Ageing via perception costs of reproduction magnifies sexual selection

Abstract

Understanding what factors modulate sexual selection intensity is crucial to a wide variety of evolutionary processes. Recent studies show that perception of sex pheromones can severely impact male mortality when it is not followed by mating (perception costs of reproduction). Here, we examine the idea that this may magnify sexual selection by further decreasing the fitness of males with inherently low mating success, hence increasing the opportunity for sexual selection. We use mathematical modelling to show that even modest mortality perception costs can significantly increase variability in male reproductive success under a wide range of demographic conditions. We then conduct a series of assays suggesting that, in Drosophila melanogaster, failure to reproduce early in life may, via perception costs of reproduction, significantly reduces the subsequent fitness of males (ca 25%), due mostly to increased reproductive ageing. Altogether, our results strongly suggest that perception costs of reproduction can magnify sexual selection in a biologically significant way. Finally, we estimate that around 29% of available studies quantify sexual selection based on short-term fitness estimates that may fail to capture these effects (if they were present in their subject species), and suggest addressing the existence and impact of perception costs of reproduction across taxa should thus be a priority.

1. Introduction

Sexual selection is a pervasive evolutionary force driving male and female adaptations and life histories [1,2], giving rise to a spectacularly wide array of morphologies, behaviours, and lifestyles across nature [3]. Beyond being one of evolution's main engines of biodiversity, sexual selection is also a fundamental determinant of population viability and evolvability [4–12], speciation processes [12], and the maintenance of sexual reproduction itself [13]. Consequently, a comprehensive understanding of the factors that modulate sexual selection intensity is central to evolutionary biology.

In a nutshell, sexual selection is fuelled by variance in the reproductive success of individuals belonging to the same sex and population due to pre- and post-copulatory competition over access to mating and fertilization [14,15]. The higher the realized variance in reproductive success, the stronger the action of sexual selection on associated phenotypic differences. In this context, biotic and abiotic factors affecting within-sex variance in reproductive success have the potential to modulate the intensity of sexual selection, and hence some of the key evolutionary processes mentioned above. In particular, it has been recently proposed that perception costs of reproduction can magnify sexual selection by differentially accelerating ageing (and hence differentially decreasing reproductive success) in low-performance individuals in a population [16]. A series of groundbreaking studies in the last couple of decades have shown that sensory perception of certain environmental cues can drastically modulate ageing [17,18], presumably by engaging physiological responses evolved to deal with the associated environmental challenges. One such environmental cue seems to be sexual pheromones. In Caenorhabditis elegans, detection of male secretions substantially increases ageing and mortality of hermaphrodite individuals [19]. Interestingly, these effects disappear when donor males are defective in pheromone production or hermaphrodites are deficient for pheromone processing, which strongly suggests they are mediated by sexual pheromones [19]. Furthermore, similar effects have been described for other Caenorhabditis species, suggesting underlying mechanisms are evolutionarily conserved within the genus [19]. In Drosophila melanogaster, two recent studies elegantly showed that male perception of female sexual pheromones triggers key survival and physiological costs in males, but only if they are subsequently not able to mate [20,21]. The fact that such perception costs are mostly realized when sensory perception of reproductive opportunities is uncoupled from mating itself suggests this phenomenon could potentially exacerbate sexual selection intensity [16]. In a local group (e.g. mating patch) or population where different males are competing over access to females, inherent differences in condition, quality ground-breaking or attractiveness will lead males with low-performance phenotypes (with respect to sexual selection) to have lower chances of mating than those with high-performance phenotypes. This, in turn, means that low-performance males will tend to suffer higher perception costs of reproduction than their rivals (because they will not mate or will do so less frequently), which would increase the variance in reproductive success among males in a positive feedback loop that would tend to increase the opportunity for sexual selection, and hence the upper limit for the action of sexual selection. Both in Drosophila and in Caenorhabditis, perception costs are realized from early on and seem substantial (e.g. approx. 20% decrease in average lifespan [19,20]) and the fact that very similar perception costs of reproduction have been described in two phylogenetically distant taxa suggests conserved underlying mechanisms [18,19,21]. Hence, ageing via perception costs of reproduction could indeed be an important yet previously ignored magnifier of sexual selection intensity [16].

