Sexual conflict promotes species coexistence through negative frequency dependence

A major challenge in community ecology is to understand the mechanisms promoting stable local coexistence. A necessary feature of local coexistence is that species show negative frequency dependence, rescuing rare species from exclusion. However, most studies have focused on ecological differences driving negative frequency dependence, ignoring non-ecological mechanisms such as reproductive interactions. Here, we combined field studies with behavioural and mesocosm experiments to investigate how reproductive interactions within and between species promote coexistence. Our results indicate that the intensity of male mating harassment and sexual conflict increases as species become more common, reducing female productivity and leading to negative frequency dependence. Moreover, field surveys reveal that negative frequency dependence operates in natural settings, consistent with our experimental results. These results suggest that sexual conflict can promote local coexistence and highlights the importance of studying reproductive interactions together with ecological differences to better understand the mechanisms promoting species coexistence. Significance statement Research on the mechanisms promoting local species coexistence have focused on canonical ecological differences that increase intraspecific over interspecific competition. However, one intrinsic factor of species that can promote coexistence are the reproductive interactions. We performed a series of behavioural and mesocosm experiments manipulating species frequencies together with field observations and show that sexual conflict can decrease female fitness when species are common and promote local coexistence. Our results suggest that reproductive interactions are an understudied mechanism that can promote species coexistence even when species are ecologically equivalent.


Introduction 35
Understanding the causes underlying species diversity in ecological communities is a major 36 challenge in both ecology and evolution. Coexistence theory predicts that negative frequency 37 dependence is necessary for local species coexistence (1). If species have a fitness advantage 38 when rare, they can increase from low abundance in a community and hence be rescued from 39 competitive exclusion (1-3). Previous research has focused on how ecological differences 40 between species can cause negative frequency dependence through rare species advantage 41 (4-11), for example through predator susceptibility (8, 9), resource competition (5, 12) and 42 phenology (11, 13). However, many communities are formed by species with little or no 43 ecological differentiation (14-18). How or do such ecologically equivalent species coexist in 44 a community? One possible answer to this question lays on an intrinsic characteristic of many 45 species that can limit species population growth rate and promote species coexistence: 46 reproductive interactions (19-22). Given how widespread sexual reproduction is in the tree of 47 The common bluetail (I. elegans; Fig. 1A) and the common bluet (E. cyathigerum; Fig. 1B) 124 are two ecologically similar and closely related damselfly species that are distributed 125 throughout Eurasia with their northernmost range limits in Scandinavia, were they are 126 commonly found in large numbers in ponds and lakes (55). These two damselfly species 127 shared a most recent common ancestor at least 12.6 million years ago (56) and overlap 128 extensively in their adult season (Fig. 1C) and are frequently locally sympatric (Fig. 1D). In 129 southern Sweden where this study took place these species are univoltine (55), and due to 130 their limited dispersal ability (less than 1Km) (57, 58), they can be found in a mosaic of 131 largely discrete populations with different environmental conditions. 132

133
The reproductive behaviours of Ischnura and Enallagma are very similar. First, males are 134 non-territorial and chase females (often several males at the time) and compete to grab the 135 females by the prothorax using the claspers situated in the tip of their abdomens. If a male is 136 able to find and subsequently clasp a female they form a tandem, after which the female can 137 respond by bending the abdomen to reach the male genitalia and copulate (55) (Fig 1A). 138 Before insemination the males remove the sperm from previous copulations (59, 60). 139 Therefore, females gain few or no benefits from multiple matings, but will experience fitness 140 costs that increase with the number of claspings (28). After copulating females oviposit in 141 emergent vegetation and the larvae grow and overwinter in the aquatic stage (55). Males and 142 females of both species are generalist predators and forage for flying insects near the water 143 (55). During the reproductive season males interact frequently with both con-and 144 heterospecific males, which can reduce male mating success (30,37,38,42,45,48). In 145 southern Sweden, where this study took place, adults of both species are found from late 146 spring to late summer (late May to August) to reproduce (Fig 1 C). 147

