Female resource limitation does not make the opportunity for selection more female biased

Environmental and physiological conditions affect how individual variation is expressed and translated into variance in fitness, the opportunity for natural selection. Competition for limiting resources can magnify variance in fitness and therefore selection, while abundance of resources should reduce it. But even in a common environment the strength of selection can be expected to differ across the sexes, as their fitness is often limited by different resources. Indeed most taxa show a greater opportunity for selection in males than in females, a bias often ascribed to intense competition among males for access to mating partners. This sex-bias could reverberate on many aspects of evolution, from speed of adaptation to genome evolution. It is unclear however, whether the sex-bias in opportunity for selection is robust to variations in environment or physiological condition that limit sex-specific resources. Here we test this in the model species C. maculatus by comparing female and male variance in relative fitness (opportunity for selection) under mate competition (i) with and without limitation of quality oviposition sites, and (ii) under delayed age at oviposition. Decreasing the abundance of the resource key to females or increasing their reproductive age was indeed challenging as shown by a reduction in mean fitness, however variance in fitness remained male-biased across the three treatments, with even an increased male-bias when females were limited by oviposition sites. This suggests that males remain the more variable sex independent of context, and that the opportunity for selection through males is indirectly affected by female-specific resource limitation.


Introduction
Variation in fitness among individuals is what natural selection acts on. It can be partitioned into 39 variation among individuals in their genetic makeup (breeding value), in their phenotypic condition 40 subjected to environmental variation, and to the interaction between the two [1]. Therefore, the 41 extent to which individual variation will be translated into variation in fitness visible to natural 42 selection depends on context, through the availability of key developmental resources and the 43 intensity of competition among individuals [2,3]. For example, under abundant resources 44 individual variation in resource acquisition should matter little to fitness, but if resources are scarce 45 even slight differences in the acquisition traits may translate into large differences in fitness. 46 To see how variation in fitness translates into opportunity for selection, it is useful to think about 47 to hatch, while there are no detectable effects of paternal age on offspring phenotype [26]. Fertility larval food resource, without which the females do not even lay eggs. Females have evolved a great 110 capacity to detect a high quality bean resource as their oviposition site [29,30]. The HT treatment 111 thus provides a challenge that can reveal variation in this crucial ability for female fitness, while 112 the AT treatment represents a situation faced by individuals required to postpone reproduction in 113 the absence of available bean resources. We estimated the strength of selection as the opportunity 114 for selection : I = σw 2 / ѿ 2 , where σw is the standard deviation in fitness and ѿ the mean fitness [5]. 115 We find that mean fitness, measured as the number of adult offspring recruited to the next 116 generation, was lower in both HT and AT treatments compared to the control, indicating they were 117 generally challenging conditions. Interestingly, individual offspring produced by older parents (i.e. 118 AT treatment) were heavier than ones from either of the other treatments, suggesting that this 119 particular stressor induced a change in the offspring resource allocation strategy. Finally, the 120 opportunity for selection was consistently higher in males than in females, and the male-bias was 121 even stronger under oviposition site limitation (i.e. HT), suggesting that this sex-specific trend is 122 not only robust to the context but that male variation can be indirectly affected through interaction 123 with females. 124 125 bean Vigna ungulata used in the present study), after which larvae burrow and develop inside the 131 beans until they emerge as reproductively mature adults. 132 The study population originates from a natural population sampled in Lome, Togo (06°10#N 133 01°13#E) in 2010. It has been kept under laboratory conditions since then (29°C, 12:12 light cycle, 134 50% humidity) with a constant population size of approximately 400-500 individuals. Fitness 135 assays were also performed under laboratory conditions (29°C, 12:12 light cycle, 50% humidity). 136

