SEARCH

SEARCH BY CITATION

Keywords:

  • copulation duration;
  • mating speed;
  • reproductive isolation;
  • sexual conflict;
  • sexual selection;
  • sexually antagonistic coevolution;
  • speciation

Abstract

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

Sexual conflict has been predicted to drive reproductive isolation by generating arbitrary but rapid coevolutionary changes in reproductive traits among allopatric populations. A testable prediction of this proposal is that allopatric populations experiencing different levels of sexual conflict should exhibit different levels of reproductive isolation. We tested this prediction using experimentally evolved populations of the promiscuous Drosophila pseudoobscura. We manipulated sexual conflict by enforcing either monogamy, maintaining natural levels of promiscuity, or elevating promiscuity. Within each treatment, we carried out sympatric and allopatric crosses using replicated populations and examined pre-zygotic (number of mating pairs, mating speed and copulation duration) and post-zygotic (hybrid inviability and sterility) indicators of reproductive isolation. After 50 generations of selection, none of the measures conformed to predictions of sexual conflict driving reproductive isolation. Our results cannot be explained by lack of genetic variation or weak selection and suggest that sexual conflict may not be a widespread engine of speciation.


Introduction

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

The notion that allopatric speciation can occur via sexual selection is appealing because sexual selection directly affects traits, such as courtship and sexual communication, involved in mate recognition (Lande, 1981; West-Eberhard, 1983; Kaneshiro & Boake, 1987; for recent reviews, see Panhuis et al., 2001; Turelli et al., 2001; Coyne & Orr, 2004). With sufficiently low gene flow, sexual selection could promote substantial and rapid differences among allopatric populations, resulting in pre-zygotic reproductive isolation (RI) when populations come into secondary contact. Pre-zygotic RI is therefore a byproduct of divergence in mating signals and preferences. While verbally appealing, the first theoretical treatment of how sexual selection might promote RI was not formalized until Lande's (1981) seminal work and empirical support has only been accumulating since the mid-1990s. Most of this evidence is indirect and only consists of positive correlations between measures of species diversification rates and indices of the strength of sexual selection (e.g. Barraclough et al., 1995; Mitra et al., 1996; Møller & Cuervo, 1998; Owens et al., 1999).

More recent models of speciation by sexual selection propose sexual conflict as an ‘engine’ of speciation (Parker & Partridge, 1998; Rice, 1998; Gavrilets, 2000; Gavrilets et al., 2001). Sexual conflict results when male and female fitness optima differ, which is suggested to be a ubiquitous feature of promiscuous mating systems (Parker, 1979). Conflict arises due to the fundamentally different ways the sexes increase their reproductive success. Males are limited by the number of females they mate with whereas females are limited by the number of progeny they can produce (Bateman, 1948; Trivers, 1972). Sexual conflict can lead to sexually antagonistic coevolution whereby iterative bouts of (typically) male manipulation of female reproductive biology to promote male reproductive success is followed by the evolution of female resistance to this coercion (Arnqvist & Rowe, 2005; Moore & Pizzari, 2005). Sexually antagonistic coevolution may target different male and female traits across allopatric populations (Pizzari & Snook, 2003, 2004) and one byproduct of these within-population processes can be the evolution of RI among populations (Gavrilets, 2000). Gavrilets (2000) mathematical model of this idea generates two primary predictions: (i) the rate of trait change within a population, and the degree of RI among populations as a consequence of that trait, increase with the intensity of conflict, and (ii) if the latter increases with population density, then stronger/faster RI will occur between higher density populations (Martin & Hosken, 2003).