Here, we test the idea above formally (by mathematical modelling) and empirically, using D. melanogaster as a model organism. First, and based on previously described mortality costs [20], we developed a mathematical model to explore the extent to which perception costs of reproduction can modulate the intensity of sexual selection across biologically relevant life history (i.e. number of reproductive events, delay between events, fecundity, etc.) and demographic (i.e. rate of population growth) conditions. Second, we set up a series of behavioural assays where we simulated male reproductive failure early in life in males that were either exposed or not to perception costs of reproduction. We then examined the consequences in terms of survival and lifetime reproductive success of these males throughout the rest of their lifetime, to test the idea that perception costs of reproduction can increase the opportunity for selection.

2. Methods

(a). Mathematical model

To formally examine the theoretical idea that perception costs of reproduction can magnify sexual selection, we developed a mathematical model based on the life history of D. melanogaster (electronic supplementary material, table S1), which we analysed using computer simulations. We developed a deterministic model with a number of assumptions to keep it as simple as possible, but able to capture the actuarial effects of perception costs on fitness.

(i). Fitness definition and demographic context

Following Tatar & Promislow [22], we define male fitness (w) as a function of age-dependent survival and fecundity schedules

2.1

where r is the intrinsic growth rate of the population, l_i(x) and m_i(x) are the proportion of surviving males and the male fecundity at age x, respectively, and i denotes the type of male. For simplicity, we consider two types of males: high-performance males (i = HPM) and low-performance males (i = LPM), where the relative frequency of HPM is f_H. Both r and f_H are assumed to be constant during the relevant time window to estimate w. Therefore, r and f_H provide the demographic context for fitness assessment.

(ii). Male interaction and reproduction

Male types differ only in relation to the probability of succeeding versus failing when they compete against another male for a female. For simplicity, we assume all interactions take the form of two males randomly encountering and competing to mate with a female, where only one achieves mating. The model does not make any assumptions as to whether the outcome of an interaction (i.e. the different probabilities of success for HPM and LPM) is due to inter- or intra-sexual processes, or a combination of both. When either two HPMs or two LPMs interact, the probability of success of each male is always set at 0.5.

When a HPM and a LPM interact, the probability of success of the HPM and the LPM are, respectively, (1 + h)/2 and (1 − h)/2. Thus, h (ranging from 0 to 1) is the interactive dominance of a HPM over a LPM. If h = 0, no dominance exists, so that HPMs and LPMs have the same probability of success. If h = 1, the HPM is always the winner when competing with a LPM.

After a male achieves his age at maturity (a₀), interactions for reproduction occur regularly (time elapsed, Δa), until a maximum number of interactions (p). p is a maximum that will not be realized by all males because a fraction of the males will die before p interactions occur. If successful in an interaction, a male gets m offspring regardless of his age; i.e. we assume no reproductive senescence in males. If unsuccessful, a male obtains no offspring, and a linearly accumulated survival cost (see below).

(iii). Survival

We used a modified version of the Gompertz survival model, extensively used to describe survival (e.g. [23]), in order to compute l_i(x). Mortality rate (μ(x), also called hazard rate) in the Gompertz model is defined as

where α is the baseline mortality (a scaling factor; mortality at x = 0) and β is the so-called ageing Gompertz parameter. The modification consists in assuming that the ‘scaling factor’ becomes a linearly increasing function of the number of failed reproductive interactions (i.e. perception costs of reproduction) accumulated by a male at his age x, f(x). The result is

2.2

where the constant s ≥ 0. We hereafter refer to s (the increase in mortality per failure in an interaction) as ‘sensitivity to perception costs'. We modelled perception costs of reproduction to weight on Gompertz's α, instead of on β, based on a previous analysis of Harvanek et al.'s [20] survival data, where we found that perception costs mainly increased α (see electronic supplementary material, figures S1 and S2 for details). Interestingly, note that in equation (2.2), if s > 0, mortality increases with age even if β = 0. Thus, due to an extrinsically driven process—interactions for reproduction—equation (2.2) adds an ageing-like acceleration component to survival. As the age-dependent distribution of the number of failures differs between HPM and LPM, equation (2.2) allows us to compute l_i(x) in (2.1) after integrating over the intervals split by the interactions for reproduction.

