Ageing via sexual perception is a by-product of male adaptive plasticity in Drosophila melanogaster

Sensory perception of environmental cues can dramatically modulate ageing across distant taxa. For example, male Drosophila melanogaster age faster if they perceive female cues but fail to mate (ageing via sexual perception). This finding has been a breakthrough for our understanding of the mechanisms of ageing, yet we ignore how and why such responses have evolved. Here, we used D. melanogaster to ask whether ageing via sexual perception may be a by-product of plastic adaptive responses to female cues, and found that while long-term sexual perception leads to reproductive costs, short-term perception increases male lifetime reproductive success in a competitive environment. Simulations under a wide range of socio-sexual and demographic scenario suggest that such plasticity as a response to sexual perception might be a widespread strategy in nature. Finally, we show that sexual perception can significantly magnify sexual selection (15-27% average increase in the opportunity for selection).

For example, male Drosophila melanogaster age faster if they perceive female cues but fail to mate 23 (ageing via sexual perception). This finding has been a breakthrough for our understanding of the 24 mechanisms of ageing, yet we ignore how and why such responses have evolved. Here, we used D. 25 melanogaster to ask whether ageing via sexual perception may be a by-product of plastic adaptive 26 responses to female cues, and found that while long-term sexual perception leads to reproductive 27 costs, short-term perception increases male lifetime reproductive success in a competitive 28 environment. Simulations under a wide range of socio-sexual and demographic scenario suggest 29 that such plasticity as a response to sexual perception might be a widespread strategy in nature. 30 Finally, we show that sexual perception can significantly magnify sexual selection (15-27% average 31 increase in the opportunity for selection). 32