Evidence supporting sexual conflict as an engine of speciation is limited. Monandrous clades of insects are less speciose than polyandrous clades, a pattern that was attributed to sexual conflict (Arnqvist et al., 2000). However, the frequency of speciation events in birds is not correlated with the intensity of sexual selection and potential conflict (Morrow et al., 2003). Interpreting comparative studies on this question can be problematic (e.g. Panhuis et al., 2001; Coyne & Orr, 2004) due to difficulties controlling for phylogenetic effects and extinction rates (e.g. Morrow et al., 2003); the latter potentially being elevated by sexual selection which imposes a reproductive load on populations (Holland & Rice, 1999). An alternative to comparative studies is to use experimental evolution in which sexual conflict has been manipulated in allopatric populations. This experimental design allows testing the prediction that mating/reproductive traits will differ between allopatric populations experiencing conflict whereas these types of traits will not diverge rapidly among monogamous allopatric populations. Experimental evolution allows these predictions to be tested within a single species and replicated. Unfortunately, even this approach has provided equivocal results. Martin & Hosken (2003) found support for greater incipient RI among populations experiencing greater conflict following 35 generations of experimental manipulation of sexual conflict in the dung fly, Sepsis cynipsea. In their selection lines, allopatric promiscuous populations appear to be more reluctant to mate compared to either sympatric promiscuous populations or monogamous populations regardless of sympatry or allopatry. However, Wigby & Chapman (2006) found no evidence of RI in experimentally evolved populations of Drosophila melanogaster after 40 generations of selection under varying levels of sexual conflict. There were no differences in the willingness of females to mate with allopatric or sympatric males, regardless of the strength of sexual selection/conflict imposed. It is difficult to determine exactly why there is this discrepancy, but these two studies tested predictions of RI via sexual conflict by examining just one potential index of RI, the propensity of females to mate with sympatric vs. allopatric males (Martin & Hosken, 2003; Wigby & Chapman, 2006). While this measure encompasses various signal/receiver traits that promote copulation, it represents just one of many pre-zygotic RI indicators. Additionally, the potential for sexual conflict to generate post-zygotic RI has so far been neglected.

Here we used experimentally evolved lines of the naturally promiscuous species, Drosophila pseudoobscura, to test the role of sexual conflict in promoting both pre- and post-zygotic RI by measuring several potential indicators of each. We imposed three treatments that manipulated the strength of sexual selection and therefore the extent of sexual conflict; enforced monogamy (M; one female housed with one male) where sexual selection was relaxed, control promiscuity (C; one female housed with three males) that mimicked natural levels of sexual selection experienced by this species (Anderson, 1974; Cobbs, 1977), and elevated promiscuity (P; one female housed with six males) which increased the variance in male reproductive success relative to the other treatments (Crudgington et al., 2005). Each treatment was replicated four times.

Our previous work on these experimental populations provides evidence that our treatments have responded as predicted for putatively sexually selected traits (Snook et al., 2005) and that sexual conflict is occurring (Crudgington et al., 2005; R. R. Snook & H. S. Crudgington, unpublished data). Courtship song evolved rapidly with P males initiating song and singing a faster song to females relative to C and M males (Snook et al., 2005). We also found that M males induced increased early productivity compared to P males (Crudgington et al., 2005) and P males were harmful to female lifetime reproductive success (R. R. Snook & H. S. Crudgington, unpublished data). One candidate mechanism underlying this finding is the elevated courtship rate of P males relative to M males (R. R. Snook & H. S. Crudgington, unpublished data). Previous experimental evolution work in D. melanogaster also found that promiscuous males courted females more than enforced monogamy males and suggested that elevated harassment rates may negatively impact female fitness (Holland & Rice, 1999). Given that our D. pseudoobscura populations are responding rapidly and as predicted by sexual selection and conflict theory we tested whether conflict was generating incipient RI.

Previous empirical research on sexual conflict driving pre-zygotic reproductive isolation assessed only the number of failed matings (Martin & Hosken, 2003; Wigby & Chapman, 2006). Here, we additionally measure mating speed (the period of time between combining the male and female to the onset of copulation) and copulation duration (time from initiation of copulation to its termination), two well-studied traits in D. pseudoobscura (Kaul & Parsons, 1965; Parsons & Kaul, 1966; Spiess et al., 1966; Kessler, 1968, 1969; Spiess, 1968; Parsons, 1974). Both traits are heritable in this species (Kessler, 1968, 1969). They also have been used to indicate reproductive isolation in other Drosophila species. Copulation duration in heterogamic matings between D. simulans females and D. mauritiana males is shorter than homogamic matings (Cobb et al., 1988; Price et al., 2001), mating speed is faster in homogamic compared to heterogamic crosses in the nasuta-albomicans Drosophila complex (Tanuja et al., 2001), and sexual isolation between D. santomea and D. yakuba is a function of both changes in mating speed and copulation duration (Coyne et al., 2002).

In allopatric populations of Drosophila, pre- and post-zygotic RI evolve at the same rate (Coyne & Orr, 1989, 1997). Post-zygotic measures of RI include hybrid inviability and sterility and in Drosophila, the initial appearance of these mechanisms follows Haldane's rule; that is, hybrid males, which is the heterogametic sex in Drosophila, are inviable and sterile (Coyne & Orr, 1989, 1997). We therefore also measured male ‘hybrid’ inviability and sterility in crosses between allopatric populations of the same experimental evolution treatment.