Fitness assays 138
Fitness was measured in lifetime competitive assays were one focal individual was placed together 139 with a competitor of the same sex and two mating partners of the opposite sex inside a 9cm petri 140 dish. The environment inside the dish varied according to the treatment (see experimental 141 treatments below). At the start of the experiment all individuals were adult virgins collected less 142 than 24 hours after emergence from the beans. The competitor individual was sterilized by gamma 143 radiation (100Gy), a commonly used method in the seed beetles that allows the competitor 144 individual to compete for matings and achieve fertilizations, but insures that zygotes fertilized by 145 the competitor will not develop due to the high number of double-stranded breaks in the embryo lifetime and offspring were counted as emerged adults of the next generation. A female fitness assay included one focal female, one sterilized female, and two male partners. The same design 149 was used for the male fitness assays, which included one focal male, one sterilized male and two 150 female partners. 151

Experimental treatments 152
Our study included three treatments, aimed to create different reproductive challenges for the sexes. 153 The control treatment (CT) represents the laboratory setting classically used in C. maculatus 154 studies: a 9cm petri dish with ad libitum black-eyed bean (27g, approximately 130 beans). While 155 male fitness variation can be manifested through pre-and post-mating sexual competition, for 156 females this environment likely represents less challenges. Their oviposition substratum, the bean, 157 is directly available, in a high and consistent quality, and in non-limiting quantity. 158 The heterogeneous environment treatment (HT) was designed to directly challenge females in their 159 ability to discriminate quality oviposition sites. Each petri dish was filled with beans of variable 160 quality: 15 high quality beans (3-4 grams) and the remainder of poor quality for a total of 27g as 161 well. The low-quality beans were produced by letting a stock population of C. maculatus use the 162 beans for larval development, resulting in bored beans that provide less resources for offspring to 163 develop on. 164 The ageing treatment (AT) was designed to challenge females in their ability to withhold their 165 reproduction until a suitable oviposition site is available. This treatment bears ecological relevance 166 to a scenario where high-quality oviposition sites are exhausted upon female hatching, requiring 167 prolonged periods of searching for suitable sites. In this treatment, the four individuals were first were added. 170 individuals: the focal individual (female for a female assay and a male for a male assay), a sterile 174 competitor of the same sex, and two potential mating partners of the opposite sex. Since all 175 individuals originate from the same population, all the non-sterile individuals (one focal individual 176 and two mating partners) will contribute to the final estimate of variance in fitness. For example, 177 variance measured from female assays will be composed of a component due to the focal female 178 present in each assay, but also of a component due to the two males present as potential mating 179 partners. We considered the contribution of mating partners for estimating the sex-specific variance 180 in fitness under the following premise. As the contribution of the mating partners is shared between 181 two individuals, but the contribution of the focal individual relies solely on one in each assay, the 182 focal sex contributes fully to the variance in fitness while the mating partners' contribution is 183 halved, so that: 184 And, 186 Where ( ) and � � are the variances estimated from female and male fitness assays 188 respectively, and and are the female and male components of these variances. This premise stems from the assumption that the contributions of both parents to fitness are additive, and that 190 breeding values of males and females are normally distributed. 191 If we call F the female breeding value and suppose that it follows a normal distribution with mean 192 1 and standard deviation ( → (1, 2 )), and call M the male breeding value and suppose → 193 (1, 2 ), we can describe the fitness of a female assay as : 194 And of a male assay as: 196 This is because the focal sex contributes fully ( The effect of experimental treatments on individual offspring weight was analyzed using a linear 218 mixed-model, as implemented in the lme4 package for R, taking into account normal distribution 219 of the data. Experimental treatment, sex of the focal individual and their interaction were specified 220 as fixed effect and the date of the fitness assay as a random effect. 221

Sex-specific variance in fitness 222
A Bayesian model, as implemented in the MCMCglmm package (version 2.26, [35]) for R, was 223 used to estimate components of variance in fitness attributed to each sex by experimental treatment 224 combination. Because opportunity for selection is the variance in relative fitness, fitness data was 225 mean standardized so that each sex by treatment subset had a mean of one prior to this analysis. 226 The model was then specified with assay date as a random effect and the total phenotypic variance estimated for each sex by experimental treatment combination (idh structure not allowing for covariances to be estimated). For each experimental treatment, the log ratio of the posterior 229 distributions for male and female variances were then computed, giving a mean log ratio and 95% 230 confidence intervals. 231