INTRODUCTION 34
Over the last decades, we have realised the potential for sensory perception to act as a modulator 35 of ageing across invertebrate and vertebrate taxa (1)(2)(3)(4). A series of ground-breaking studies have 36 shown that sensory perception can trigger physiological changes at multiple levels, from 37 homeostasis to tissue physiology, that significantly accelerate ageing (3, 5-8). For instance, 38 chemosensory perception of food impairs the extended lifespan conferred by dietary restriction in 39 both the vinegar fly Drosophila melanogaster (3) and in the nematode Caenorhabditis elegans (1,9). 40 Similarly, in these two model species, exposure to different types of conspecific cues decreases 41 lifespan and alters critical physiological traits (5,7,8,10), suggesting that ageing via sensory 42 perception operates across distant taxa and functional contexts. 43 Organisms respond to environmental cues through a host of plastic physiological, 44 anatomical or morphological changes that allow them to adjust their behaviour and life-history 45 strategy so as to improve their fitness (11). For example, in some insect species, males respond to 46 socio-sexual cues (i.e. density and diversity of rival male odours indicative of varying levels of sperm 47 Survival -A Cox proportional hazards survival model revealed a significant exposure x treatment 114 interaction (LRχ 2 1 = 5.14, P= 0.023). 115 Reproductive lifespan and reproductive ageing-We did not find significant effects on either 116 reproductive lifespan (i.e. treatment x exposure effect, LR χ 2 1 = 0.09, P= 0.767; sensory treatment 117 effect, LR χ 2 1 = 0.35, P= 0.555; exposure time effect, LR χ 2 1= 0.145, P= 0.704) or reproductive ageing 118 (using net decrease in male relative reproductive success between the first and the second female 119 they competed over -henceforth Δ♀ 1 ♀ 2 -, F1,393 = 0.22, P= 0.636; using net decrease in male 120 relative reproductive success between the first and the third female they competed over -121 henceforth Δ♀1♀3-, F1,359 = 1.38, P= 0.240). Sensory treatment had no effect on reproductive 122 ageing (using Δ♀1♀2: F1,393 =0.33, P=0.569; using Δ♀1♀3: F1,359 = 2.00, P= 0.158), but exposure 123 time significantly affected reproductive ageing (Δ♀1♀2: F1,393 =7.56, P= 0.006; using Δ♀1♀3: F1,359 124 =8.80, P= 0.003) as the decrease in relative reproductive success over the course of two (and three) 125 females was of higher magnitude with increasing exposure (see SM for further details). 126 Model and simulations-We built a computer model to estimate male fitness based on the 127 assumptions that (1) the log-reproductive function (l(x) m(x); standard demographic notation) 128 modelling realized reproductive lifespan (i.e. from the onset of mating (b) onwards) is a 90-degree 129 rotated sigmoid curve (i.e. monotonously decreasing, levelling-off at intermediate ages) and that 130 (2) the elevation (E) of that curve declines linearly with b. The features and additional details of our 131 computer model were largely based on the fitting of ad hoc Generalized Additive Models to our 132 empirical results. Another assumption of the computer model is that (3) the relationship between E 133 vs. b differs according to two alternative strategies: a) a "spendthrift" strategy (henceforth S 134 strategy), characterised by males that always respond as if perceiving female cues (i.e. always 135 engage the physiological responses associated with perception of reproductive opportunities), and 136 b) a "thrifty" strategy (henceforth T strategy), characterised males that never respond to female 137 cues (i.e. never engage the physiological responses associated with perception of reproductive 138 opportunities; see SM for further detail). Note that presence/absence of female cues is irrelevant 139 to both these strategies, as they are unconditional fixed strategies. Our aim with these simulations 140 was to explore what conditions favour the expression of the S fixed phenotype vs. the T fixed 141 phenotypes vs. the evolution of a plastic phenotype that could display either of S or T. In our 142 simulations, a randomly sampled b was assigned to each individual, and this allowed us to compute 143 the effects of the S and T strategies on within-male population variation in fitness. We explored the S and T strategies in scenarios resulting from the combination of three factors: (1) daily population 145 growth rate (r: -0.1, 0, 0.1), (2) mean value of b (from 1 to 15 d), and (3) standard deviation of b (sb: 146 0.01, 0.1, 1d). We envision mean b as a direct consequence of female density, and we assume 147 standard deviation of b to be correlated to the heritable variation in male mating performance. 148 The effect of the mean value of b on simulated fitness followed a pattern tightly consistent with 149 variation in E (Figure 3). This points out the prevalent role of E in mediating the effect of b on 150 fitness. Our simulations showed that the S strategy was advantageous (compared to the T strategy) 151 whenever mean b was lower than 7-8 days, and disadvantageous for higher mean b (Figure 3). 152 More precisely, within a range of b between 1 to 7-8 days, the relative advantage of the S strategy 153 increased with decreasing mean b. Contrastingly, within a range of b between 8 to 15 days, the 154 disadvantage the S strategy increased with increasing mean b. Standard deviation in male fitness 155 declined steeply with mean b with the S strategy, and was rather independent of mean b with the T 156 strategy. This results in the S strategy leading to higher fitness variation than the T strategy for 157 mean b lower than 7-8 days, but causing lower fitness variation for mean b higher than 7-8 days. 158 For example, in a stationary population (r = 0), fitness was higher for the T strategy (38% in 159 average) across average mean b equal or above 7 days. Contrastingly fitness was lower for the T 160 strategy (35% in average) across average mean b below 7 days. With respect to the increase in male 161 fitness variation (i.e. opportunity for selection), we found that responding to female cues (i.e. 162 plastic male strategy switching between the S and the T strategies accordingly to what strategy is 163 advantageous) resulted in an overall average increase in standard deviation of 15 to 27%. Within-164 population standard variation in onset of mating (male mating rate; i.e. a proxy for variation in male 165 quality and hence starting opportunity for selection) had little effect on mean fitness and standard 166 deviation of fitness, so estimates above are given for the median across the whole range explored 167 (see SM). These patterns are qualitatively the same for the population growth rates and the 168 standard deviation of b explored in all simulations (see SM for further details). 169

DISCUSSION 171
In this study, we found evidence strongly suggesting that the so far reported ageing effects induced 172 by male perception of female cues in fact derive from adaptive male plastic responses. Sensing 173 female cues for a short period of time (1 day) prior to reproduction triggered plastic responses in 174 males that made them better competitors, and ultimately increased their lifetime reproductive success. Perception of female cues was neutral following intermediate exposures (between 3 and 7 176 days) and only resulted in net fitness costs if males perceived females for at least 15 days of 177 simulated reproductive failure. Simulations showed that such plastic response can be advantageous 178 under a wide range of socio-sexual and demographic contexts and significantly increase the 179 opportunity for selection, and thus the intensity of sexual selection. 180