We provide specific predictions for these traits regarding the role of sexual conflict in generating RI that can be tested using experimentally evolved lines (Table 1). If sexual conflict promotes RI, then we predict that all measured variables in the C and P replicate populations will differ in allopatry compared to sympatry whereas responses should not differ between sympatric and allopatric M crosses. Following from other work on sexual conflict as a driver of RI (Martin & Hosken, 2003; Wigby & Chapman, 2006), different treatments should also vary in their responses as a consequence of sexual conflict, with predicted responses depending on the variable being measured (Table 1). Following 50 generations of selection, we found no evidence in any of our measured traits supporting these predictions for reproductive isolation driven by sexual conflict.

Table 1.   Predictions and their rationale for each measured index of reproductive isolation relative to the manipulated level of sexual conflict (‘Treatment Response’) and whether crosses occur between sympatric or allopatric individuals of the same treatment (‘Sympatric and Allopatric Populations’)*.
RI indicator traitPrediction for treatment responsePrediction for comparisons of sympatric (s) and allopatric (a) populationsRationale if conflict is a Driver of RI
  1. Testing both columns of predictions requires a fully factorial (two-way) anova as it requires a test of the interaction. Only the column ‘Predictions for comparisons of Allopatric and Sympatric Populations’ can be tested using nested anova. To date, all studies, including ours, have used a nested design even where ‘Predicted Treatment Responses’ have been erroneously tested. See text for description of treatments and cross types.

  2. M, enforced monogamy; C, control promiscuity; P, elevated promiscuity; a, allopatric cross; s, sympatric cross.

  3. *The support for these predictions is discussed in the Introduction.

Mating speedM < C < PMs = Ma; Cs < Ca; Ps < PaDue to the absence of conflict, females exposed to and mating with only a single male (M treatment) should be less reluctant to mate and so mating should be initiated faster than females from the moderate promiscuity (C), which is predicted to be faster than females from extreme promiscuity (P) populations. If conflict drives RI, then in the promiscuous treatments (C, P) allopatric females will not recognize males and so mating speed will be greater in allopatric crosses compared to sympatric crosses.
Proportion of failed matingsM < C < PMs = Ma; Cs < Ca; Ps < PaDue to the absence of conflict, M females should be more willing to mate and so the proportion failing to mate should be less than C, then the P treatments. If conflict drives RI, then in the promiscuous treatments, C and P, allopatric females will not recognize males and so the proportion of failed matings will be greater in allopatric crosses compared to sympatric crosses.
Copulation durationM < C < PMs = Ma; Cs > Ca; Ps > PaBecause copulation duration appears to be controlled by males in this species (Kaul & Parsons, 1965; Parsons & Kaul, 1966), sexual conflict associated with greater promiscuity should promote copulation durations that are longer. If conflict drives RI, then promiscuous allopatric males will not be able to manipulate females and therefore their copulation duration should be shorter compared to sympatric promiscuous males.
F1 male inviabilityM = C = PMs = Ma; Cs < Ca; Ps < PaInviability should not differ between the treatments per se, however, if conflict generates RI, then in the promiscuous treatments allopatric crosses will result in post-zygotic RI.
F1 male sterilityM = C = PMs = Ma; Cs < Ca; Ps < PaSterility should not differ between the treatments per se, however, if conflict generates RI, then in the promiscuous treatments allopatric crosses will result in post-zygotic RI.

Methods

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

Experimental evolution of Drosophila psuedoobscura and general experimental design

Our selection regime has been described in detail elsewhere (Crudgington et al., 2005) so we only highlight the design details here. We set up four replicate populations of each of three experimental evolution treatments; monogamous lines (M) in which one female is randomly housed with one male, control promiscuous treatment (C) in which one female is housed with one males, and an elevated promiscuity treatment (P) in which one female is housed with six males. Each experimental evolution population is maintained in family sizes of either 80 (M) or 40 (C, P). Doubling the family size within M lines was done to minimize inbreeding effects (Snook, 2001) and has been successful; there is no difference in microsatellite diversity between the treatments suggesting that Ne does not substantially differ between treatments (R. R. Snook, unpublished data). Replicate populations of each treatment were staggered when set up with replicate 1 oldest and replicate 4 youngest. Within the replicate populations of each treatment, 5-day-old virgin males and females are housed together for 10 days (one vial change after 5 days), after which adults are discarded. Subsequent virgin progeny are collected and used to propagate the next generation. Overall, generation time in these populations is 27 days. The replicate populations were derived from an original wild-caught population that we maintain in large equal sex ratio breeding groups in the laboratory. We refer to this as the ancestral population. Experimental virgin flies from these populations were collected by CO2 anaesthesia and housed in single sex groups, ten flies per food vial for 5 days until reproductive maturity.