(iv). Model computation

Computations were performed using Wolfram Mathematica 10 (Wolfram Research, Inc., Mathematica, v.10.4, Champaign, IL, 2016), and the values in electronic supplementary material, table S1. The effect of model parameters on fitness was explored varying one or two parameters at a time (see electronic supplementary material, table S1 for reference values and range of values explored).

Using equation (2.1), the relative fitness of LPM was computed as w_LPM/w_HPM. This relative fitness was computed for s ≥ 0 and for s = 0, which correspond respectively to the assumption of possible perception costs (PC), and with the assumption of no perception costs (no PC), where the decrease due to PC in the relative fitness of LPM is:

(b). Experiment: the effect of perception costs of reproduction on fitness in Drosophila melanogaster

(i). Fly stocks and general husbandry

Laboratory-adapted D. melanogaster flies from a wild-type stock (wt) maintained outbred since 1970 were used as focal and donor flies in this experiment [24]. Flies were maintained at 25°C in 12 L : 12 D conditions, with overlapping generations. Flies for experiments were reared using a standard larval density method [25], by placing approximately 200 eggs on 50 ml of food in 250 ml bottles. Virgins were collected on ice anaesthesia within 6–7 h of eclosion.

(ii). Experimental design

We proceeded to set up an experimental assay to test the theoretical idea above that perception costs will significantly decrease the fitness of LPM, and hence tend to increase the overall variability in male fitness. We simulated two sets of LPM (in our model) by ageing males for three weeks without access to matings, simulating repeated failures in reproduction, while in the presence (perception costs) or absence (no perception costs) of female pheromones.

We then transferred these males into competition vials (i.e. with two rival males and a female) to study their fitness for the rest of their lifetime. We employed laboratory-adapted wild-type (wt) Dahomey stock flies as focal and donor flies, while male rivals and females were homozygous for a recessive eye marker (sparkling poliert), which allowed us to directly estimate relative reproductive success across time.

All flies were raised following the same procedures. Yeasted grape juice agar plates were introduced into stock cultures to allow for oviposition, and then collected eggs were placed in bottles with standard food, at a controlled density [25] of ca 200 eggs per bottle containing ca 75 ml of food. Emerging virgin flies were sexed and isolated 2–3 days prior to experiments. Isolated focal male flies of 48–72 h of age were individually allocated to either of two treatments: a) males exposed to female pheromones (‘PC’; N_PC = 80) or b) males not exposed to female pheromones (‘NPC’; N_NPC = 80). Vials in the first treatment were connected to another vial containing three 48 to 72 h-old donor females (electronic supplementary material, figure S3a). Importantly, these two vials were divided by a mesh partition, so that focal males could perceive female odours, but not directly interact with them. Vials in the second treatment were kept in the exact same conditions as in the previous treatment, but without any donor females in adjacent vials (electronic supplementary material, figure S3b). Flies were kept under these conditions for three weeks, during which time they were transferred to fresh vials twice a week using gentle CO₂ anaesthesia (i.e. a short, approx. 2 s puff of CO₂ from which flies recovered within the minute). After three weeks, donor females were discarded and focal males from both treatments were placed in competition vials containing a triplet of 48 to 72 h-old sparkling poliert (spa) virgin flies consisting of 2 rival males and 1 female (electronic supplementary material, figure S3, step 2). Focal males were left to compete for the rest of their lifetime, during which time we checked vials daily for focal male deaths (i.e. to score lifespan). Additionally, we transferred flies to fresh vials twice a week using gentle CO₂ anaesthesia (see above) and incubated ‘old’ vials containing female eggs at 25°C for 16 days, after which we froze them at −21°C and eventually counted each vial (i.e. no. of wt/spa offspring) to estimate focal male fitness. During interactions, we replaced the triplet of spa flies (i.e. female and rival males) with young 2–4d-old virgin flies every 10–12 days (at the same time for all treatments). We kept a stock of spa flies during each round of spa collection to replace dead spa rivals/females, which only occurred for 11 females and two rival males during the whole experiment. Initial sample sizes were n = 80 per treatment, but we discarded four individuals from fitness and survival analyses because they escaped during manipulation in the first round, and four more from fitness analyses and right-censored for survival analyses because they escaped during manipulation in subsequent rounds.