181
We found that, in Drosophila melanogaster, short-term (1 day) sensory exposure to female 182 cues (simulating a short lag between perception of reproductive opportunities and the onset of 183 mating) increases male lifetime reproductive success in a biologically relevant context (i.e. in 184 competition against rival males over a series of different females; Figure 1). In contrast, we did not 185 find evidence of net perception effects for intermediate sensory exposure treatments simulating a 186 lag of 3d and 7d between the perception of reproductive opportunities and the onset of mating. 187 Finally, and in line with previous evidence, we found that extended exposure to female cues prior 188 to mating (i.e. 15d lag) led to net fitness costs in D. melanogaster males (5,7,22). Such fitness costs 189 were due to a decrease in reproductive success (see also 22) but, contrarily to previous studies 190 (5,7,22), we did not find that sexual perception reduced survival. This is likely due to the fact that 191 our longest exposure (15 days) was considerably shorter than those in previous studies (see 5,7,22). 192 To summarise, these results show that sexual perception is beneficial if males rapidly access mating 193 after perceiving reproductive opportunities, and that sexual perception only leads to net fitness 194 costs and accelerated ageing in males when the lag between perception of reproductive 195 opportunities and the onset of mating is relatively long (i.e. 15 days long or more). 196 197 As population demography modulates the relative importance of reproductive timing (32-198 35), we complemented our analysis by calculating the rate-sensitive fitness consequences of male 199 exposure to female cues. Our rationale was both to assess fitness effects in a range of demographic 200 backgrounds and to examine the relative importance of early-versus late-life reproduction at 201 modulating the aforementioned effects of sexual perception. While qualitatively similar across 202 population dynamics, we found the potential for sexual perception to affect individual fitness to be 203 more marked in decreasing populations, across all exposure lengths (Figure 2). Given that late-life 204 reproduction is more important in decreasing populations, relative to stable or increasing 205 populations (34,35), this result highlights that sexual perception effects accumulate through life. 206 Interestingly, post-hoc exploration of the data indicated that benefits linked to short-term (1d) 207 perception were rapidly observable (as soon as over the 24 hours following the onset of mating) 208 and persisted over the whole life of males (see SM for further details), implying that the 209 physiological changes triggered by early-on perception of female cues were conserved in the long 210 term. In other words, males do not seem to experience transient changes, but rather long-lasting 211 responses that impact their life history. This is, in itself, a remarkable finding, and we suggest a 212 priority for future studies should be to address the mechanism underlying this phenomenon. 213 Understanding what specific fitness benefits (e.g. pre-vs. post-copulatory competition) are 214 involved in these effects might offer valuable information about reproduction-survival trade-offs 215 and the evolution of ageing. 216

