. 2012). As the detection of significant curvature in the fitness surface does not necessarily represent the presence of an intermediate fitness optima (or minima), we used the methods outlined by Mitchell-Olds and Shaw (1987) to test for significant stabilizing and disruptive selection on our canonical axes. This method involves two tests that constrain fitness to be at its maximum at one or the other extreme of the phenotypic range and tests whether this provides a significantly worse fit compared to the unconstrained model (see Chenoweth et al. (2007) for further details). Again, permutation testing was used to assess the significance of the relevant model term. To determine how the inclusion of the behavioral trait, courtship rate, influenced the selection gradient estimates for our CHC PCs, we repeated the first- and second-order multiple regressions, but this time excluding courtship rate. We then compared these linear and nonlinear selection gradients with those on our CHC PCs estimated from the model which included courtship rate. We calculated the angle between our two vectors of selection gradient estimates to compare how the overall estimated direction of selection differed between them (an angle of 0° would indicate perfect alignment, 90° that the overall direction of selection estimates were orthogonal, and 180° that selection was estimated to be in the opposite direction). Similarity between our two γ matrices was estimated using the matrix comparison method of Krzanowski (1979), which calculates the similarity of the matrix subspaces that describe the majority of the variation (see Rundle et al. (2008) for an example of how these point estimates can be used for the comparison of multiple γ matrices). In our case, we retained the first three eigenvectors of both matrices, with a score of 0 indicating that the matrices were completely unaligned, whereas a score of 3 would show perfect alignment. All analyses were performed in R (R Core Team 2017) using the FactoMineR package for the PC analysis (Lê et al. 2008) and the fields package for thin plate spline visualization (Nychka et al. 2015). RESULTS We analyzed hexane washes from a total of 119 male O. taurus, 21 (18%) of which were successful in securing a mating. Results from the GCMS analysis are presented in Table 1 and Figure 1 shows a typical chromatogram. The vast majority of the CHCs detected were methyl-branched compounds, with alkanes less common but representing some of the larger peaks. Our PC analysis returned 7 PCs with eigenvalues greater than 1 that cumulatively explained 83.13 % of the total variation (Table 1). PC1 was weighted positively by all but one compound and most strongly by longer chained compounds. We therefore interpret this axis as describing the relative abundance of heavier compounds. PC2 contrasts straight-chained alkanes with methyl-branched compounds that have a chain-length of 26 carbons or less. PC3 contrasts four methyl-branched compounds and two alkanes with the three longest-chained compounds. PC4 was positively loaded by two 2-methylalkanes, two alkanes, and two longer chained 15-methylalkanes. PC5 contrasts two 2-methylalkanes with an 8-methylalkane. PC6 contrasts a 2-methylalkane with a 7-methylalkane and PC7 is weighted by two short-chained alkanes. Though the patterns of variation explained by these PCs are not completely clear, and our description is necessarily a simplification, it is interesting to note that the 2-methylalkanes contribute to the phenotypic variation described by 5 of the 7 PCs. Figure 1 View largeDownload slide Chromatogram of a typical Onthophagus taurus CHC profile. The x-axis shows the retention time and the y-axis peak abundance. Numbers above peaks correspond to those in Table 1. Figure 1 View largeDownload slide Chromatogram of a typical Onthophagus taurus CHC profile. The x-axis shows the retention time and the y-axis peak abundance. Numbers above peaks correspond to those in Table 1. Table 1 Cuticular hydrocarbons of Onthophagus taurus, their mean relative amounts, and the results of the PC analysis PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 Only PCs with eigenvalues greater than 1 are shown. Loadings are shown in the body of the table and those in bold have values that are equal to or greater than 70% of the highest loading. View Large Table 1 Cuticular hydrocarbons of Onthophagus taurus, their mean relative amounts, and the results of the PC analysis PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 PC1 PC2 PC3 PC4 PC5 PC6 PC7 Eigenvalue 16.221 5.287 2.787 1.991 1.677 1.566 1.230 % variance 43.839 14.290 7.533 5.381 4.533 4.232 3.326 Peak Compound Mean SE 1 n-C22 0.10 0.00 0.083 −0.081 0.091 −0.163 −0.089 −0.032 0.551 2 n-C23 1.22 0.04 3 11-MeC23 0.86 0.05 0.067 0.343 0.025 0.012 −0.302 0.086 −0.010 4 2-MeC23 0.17 0.01 −0.034 0.249 −0.046 0.355 0.242 −0.158 0.247 5 3-MeC23 0.35 0.02 0.007 0.338 0.010 0.063 0.146 0.109 −0.110 6 n-C24 0.44 0.01 0.137 0.025 −0.036 0.034 −0.102 0.245 0.534 7 2-MeC24 0.19 0.01 0.008 0.267 0.067 0.310 0.258 −0.100 