We measured the traits outlined in Table 1 after 48–52 generations of experimental selection. Our experimental design mimicked the two previous experimental evolution studies testing whether sexual conflict drives RI (Martin & Hosken, 2003; Wigby & Chapman, 2006). This design is nested (Fig. 1) such that within each experimental evolution treatment (e.g. M, C or P), we either placed a 5-day-old virgin female and 5-day-old virgin male together from the same replicate population (‘sympatric’; e.g. ♀M1 with ♂M1), or from different replicate populations (‘allopatric’; e.g. ♀M1 with ♂M2, or ♂M3, or ♂M4) (Fig. 1). Therefore, within each experimental evolution treatment we have four crosses (one sympatric and three allopatric) replicated four times (Replicate populations 1–4). The sample size for sympatric crosses was 60 and each allopatric cross had 20 pairs for a total of 60 (Fig. 1). We refer to sympatric and allopatric crosses as ‘cross-type’.

image

Figure 1.  Experimental nested design used to test the role of sexual conflict as a driver of reproductive isolation. For each selection treatment (M, C or P), two types of crosses (‘cross-type’) were performed: sympatric and allopatric. Sympatric crosses were carried out between females and males from the same replicate population of an experimental evolution treatment. Allopatric crosses were carried out between females from one replicate against males from all the others for a given experimental evolution treatment. Different experimental evolution treatments were never tested against each other. M, monogamy; C, control promiscuity; and P, promiscuity. Numbers (1–4) represent the replicate population for each treatment. Each replicate population is allopatric to each other. See text for description of treatments.

Download figure to PowerPoint

Pre-zygotic measures of reproductive isolation

In natural populations, it is likely that females are free to abandon costly or undesirable mating interactions. This scenario contrasts with laboratory trials in which each female is confined in a vial with a male. We therefore determined a threshold time within which to observe matings, which we defined as the time until 95% of a preliminary sample had mated. We determined the threshold by measuring mating speed (i.e. time to initiation of mating, measured in seconds) in a subsample of our populations and treatments using 27–32 pairs of 5-day-old virgin males and females. We placed one male and one female from the same treatment and replicate population together in a food vial and watched until mating occurred. One-way anova on log10-transformed data showed no difference among the different treatments in mating speed (F2,84 = 2.015, P = 0.140). The threshold was 586 s, approximately 10 min. In our experiment, we therefore watched pairs for 10 min to assess the number of successfully copulating pairs defined as those with a mating speed of 10 min or less, and classified those that failed to mate within this period as noncopulating.

Using our experimentally derived threshold, we assessed pre-zygotic measures of RI by placing a single 5-day-old virgin female and then a single 5-day-old virgin male in a food vial following the design outlined in Fig. 1. We performed these experiments within 60 min of the lights coming on in the incubator housing the flies. We determined the number of pairs copulating within the 10 min exposure threshold. When copulation occurred, we also measured mating speed and copulation duration. Pairs that had successfully mated were retained for 5 days and any productive vials were used subsequently to gain measures of post-zygotic RI.

Post-zygotic measures of reproductive isolation

Vials from all successful matings in the above experiment were kept and then scored subsequently for the presence or absence of male progeny (F1 male inviability). Of those vials in which male progeny enclosed, we took a random sample of 30 vials, collected one virgin male from each, and when they were 5 days old mated them to 5-day-old virgin females from the ancestral (noncoevolved) population. To determine F1 male sterility, we subsequently checked these vials for the presence or absence of larvae.

