Influences of male age, mating history and starvation on female post mating aggression and feeding in Drosophila

Mating changes female behaviour and physiology across a wide range of taxa, with important effects for male and female fitness. These changes are often induced by components of the male ejaculate, such as sperm and seminal fluid proteins. However, males can vary significantly in their ejaculates, due to factors such as age, mating history, or nutritional status. This male variation may therefore lead to variation in the strength of responses males can stimulate in females, with alterations in fitness outcomes for both sexes. Using the fruit fly, Drosophila melanogaster, we tested whether three aspects of male condition shape an important, but understudied, post-mating response – increased female-female aggression. We found that females mated to old males fought less than females mated to young males. This effect was exacerbated in mates of old, sexually active males, but there was no effect of male starvation status on mating-induced female aggression. There was also a significant effect of age and mating history on female post-mating feeding duration. Our results add to a growing body of literature that variation in male condition can shape sexual selection through post-mating responses in females, including female-female interactions. Studying such variation may therefore be useful for understanding how the condition of sex affects the behaviour of the other.


Methods 157
Fly stocks and culture 158 We used the Dahomey wild-type stock, which was first collected in Benin, Africa in 1970 159 (Wigby & Chapman, 2004). Flies have been maintained in large, outbred populations in cages 160 with overlapping generations. All flies were cultured and experiments were conducted at 25°C 161 on a 12:12 light: dark cycle in a non-humidified room. Except where stated, adult flies were 162 kept on standard Lewis medium (Lewis, 1960), with no access to live yeast. ('unmated': U) or of 3 virgin males and 9 virgin females ('frequently-mated': F). Males from 173 two age classes were used: 1 week (1w, young) and 5 weeks (5w, old) old. Flies in the U 174 treatment were transferred to new vials once a week and those in the F treatment were 175 transferred twice a week using light CO2 anesthesia at each transfer. Dead or escaped females 176 were replaced with similarly aged females at each transfer. At 3 weeks, females from the 5-177 week group were replaced with virgin 3-5 day old females, to reduce co-ageing effects. The 178 males from F were merged into single-sex groups of 10-12 males four/five days before 179 assaying, in order to provide a consistent period of sexual rest prior to the assay point.

181
To generate the starvation treatments, males were randomly assigned to the 'fed' or 'starved' 182 treatment upon collection directly after eclosion. Males from the 'fed' treatment were kept 183 in vials with standard fly medium, while males in the 'starved' treatments were kept in vials 184 containing only damp cotton wool, which provides no nutritional value. Males in both 185 treatments were kept in groups of 15-20 for three days prior to mating. Females were either 186 mated to the 'fed' or 'starved' males, or were kept as virgins. This experiment was conducted 187 in two blocks. 188 189

Mating assays 190
Mating assays took place when the experimental females were 3-5 days old. Females were 191 randomly assigned to a treatment group and an individual male was aspirated into each 192 female vial. Pairs were given five hours to mate. Latency to mating and mating duration were 193 recorded for each pair. If pairs did not mate during this time, they were discarded. However 194 unmated males from the 5w-F, sexually active treatment were tested again the following day 195 to increase sample size as such males have been shown to be less likely to mate than the 196 other treatments (Sepil et al., 2020). Only 6 females in our aggression analyses mated with 197 reused males (n = 3 dyads), and results remained the same regardless of whether these 3 198 dyads were included in analyses or not. We therefore kept these 3 dyads in our main analyses. 199 Once females had mated, they were removed and placed into new vials containing regular fly 200 food medium and no live yeast. We also set up vials with virgin females as controls, where 201 virgin females were switched to new vials at the same time as mated females. Females 202 remained in these vials overnight until they were used in aggression assays. These vials were 203 10 subsequently frozen to count eggs laid by each female. The mating assays were performed 204 over three days. 205 206 Aggression assays 207 Females from the same treatment were used in contests 24 hours after mating. For the two 208 hours directly before being used in a contest, females were kept in a vial with damp cotton 209 wool but no food. This brief period of starvation has been shown to increase how much stopped physically touching each other for three seconds. We also recorded the amount of 226 11 time each fly spent on the food cap, which we use as a proxy for feeding duration, as we were 227 unable to detect when flies were actually feeding from the food cap. In all analyses, the age and mating history experiment data was analysed using two separate 244 models. The first only included mated females so that we could investigate whether it was 245 age, mating history, or an interaction between the two that alter the behaviour. As 246 behavioural assays took place over three days, we also included day as a fixed effect in the 247 model. This model allowed us to investigate which aspects of age and mating history 248 As most female aggression takes place on the food cap over access to food, we measured the 258 amount of time females spent on the food cap, and use this as a proxy for female feeding 259 duration. In both age and mating history models, a linear model with no transformations was 260 the best fit for the data. In the starvation experiment, the model that best fit the data was a 261 GLM with a Gamma distribution and inverse link. To avoid non-independence of data points, 262 we used dyad as the unit of replication for analyses of contest behaviour and feeding 263 duration, rather than individual flies. We analysed latency to mate using Cox proportional hazards models. We analysed the data 273 using the R package 'survival' (Therneau, 2015). To analyse mating duration, we used a linear 274 model with mating duration as the response variable. There was a significant effect of age on fencing duration, with females mated to old males 313 spending less time fencing than those mated to young males (F1,101 = 4.72, p = 0.03; Fig. S2.a). 314 There was no effect of mating history (F1,101 = 0.14, p = 0.71), or the interaction (F1,101 = 0.25, 315 p = 0.61). There was no effect of treatment on fencing duration when looking at all females 316 (F4,124 = 1.54, p = 0.2). There was no effect of treatment on fencing duration in the starvation 317 experiment (GLM: X 2 2,77 = 2.41, p = 0.3; Fig. S2.b). 318 319 Food cap duration 320 When looking solely at mated females in the age experiment, there was a significant 321 interaction between age and mating history (F1,101 = 6.85, p = 0.01, Fig. 2.a).. When looking at 322 all females, there was a significant effect of treatment (F4,124 = 8.41, p < 0.001; Fig. 2.a). 323 Females mated to Old-F males were similar to virgin females, with both spending less time on 324 the food cap than the rest (pairwise comparisons in Supplementary Table 3). Old males and F males were significantly slower and less likely to mate than young and U 344 males (Cox proportional hazards model: Age: X 2 1,249 = 16.41, p < 0.001; Mating history: 345 X 2 1,249 = 5.2, p = 0.02). This was primarily driven by Old-F males being significantly slower 346 and less likely to mate than all other types of males, as demonstrated by a marginally non-347 significant interaction between age and mating history (X 2 1,249 = 3.71, p = 0.05; Fig. S3.a).