0.281 8 n-C25 5.50 0.13 0.178 −0.110 0.029 0.121 −0.144 0.327 0.191 9 11-MeC25 5.53 0.21 0.114 0.319 0.070 0.049 −0.106 0.201 −0.067 10 7-MeC25 0.64 0.03 0.057 0.251 0.121 −0.078 0.101 0.428 −0.221 11 2-MeC25 1.33 0.07 0.091 0.238 0.300 −0.071 −0.320 −0.181 −0.014 12 3-MeC25 1.86 0.07 0.118 0.317 0.196 0.104 −0.008 0.096 −0.029 13 n-C26 0.66 0.02 0.139 −0.226 0.166 0.271 −0.147 0.166 0.107 14 8-MeC26 1.41 0.05 0.171 0.052 0.085 −0.071 0.457 0.088 −0.038 15 8,14-diMeC26 0.68 0.03 0.159 0.066 0.250 −0.090 0.272 0.021 0.004 16 2-MeC26 0.59 0.02 0.103 −0.013 0.252 0.221 0.064 −0.496 −0.042 17 n-C27 6.23 0.26 0.111 −0.257 0.221 0.270 −0.040 0.168 −0.166 18 13-MeC27 5.62 0.16 0.190 0.090 0.176 −0.161 −0.066 0.085 −0.038 19 2-MeC27 2.63 0.11 0.129 0.069 0.293 −0.065 −0.378 −0.256 0.020 20 3-MeC27 3.10 0.08 0.190 −0.102 0.161 −0.007 0.090 −0.157 −0.064 21 n-C28 0.62 0.02 0.140 −0.216 0.222 0.161 0.125 0.153 −0.062 22 14-MeC28 1.26 0.04 0.195 −0.024 0.160 −0.231 0.154 −0.107 0.020 23 n-C29 1.99 0.10 0.119 −0.259 0.168 0.203 0.051 0.067 −0.075 24 15-MeC29 8.76 0.18 0.218 −0.031 0.035 −0.113 −0.057 −0.105 −0.082 25 11,15-diMeC29 1.62 0.04 0.223 0.025 −0.049 −0.164 0.130 −0.016 −0.008 26 19,23-diMeC29 3.25 0.08 0.228 −0.050 0.018 −0.175 0.058 −0.044 −0.030 27 5,15-diMeC29 2.98 0.08 0.211 −0.067 −0.052 −0.183 0.104 0.035 0.161 28 15-MeC30 1.32 0.03 0.228 −0.017 −0.083 −0.025 0.098 −0.071 −0.031 29 15-MeC31 12.81 0.25 0.220 −0.082 −0.093 0.110 −0.054 0.057 −0.093 30 13,17-diMeC31 1.80 0.05 0.206 0.073 −0.191 0.031 −0.012 −0.052 −0.028 31 9,21-diMeC31 3.36 0.12 0.204 −0.016 −0.193 −0.173 −0.049 −0.002 −0.015 32 7,17-diMeC31 2.07 0.05 0.221 0.015 −0.148 −0.047 0.017 −0.082 −0.036 33 5,17-diMeC31 2.11 0.06 0.205 0.016 −0.199 0.037 0.074 −0.063 0.166 34 15-MeC32 0.55 0.02 0.200 0.058 −0.154 0.147 −0.032 −0.146 −0.054 35 15-MeC33 3.33 0.10 0.196 −0.031 −0.158 0.269 −0.146 0.034 −0.116 36 11,21-diMeC33 8.36 0.28 0.205 0.018 −0.244 −0.077 −0.063 −0.003 −0.068 37 15-MeC35 0.54 0.03 0.133 0.052 −0.240 0.308 −0.083 −0.076 −0.108 38 11,21-diMeC35 4.17 0.18 0.183 0.053 −0.237 0.048 −0.070 −0.026 −0.065 Only PCs with eigenvalues greater than 1 are shown. Loadings are shown in the body of the table and those in bold have values that are equal to or greater than 70% of the highest loading. View Large The multivariate selection analyses revealed that courtship rate was subject to directional selection, with nonlinear selection acting on the CHC profile. We detected overall significant linear selection (χ28 = 34.938, P < 0.001, r2(adjusted) = 0.258), with this effect driven by courtship rate as no significant linear selection was found for our CHC PCs (Table 2). We detected significant negative correlational selection between PC2 and PC5 (Table 2) and overall significant nonlinear selection on the canonical axes (χ216 = 83.140, P < 0.001). Significant concave selection was found on axes m1 and m2, along with significant convex selection on axes m7 and m8 (Table 3). Selection acting upon axes m1 and m2 was disruptive, with models constraining the highest fitness to be at only the minimum or maximum of the phenotypic variation providing a significantly worse fit (P <0.001 in both cases for m1, P = 0.014 and P < 0.001 for m2). We did not detect significant stabilizing selection on axis m8 (P = 0.164 and P = 0.422) nor on axis m7 (P = 0.078 and P = 0.066). Table 2 Results of the multiple regression analyses including courtship rate (CR) β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05, ***P < 0.001 View Large Table 2 Results of the multiple regression analyses including courtship rate (CR) β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR PC1 0.182 0.268 −0.212 0.495 0.050 0.295 0.193 0.039 −0.054 PC2 −0.091 0.341 −0.174 −0.055 −0.639* −0.289 −0.104 0.355 PC3 −0.044 −0.316 −0.392 0.197 −0.224 −0.420 0.245 PC4 −0.141 −0.127 −0.393 0.127 0.050 −0.475 PC5 −0.017 −0.003 0.113 −0.038 −0.447 PC6 0.152 −0.646 0.104 0.538 PC7 0.220 0.567 −0.053 CR 1.137*** −0.014 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05, ***P < 0.001 View Large Table 3 Results of the canonical analysis derived from the multiple regression including PCs 1–7 and courtship rate (CR) mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01, ***P < 0.001 View Large Table 3 Results of the canonical analysis derived from the multiple regression including PCs 1–7 and courtship rate (CR) mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 CR m1 −0.183 1.303** 0.418 −0.611 0.244 −0.088 0.555 0.054 −0.038 −0.270 m2 0.346 0.943* 0.161 0.207 0.451 −0.376 −0.003 −0.021 −0.656 0.395 m3 −0.918*** 0.515 −0.290 0.005 −0.078 0.314 0.022 −0.314 −0.631 −0.560 m4 −0.063 0.188 0.742 0.218 0.090 0.525 −0.326 0.044 −0.043 −0.091 m5 0.404* −0.006 −0.164 −0.497 −0.196 0.353 −0.178 0.537 −0.329 0.367 m6 0.011 −0.516 0.157 0.461 −0.519 −0.024 0.560 0.383 −0.179 −0.009 m7 0.049 −0.992* −0.313 0.239 0.560 0.540 0.418 0.103 0.166 0.161 m8 −0.512* −1.363* −0.123 0.141 0.319 −0.245 −0.255 0.671 0.030 −0.537 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01, ***P < 0.001 View Large Further inspection of the major canonical axes revealed that the disruptive selection was most strongly associated with the CHC profile. Visualizing selection acting on the major canonical axes, m1 and m8, revealed a fitness trough at intermediate values of m1 (as indicated above by the significant disruptive selection) and a peak at decreasing values of m1 and m8 (Figure 2). An examination of the eigenvectors of axis m1 revealed that the fitness trough was associated with intermediate values of PC2 and PC5, representing a lack of contrast between short-chained methyl-branched compounds with n-alkanes (PC2), as well as a lack of contrast between two 2-methyl-branched compounds and 8-methylhexacosane (PC5). Inspecting the eigenvectors for axes m1 and m8 together revealed that the fitness peak was associated with high values for PC2 and low values for PC5 (m1), along with high courtship rate and low values of PC6 (m8). An inspection of the loadings of the original CHC logcontrasts on these PCs revealed an association between mating success and investment in short-chained methyl-branched compounds (PC2) and 2-methyalkanes (PC5 and PC6). Figure 2 View largeDownload slide Thin plate spline visualization in 3 dimensions (a) and as a contour plot (b) of selection acting on the major canonical axes, m1 and m8. Relative fitness is represented by the vertical axis in (a) and through color in both (a) and (b) with red representing high fitness and blue representing low fitness. The points on (b) are the raw data points. Figure 2 View largeDownload slide Thin plate spline visualization in 3 dimensions (a) and as a contour plot (b) of selection acting on the major canonical axes, m1 and m8. Relative fitness is represented by the vertical axis in (a) and through color in both (a) and (b) with red representing high fitness and blue representing low fitness. The points on (b) are the raw data points. A comparison of analyses that either included or excluded courtship rate revealed that nonlinear selection on the CHC profile was largely independent of the effect of courtship rate on mating success. Removing courtship rate and repeating the multiple regression analyses resulted in the model of directional selection no longer explaining a significant amount of variation in male mating success (χ27 = 9.275, P = 0.234, r2(adjusted) = 0.017). Interestingly, of the small amount of variation that was explained, there was significant directional selection on PC6 (Table 4), which out of all the PCs was found to have the strongest correlational selection with courtship rate (Table 2), although this was not significant (P = 0.094). Comparing the vector of directional selection gradients on the CHC PCs from this analysis with that obtained from our analysis that included courtship rate revealed that the two vectors were oriented 48° from each other and that the individual β estimates were not significantly correlated (r = 0.689, P = 0.087). The exclusion of courtship rate had little effect on the nonlinear selection estimates, with a comparison of the γ matrices returning a Krzanowski value of 2.513 out of a maximum of 3 (or 84% of the score for complete similarity), and a significant correlation between the individual γ estimates (r = 0.869, P <0.001). This was reflected again in finding significant nonlinear selection on the canonical axes derived from the multiple regression that excluded courtship rate (χ214 = 59.431, P <0.001). Although the two sets of canonical axes are not comparable (as they were based on two different sets of traits), it is notable that the CHC PC loadings on the two m1 axes bear some resemblance (Tables 3 and 5) and that this axis displays the greatest curvature in our canonical rotation that excluded courtship rate (Table 5). Table 4 Results of the multiple regression analyses excluding courtship rate β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05 View Large Table 4 Results of the multiple regression analyses excluding courtship rate β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 β PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC1 −0.029 0.154 −0.251 0.143 0.134 0.431 0.197 0.145 PC2 −0.094 0.148 −0.056 0.163 −0.690* −0.396 −0.281 PC3 0.108 −0.281 −0.279 0.081 −0.278 −0.454 PC4 −0.321 0.050 −0.486* 0.171 0.226 PC5 −0.207 0.131 −0.023 −0.047 PC6 0.393* −0.096 0.129 PC7 0.175 0.367 The vector of standardized directional selection gradients (β) and the γ matrix are shown. *P < 0.05 View Large Table 5 Results of the canonical analysis derived from the multiple regression of PCs 1–7 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01 View Large Table 5 Results of the canonical analysis derived from the multiple regression of PCs 1–7 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 mi θ λ PC1 PC2 PC3 PC4 PC5 PC6 PC7 m1 0.106 1.277** 0.399 −0.601 0.045 −0.226 0.599 0.199 0.167 m2 0.099 0.963* 0.102 −0.089 −0.416 0.473 −0.236 0.322 0.652 m3 −0.141 0.140 0.698 0.153 0.215 0.470 −0.096 0.190 −0.421 m4 −0.220 −0.046 0.407 0.322 0.140 −0.045 0.115 −0.681 