217
The arising question is whether such male plasticity is likely to be adaptive for D. 218 melanogaster males in nature, and available evidence strongly suggests it will. Virgin D. 219 melanogaster females generally mate soon when presented to males, with average mating rates of 220 one mating every 1 to 3 days in lab and wild populations (36-43). This implies that, despite strong 221 male-male pre-copulatory competition and high variance in male mating success (44), most of the 222 males that will ever reproduce are likely to start mating within their first two weeks of life. 223 However, the aforementioned mating rates reflect situations of relatively high density that are 224 more likely to represent maximum mating rates rather than to be indicative of average mating 225 rates over space and time in the wild. While little is known about fine-grained population density 226 dynamics in wild D. melanogaster (45), there is indirect evidence to suggest that density 227 fluctuations are probably common in the field. This species' ecology is closely linked to food sources 228 (e.g. orchards) whose availability exhibits drastic spatiotemporal variation, which is inevitably 229 bound to modulate local density. Accordingly, in the field D. melanogaster larvae maintain a stable 230 polymorphism (largely driven by a single locus -the for gene-) with two foraging variants: a) 231 rovers, characterized by long foraging trips and pupation away from the food source, are 232 preferentially selected under high densities, while b) sitters, characterized by short foraging trips 233 and pupation in the food source, are preferentially selected in low densities (46-47). Thus, there is 234 suggestive evidence that natural populations of this species are subject to frequent fluctuations in 235 local density. Under low densities and/or during phases where dispersal in search of food sources is 236 likely to be common, finding (and mating with) females may be less frequent. This implies that 237 average mating rates in the wild are almost certainly variable, as a consequence of fluctuations in 238 local density. Under this context, male plastic responses such as those reported in this study are 239 likely to be adaptive because they will allow males to engage the physiological machinery that 240 allows them to maximize their competitive ability, but only in the presence of socio-sexual cues 241 indicative of mating opportunities. Thus, a plastic response based on the presence of reliable socio-242 sexual cues seems to allow males to accrue the benefits of engaging in responses that condition 243 them for competition over reproduction when in a high-density scenario, while avoiding the long-244 term costs that would ensue from unconditionally engaging such responses in a low-density 245 environment. about what contexts we may expect these mechanisms to evolve in. In the case of ageing via sexual 254 perception, an interesting facet of the costs involved in male plastic responses of this sort is that 255 they are contingent on mating, so that costs are only paid if there is a relatively long lag between 256 male perception of reproductive opportunities and the onset of mating (5,7,22). This has important 257 implications for both the evolution of this strategy and its impact on related evolutionary 258 processes. We suggest that male plastic behaviour similar to that reported here for D. 259 melanogaster will arise frequently in nature in response to fluctuations in the availability of 260 reproductive opportunities, whenever these correlate reliably with environmental cues (e.g. female 261 odours). Implicit in this hypothesis are two predictions. First, that such changes must be, on 262 average, beneficial to males facing imminent competition over reproduction, as strongly suggested responses to female cues will be favoured under a wide array of socio-sexual contexts. We found mating rate, as captured by the average b (onset of mating) in our simulations, to be the main 272 determinant for the evolution of male plastic responses to reproductive cues. Consistently high 273 mating-rates benefited a fixed "spendthrift" (i.e. always respond as if females were present) 274 strategy (30-85% fitness advantage for mating rates of one every 1 to 4 days), while consistently 275 low mating rates benefited a "thrifty" (i.e. always respond as if females were absent) strategy (38-276 56% fitness advantage for mating rates of one every 12 to 14 days; see SM for further details). In 277 contrast, a plastic response would have higher fitness whenever average mating rates vary within 278 an intermediate range of mating rates ( Figure 4). Thus, we predict that in species where average 279 mating rates are consistently high, such as promiscuous species with little fluctuation in density, a 280 fixed spendthrift strategy will be favoured ( Figure 4). This will be the case of species with very short 281 lifespans, where long-term costs are likely to be negligible (mayflies are a good, albeit quite 282 extreme, example). In contrast, we predict that a fixed thrifty strategy will be favoured in species 283 with consistently low mating rates (Figure 4), such as iteroparous species with low density and/or 284 prolonged reproductive seasons. Finally, male plastic responses are expected to evolve in species 285 where mating rates fluctuate in accordance with changes in population density ( Figure 4). Given 286 that environmental stochasticity effects on population density are very frequent, we suggest that 287 plastic responses to sexual perception, such as those reported here, might actually be a common 288 strategy in nature; at least in promiscuous species with reproductive lifespans and demographic 289 parameters within the range of D. melanogaster (modelled in this study), which are typical of many 290 insects. Our model shows that under the latter condition, male plastic behaviour will be favoured 291 almost irrespective of population demography (i.e. growth rate) and inter-individual variability in 292 male mating rates. We thus suggest a priority for future studies should be to study the 293 phenomenon of ageing via sexual perception across species with contrasting life histories. 294 295 A second implication of the fact that male plasticity costs are contingent on mating is that 296 this phenomenon can magnify sexual selection. The idea is that male responses to sexual 297 perception can magnify sexual selection by further reducing the reproductive success of low-quality 298 males. Briefly, if perception costs result from a decoupling between perceived and realised mating 299 opportunities, it follows that low-condition males will tend to disproportionally pay such costs 300 simply because they have lower mating success, and hence will take longer to mate (22,23). In 301 addition to this, our results suggest that males able to mate soon after perceiving female cues 302 accrue a lifetime reproductive advantage over rival males. Given that high-quality males (i.e. good at intra-sexual competition) will tend to mate quicker than the average male, this means they are 304 expected to disproportionally harvest perception benefits. As a consequence, male plastic 305 responses to female cues are bound to increase the overall variability in male reproductive success 306 (i.e. opportunity for selection), hence potentially magnify sexual selection beyond previously 307 surmised. Our simulations show that this modulation can be biologically meaningful, with an 308 average increase in the opportunity for selection estimated of between 15 and 27% for the whole 309 range of average mating rates explored (Figure 4 & SM). Given that plastic male responses to 310 female cues are expected to be favoured in promiscuous species, where sexual selection is already 311 expected to be intense, this could lead to eco-evolutionary feedback that further magnifies sexual 312 selection. 313