Are reproductive interactions frequency dependent? 149
To investigate if any of the four types of reproductive interactions shows frequency 150 dependence, we carried out a mating trial experiment where we varied these two species 151 relative frequencies: common (75%) and rare (25%). We used adult males and females (aged 152 by the stiffness of the wings) from natural populations with no visible signs of external 153 physical harm such as wing damage. We separated the captured individuals by sex and kept 154 them at a density of 10 individuals in netted containers (10.2 cm diameter and 22.9 cm 155 height) during transportation to Stensoffa Ecological Field Station, southern Sweden, where 156 the experiments took place. At the field station we set up males and females in larger netted 157 cages (45 cm diameter and 50 cm height). We added twigs and grasses to each cage to mimic 158 natural vegetation and allow individuals to perch or rest, and a plastic cup with water to 159 prevent desiccation. In each cage we put six individuals of one species (three males and three 160 females) and two individuals of the other species (one male and one female). Thus, in these 161 cages, we had two frequency treatments, both with equal sex ratios: "common" (75%) for the 162 most abundant species and "rare" (25%) for the less abundant species (Supplementary Table  163 1A). 164 165 We marked all males in each cage with individual fluorescent colour powder in the genital 166 area at the base of the abdomen and on the claspers (Fig 1B). After 24 hours, we terminated 167 the experiment and searched for traces of colour dust on the genitalia and prothorax of the 168 females. This technique allowed us to identify how many and which type of males (i.e., con-169 or heterospecifics) attempted to mate (i.e., clasped) or mated a given female, as these marked 170 males left traces of colour dust in the female prothorax (i.e., mating attempt or clasping) and 171 genitalia (i.e., successful mating) visible under UV-light. This method has previously been 172 successfully used to quantify the degree of short-term male mating harassment (number of male claspings) and female mating rates in I. elegans, E. cyathigerum and other damselfly 174 genera (28,30,37,38). 175

176
To estimate the intensity of intraspecific reproductive interactions between sexes, and hence 177 the potential for sexual conflict, we counted the number of male mating attempts (i.e., 178 number of claspings per female) on conspecific females in 24 hours. We underscore that this 179 rate of claspings does not take into account mating attempts that did not end up in claspings 180 (i.e., chasing of females) or repeated claspings of females by the same male. Therefore, our 181 measure of sexual conflict is conservative and will underestimate the total costs of male 182 mating harassment to female fitness. To estimate interspecific reproductive interactions 183 between sexes we counted the number of male mating attempts of heterospecific females, 184 using the same procedure (i.e., remnants of coloured dust on the female prothorax or 185 genitalia) in 24 hours. This measure is also a conservative measure of male mating 186 harassment, as it does not take into account heterospecific mating attempts that did not end 187 up in claspings. Finally, we quantified the costs of intra-and interspecific interactions within 188 sexes as male mating success (mated = 1; not mated = 0). Because male-male competition 189 can reduce male mating success (38), and if conspecific competition is stronger than 190 heterospecific competition, male mating success is expected to decrease when species are 191 common (i. e., negative frequency dependence). Conversely, if heterospecific male-male 192 competition is stronger, male mating success is expected to decrease when species are rare (i. We performed a series of mesocosm experiments under semi-natural conditions in eight large 199 square outdoor cages (3m per side; total volume 27 m 3 ) at the field station ( Fig 1E). The aim 200 of this mesocosm experiment was to quantify adult female longevity and per capita female 201 productivity (i. e., the number of emerging female offspring in the next generation per female 202 in the previous generation, a measure that should closely reflect population mean fitness or 203 mean per capita growth rateunder different species frequency treatments (common, 75% and 204 rare, 25%). Each cage contained a large water container (600L) with natural vegetation to 205 resemble natural conditions and facilitate oviposition (Fig 1F). Each water container was 206 inoculated repeatedly in the spring preceding these experiments with zooplankton (mainly 207 copepods and Daphnia) obtained from nearby ponds and macrophytes obtained from an 208 aquarium shop. This ensured that the damselfly larvae in our experiments would have enough 209 food to forage and grow. A few weeks after inoculations we confirmed by visual inspections 210 that these water containers had reproducing populations of zooplankton in the water. We 211 added six coffee filter papers and small pieces of floating vegetation (Phragmites australis) 212 to provide a resting substrate and to facilitate oviposition in these water containers. The 213 outdoor cages were covered with mesh small enough to keep damselflies in and predators 214 out, but wide enough to let smaller insects necessary as food for the foraging adults to enter 215 (25, 30, 37). Importantly, these cages had no predators as we aimed to investigate if intra-216 and interspecific interactions could cause negative frequency dependence and potentially 217 promote species coexistence. We have showed in previous studies that adult damselfly 218 survival is not affected by total adult density, indicating that prey availability is not an issue 219 in this experimental set up (30). 220