0.481 m5 0.088 −0.449 0.245 0.553 −0.555 −0.442 0.188 0.300 −0.076 m6 −0.505** −0.494 −0.185 −0.019 −0.479 0.504 0.526 −0.358 −0.278 m7 −0.027 −0.920* 0.288 −0.445 −0.471 −0.230 −0.500 −0.371 −0.237 The strength of multivariate linear selection (θ), multivariate nonlinear selection (λ), and the loadings of the original traits on the mi axes are shown. *P < 0.05, **P < 0.01 View Large Discussion Using multivariate selection analyses that incorporated traits from different sensory modalities, we have found significant nonlinear selection on the axes of major CHC phenotypic variation (CHC PCs) and significant directional selection on a behavioral sexual display (courtship rate) in the dung beetle, O. taurus. Our finding of significant directional selection on courtship rate is consistent with previous studies on mate choice using this species (Kotiaho et al. 2001; Kotiaho 2002; Simmons and Holley 2011; McCullough and Simmons 2016). What is novel about our results is the finding that CHCs also contribute to mating success in O. taurus. Furthermore, the high level of similarity between our two γ matrices (84% of the maximum similarity score) indicates that nonlinear selection on the CHC PCs does not depend on an individual’s courtship rate. Our study therefore provides an example of mate-choice mediated sexual selection acting on an insect’s CHC profile independently of the influence of a well-characterized courtship display. The fact that we have shown this in the laboratory using beetles collected from the field, that would have varied in a variety of life-history traits including age and previous mating history, gives support to the view that sexual selection is likely to be a persistent driver of the evolution of CHCs in natural populations of insects. Although nonlinear selection on the CHC PCs appeared to act independently of courtship rate, the estimates of directional selection gradients were influenced to some extent by whether courtship rate was incorporated into the analysis. This was most evident in the doubling of the selection gradient (and consequently altering its statistical significance) for PC6, the PC that showed the strongest correlational selection with courtship rate. Similarly, comparing the overall direction of selection acting on the CHC PCs estimated from the two analyses which included or excluded courtship rate resulted in β vector estimates that were angled moderately (48°) away from each other. Although total linear selection on the CHC PCs was not significant, the ability of a non-CHC trait to influence the estimate of the overall direction of selection has important implications. For example, such an influence will affect the estimated alignment of selection with genetic variation, potentially altering interpretations of the amount of genetic variation available for selection. The potential for unmeasured characters to influence selection gradients is well known (Lande and Arnold 1983; Mitchell-Olds and Shaw 1987), and we have shown that estimates of selection on CHCs are no exception to this rule. As CHCs are commonly used in studies of sexual selection on multivariate traits (Chenoweth and Blows 2003; Van Homrigh et al. 2007; Thomas and Simmons 2009b, 2010; Curtis et al. 2013; Steiger et al. 2013; Ingleby et al. 2014; Steiger et al. 2015; Lane et al. 2016), our results indicate that a fuller understanding of mate choice will be gained by incorporating non-CHC traits in future studies. This will of course likely necessitate much larger sample sizes than we have included here, an admittedly challenging logistical proposition. Here, we were interested in assessing selection imposed on male O. taurus CHC profiles through the action of female mate choice. We emphasize that this is highly unlikely to represent the total selection acting on CHCs in this species. The role CHCs play in desiccation resistance (Gibbs and Rajpurohit 2010), along with their production utilizing the shared resource of internal hydrocarbons (Schal et al. 1994; Wicker and Jallon 1995), are both likely to impose strong selection on CHCs. For example, experimental evolution studies have found that altering the environmental temperature (Sharma et al. 2012) and imposing fecundity selection (Blows 2002) lead to an evolutionary response in the CHC profile. The CHCs of some species are also important in male dominance displays (Roux et al. 2002; Kortet and Hedrick 2005; Thomas and Simmons 2009a, 2011b), and Lane et al. (2016) have recently shown that selection on CHCs imposed by female mate choice can differ to that imposed by male–male competition. In investigations of selection via female choice on male CHC profiles, the use of no-choice mating assays is a useful experimental technique that avoids the potential effect of male–male competition. Our study adds O. taurus to the limited number of species for which no-choice mating assays have been used to assess selection from mate choice on male CHCs (broad-horned flour beetle, Gnatocerus cornutus (Lane et al. 2016); Australian field cricket, T. oceanicus (Thomas and Simmons 2009b); decorated cricket, Gryllodes sigillatus (Steiger et al. 2015); and D. simulans (Ingleby et al. 2014)). The fact that male O. taurus CHCs are subject to selection from female choice is evidence that females of this species use CHCs in their assessment of potential mates. Although our study was not designed to test what benefits such assessment may provide, our results suggest at least one potential hypothesis. Chemical traits have been suggested as good candidates for genetic compatibility-based mate choice (Mays and Hill 2004), and there is some evidence that CHCs act in this manner. For example, female cucumber beetles, Diabrotica undecimpunctata, prefer males with more dissimilar CHC profiles and produce offspring with higher immunocompetence when mated to their preferred mates (Ali and Tallamy 2010). Mating is more likely to occur between T. oceanicus pairs that have more dissimilar CHC profiles, and genetic distance seems to be correlated with CHC dissimilarity in this species (Thomas and Simmons 2011a). This, and the finding that CHC attractiveness is uncorrelated with a potential signal of good genes (courtship song), suggests that compatibility-based mate choice is facilitated through CHCs in these crickets (Simmons et al. 2013). Compatibility and “good” genes benefits are theoretically uncorrelated (Puurtinen et al. 2009), such that individuals can use one trait to assess good genes benefits, and another trait to assess the genetic compatibility of a potential mate. Such a view of mate choice falls under the “multiple messages” hypothesis which states that multicomponent/multimodal signals evolve to communicate different mate characteristics (Candolin 2003; Hebets and Papaj 2005; Partan and Marler 2005). Our finding of disruptive selection on male O. taurus CHCs acting independently of courtship rate is consistent with this scenario. There is strong evidence that courtship rate signals good genes in this species (Kotiaho et al. 2001; Simmons and Holley 2011) and the disruptive selection on the CHC profile we detected is consistent with compatibility-based mate choice. Alternatively, variation in courtship rate and the CHC profile in our study sample may be associated with age or other life-history traits that would have varied among individuals collected from the field, and females may be using this variation to select mates based on these characteristics. Although our data are suggestive of a multiple-messages/good-genes/compatible-genes scenario, alternative hypotheses are plausible, and at this point further experimentation is required before firm conclusions can be drawn. Due to its correlative nature, multiple regression is essentially a hypothesis-generating tool that requires experimental manipulation of traits to test its predictions (Chenoweth et al. 2012). Here, we have provided initial evidence that CHCs play a role in the mating system of O. taurus and suggest hypotheses for future research. Results from multivariate selection studies on CHCs in other species have proven fruitful for providing a basis for further investigation. For example, artificial selection on the linear combination of CHCs found to be subject to sexual selection in D. serrata resulted in a correlated increase in male mating success (Hine et al. 2011). Similarly, a positive selection gradient on 2-methylhexacosane suggested an attractiveness function for this compound in D. serrata (Chenoweth and Blows 2005), with this role experimentally validated using a perfuming experiment (Chung et al. 2014). These findings indicate that multivariate selection analyses on CHCs can provide useful insights for further experimental work. We suggest that incorporating different sensory modalities will provide further insights into how CHCs influence attractiveness alongside other traits, leading to a fuller understanding of the multivariate nature of mate choice. Conducting these studies across a broad range of taxa will contribute to elucidating general patterns in the relative importance of CHCs in mate choice, as well as consistencies (or otherwise) in the information gained by mate assessment based on CHCs in different taxa. We hope that our findings encourage further research on CHC mediated mate choice across a broader range of species that incorporate traits from other sensory modalities. Funding This work was supported through an Australian Government Research Training Program Scholarship to J.D.B. and an ARC Discovery Project to L.W.S. Data accessibility Analyses reported in this article can be reproduced using the data provided by Berson and Simmons (2018). We thank Jessica Horn for performing the mating trials and assisting with CHC extraction, Anchal Gupta for integrating the CHC peak areas, and Rowan Lymbery for statistical advice. The quality of the manuscript was improved by comments from 2 anonymous referees. The authors acknowledge the facilities and the scientific and technical assistance of the Metabolomics Australia Facility at the Centre for Microscopy, Characterisation & Analysis, The University of Western Australia, a facility funded by the University, State, and Commonwealth Governments. References Ali JG , Tallamy DW . 