Statistical issues and analyses

Our experimental design mimics previous research that has used experimentally evolved populations to test incipient speciation arising from sexual conflict (Martin & Hosken, 2003; Wigby & Chapman, 2006). In these studies, all crosses were either within the same treatment and the same replicate population (which we refer to as the sympatry cross-type) or within the same treatment but between replicate populations (which we refer to as the allopatry cross-type). Importantly, for statistical interpretation, in all of these studies, crosses were always nested within a mating treatment (e.g. MxM, CxC and PxP for each replicate population; see Fig. 1) and never crossed among treatments. Given this design, we can ask if evolution under the same selection regime results in the same outcome; that is, we can compare the pattern between sympatry and allopatry within a treatment. However, because the design is nested (replicate population under cross-type), a statistical investigation of whether variation in sexual conflict yields a different effect on RI cannot be performed (i.e. you cannot statistically examine an interaction among selection regimes and cross-type). Previous studies (Martin & Hosken, 2003; Wigby & Chapman, 2006) have tested for this interaction despite it not having been experimentally performed and therefore the statistical model used in those studies is incorrect.

We used the following linear model to analyse our nested experimental design for mating speed and copulation duration (Fig. 1):

  • image

where yijk is the kth observation from the jth replicate population within the ith cross-type, μ is the overall mean of the response variable, αi is the effect of the sympatric or allopatric cross-type, and βj(i) is the replicate nested within cross type. Since cross-type is a fixed effect, αi is the difference between the overall mean and the mean for each cross-type. The nested factor, βj(i) is a random factor with zero mean and variance σ2, and ɛijk is the unexplained error associated with the kth observation from the jth replicate population within the ith cross-type. Regarding the potential role of sexual conflict as a driver of reproductive isolation, the hypothesis that we can test for all our variables is at the level of cross-type (see Table 1) and therefore the F-ratio for testing it has 1 degree of freedom in the numerator and 6 in the denominator. We cannot statistically test either the differences between the treatments or the interaction between treatments and cross-type. Sample sizes for mating speed and copulation duration varied among evolution treatments and cross-types as not all matings resulted in successful copulations. However, the level of imbalance was minor and the results from least squares analyses do not differ from restricted maximum likelihood and so we present the former. Each evolution treatment (M, C, or P) was analysed separately.

We used a G-test of independence (Sokal & Rohlf, 1995, pp. 715–724) to examine whether the number of failed matings depended on the cross-type for a given experimental evolution treatment, rather than the two-way anovas used incorrectly in previous studies. This test examines whether the different replicates within an evolution treatment vary in outcome and whether there are differences between cross-type within a given evolution treatment. Our tests of post-zygotic measures of RI, F1 male inviability and sterility, were designed to be tested using replicated G-tests as in the pre-zygotic mating success experiment above. However, for these traits, the level of failure was either non-existent (F1 male inviability) or low (F1 male sterility) so our analyses are limited.

Results

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

Pre-zygotic reproductive isolation

We analysed the number of pairs that failed to mate using replicated G-tests. For each of the four replicate populations, we predicted that within a given evolution treatments (M, C, or P) the number of failed pairs should depend on cross-type (Table 1). This pattern is not observed (Table 2). In the M treatment, there was significant heterogeneity among the four replicates. Contrary to sexual conflict predictions (Table 1), there were significant differences between sympatric and allopatric crosses in two M replicates and furthermore, the two significant observations were opposite in direction (there were more failed matings in sympatry in replicate 2 whereas there were more failed matings in allopatry in replicate 4). In the C treatment, there was no significant heterogeneity indicating that all replicates responded similarly. However, contrary to predictions, there was no difference in the number of failed allopatric compared to sympatric matings (Table 1). In the P treatment, similar to M, there was significant heterogeneity, indicating no consistent pattern among the replicate populations and differences in response between cross-type that were opposite to those predicted if sexual conflict drives RI (Table 1). In none of the tests (M, C, or P) did the within-treatment pooled analyses differ significantly between sympatric and allopatric crosses (although C was marginally nonsignificant). Overall, there was no support for a greater likelihood of failed matings in allopatry than in sympatry for any of the evolution treatments.

Table 2.   Replicated G-test examining the pattern of failed matings out of 60 within a 10 min observation period.
PopulationMating type G d.f. P-value
SympatricAllopatric
  1. Females were mated to sympatric (within replicate population) or allopatric (between replicate population) males (‘cross type’) within an experimental evolution treatment. The ‘Population’ column indicates the evolution treatment (M, C or P) and the numbers after the letter indicate the replicate population (1–4). See text for description of treatments and Fig. 1 for experimental design.