Likelihood of laying any eggs 364
In the age and mating history experiment, we first restricted our analysis to mated females. 365 There were significant effects of both age and mating history on the likelihood of mated 366 females laying any eggs (Age: X 2 1,261 = 12.08, p < 0.001; Social: X 2 1,261 = 7.82, p = 0.005; Fig.  367 3.a). Females mated to old males were less likely to lay any eggs, as were females mated to F 368 males. There was no significant interaction between age and mating history (X 2 1,261 = 1.41, p 369 = 0.24). When considering all females in the age experiment, there was a significant effect of 370 treatment (X 2 4,324 = 96.91, p < 0.001; Fig. 3.a; pairwise comparisons in Table S.4). Virgin 371 females and females mated to Old-F males were less likely to lay any eggs. Only two virgin 372 18 females laid eggs, while the majority of mated females laid at least one egg, making mated 373 females much more likely to lay eggs overall. 374

375
In the starvation experiment, mated females were more likely to lay eggs than virgins (X 2 2,327 376 = 177.85, p < 0.001, Fig. 3.b), however there was no effect of male feeding status on a female's 377 likelihood to lay eggs (X 2 1,222 = 0.01, p = 0.93). We investigated how three aspects of male condition influenced the stimulation of post-403 mating female aggression. We found that old males stimulated less aggression in females, 404 less time on the food cap, and were less likely to stimulate egg production. This effect of age 405 was particularly strong for mates of old sexually-active males, with females mated to Old-F 406 males being the most similar to virgin females in all our behavioural and fecundity measures. 407 20 Our results suggest that these differences in female behaviour due to male condition are likely 408 predominantly driven by changes in Sfp quality, but that age and mating related changes in 409 sperm numbers transferred to females also contribute. Interestingly, different PMRs respond differently to aspects of male variation. While Old-F and 424 Old-U males are both poor sperm competitors, Old-U males do not differ from young males 425 in their reproductive output, fertility, and ability to suppress female remating (Sepil et al., 426 2020). These differences are primarily due to differences in the ageing of sperm and seminal 427 fluid proteins (Sfps). Sperm transfer declines in Old-F males, but not in Old-U males, but at 428 least some Sfps undergo qualitative changes with age (Sepil et al., 2020). For post-mating 429 female aggression, male age was the only significant factor, with both Old-U and Old-F males 430 stimulating less aggression than young males, suggesting this age difference was potentially 431 21 driven by a decline in Sfp quality (Sepil et al., 2020). However, females mated to Old-F males 432 showed values closest to virgin females, suggesting a potential role for sperm quantity in 433 determining their level of aggression as well. 434 435 Sperm is necessary for females to display a full increase in aggression after mating (Bath et 436 al., 2017). It therefore seems likely that the significant reduction in the amount of sperm 437 transferred to females by Old-F males is primarily responsible for the corresponding reduction aggression. If aggression is stimulated by the sperm storage organs filling with sperm, for 505 example, then if the sperm storage organs are not filled due to males transferring less sperm, 506 then females will not get as aggressive, even if it would benefit them to do so. Our results are 507 more consistent with the second hypothesis as females mated to Old-F males were less likely 508 to lay any eggs in our experiment, and showed less aggression. However, when they did lay 509 eggs, they laid as many as all other mating treatments, suggesting it is the initial receipt of 510 sperm that is related to the stimulation of aggression. 511 512 Potentially, females use mating as a cue to upregulate their aggression. This response may 513 represent more of an 'on-off' switch with the receipt of male ejaculate components turning 514 on the aggressive pathway, rather than females increasing their aggression in proportion to 515 the amount of ejaculate they receive. It is possible that females mated to Old-F males did not 516 receive enough ejaculate to switch on this pathway. The idea of an 'on-off' switch is consistent 517 with another study which found that different genotypes of males do not stimulate different 518 levels of post-mating aggression in females, despite potential differences in their ejaculates 519 .