2010 . Female spotted cucumber beetles use own cuticular hydrocarbon signature to choose immunocompatible mates . Anim Behav . 80 : 9 – 12 . Google Scholar CrossRef Search ADS Andersson M . 1994 . Sexual selection . Princeton (NJ) : Princeton University Press . Bentsen CL , Hunt J , Jennions MD , Brooks R . 2006 . Complex multivariate sexual selection on male acoustic signaling in a wild population of Teleogryllus commodus . Am Nat . 167 : E102 – E116 . Google Scholar CrossRef Search ADS PubMed Berson JD , Simmons LW . 2018 . Data from: sexual selection across sensory modalities: female choice of male behavioral and gustatory displays . Dryad Digital Repository . https://doi.org/10.5061/dryad.8p81b6h. Blomquist GJ , Bagnères AG . 2010 . Insect hydrocarbons: biology, biochemistry, and chemical ecology . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Blows MW . 1998 . Evolution of a mate recognition system after hybridization between two Drosophila species . Am Nat . 151 : 538 – 544 . Google Scholar CrossRef Search ADS PubMed Blows MW . 2002 . Interaction between natural and sexual selection during the evolution of mate recognition . Proc R Soc B . 269 : 1113 – 1118 . Google Scholar CrossRef Search ADS PubMed Blows MW . 2007 . A tale of two matrices: multivariate approaches in evolutionary biology . J Evol Biol . 20 : 1 – 8 . Google Scholar CrossRef Search ADS PubMed Blows MW , Brooks R . 2003 . Measuring nonlinear selection . Am Nat . 162 : 815 – 820 . Google Scholar CrossRef Search ADS PubMed Blows MW , Chenoweth SF , Hine E . 2004 . Orientation of the genetic variance-covariance matrix and the fitness surface for multiple male sexually selected traits . Am Nat . 163 : 329 – 340 . Google Scholar CrossRef Search ADS PubMed Brooks R , Endler JA . 2001 . Direct and indirect sexual selection and quantitative genetics of male traits in guppies (Poecilia reticulata) . Evolution . 55 : 1002 – 1015 . Google Scholar CrossRef Search ADS PubMed Byers J , Hebets E , Podos J . 2010 . Female mate choice based upon male motor performance . Anim Behav . 79 : 771 – 778 . Google Scholar CrossRef Search ADS Callander S , Jennions MD , Backwell PRY . 2012 . The effect of claw size and wave rate on female choice in a fiddler crab . J Ethol . 30 : 151 – 155 . Google Scholar CrossRef Search ADS Candolin U . 2003 . The use of multiple cues in mate choice . Biol Rev . 78 : 575 – 595 . Google Scholar CrossRef Search ADS PubMed Chenoweth SF , Blows MW . 2003 . Signal trait sexual dimorphism and mutual sexual selection in Drosophila serrata . Evolution . 57 : 2326 – 2334 . Google Scholar CrossRef Search ADS PubMed Chenoweth SF , Blows MW . 2005 . Contrasting mutual sexual selection on homologous signal traits in Drosophila serrata . Am Nat . 165 : 281 – 289 . Google Scholar CrossRef Search ADS PubMed Chenoweth SF , Hunt J , Rundle HD . 2012 . Analyzing and comparing the geometry of individual fitness surfaces . In: Svensson EI , Calsbeek R , editors. The adaptive landscape in evolutionary biology . UK : Oxford University Press . p. 126 – 149 . Google Scholar CrossRef Search ADS Chenoweth SF , Petfield D , Doughty P , Blows MW . 2007 . Male choice generates stabilizing sexual selection on a female fecundity correlate . J Evol Biol . 20 : 1745 – 1750 . Google Scholar CrossRef Search ADS PubMed Chung H , Loehlin DW , Dufour HD , Vaccarro K , Millar JG , Carroll SB . 2014 . A single gene affects both ecological divergence and mate choice in Drosophila . Science . 343 : 1148 – 1151 . Google Scholar CrossRef Search ADS PubMed Cole GL , Endler JA . 2015 . Variable environmental effects on a multicomponent sexually selected trait . Am Nat . 185 : 452 – 468 . Google Scholar CrossRef Search ADS PubMed Coleman SW . 2009 . Taxonomic and sensory biases in the mate-choice literature: there are far too few studies of chemical and multimodal communication . Acta Ethologica . 12 : 45 – 48 . Google Scholar CrossRef Search ADS Curtis S , Sztepanacz JL , White BE , Dyer KA , Rundle HD , Mayer P . 2013 . Epicuticular compounds of Drosophila subquinaria and D. recens: identification, quantification, and their role in female mate choice . J Chem Ecol . 39 : 579 – 590 . Google Scholar CrossRef Search ADS PubMed Darwin C . 1871 . The descent of man, and selection in relation to sex . London : J. Murray . Garcia-Gonzalez F , Simmons LW . 2011 . Good genes and sexual selection in dung beetles (Onthophagus taurus): genetic variance in egg-to-adult and adult viability . Plos One . 6 : e16233 . Google Scholar CrossRef Search ADS PubMed Gerhardt HC , Brooks R . 2009 . Experimental analysis of multivariate female choice in Gray treefrogs (Hyla versicolor): evidence for directional and stabilizing selection . Evolution . 63 : 2504 – 2512 . Google Scholar CrossRef Search ADS PubMed Gershman S , Delcourt M , Rundle HD . 2014 . Sexual selection on Drosophila serrata male pheromones does not vary with female age or mating status . J Evol Biol . 