  2. Exact P-values are provided. Bold values are significant.

  3. M, enforced monogamy; C, control promiscuity; P, elevated promiscuity.

M111161.20010.273
M22187.9041 0.005
M314160.17810.673
M4112710.0941 0.002
Pooled57671.08810.297
Heterogeneity  18.2873 0.0004
C112100.22310.637
C218103.01410.083
C3560.10010.752
C422142.55610.110
Pooled57403.75010.053
Heterogeneity  2.14430.543
P11180.56510.452
P210183.01410.083
P31354.3111 0.038
P41161.73510.188
Pooled45370.94310.332
Heterogeneity  8.6833 0.034

We used separate mixed-effect nested anovas for each evolution treatment to test whether mating speed (time to initiation of mating) and copulation duration differed between allopatric and sympatric crosses (Table 1). Cross-type had no significant effect on mating speed in the M or C treatments (Fig. 2; Table 3a). In contrast, there was a significant difference in sympatric vs. allopatric matings in the P treatment (F1,6 = 7.525, P = 0.033; Table 3a), with faster mating speeds in allopatry than in sympatry (Fig. 2), contrary to what is predicted if sexual conflict drives RI. Cross-type had no significant effect on copulation duration for any treatment (Fig. 3; Table 3b).

image

Figure 2.  Effects of experimental evolution treatment and sympatric or allopatric crosses on mating speed (mean seconds + 95% upper confidence interval). M, enforced monogamy; C, control promiscuity; P, elevated promiscuity. See text for description of treatments.

Download figure to PowerPoint

Table 3.   Nested anova examining the effect of sympatric (within replicate population) and allopatric (between replicate populations) crosses (cross-type) on mating speed (a) and copulation duration (b) within an experimental evolution treatment (M, C or P).
 Sourced.f.MS F P-value
  1. See text for description of treatments and Fig. 1 for experimental design. Bold values are significant.

  2. M, enforced monogamy; C, control promiscuity; P, elevated promiscuity.

(a) Mating speed
MCross-type10.00260.00710.9356
Replicate (Cross)60.3676
Residual3480.1405
CCross-type10.56100.88240.3838
Replicate (Cross)60.6358
Residual3750.1512
PCross-type12.23797.5072 0.0334
Replicate (Cross)60.2981
Residual3900.1453
(b) Copulation duration
MCross-type10.30640.36320.5688
Replicate (Cross)60.8435
Residual3220.0514
CCross-type10.27121.02610.3502
Replicate (Cross)60.2643
Residual3420.0431
PCross-type10.10940.47090.5182
Replicate (Cross)60.2323
Residual3700.0496
image

Figure 3.  Effects of experimental evolution treatment and sympatric or allopatric crosses on copulation duration (mean seconds + 95% upper confidence interval). M, enforced monogamy; C, control promiscuity; P, elevated promiscuity. See text for description of treatments.

Download figure to PowerPoint

Post-zygotic reproductive isolation

Contrary to predictions regarding male inviability (Table 1), all matings resulted in the production of F1 male offspring thus no statistical test is necessary. This result provides no evidence of male inviability arising from allopatric crosses within any of the evolution treatments. Also contrary to predictions regarding male sterility (Table 1), the absence of larvae resulting from crosses between F1 males mated to ancestral females was observed in only 15 out of 720 vials, and were randomly distributed both among treatments and replicate populations and whether F1 males had derived from sympatric or allopatric crosses. Due to the small number of vials failing to produce larvae, a statistical analysis was not appropriate.

Discussion

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

Our experimental manipulations of the strength of sexual selection and conflict have successfully shown the action of both (Crudgington et al., 2005; Snook et al., 2005; R. R. Snook & H. S. Crudgington, unpublished data). Therefore, following the predictions of a recent model (Gavrilets, 2000) and using the experimental designs of previous studies testing this idea (Martin & Hosken, 2003; Wigby & Chapman, 2006), we expected that a signature of incipient reproductive isolation would be associated with sexual conflict. However, after 50 generations of selection, which is longer than either of the previous studies, we see no evidence of either pre- or post-zygotic reproductive isolation as predicted by sexual conflict theory. In fact, when we do see variation in response between sympatric and allopatric crosses (Table 2; Fig. 2), they directly contradict the predictions of such theory. Our tests have been conservative, appropriately analysed, and we have examined more potential indices of RI than the previous studies (Martin & Hosken, 2003; Wigby & Chapman, 2006); we find nothing to indicate that rapid random changes have caused divergence in our populations experiencing higher levels of conflict compared to those with either lower levels of or absent conflict.

Several experimental factors, such as weak selection, insufficient evolutionary time, inbreeding, or not measuring the correct variable, could explain the lack of support for sexual conflict driving reproductive isolation in our study. However, we observed the action of sexual conflict by at least 21 generations of selection (Crudgington et al., 2005; R. R. Snook & H. S. Crudgington, unpublished data). Moreover, sexual selection on reproductive signalling appears to be strong. When we examined courtship song in our M, C and P treatments after 25 generations of selection, there were changes consistent with the action of sexual selection (Snook et al., 2005). We have controlled for differential levels of inbreeding among our evolution treatments through the experimental design (see Methods). Furthermore, not only we have measured more traits in more replicate populations than other studies but also we have measured those traits typically examined in studies of Drosophila reproductive isolation (Cobb et al., 1988; Coyne & Orr, 1989, 1997; Price et al., 2001; Tanuja et al., 2001; Coyne et al., 2002).

If sexual conflict is a strong engine of speciation, why then did we not observe incipient RI? We outline two potential explanations. First, if the mathematical model of sexual conflict and speciation is altered to incorporate costs associated with male manipulation and female resistance, then there is reduced potential for allopatric speciation to occur (Gavrilets et al., 2001). There are costs of reproduction for D. melanogaster males associated with increased mating activity, including wing vibrations and copulation attempts (Partridge & Farquhar, 1981; Cordts & Partridge, 1996). Male courtship costs are likely to be greater in populations undergoing more intense sexual conflict and indeed P D. pseudoobscura males court females more than M males (R. R. Snook & H. S. Crudgington, unpublished data). Such costs may mitigate the potential for conflict to generate RI in D. pseudoobscura. If such costs easily undermine any tendency to generate RI, then sexual conflict may not be as pervasive an engine of speciation as anticipated.

Second, the evolutionary trajectories of reproductive traits may not be arbitrary which would promote gene flow between populations. We see parallel evolution of courtship song in our replicate populations; that is the responses across all four replicate populations of each evolution treatment are qualitatively the same (Snook et al., 2005). The idea that sexual selection and conflict promotes RI is based on these selection pressures causing arbitrary (not parallel) trajectories in reproductive traits of allopatric populations. If male mating signals and coercion traits in replicate populations of the same sexual selection treatment fail to diverge along different trajectories as predicted, then individuals from these populations will not have different reproductive signals and therefore not have reproductive barriers to mating. We therefore would not expect to see RI in allopatric crosses. Parallel evolution of sexually selected traits has also been found in sticklebacks (Boughman et al., 2005). If parallel evolution of reproductive communication occurs frequently among allopatric populations then this will reduce the ability of sexual conflict, and sexual selection in general, to generate divergence and consequently RI.

In conclusion, understanding the forces that result in diversification has been problematic and burdened by the challenges of disentangling them (e.g. Coyne & Orr, 2004). Despite the current popularity of the idea that sexual conflict promotes speciation (e.g. Rice, 1998; Arnqvist et al., 2000; Gavrilets, 2000; Gavrilets et al., 2001), the available empirical evidence does not support this enthusiasm. While a comparative analysis has found correlative support for sexual conflict driving speciation (Arnqvist et al., 2000), there are interpretation issues with this result (Panhuis et al., 2001; Coyne & Orr, 2004) and other comparative work on birds (Morrow et al., 2003) and Goodeid fish (Ritchie et al., 2005) does not indicate that sexual conflict influences patterns of divergence. Although no prior study could compare across sexual selection treatments, within a treatment there are clear predictions regarding the trait response to sexual conflict as a driver of reproductive isolation (Table 1). In the three experimental evolution studies three different patterns of response have been seen; supportive (Martin & Hosken, 2003), no variation in response (Wigby & Chapman, 2006), and when responses are present, they are in opposite directions to those predicted by sexual conflict as a driver of RI (this study). Even if sexual conflict is a universal feature of promiscuous mating systems, it is not necessarily a widespread engine of speciation.

Acknowledgments

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

We thank NSF and NERC for funding, and Anthony Turner, Kate Hutchence, Daniel Bradshaw, Claire Cooper, Hannah Dixson, and Jen Stockdale for help during the experiment. The reviewers provided valuable comments that helped clarify our presentation.

References

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