221
In each of these eight outdoor cages we manipulated species frequencies in two treatments 222 with the same frequencies as in the mating trials described above: common (75%) and rare 223 (25%). In each cage, we included 18 conspecifics (six females and 12 males) and six 224 heterospecifics (two females and four males) for a total of 24 individuals per cage. We 225 carried out a total of nine replicates (five for treatment with I. elegans being common and 226 four for the treatment with E. cyathigerum being common) during the reproductive season 227 (June and July) (Supplementary Table 2A). We also carried out two additional control 228 treatments that would allow us to assess if there could be contamination in our water tanks  Table 3). These sites varied in 250 relative species frequencies (Fig 1D). To quantify species frequencies and densities we 251 visited each site between three and five times per season (May-July) during warm (>15˚C) 252 days with no rain or strong wind, the most favourable conditions for these damselflies 253 (Supplementary Table 3 factors. All models were performed using the packages "lme4" (62) and "car" (63) in R (64). 277 278

Results 279
Are reproductive interactions frequency dependent? 280 We quantified the number of conspecific and heterospecific claspings from 87 females and 281 mating success of 89 males in our mating trials. We found a significant effect of species 282 frequency on the number of con-and heterospecific claspings of females, but in opposite 283 directions. Females experienced more conspecific claspings when they were common than 284 when they were rare (χ 2 = 4.61, p = 0.031; Fig. 2A) but more heterospecific claspings when 285 they were rare than when they were common (χ 2 = 11.12, p < 0.001; Fig. 2B). In contrast, we 286 found no effect of species frequency on male mating success. In all the models we found no 287 effect of species nor the interaction between species and frequency (Supplementary Table  288 1B-D). 289 290 Are reproductive interactions costly and do they result in rare species advantage? 291 We quantified adult longevity for 128 females (64 of each species) in our mesocosm 292 experiments. We found no main effect of frequency treatment on female longevity nor a 293 significant interaction between species and frequency, suggesting that neither species 294 longevity was affected by changes in the species frequency. However, we found a significant 295 main effect of species on female longevity (χ 2 = 24.72, p < 0.001), with shorter longevity (> 296 50%) of E. cyathigerum compared to I. elegans in these mesocosm cages (Supplementary  297   Table 2B). 298 299 Next, we analysed female productivity (i.e., the number of female offspring in the next 300 generation per adult female in the initial generation) in the mesocosm experiments. Female 301 productivity differed significantly between the two species (F = 15.63, p = 0.028), with I. 302 elegans females being on average more productive than E. cyathigerum. Importantly, we 303 found a significant and negative effect of species frequency (F = 53.55, p = 0.005) on female 304 productivity, with lower female productivity in the common compared to the rare frequency 305 treatment (i.e., negative frequency dependence) (Fig. 3). We found no significant interaction 306 between species identity and species frequency (Supplementary Table 2C), suggesting that 307 the strength of negative frequency-dependence was similar in both species. The results were 308 similar when we analysed the total number of emerging individuals in the offspring 309 generation and the number of emerging male offspring (Supporting Table 2D). We found 310 only minor contamination in our control cages, and our results above remain qualitatively 311 similar after correcting for such contamination (Supplementary Analysis 1). 312 313 Do these species show negative frequency dependence in nature? 314 Finally, we analysed species density changes across two generations at the18 natural 315 sympatric sites of I. elegans and E. cyathigerum. We found a significant effect of initial 316 species frequency (we present results on a logarithmic scale as they show better fit, although 317 untransformed data was also significant) on species density change (F = 14.95, p < 0.001; 318 Fig. 4). There was no significant effect of species identity nor the interaction between species 319 identity and initial frequency (Supplementary Table 4), suggesting that these two species 320 respond similarly to changes in relative frequencies in nature. These between-generation 321 changes indicate that the higher species frequency was at a site in 2018, the more it declined 322 in abundance the following year. These results suggest negative frequency dependence, 323 consistent with the findings in the mesocosm experiment (Fig. 3). 324 325