Mean fitness 233
Mean fitness (offspring number) differed among all experimental treatments (Table 1) Bonferroni method). A weak main effect of sex was also detected (Table1, Figure1) with males 238 having slightly overall higher mean offspring number but there was no sex by treatment interaction. 239 These result indicate that the HT and AT treatments were indeed challenging, with respectively 240 14% and 36% reduction in mean fitness compared to the control, and that the AT was more stressful 241 than the HT.

Sex-specific variance in fitness 254
Variance was calculated from mean standardized fitness. It is therefore the variance in relative 255 fitness, which represents the opportunity for selection. Variance in relative fitness was larger in 256 males than in females in all three treatments ( Figure 3 a and b). The male-bias was largest in the 257 HT, while the CT and AT did not differ from each other (HT-CT:p=0.039, HT-AT: p=0.039, AT-258 CT=0.45, p-values were obtained from Bayesian posterior distributions, correction for multiple 259 testing was done using the Bonferroni method). 260

261
In sexually reproducing species, selection is often measured to be stronger on males that on 262 females, and this sex-bias has often been ascribed to sexual selection acting more on males [21, 263 22]. This general sex-bias can play an important role in evolution by shaping sexually reproducing 264 populations in many ways, from genetic architecture to mutation load and speed of adaptation. Yet, 265 it is not clear how robust this pattern is to variation in ecological conditions; because the sexes are 266 limited by different resources, variation in sex-specific limiting resources should alter the sex-bias 267 in selection. Here, we used the model species C. maculatus to test the hypothesis that limiting 268 female-specific resources should cause a shift towards more female-biased opportunity for 269 selection. However, after challenging females by limiting high-quality oviposition sites (HT) or by 270 delaying age at oviposition (AT), we found that selection remained male-biased and in one case trend of male-biased opportunity for selection is robust to variation at least regarding 273 environmental variables studied here. One possible explanation that we discuss below is that 274 selection on males is partly mediated by female choice and therefore reflects selection acting on 275 females as well. Additionally, variance in fitness may not consistently increase in response to 276 stress, which further complicates predictions of how sex-specific selection should behave under 277 stress. 278 The two experimental treatments, HT and AT, were designed to be challenging and this is 279 confirmed by our results that show how these stressors decrease the mean fitness (adult offspring 280 count) compared to the CT. A general expectation is that variance in fitness should increase under 281 such stressful conditions, as the population is pushed away from its fitness peak [36] and 282 differences among individuals are revealed or magnified [37]. However, as outlined by Hoffmann 283 and Merilä [38], there are scenarios such as severe resource limitation that prevents individuals 284 from expressing their full potential, which allows for a reduction instead of an increase in the 285 opportunity for selection under stress, a prediction that has found some empirical support (reviewed 286 in [37]). This is what we also find here: both male and female opportunity for selection decreased 287 under the HT compared to CT, and female variance decreased proportionally more than male 288 variance resulting in a more male-biased opportunity for selection in that treatment. It is possible 289 that limiting good-quality larval environment in the HT prevented individuals from achieving their 290 full reproductive potential, thereby decreasing variance in relative fitness at the population level, 291 as predicted by Hoffman and Merilä [38]. However, if environmental conditions had imposed a 292 ceiling on reproductive performance, we would have expected to see this reflected in the fitness 293 distributions that should have been more negatively skewed in the HT treatment. We did not observe this (skewness score: CT= -0.38, HT= -0.09, AT= 0,17). In fact, the HT treatment of female oviposition strategy and ultimately lowered mean fitness, it may also have removed some 299 of the constraints presented to females in the CT. C. maculatus is known for pervasive interlocus 300 sexual conflict, where male mating behavior can substantially lower female lifespan and 301 reproductive success [39]; it is possible that the beans filled with cavities (constituting the majority 302 of the substrate in the