(iii). Statistics

Briefly, perception costs effects on survival, lifetime reproductive success, and rate-sensitive fitness estimates (W_ind) were estimated fitting linear models with treatment as the only fixed factor. We also ran a Cox proportional hazards survival model to test for differences in mortality risk across treatments. Reproductive senescence was tested by fitting a binomial generalized linear mixed models (GLMM) (i.e. number of wt offspring versus number of spa offspring) with treatment, time (week 1, 2, or 3), and the interaction between time and treatment as fixed factors and individual ID as a random factor. Significance of terms was explored by comparing models with and without the term of interest [26]. W_ind data was standardized (z-scores, using global mean and s.d. across treatments [27]) prior to analysis, and α-winsorized at 0.05 due to the presence of a few extreme outliers (but note that results on raw data are qualitatively identical to those reported). Data and models were checked to meet necessary assumptions. We performed all of our statistical analyses in R [28].

(c). Systematic bibliographic search

In order to explore the degree to which existing studies that have measured the intensity of sexual selection are adequate to detect the effects on fitness of perception costs of reproduction, we performed a systematic search of the literature. We used the following search string [(‘sexual selection’ AND ‘selecti* intensity’) OR (‘sexual selection’ AND ‘selecti* strength’)] in Scopus® (56 hits), PUBMED (535 hits), and Web of Science™ (105 hits). All three searches were conducted on the 6 February 2018. The combined dataset of 696 studies was purged for duplicates (79 erased) and imported into Rayyan [29]. In Rayyan, we deleted an additional 23 duplicates and then scanned the titles and abstracts of these papers to select 140 papers that explicitly stated estimating selection. We then carefully read this subset of papers and selected a final database of papers that reported one (and in a few cases more) studies with unambiguous estimates of sexual selection intensity based on direct or indirect measures of reproductive and/or mating success (i.e. a total of 100 papers including 105 studies overall). Finally, we classified these papers into one of the following three categories, according to whether fitness estimates (and hence sexual selection estimates) would have captured the effects of perception costs of reproduction, were they to exist in the species studied (see electronic supplementary material, table S2 for details of each study):

• (A) Papers that would fully capture PC. Here, we included all studies that followed a cohort of individuals over most or all of their lifespan (e.g. lifetime reproductive success estimates) and/or studies that followed a natural subset of individuals of mixed age (i.e. cross-sectional study) over most or all of their lifespan. Hence, all studies included in this category provided long-term (relative to the species lifespan) longitudinal fitness estimates.

• (B) Papers that would partly capture PC. Here, we included studies that either followed a cohort across a significant part of their reproductive lifespan (longitudinal studies), or more commonly studies calculating short-term fitness in a natural subset of a population including individuals of different ages. Such studies would tend to capture most, but probably not all perception costs effects on fitness; for example due to changes in the age-class structure throughout the reproductive season.

• (C) Papers that would not capture PC. Here, we included studies that provide short-term term fitness estimates (with respect to the species lifespan) of a cohort and/or a group of young individuals of approximately the same age.

3. Results

(a). Mathematical model

The results of our model show that even modest increases in actuarial senescence via perception costs of reproduction, in line with those described in D. melanogaster and C. elegans [19–21], will act to increase the fitness differences between males (i.e. opportunity for selection) as soon as high-performance individuals have a small reproductive advantage over low-performance individuals (figure 1). Furthermore, we show that these effects are consistent irrespective of initial frequencies of HPM versus LPM and across a wide range of population growth rates and life-history traits (see electronic supplementary material figures S4–S5 and table S1). As expected, we found that perception costs effects on fitness become more important as reproductive parameters shift towards more iteroparous schedules (e.g. increased number of reproductive episodes, and interval among them; electronic supplementary material, figure S5). This makes sense given that iteroparity increases the scope for ageing effects, and provides a general rule-of-thumb to identify species where perception costs may be expected to exacerbate sexual selection. It is important to note that in this model we restricted potential perception costs to survival effects described so far in C. elegans and D. melanogaster [19,20]. Thus, given that previous studies have not addressed potential fitness effects due to reproductive senescence, our mathematical analysis must be taken as a conservative estimation of the potential role of perception costs of reproduction as a magnifier of sexual selection.

Figure 1.

Model results. Relative change in the fitness of low-performance males (LPM) with respect to high-performance males (HPM; note this is a measure of the increase in the opportunity for selection) in relation to h (the dominance of HPM over LPM), r (the daily growth rate of the population within which HPM and LPM compete), and s (the sensitivity to perception costs in terms of increased mortality; i.e. the degree to which mortality increases with each successive failure to reproduce).

(i). Empirical study in Drosophila melanogaster

Consistent with Harvanek et al. [20], we found that focal males exposed to perception costs during their early life showed a significant decrease in lifespan (F_1,154 = 11.42, p < 0.001, estimate = −4.0 ± 1.2; figure 2a). Survival analysis by means of a Cox proportional hazards model evidenced significant effects of perception costs in focal male mortality risk (d.f. = 1, z = 3.636, p < 0.001). Importantly, we also found that perception costs accelerated reproductive ageing, whereby offspring production declined significantly faster for males exposed to perception costs (i.e. significant treatment:time interaction in a GLMM; Chi = 109.97, d.f. = 1, p < 0.001; figure 2b). Altogether, survival and reproductive ageing differences translated into significant differences in lifetime reproductive success (LRS; F_1,158 = 6.22, p = 0.01, estimate = −3.6 ± 1.5; figure 2c). In turn, LRS differences were reflected in significant variation in standardized rate-sensitive fitness indexes (w_ind [27]; F_1,158 = 6.995, p = 0.009, estimate = −0.360 ± 0.136), and these effects were qualitatively invariant across different population growth rates (figure 2d; electronic supplementary material, Results). Finally, fitness estimations of overall population effects (w_pop) showed that the impact of perception costs of reproduction on fitness (Cr, see Edward et al. [27]) was mainly due to its effects on reproductive ageing, and not on actuarial ageing (figure 2e).

Figure 2.

Effects of perception costs of reproduction on different fitness components/measures in D. melanogaster: (a) survival and average lifespan, (b) reproductive senescence (i.e. decrease in offspring production with age), (c) average reproductive success (i.e. average proportion of offspring sired in competition vials), and (d) standardized male fitness differences over a range of population growth rates (r). In addition, (e) shows the relative population fitness costs (Cr; Edward et al. [27]) that result from perception costs due to actuarial senescence (long-dashed green line), reproductive senescence (short-dashed magenta line), or both (solid black line). The top solid red line marks the absence of population costs (i.e. Cr = 1).

4. Discussion

Here, we provide formal theoretical grounding to the idea that accelerated ageing via perception costs of reproduction can accentuate sexual selection, and empirical results suggesting this can actually happen in D. melanogaster.

First, our mathematical models show that even modest mortality perception costs can significantly increase the differences in reproductive success under a wide range of demographic conditions, via actuarial senescence, and that these effects become more important as reproductive parameters are shifted towards iteroparity. Even though this model is conservative, because it only encompasses actuarial costs, it clearly shows that perception costs of reproduction in line with those described previously in D. melanogaster and C. elegans [19–21] will act to increase the fitness differences between males (i.e. opportunity for selection) as soon as high-performance individuals have a small reproductive advantage over low-performance individuals (figure 1). Interestingly, these effects were robust irrespective of initial frequencies of HPM versus LPM, population growth rates, and across life-history traits (figure 1; electronic supplementary material, figures S4, S5 and table S1). As expected, we found that perception costs effects on fitness are expected to be more important in iteroparous species, which makes sense given that iteroparity increases the scope for ageing effects.

Second, we found that a failure to reproduce early in life drastically decreased subsequent survival and lifetime reproductive success of males when we allowed for perception costs of reproduction. Our results hence offer empirical support to the idea that restricted access to reproduction early in life significantly decreases subsequent male fitness via perception costs of reproduction in D. melanogaster. Although we didn't measure opportunity for selection directly in this study, the above will in practice always tend to increase the opportunity for selection whenever male performance (i.e. quality/condition) correlates with mating success. Interestingly, we found that perception costs effects on fitness were mainly mediated by reproductive ageing, much more so than by actuarial ageing. Relative population fitness costs considering survival effects only were estimated to be around 5%, whereas reproductive senescence costs alone were estimated to be around 23% (figure 2e). Overall, total fitness population costs considering both survival and reproductive senescence costs were estimated to be very high (ca 25%; figure 2e). The latter results are particularly interesting because studies describing perception costs have so far focused on its effects on survival [20,21], and have not measured its actual consequences on reproductive senescence or fitness at large.