Stocks and maintenance 316
Unless when stated otherwise, all flies used in this experiment were laboratory wild type (wt) 317 Dahomey D. melanogaster. We used homozygous recessive spa mutants (sparkling poliert) as 318 competing males and reproducing females in order to assess paternity of focal wt individuals. 319 Homozygous spa flies have a distinguishable rough-eye phenotype that allowed us to distinguish all 320 offspring from our focal males, and have the added advantage of being slightly worse competitors 321 than wt males, which ensured focal males would eventually mate. Stock populations are 322 maintained outbred, with overlapping generations, at 25°C on a 12h light/12h dark cycle fed with 323 standard food (solidified aqueous mix containing 60g.L -1 corn flour, 50g.L -1 white sugar, 40g.L -1 fresh 324 baker's yeast, 10g.L -1 soy flour, 10g.L -1 industrial agar, 3g.L -1 Methyl 4-hydroxybenzoate (nipagin), 325 10mL.L -1 96 % EtOH, 5mL.L -1 99% propionic acid). We obtained all flies by collecting eggs on yeasted 326 grape juice agar plates (FlyStuff grape agar premix, Genesee Scientific) from stock populations. We 327 reared all flies used in this experiment at a controlled density of ca. 200 individuals per 250mL 328 bottle filled with ca. 75mL of food, and isolated them by sex within 6 hours of emergence (i.e. as 329 virgins) at standard densities of 15 females and 20 males per vial, using ice anaesthesia. 330 331