27 : 1279 – 1286 . Google Scholar CrossRef Search ADS PubMed Gibbs AG , Rajpurohit S . 2010 . Cuticular lipids and water balance . In: Blomquist GJ , Bagnéres AG , editors. Insect hydrocarbons: biology, biochemistry, and chemical ecology . Cambridge : Cambridge University Press . p. 100 – 120 . Google Scholar CrossRef Search ADS Hebets EA , Papaj DR . 2005 . Complex signal function: developing a framework of testable hypotheses . Behav Ecol Sociobiol . 57 : 197 – 214 . Google Scholar CrossRef Search ADS Hill GE . 1991 . Plumage coloration is a sexually selected indicator of male quality . Nature . 350 : 337 – 339 . Google Scholar CrossRef Search ADS Hine E , Chenoweth SF , Blows MW . 2004 . Multivariate quantitative genetics and the lek paradox: genetic variance in male sexually selected traits of Drosophila serrata under field conditions . Evolution . 58 : 2754 – 2762 . Google Scholar CrossRef Search ADS PubMed Hine E , McGuigan K , Blows MW . 2011 . Natural selection stops the evolution of male attractiveness . Proc Natl Acad Sci USA . 108 : 3659 – 3664 . Google Scholar CrossRef Search ADS PubMed Ingleby FC , Hosken DJ , Flowers K , Hawkes MF , Lane SM , Rapkin J , House CM , Sharma MD , Hunt J . 2014 . Environmental heterogeneity, multivariate sexual selection and genetic constraints on cuticular hydrocarbons in Drosophila simulans . J Evol Biol . 27 : 700 – 713 . Google Scholar CrossRef Search ADS PubMed Johansson BG , Jones TM . 2007 . The role of chemical communication in mate choice . Biol Rev . 82 : 265 – 289 . Google Scholar CrossRef Search ADS PubMed Kortet R , Hedrick A . 2005 . The scent of dominance: female field crickets use odour to predict the outcome of male competition . Behav Ecol Sociobiol . 59 : 77 – 83 . Google Scholar CrossRef Search ADS Kotiaho JS . 2002 . Sexual selection and condition dependence of courtship display in three species of horned dung beetles . Behav Ecol . 13 : 791 – 799 . Google Scholar CrossRef Search ADS Kotiaho JS , Simmons LW , Tomkins JL . 2001 . Towards a resolution of the lek paradox . Nature . 410 : 684 – 686 . Google Scholar CrossRef Search ADS PubMed Krzanowski WJ . 1979 . Between-groups comparison of principal components . J Am Stat Assoc . 74 : 703 – 707 . Google Scholar CrossRef Search ADS Lande R , Arnold SJ . 1983 . The measurement of selection on correlated characters . Evolution . 37 : 1210 – 1226 . Google Scholar CrossRef Search ADS PubMed Lane SM , Dickinson AW , Tregenza T , House CM . 2016 . Sexual selection on male cuticular hydrocarbons via male-male competition and female choice . J Evol Biol . 29 : 1346 – 1355 . Google Scholar CrossRef Search ADS PubMed Lê S , Josse J , Husson F . 2008 . FactoMineR: an R package for multivariate analysis . J Stat Softw . 25 : 1 – 18 . Google Scholar CrossRef Search ADS LeBas NR , Hockham LR , Ritchie MG . 2003 . Nonlinear and correlational sexual selection on ‘honest’ female ornamentation . Proc R Soc B . 270 : 2159 – 2165 . Google Scholar CrossRef Search ADS PubMed Leonard AS , Hedrick AV . 2010 . Long-distance signals influence assessment of close range mating displays in the field cricket, Gryllus integer . Biol J Linn Soc . 100 : 856 – 865 . Google Scholar CrossRef Search ADS Mardia KV , Kent JT , Bibby JM . 1979 . Multivariate analysis . London : Academic Press . Mays HL , Hill GE . 2004 . Choosing mates: good genes versus genes that are a good fit . Trends Ecol Evol . 19 : 554 – 559 . Google Scholar CrossRef Search ADS PubMed McCullough EL , Buzatto BA , Simmons LW . 2017 . Benefits of polyandry: molecular evidence from field-caught dung beetles . Mol Ecol . 26 : 3546 – 3555 . Google Scholar CrossRef Search ADS PubMed McCullough EL , Simmons LW . 2016 . Selection on male physical performance during male-male competition and female choice . Behav Ecol . 27 : 1288 – 1295 . Google Scholar CrossRef Search ADS McGuigan K . 2009 . Condition dependence varies with mating success in male Drosophila bunnanda . J Evol Biol . 22 : 1813 – 1825 . Google Scholar CrossRef Search ADS PubMed Mitchell-Olds T , Shaw RG . 1987 . Regression analysis of natural selection: statistical inference and biological interpretation . Evolution . 41 : 1149 – 1161 . Google Scholar CrossRef Search ADS PubMed Nychka D , Furrer R , Paige J , Sain S . 2017 . fields: tools for spatial data. Version 9.6. Boulder (CO): University Corporation for Atmospheric Research. doi : Partan SR , Marler P . 2005 . Issues in the classification of multimodal communication signals . Am Nat . 166 : 231 – 245 . Google Scholar CrossRef Search ADS PubMed Phillips PC , Arnold SJ . 1989 . Visualizing multivariate selection . Evolution . 43 : 1209 – 1222 . Google Scholar CrossRef Search ADS PubMed Pryke SR , Andersson S , Lawes MJ . 2001 . Sexual selection of multiple handicaps in the red-collared widowbird: female choice of tail length but not carotenoid display . Evolution . 55 : 1452 – 1463 . Google Scholar CrossRef Search ADS PubMed Puurtinen M , Ketola T , Kotiaho JS . 