Discussion 326
Negative frequency dependence is an fundamental requirement for species coexistence, as a 327 species that has a fitness advantage when rare can recover from low abundance and 328 competitive exclusion can be prevented (1, 2). However, our understanding of the underlying 329 mechanisms responsible for negative frequency dependence and stable coexistence is still 330 poor (2, 65). Many previous studies have focused on the ecological niche differences by 331 which negative interspecific interactions can be reduced, although reproductive interactions 332 alone can also cause negative frequency dependence and promote species coexistence (19-333 22, 32). Our results suggest that intraspecific male mating harassment and the resulting 334 sexual conflict it generates has the potential to reduce female productivity, causing negative 335 frequency dependence and promoting species coexistence. To the best of our knowledge, our 336 study is the first empirical example of how a mechanism not based on ecological niche 337 differences can promote stable species coexistence. 338 339 Sexual conflict can have severe negative effects on female fitness and by extension reduce 340 population growth (23-27, 66-70). Moreover, sexual conflict is expected to increase at 341 higher densities (25, 27, 28, 30), as high densities should increase encounter rates between 342 the sexes and thereby elevate male mating harassment on females (25, 28). If the negative 343 fitness effects of sexual conflict on females are larger when species are common and reduced 344 when rare, sexual conflict could lead to negative frequency-dependence and rescue rare 345 species from competitive exclusion (19, 20). Consistent with these predictions, we found 346 evidence for negative frequency-dependence on female productivity (Fig. 3). Our mating experiments suggest that sexual conflict could be the driving mechanism causing negative 348 frequency dependence. Although other mechanisms (e.g., predation, cannibalism), especially 349 during larvae stage could also influence species relative frequency changes in the wild. 350 Conversely, heterospecific mating attempts are likely to be shorter in duration, given that 351 females reject heterospecific males and given that heterospecific male claspers do typically 352 not match female prothorax structures (71, 72). Therefore, heterospecific mating attempts are 353 likely to be less costly than conspecific male mating attempts. Trend Ecol Evol (2020). 466 Figure 2. We perform a series of mating experiments in which we manipulated species 589 frequencies, "rare" (25%) and "common (75%) to test the intensity of intra and interspecific 590 reproductive interactions (Supplementary Table 1). We found that the intensity of sexual 591 conflict, measured as the number of mating attempts (i.e., claspings), was more intense when 592 species were common than when rare (A). Heterospecific matings attempts followed the 593 opposite pattern, females experiencing more mating attempts by heterospecifics when rare 594 than when common (B). Points show the means and error bars the standard error. 595 Figure 3. We used mesocosm experiments to quantify the costs of sexual conflict in female 597 fitness. We found strong negative frequency-dependence in female productivity (measured as 598 the number of female offspring that emerge in the following generation per female in the 599 initial generation), having higher fitness when rare over common. Similar results were found 600 when we analyzed total productivity (i.e., number of offspring emerged per adult female in 601 the initial generation; Supplementary Table 2   outdoor cages with water containers across the entire life-cycle of damselflies (Fig. 1). We  The source of this contamination might have come from larvae or eggs attached to the vegetation used to inoculate the water containers or adult of either species entering the cages 684 by mistake. The mean contamination in the two E. cyathigerum cages was 18% (min = 5%, 685 max = 31%), and in the two I. elegans cages 2% (min = 0.8%, max = 4%). Assuming similar 686 levels of contamination occurred in the other cages, we performed two corrections to see how 687 this might have confounded our results. First, we removed from the female productivity of 688 each cage the mean percentage of contamination of each species, (-18% and -2% of E. 689 cyathigerum and I. elegans, respectively) (A). Second, we performed a more conservative 690 correction, by instead removing the maximum level of contamination found of each species (-691 31% and -4% of E. cyathigerum and I. elegans, respectively), (B). The two models below (A 692 and B) were performed using the corrected value of female productivity as response variable, 693 species frequency, species identity and their interaction as fixed factors in a generalized 694 linear model. In both cases the results are qualitatively the same as in the uncorrected 695 emergences and the effects of both "frequency" and "species" remain highly significant (see 696 also Fig. 3