Furthermore, in this study, we measured male fitness in competition against other males, which means we estimated ageing effects on the whole suite of intra- and inter-sexual processes involved at both the pre- and post-copulatory level. In nature, these effects would tend to penalize males that are relatively unsuccessful in their struggle to reproduce early in life, hence potentially broadening the inherent fitness gap between LPM versus HPM in a group/population, and with it the potential intensity of sexual selection processes [16]. We found this effect to be important (approx. 25% decrease in relative fitness) and relatively invariant for a wide range of population growth rates, suggesting biologically relevant effects irrespective of population dynamics.

As a cautionary note, our empirical study was meant as a proof-of-concept and, as such, we suggest its findings should be expanded by future studies in three directions. First, in our empirical study we did not manipulate male condition, and hence differential ageing (and ensuing fitness costs) between treatments is exclusively due to perception costs arising from a failure to reproduce (i.e. performance, in our model). This has the advantage of eliminating any confounding effects of male condition and focusing on the effects of mating history, which is what we intended, but we suggest future studies may wish to complement our findings by manipulating both mating history and male condition. Second, for logistics reasons we were only able to implement a single mating history treatment (i.e. three weeks of reproductive failure), while expanding this range (e.g. 3d, 7d etc.) will be crucial to ascertain how likely is it that perception costs effects are biologically relevant. Finally, we did not test here for fitness differences in males with early access to reproduction (i.e. males in a simulated high-performance treatment) when exposed or not to perception costs. Original studies in D. melanogaster did not find any survival perception costs in males that had access to matings [20,21]. However, because these studies didn't focus on reproductive fitness, we acknowledge that there could, in principle, be effects unaccounted for in this study. If anything, though, we would expect any differential effects here to reinforce our findings given that, as already mentioned, previous studies have failed to find survival costs in HPM. It is also very likely that in nature high condition males are, in addition to being less likely to fail to reproduce early in life (i.e. high performance, in our model), frequently more resilient to ageing due to perception costs. Similarly, an obvious yet extremely interesting arising question has to do with the adaptive significance that perception costs of reproduction may have for males. Presumably, perception costs of reproduction may be the result of behaviourally plastic changes that ‘prepare’ males for competition over reproduction at the pre- and/or post-copulatory level, and that are therefore adaptive on average because they generally increase a male's chances to be successful in its struggle to reproduce. Under this light, male physiological responses to female pheromones may be conceived as having antagonistic pleiotropic effects in that they may increase a male's chances to reproduce in the short term, but decrease its chances to reproduce in the long term if it fails to do so in the short term. In any case, perception of female pheromones is hypothesized to always increase the competitive ability of HPM (who mate), and hence their reproductive fitness. All of the above considerations hence predict a further increase in the fitness gap between high and low condition males, accordingly further increasing the opportunity for selection. Future studies in the directions proposed above will be crucial to further our insight about the general importance of this phenomenon, and in particular about its true biological relevance in D. melanogaster.

The finding that perception costs of reproduction can magnify sexual selection may also have bearing for life-history theory concerning survival-reproduction trade-offs. This is mainly because perception costs of reproduction will tend to favour early reproduction disproportionally (i.e. given population growth rates), due to its cascading effects over subsequent lifetime fitness. It also suggests that, in mating systems when perception costs are high and paid from very early on (e.g. after just two or three failures to mate), random variation in early-life reproductive success may be more important than expected given its multiplicative effects over subsequent fitness. We hence suggest future theoretical and empirical work on this area may offer interesting insight into our understanding of survival-reproduction trade-offs. Finally, we performed a systematic search and thorough review of papers measuring sexual selection (i.e. in the form of univariate or multivariate selection gradients, selection differentials, the opportunity for sexual selection, and/or Bateman gradients), and categorized them into studies that, if perception costs of reproduction were to exist in the target species, would: a) fully capture perception costs effects on fitness, b) partly capture perception costs effects on fitness, or c) fail to capture perception costs effects on fitness. Our aim was to estimate the degree to which available studies would capture perception costs effects were they to exist in their subject species, simply because failing to do so could yield estimates that underestimate the true strength of sexual selection. Out of 105 studies, we found that 29% were designed in a way that would completely fail to capture perception costs of reproduction, and may hence offer biased sexual selection estimates (figure 3) if perception costs of reproduction exist in their subject species.

Figure 3.