2009 . The good-genes and compatible-genes benefits of mate choice . Am Nat . 174 : 741 – 752 . Google Scholar CrossRef Search ADS PubMed R Core Team . 2017 . R: a language and environment for statistical computing. Version 3.3.3 . Vienna (Austria) : R Foundation for Statistical Computing . Rebar D , Bailey NW , Zuk M . 2009 . Courtship song’s role during female mate choice in the field cricket Teleogryllus oceanicus . Behav Ecol . 20 : 1307 – 1314 . Google Scholar CrossRef Search ADS Reynolds RJ , Childers DK , Pajewski NM . 2010 . The distribution and hypothesis testing of eigenvalues from the canonical analysis of the gamma matrix of quadratic and correlational selection gradients . Evolution . 64 : 1076 – 1085 . Google Scholar CrossRef Search ADS PubMed Roux E , Sreng L , Provost E , Roux M , Clement JL . 2002 . Cuticular hydrocarbon profiles of dominant versus subordinate male Nauphoeta cinerea cockroaches . J Chem Ecol . 28 : 1221 – 1235 . Google Scholar CrossRef Search ADS PubMed Rundle HD , Chenoweth SF , Blows MW . 2008 . Comparing complex fitness surfaces: among-population variation in mutual sexual selection in Drosophila serrata . Am Nat . 171 : 443 – 454 . Google Scholar CrossRef Search ADS PubMed Rybak F , Sureau G , Aubin T . 2002 . Functional coupling of acoustic and chemical signals in the courtship behaviour of the male Drosophila melanogaster . Proc R Soc B . 269 : 695 – 701 . Google Scholar CrossRef Search ADS PubMed Schal C , Gu X , Burns EL , Blomquist GJ . 1994 . Patterns of biosynthesis and accumulation of hydrocarbons and contact sex pheromone in the female German cockroach, Blattella germanica . Arch Insect Biochem Physiol . 25 : 375 – 391 . Google Scholar CrossRef Search ADS PubMed Sharma MD , Hunt J , Hosken DJ . 2012 . Antagonistic responses to natural and sexual selection and the sex-specific evolution of cuticular hydrocarbons in Drosophila simulans . Evolution . 66 : 665 – 677 . Google Scholar CrossRef Search ADS PubMed Simmons LW , Holley R . 2011 . Offspring viability benefits but no apparent costs of mating with high quality males . Biol Lett . 7 : 419 – 421 . Google Scholar CrossRef Search ADS PubMed Simmons LW , Thomas ML , Simmons FW , Zuk M . 2013 . Female preferences for acoustic and olfactory signals during courtship: male crickets send multiple messages . Behav Ecol . 24 : 1099 – 1107 . Google Scholar CrossRef Search ADS Steiger S , Capodeanu-Nagler A , Gershman SN , Weddle CB , Rapkin J , Sakaluk SK , Hunt J . 2015 . Female choice for male cuticular hydrocarbon profile in decorated crickets is not based on similarity to their own profile . J Evol Biol . 28 : 2175 – 2186 . Google Scholar CrossRef Search ADS PubMed Steiger S , Ower GD , Stökl J , Mitchell C , Hunt J , Sakaluk SK . 2013 . Sexual selection on cuticular hydrocarbons of male sagebrush crickets in the wild . Proc R Soc B . 280 : 20132353 . Google Scholar CrossRef Search ADS PubMed Steiger S , Stökl J . 2014 . The role of sexual selection in the evolution of chemical signals in insects . Insects . 5 : 423 – 438 . Google Scholar CrossRef Search ADS PubMed Stinchcombe JR , Agrawal AF , Hohenlohe PA , Arnold SJ , Blows MW . 2008 . Estimating nonlinear selection gradients using quadratic regression coefficients: double or nothing ? Evolution . 62 : 2435 – 2440 . Google Scholar CrossRef Search ADS PubMed Tanner JC , Ward JL , Shaw RG , Bee MA . 2017 . Multivariate phenotypic selection on a complex sexual signal . Evolution . 71 : 1742 – 1754 . Google Scholar CrossRef Search ADS PubMed Thomas ML , Simmons LW . 2009a . Male dominance influences pheromone expression, ejaculate quality, and fertilization success in the Australian field cricket, Teleogryllus oceanicus . Behav Ecol . 20 : 1118 – 1124 . Google Scholar CrossRef Search ADS Thomas ML , Simmons LW . 2009b . Sexual selection on cuticular hydrocarbons in the Australian field cricket, Teleogryllus oceanicus . BMC Evol Biol . 9 : 12 . Google Scholar CrossRef Search ADS Thomas ML , Simmons LW . 2010 . Cuticular hydrocarbons influence female attractiveness to males in the Australian field cricket, Teleogryllus oceanicus . J Evol Biol . 23 : 707 – 714 . Google Scholar CrossRef Search ADS PubMed Thomas ML , Simmons LW . 2011a . Crickets detect the genetic similarity of mating partners via cuticular hydrocarbons . J Evol Biol . 24 : 1793 – 1800 . Google Scholar CrossRef Search ADS Thomas ML , Simmons LW . 2011b . Short-term phenotypic plasticity in long-chain cuticular hydrocarbons . Proc R Soc B . 278 : 3123 – 3128 . Google Scholar CrossRef Search ADS Van Homrigh A , Higgie M , McGuigan K , Blows MW . 2007 . The depletion of genetic variance by sexual selection . Curr Biol . 17 : 528 – 532 . Google Scholar CrossRef Search ADS PubMed Wicker C , Jallon JM . 1995 . Influence of ovary and ecdysteroids on pheromone biosynthesis in Drosophila melanogaster (Diptera: Drosophilidae) . Eur J Entomol . 92 : 197 – 202 . © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Behavioral EcologyOxford University Press

Published: Jun 6, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off