HT) offered more hiding opportunities for females to avoid male mating 303 attempts, than fresh beans, as adults easily fit in the bean holes made by previous generations 304 (personal observation). There is previous evidence suggesting that more complex laboratory 305 environments could reduce the impact of sexual conflict in Drosophila melanogaster [40]. If that 306 is the case here, the HT may have presented females with oviposition challenges but removed or 307 alleviated selection pressure from interlocus sexual conflict. In turn, if the HT made it more difficult 308 for males to find mating partners, this could also explain the stronger male-bias in opportunity for 309 selection in that treatment. 310 In the AT, the opportunity for selection on females increased, as we expected when imposing a 311 challenge on female oviposition strategy (here, age-at-reproduction). However it also increased 312 proportionally in males, which resulted in a sex-bias similar to the one measured in the CT. We 313 consider several alternative explanations for this result. 314 Males and females were interacting throughout their lifetime in all of the three treatments, however 315 in the AT, the oviposition was only possible after 48h imposing a constraint particularly to the 316 female reproduction. In a related seed beetle species (Acanthoscelides obtectus), experimental work has shown how a selection for a delayed oviposition has resulted in sex-specific evolution of However, even in that case the different reproductive functions are under selection in the sexes, 323 and the effects of ageing are still expected to be sex-specific with females being more sensitive 324 than males [27,28]. 325 Another possibility is that the challenge imposed on females by the AT was reverberated onto 326 males through mate choice mechanisms if females confronted to a stressful environment became 327 choosier. The impact of female condition on mate choice has been studied in many systems, 328 however the observations mainly support a weaker mate choice for females in poor conditions 329 (reviewed in [42], and supported by more recent empirical studies [43,44]). Similarly, in the A. 330 obtectus seed beetles mate choice becomes relaxed in females when tested in stressful conditions 331 [45]. These studies indicate that female-specific stress reduces rather than increases the strength of 332 selection imposed on males by female choice. However, a different response could be expected if 333 males can contribute to improve female condition through direct benefits such as nuptial gifts or 334 parental care. In C. maculatus, male ejaculate represents a large amount of water, carbohydrates, 335 proteins and peptides, and is sometimes considered a nuptial gift [46,47] in this aphagous species. 336 It is possible that ageing females would rely more on nutrition and hydration from the contributions 337 of male ejaculate to sustain their reproductive capacity. By imposing selection on delayed 338 reproductive ageing, the AT treatment could have resulted in more stringent mate choice imposed 339 on males that could in turn explain the proportional increase of both the male and female variance under the limitation of female-specific resources at least in species where mating provides direct 342 resource benefits to females. 343 344

Conclusions: 345
We have shown that there are sex-specific changes in the opportunity for selection in response to 346 different ecological challenges. Although this has been tested before (e.g. [48,49]), in the present 347 study we placed particular focus on female-specific resource limitation, with the prediction that it 348 would lead to a more female-biased opportunity for selection. This prediction relied on the 349 assumption that resource limitation would generally increase opportunity for selection, which has 350 not been the case for all treatments. Despite the variety of ways in which sex-specific selection 351 responded to our different treatments, selection remained male-biased in all cases, which suggests 352 that this pattern is in fact relatively robust. Moreover, our results from the HT showed that a male-353 bias in the opportunity for selection can also be driven by a response of females to changes in 354 environmental conditions, which challenges the view that male-bias in selection is generally driven 355 by intense sexual competition in males. While it is not surprising that manipulating variance in 356 fitness of one sex should trigger a response in the other because of the many levels of interactions 357 involved in sexual reproduction, it is rather striking that males remained the more variable sex 358 regardless of the degree of stress on females.