Experimental design 332
We individually exposed wt males to either donor females (treatment) or not (control) following 333 the method as described in García-Roa et al. (22). Briefly, we connected two vials containing food to each other and placed a mesh partition between them. Males were then singly isolated on one 335 side of the mesh, while the other contained either: a) three wt females (i.e. treated males) or b) no 336 females (i.e. control). This mounting allowed treated males to be exposed to female odours while 337 ensuring they would not mate. To explore whether sensory perception effects are contingent on 338 the time of exposure to female cues, we exposed experimental males to sensory stimuli (i.e. female 339 cues or control) for one day, three days, seven days or 15 days; n = 60 per treatment combination 340  We fitted a general linear model (LM) in order to analyse relative reproductive success (using 361 wt/total number of F1 -#F1-).To analyse reproductive ageing, we first calculated the average 362 reproductive success of each focal male over each one of the first three females it competed for 363 (only for individuals surviving this period of time). Then, we calculated the net decrease in male 364 relative reproductive success between the first and the second female (Δ♀1♀2), as well as between the first and the third female (Δ♀1♀3). We fitted general linear models to Δ♀1♀2 and 366 Δ♀1♀3. 367 368 All these models incorporated treatment (categorical variable, two levels: sensory sexual 369 perception and control), exposure time (continuous variable) and the interaction between 370 treatment and exposure as fixed factors. We checked model assumptions (residuals normality, 371 homoscedasticity, homogeneity of variances, absence of outliers and influential points, absence of 372 autocorrelation of factors) using the "performance" package (50). We assessed model term 373 significance with α=0.05. Models were run computing type III ANOVA using the "car" package (51), 374 in R studio 1.1.456. No clear outlier was detected in the dataset, and no datapoint was excluded of 375 statistical analyses. All significant reported p-values remain so after correcting for inflation of type I 376 error rate due to multiple testing (using the Benjamini-Hochberg procedure for a false discovery 377 rate of 0.05). 378 In order to place lifetime reproductive success effects in a demographic context, we 379 additionally estimated individual rate-sensitive fitness estimates of treated versus control where r is the intrinsic growth rate, and k(x) is the so called reproductive function (l(x) m(x); i.e. 390 survival times fertility at age x; e.g. 33). Additionally, x = 0 is the age at maturity, so that 391 developmental times are not considered to make a difference when fitness is compared between 392 scenarios or individuals. Our computer simulation model is informed by our experimental design, 393 but making the biological mechanisms as explicit as possible and using directly interpretable factors 394 that act on k(x). In this way, the values of model parameters can be obtained from the estimates 395 from experimental data. 396 Therefore, in order to build our computer simulation model, we performed data analysis to 397 describe and extract model components. In that analysis, k(x) was analysed on an individual basis; 398 i.e., ki(x), where i identifies the i th experimental individual. The realized ki(x) is a count; i.e., the 399 number of offspring at age x (i.e. from x to x + ∆xx; the subscript stresses that the experimental time 400 lag between consecutive observations is not constant) of the individual i th . Commonly in Mixed 401 Effects Models the expectation of the log-count is modelled additively; in our case 402 As ui(x) may be a rather complex function, we used Generalized Additive Models (GAMs). 404 We assumed a negative binomial (NB) error distribution and applied GAM separately for each 405 experimental condition (four treatments and four controls). In addition to a structural relationship 406 between ui(x) and x (a more or less elevated smooth function to be found), our GAM analysis 407 assumed two random components (random slope and random intercept; both following 408 uncorrelated Gaussian distributions) to account for among-individual variation in the experiments. 409 We used packages mgcv (54) and nlme (55). GAMs found smooth functions (expected ui(x) vs. x) 410 with similar shapes (approximately, an asymmetric a 90-degree rotated sigmoid curve) for the eight 411 experimental conditions (see SM for further details. The shape was compressed for the life window 412 with potential access to females; that is, between the onset of access to females (B; 1, 3, 7 and 15 413 days) and the end of the reproductive life. The rotated sigmoid functions were more or less 414 elevated depending on the experimental condition, and we found a clear pattern whereby 415 elevation decreases with the female access onset. Using least squares, we fitted lines (treatment 416 and control separately) to the relationship between the elevation and the female access onset (E 417 vs. B). Our GAMs did not find random intercepts to be significant but did find a significant variation 418 in random slopes. 419 420 After this analysis, for our computer model the i th individual is assumed to have the expected 421 exp(ui(x)) (i.e. randomness from NB distribution was neglected). Our computer model uses the 422 functional shape showed in SM with the elevation describes by SM). However, in order to compute 423 elevation of ui(x) now the mating onset is β (not B, fixed experimentally) and is a log-normal 424 variable with values assigned randomly to males. Additionally, a male-dependent Gaussian random 425 slope is added to get ui(x). We envisage the random slope as being due to demographic 426 stochasticity, and assumed it to be independent on the timing of the mating onset. Simulations experienced by the population. Average mating rates that are consistently high will favour a fixed 723 "spendthrift" strategy that taps on the short-term benefits of engaging a maximum male 724 physiological response in preparation to competition for reproduction. In turn, average mating 725 rates that are consistently low will favour a fixed "thrifty" strategy that avoids the long-term costs 726 of engaging a maximum male physiological response in preparation to competition for 727 reproduction. Finally, variation in mating rates within a low-to-moderate range (i.e. uncertainty as 728 to whether males will be, on average, quick or slow to mate) will favour plastic male strategies 729 whereby males engage (or not) maximum physiological responses depending on the presence of 730 female cues (i.e. local mating patch density). This analysis assumes that local mating patch density 731 correlates positively with average mating rates in the mating patch.

Distribution of Perception of Females
Assumed values for the model parameters if otherwise is not explicitally stated Parameters assumed to be indenpendent of the occurrence of percetion effects:

Distribution of Perception of Females
Assumed values for the model parameters if otherwise is not explicitally stated Parameters assumed to be indenpendent of the occurrence of percetion effects:

Distribution of Perception of Females
Assumed values for the model parameters if otherwise is not explicitally stated Parameters assumed to be indenpendent of the occurrence of percetion effects:

Distribution of Perception of Females
Assumed values for the model parameters if otherwise is not explicitally stated Parameters assumed to be indenpendent of the occurrence of percetion effects:

Distribution of Perception of Females
Assumed values for the model parameters if otherwise is not explicitally stated Parameters assumed to be indenpendent of the occurrence of percetion effects: