Allen J. Moore, School of Biological Sciences, University of Manchester, 3.614 Stopford Building, Oxford Road, Manchester M13 9PT, UK. Tel.: +44(0) 161 275 7221; fax: +44(0) 161 275 3938; e-mail: email@example.com
Abstract Female mate choice has been demonstrated in a wide variety of species and is now accepted as an important factor in sexual selection. One of the remaining questions, however, is why females prefer specific males. Do females or their offspring benefit from their choice? Or do females choose mates to minimize costs of mating? Here we show that, in the ovoviviparous cockroach Nauphoeta cinerea, where sexual selection has been well documented, females chose mates to avoid costly male manipulation. Females were partnered with preferred or nonpreferred mates, and fitness of the females measured. We found that females lived longer when they mated with preferred males. Female lifespan depended on the rate at which offspring developed from egg to parturition: slower development led to longer life. We manipulated the male pheromone and showed that the component of the pheromone blend that makes males attractive to females also delayed parturition. Thus, like other aspects of sexual conflict in this species, offspring development and thereby the mother's lifespan depended on exposure of females to specific components of the male pheromone. Males benefit from manipulating offspring development because females with accelerated parturition remained unreceptive whereas females with slower developing offspring readily remated after giving birth to their offspring. Our results suggest a hormone-like role for the male pheromone in N. cinerea and provide the first direct evidence of mate choice to avoid male manipulation. This study shows that dominant males may not be preferred males if they are manipulating females, why multiple components with contrasting effects can exist in a sexual signal, and emphasizes the complex fitness relationships that can arise in species with sexual conflict.
There have been two main approaches to investigating sexual conflict: comparative studies among species (Eberhard, 1996; Arnqvist et al., 2000; Arnqvist & Rowe, 2002) and investigations of fitness effects of social interactions within species (Holland & Rice, 1998; Moore et al., 2001). Examining sexual conflict within a species shares some of the difficulties encountered by studies of mate choice; it is often difficult to separate male and female control of mating outcomes, and therefore particularly difficult to document fitness differences between the sexes (Parker, 1979). The most successful studies of sexual conflict capitalize on species that can be studied in the laboratory, and have imposed or eliminated conflict by artificial selection (Holland & Rice, 1998). The limit to this approach is that artificial selection studies can answer how sexual conflict will result in evolution, but not why sexual conflict should exist. An alternative approach that can address the latter question is to use laboratory studies to examine a species where male and female behaviour can be studied in isolation, and where both male–male competition and female mate choice can be experimentally investigated. Despite the large number of studies investigating both sexual selection and sexual conflict, such conditions exist for very few species. An exception is the cockroach Nauphoeta cinerea, where males form dominance hierarchies, males attract females by a pheromone, and females discriminate amongst potential mates using these pheromonal cues.
Previous studies of sexual selection in N. cinerea allow us to unambiguously separate and identify the different mechanisms of sexual selection. In N. cinerea males form dominance hierarchies, females search for males, and females discriminate among males. Unusually for insects in general and cockroaches in particular, females do not produce attractant pheromones. Instead, all of the mechanisms of sexual selection in N. cinerea are influenced by a male-produced social pheromone that has three components (Moore et al., 1997, 2002; Moore, 1997; Moore & Moore, 1999). High levels of two components that are genetically coupled (Moore, 1997), 4-ethyl-2-methoxyphenol and 2-methylthiazolidine, result in high social status for males (Moore et al., 1997; Moore, 1997; Moore et al., 2002). Increased levels of the genetically independent third component, 3-hydroxy-2-butanone, result in subordinate status for males (Moore, 1997; Moore et al., 1997; Moore et al., 2002). These effects are additive, so that the ratio of the various components influences male social status. Females use this same pheromone to discriminate among males; females prefer males that have significant levels of 3-hydroxy-2-butanone regardless of the levels of the other two components (Moore & Moore, 1999; Moore et al., 2002). Thus there appears to be sexual conflict in N. cinerea, with males and females having different fitness optima with regard to the heritable pheromone of males. Further, as is expected in sexual conflict, there is a demographic cost associated with male and female conflict in N. cinerea (Moore et al., 2001). When females are exposed to the social pheromone and the component 3-hydroxy-2-butanone in particular, they produce fewer male offspring and therefore fewer total offspring in their first clutch (Moore et al., 2001). These results support the hypothesis that there is a sexual conflict in this species, but why do females preferentially expose themselves to a compound that reduces their reproductive output? Furthermore, why does the behaviour of females result in selection to increase this pheromone component?
We addressed these questions by first allowing socially naïve virgin females to discriminate between males and then mating them with their preferred (P) or nonpreferred (NP) mates (Moore et al., 2001). We then allowed females to remate with the same or the other male when they regained receptivity resulting in four treatments: P/P, P/NP, NP/P and NP/NP (reflecting the category of the first male then the second male to mate). In N. cinerea, females receive sufficient sperm in a single mating to fertilize all of their eggs and remating opportunities are limited. Immediately after mating and during pregnancy, females are unreceptive and will not remate. This is initially because of physiological mechanisms first induced by the presence of a male spermataphore (Roth, 1962, 1964a), and then by pregnancy (Roth, 1964b). After females give birth to their first clutch, receptivity is regained but only for 24–48 h (Roth, 1962, 1964a, b). It was during this period we provided females with an opportunity to remate.
Our experimental design allowed us to discriminate between female mating preferences that result in fitness benefits, and female mating preferences avoiding manipulation. If mate choice produces benefits, we predicted that females that mated with preferred males at any time would have greater fitness for themselves or their offspring than females that mated with only nonpreferred males. If mate choice evolved to avoid male manipulation, we predicted that the order in which females mated with preferred or nonpreferred males would matter. Fitness was scored as longevity (for females) and number of offspring produced. Duration of offspring development in N. cinerea is correlated with attractiveness of males to females as well as with male social status (Moore, 1994; Moore et al., 2002) so we also examined how offspring development and female survival were related. To determine male manipulation, we measured willingness of females to remate. Finally, given the importance of the male-produced pheromone in sexual selection and its putative role in sexual conflict (Moore & Moore, 1999; Moore et al., 2001, 2002), we conducted a second experiment where females were exposed to different pheromone components.
The cockroaches used in this study were derived from a laboratory population that has been established for over 50 years and within which we have studied sexual selection for over 15 years. This population is maintained with many thousands of individuals and, as assessed by starch gel electrophoresis scoring allozyme variation, there is little inbreeding (Corley et al., 2001). Individuals were isolated from mass colonies as nymphs and reared in separate sex containers under constant temperature, humidity and lighting conditions as in previous studies (Moore, 1997; Moore & Moore, 1999). Both males and females were transferred to individual containers with ad libitum food and water on the day they emerged to adulthood. Ten days past adult emergence, individuals were used in the first olfactometer trials. Details of rearing procedures are presented in detail elsewhere (Moore & Moore, 1999; Moore et al., 2001).
Mate choice trials and fitness
Preference tests were carried out in an olfactometer following previously established protocols (Moore, 1988, 1989; Moore et al., 1997; Moore & Moore, 1999; provides a diagram of the apparatus). Briefly, under red light, males are placed in isolated chambers and filtered air blown over them. Females are introduced downstream and their behaviour observed. The olfactometer allows us to score female choice because females walk upstream in a partitioned maze that requires them to follow one or the other odour stream if they continue. Females typically pause at the point where both male odours can be sampled simultaneously before proceeding along the olfactometer following a single male odour. The advantages of using this apparatus is that it allows females to perceive the odour of the two males but prevented the males from perceiving the female or each other. In addition, it isolates the cue used for mate choice to (primarily) odours, as males are not visible to females under the experimental conditions.
Females were allowed to choose between two naïve males. Choices scored under these conditions correspond with preferences exhibited in courtship and mating (Moore & Moore, 1999) and are highly repeatable (Moore, 1989). Once a female had exhibited a preference for one male over the other, the female was randomly assigned to mate with the preferred male (P mating, n = 61) or the nonpreferred male (NP mating, n = 61). Once mated, females were returned to their individual containers, held in an incubator, provided ad libitum food and water, and allowed to give birth. Females were checked twice a day for the presence of first instar larvae. Males were returned to individual containers and provided food and water ad libitum. Males and females were held in separate incubators to prevent exposure of females to additional male odours during gestation.
Twenty-four hours after giving birth, females were again mated to one of the two males, again randomly assigned to the P or NP male. Thus, there were four treatments in total: P mating followed by P mating (P/P, n = 31); P mating followed by NP mating (P/NP, n = 30); NP mating followed by P mating (NP/P, n = 30); NP mating followed by NP mating (NP/NP, n = 31). Females were not run through the olfactometer trials again, but placed directly with a male from the previous trial whose rank reflected the original preference of the female. If females did not mate within 10 min, a time sufficient to indicate long-term refusal to mate (Moore, 1990), the male and female were returned to their original containers. Twenty-four hours later, the male and female were provided another opportunity to mate. Only two females who previously refused to mate eventually mated during this second opportunity.
Females were returned to their individual containers after the second mating. These females were maintained in the incubator, provided with fresh food and water ad libitum and checked for additional clutches or death twice a day. When females produced clutches, the offspring were removed 24 h after birth. For all clutches the interbrood interval and number of offspring produced was recorded on the day of birth.
Females were mated to males that had artificially enhanced pheromone profiles as in previous studies (Moore et al., 1997, 2001; Moore & Moore, 1999). Pheromone quantities were manipulated by adding components in acetone solvent to a filter paper disc glued to the male pronotum. Manipulated males were placed with females within 5 min of having pheromone added to a disc. Males were manipulated so that genetically independent components were enhanced (2-methylthiazolidine + 4-ethyl-2-methoxyphenol or 3-hydroxy-2-butanone), the full blend was enhanced in a ratio representative of the population mean (Moore, 1997), or an acetone (solvent) control was present.
Mate choice trials and fitness
Total lifespan of the female depended on category of males with which she had mated (Table 1), reflecting a statistically significant effect of female mating experiences on probability of survival (Fig. 1; Cox proportional hazards regression with mating treatments as strata, = 7.8212, P < 0.05). The effects of treatment on the pattern of survivorship became especially pronounced immediately following the birth of the second clutch (post day 90, Fig. 1). Although females were exposed to both types of males in the olfactometer, females that were allowed to mate (and therefore interact more extensively) exclusively with preferred males were more likely to survive than females that were limited to mating only with nonpreferred males. In addition, the effects may interact, as a subjective inspection of the figure, suggests that females exposed to different category males for the second mating had intermediate patterns of survivorship. These intermediate values corresponded to a period after the second clutch was born (Table 1). Finally, there was a trade-off between survivorship and reproduction. Female lifespan was significantly related to average size of the clutch that was produced, with the expected negative trade-off between total lifespan and reproduction (Fig. 2). Females with larger clutches had significantly shorter lifespan (regression; t = −2.012, d.f. = 120, P < 0.05).
Table 1. Effect of mating treatments on female life history. Data presented are means with standard errors in parentheses.
Category of mate in first mating (P, preferred male; NP, nonpreferred male) followed by category of mate in second mating.
Previous work has shown a relationship between male attractiveness and rate of offspring development (Moore, 1994) so we examined the relationship between offspring development and mother's lifespan. Females that produced offspring that developed slower had a longer lifespan (Fig. 3; regression: t = 4.029, d.f. = 120, P < 0.001). Faster development may have been associated with shorter lifespan for females because giving birth is costly for females. We found that death was often associated with parturition. Forty-three of the 122 females died within 24 h of parturition of a clutch, including 27 that died in the process of giving birth. A further 11 females died within 1 week of parturition. The remainder was clustered around the expected date-of-next birth, based on average development rate, of the next clutch. In addition, the effects on female survival may not depend on actually mating a second time: social exposure to males and their pheromones was sufficient to observe an effect. Although our sample sizes precluded meaningful statistical analysis, the survivorship pattern of those females that did not regain receptivity was the same as those that did regain receptivity. Of those that were not receptive, P/P lived longest (n = 7; ± SD = 154.86 ± 10.36 days), P/NP (n = 5; 134.00 ± 12.26) and NP/P (n = 5; 138.00 ± 12.26) lived an intermediate length of time, and NP/NP (n = 8; 121.75 ± 9.69) lived shortest.
Although there was a fitness costs to females, males benefited by manipulating offspring development. Females with clutches that developed more quickly were significantly less likely to regain receptivity to courting males. Length of offspring development, measured as number of days to parturition, had a significant effect on likelihood of females remating and thus their receptivity (length of development for first clutch logistic regression coefficient = 0.212, SE = 0.104; log-odds = 4.656, d.f. = 1, P < 0.05). There was also a significant difference between length of development of the second clutch and receptivity (logistic regression coefficient = 0.415, SE = 0.120; log-odds = 16.696, d.f. = 1, P < 0.001). Thus, by manipulating offspring development males can further reduce the likelihood of their mates remating.
The effects on development are directly affected by the social pheromone. Exposure to the components or blends of the male pheromone that influence status and attractiveness had a significant effect on duration of offspring development (Fig. 4; F2,72 = 3.177, P < 0.05). Females that were exposed to the component that makes males attractive but subordinate, 3-hydroxy-2-butanone, had offspring that took significantly longer to develop. If the effects of different pheromone components are additive, as in the effects on dominance but not on attractiveness, then increasing all three components simultaneously should alter development less than increasing 3-hydroxy-2-butanone by itself. Exposure of females to increased concentrations of the full blend resulted in intermediate offspring development duration, with a significant linear trend in duration of development from the mixture of 2-methylthiazolidine + 4-ethyl-2-methoxyphenol, to the full blend, to 3-hydroxy-2-butanone (F1,93 = 4.917, P < 0.05).
In N. cinerea, life-history characters of females and her offspring clearly depend on the male with whom she interacts (Table 1; Moore, 1994; Moore et al., 2001, 2002) and there is a benefit of exposure to 3-hydroxy-2-butanone to females (Moore et al., 2001). Why then are dominant males deficient in this component (Moore, 1997; Moore & Moore, 1999; Moore et al., 2001, 2002)? Furthermore, the male pheromone acts as a badge of status; males manipulated to have high levels of 3-hydroxy-2-butanone are subordinate (Moore et al., 1997). As expected given the benefits of 3-hydroxy-2-butanone to females, dominant males are not always preferred as a mate when females have unrestricted choice (Moore & Moore, 1999; Moore et al., 2002). Yet, in social interactions, dominant males behaviourally prevent subordinate males from releasing pheromone (Moore et al., 1995) reducing the likelihood that females will be exposed to this compound and preventing even indirect benefits to females. Such behaviour supports our previous interpretation (Moore, 1988; Moore et al., 2001) that male and female evolutionary interests do not match. Both mate choice and the dominant male strategy can be successful; dominant males are more likely to mate than subordinate males (Moore et al., 2001) but, ultimately, males can only mate if females allow males to assume the necessary position (Moore, 1990). Dominance by males may limit the ability of females to exert unconstrained mate choice given the potential for injury arising from aggressive males (Moore, 1990).
The behavioural advantages of dominance for males would be limited if females could remate. Remating by females would be predicted given the reduced fecundity of females mating with dominant males. However, remating is related to duration of offspring development (Roth, 1964b), and dominant males also manipulate female remating opportunities by manipulating the rate of offspring development. Females with clutches that developed more quickly were significantly less likely to regain receptivity to courting males. Males that manipulate offspring development ensure that they father all of the offspring produced by a female in her lifetime. Therefore, although fewer offspring may be produced per clutch (Moore et al., 2001), the total reproductive output of manipulative males is greater, albeit at a cost to females.
In N. cinerea, either the male pheromone or a factor stimulated by the pheromone appears to be directly responsible for the manipulative effects on reproduction and social behaviour we observe. Previous research showed that direct applications of the male pheromone influences both male aggression and dominance (Moore et al., 1997) and sexual conflict (Moore et al., 2001). Combined with the present research linking the pheromone, offspring development, and female lifespan, this suggests a hormone-like role for the male-produced pheromone of N. cinerea. Although additional effects of male-produced seminal fluids on female reproduction and longevity, such as seen in Drosophila (Chapman et al., 1995), and mechanical damage during copulation (Crudington & Siva-Jothy, 2000; Stutt & Siva-Jothy, 2001), cannot be ruled out completely, neither are they sufficient explanation as the manipulative effects of males can occur with social exposure alone. Offspring development and female survivorship depended on exposure to, but not necessarily mating with, preferred or nonpreferred males. Although hormonal effects of pheromones are not often studied in insects outside of queen pheromones in social insects (Keller & Nonacs, 1993), pheromones affect reproductive synchrony of humans and mammals (Stern & McClintock, 1998) and perception of odours, presumably pheromones in the social environment, reduces lifespan of Caenorhabditis elegans (Apfield & Kenyon, 1999). Hormone-like effects of pheromones are therefore not unexpected, and pheromones are likely to be important mediators of sexual conflict in other species.
Mate choice by female N. cinerea therefore reflects attempts to avoid male manipulation as much as preference for specific types of males. Avoidance of the cost of male manipulation by mate choice is predicted by models of sexual conflict (Gowaty, 1996, 1997; Holland & Rice, 1998), but has never before been directly demonstrated. Exposure to male pheromones not only limits a female's opportunities for remating but also carries the further cost of decreased female survivorship. Conflicts such as these are expected to result in irresolvable evolutionary chases (Parker, 1979), or chase-away selection (Holland & Rice, 1998), when neither sex predominates. This appears to be true for N. cinerea, as male–male competition, female mate choice, or a combination can influence mating outcomes (Moore, 1988; Moore et al., 1995, 2001, 2002; Moore & Moore, 1999). Thus, neither females nor males can ‘win’ the conflict. Furthermore, there are costs and benefits associated with both mate choice and male–male competition and complex multivariate fitness relationships expected under models of sexual conflict (Parker, 1979; Eberhard, 1996; Gowaty, 1996, 1997; Holland & Rice, 1998; Gavrilets et al., 2001). Finally, these results support the hypothesized balancing role for different mechanisms of sexual selection in N. cinerea (Moore & Moore, 1999), maintaining variation among males in sexually selected characters in the population when the two mechanisms work in opposition because of sexual conflict.
The importance of this finding for these cockroaches in nature is unfortunately unknown. Little or no work has been done on cockroaches outside the laboratory (Schal et al., 1984), and N. cinerea have never been studied in their native African habitat. Thus, as is true for other more familiar model systems like Drosophila or C. elegans, our interpretations can only indicate what can happen in this species. Nevertheless, such studies provide general insights into sexual selection and sexual conflict. It is clear that sexual conflict can lead to complicated behavioural interactions, many of which will, at first glance, appear to be counter-intuitive. Inferences of sexual conflict from behaviour alone are therefore difficult. Measurements of the fitness consequences of behaviour are needed to clarify how sexual conflict may affect a species.
We thank Marie Goldrick and Melanie Gibbs for technical assistance; W. Anderson, C. Bluhm, L. Drickamer and G. Rosenqvist for discussion; and R. Preziosi and J. Sadowski for comments on the manuscript. Our cockroach research is supported by NSF grants to A.J.M and P.A.G., and NERC to P.J.M.
Received 12 September 2002;revised 13 October 2002;accepted 17 October 2002