Breakdown of studies explicitly measuring sexual selection estimates reviewed in our review of the literature. We classified studies according to whether they capture (i.e. fully, partly, or no) the effects that perception costs of reproduction would have on fitness (and hence selection) estimates, were they to exist in each of the species studied. Studies were also classified according to whether they were short term (ST) or long term (LT), and to whether they were conducted on a cohort/s of individuals versus a natural cross-section of a population.

Ageing effects can make short-term measures of the opportunity for selection biased whenever the relationship between age and fitness changes across individuals according to performance/condition. For instance, in the context of age-dependent female choice, when high-quality males tend to survive more or less than low-quality males [30]. In these cases, the relative reproductive success of high- versus low-quality males will change through time, and hence short-term estimates of the opportunity for selection may be somewhat biased, under- or over-estimating the net fitness differences across males in a group. However, such effects would act either to increase or decrease the overall variability in lifetime reproductive success with respect to short-term estimates, depending on each species' ecology (i.e. the direction in which fitness differences change through time). Intra-specifically, such effects may even change qualitatively across populations and/or local groups depending on local conditions (e.g. predation pressure), resulting in soft selection. By contrast, perception costs of reproduction are, as far as we currently understand them, always expected to increase the variability in lifetime reproductive success across males, and lead to a monotonic increase in ‘hard’ selection [31].

Needless to say, there is currently no evidence that perception costs of reproduction do exist in species other than C. elegans and D. melanogaster, the only two species where it has been investigated to date. The fact that perception costs are present in these two phylogenetically distant species does suggest underlying mechanisms might be conserved, and thus present in other taxa. Of course, even if perception costs of reproduction were to exist in many species, the degree to which they may be biologically important would likely vary substantially from species to species, and they may well be absent or weak in many cases. Thus, our aim is not to imply that a portion of available studies are in any way deficient per se. Rather, we simply point out that given standard protocols will not necessarily always capture perception costs effects (e.g. short-term studies on cohorts; figure 3), this phenomenon should be studied in other systems and taken into account in sexual selection studies, where necessary. At the very least, studies aiming to measure sexual selection in Caenorhabditis and Drosophila, two model systems in the study of sexual selection and ageing. For example, most of the sexual selection studies we reviewed for Drosophila species (i.e. 85%, see electronic supplementary material, table S2 for details) focus on cohort/s of individuals over a relatively short window of their reproductive lifespan, and thus would fail to capture perception costs of reproduction. We thus contend exploring the degree to which perception costs of reproduction exist in these and other taxa, and if so their effects on both actuarial and reproductive senescence, should be a priority for the future.

To conclude, in this paper we identify a potentially important new modulator of sexual selection that could in principle operate across a wide range of taxa. Correctly parameterizing sexual selection is critical to properly gauge its role as a driver of male and female adaptations and life histories, in speciation, or in the maintenance of sexual reproduction itself [12–14]. Because perception costs of reproduction increasingly impact fitness through time (i.e. via reproductive and actuarial ageing), sexual selection studies relying on short-term estimations of fitness may not capture most of these effects (particularly given that positive feedbacks usually result in nonlinear responses), potentially underestimating the true strength of sexual selection.

Data accessibility

Data available from the Dryad Digital Repository at: http://dx.doi.org/10.5061/dryad.mh4240m [32].

Authors' contributions

P.C. conceived this study; M.S. developed the mathematical models and ran simulations; P.C., M.S., and R.G.R. designed the empirical study in D. melanogaster; R.G.R. conducted the experiments with help from P.C.; P.C. and R.G.R. analysed the results; P.C. wrote the manuscript with considerable feedback from R.G.R. and M.S.

Competing interests

We declare we have no competing interests.

Funding

P.C. was supported by a RyC Research Fellowship from the Spanish Government (RYC-2013-12998) and by a Spanish Plan Nacional de Investigación Científica, Desarrollo e Innovación Tecnológica (I + D + I) from the Spanish Ministry of Economy and Competitiveness grant (no. CGL2017-89052-P) to PC (co-financed by FEDER funds, European Union), which also supported R.G.R. M.S. was supported by a Spanish Plan Nacional de Investigación Científica, Desarrollo e Innovación Tecnológica (I + D + I) from the Spanish Ministry of Economy and Competitiveness grant no. CGL2015-65422-P (co-financed by FEDER funds, European Union).