Conservation of social effects (Ψ) between two species of Drosophila despite reversal of sexual dimorphism

Abstract Indirect genetic effects (IGEs) describe the effect of the genes of social partners on the phenotype of a focal individual. Here, we measure indirect genetic effects using the “coefficient of interaction” (Ψ) to test whether Ψ evolved between Drosophila melanogaster and D. simulans. We compare Ψ for locomotion between ethanol and nonethanol environments in both species, but only D. melanogaster utilizes ethanol ecologically. We find that while sexual dimorphism for locomotion has been reversed in D. simulans, there has been no evolution of social effects between these two species. What did evolve was the interaction between genotype‐specific Ψ and the environment, as D. melanogaster varies unpredictably between environments and D. simulans does not. In this system, this suggests evolutionary lability of sexual dimorphism but a conservation of social effects, which brings forth interesting questions about the role of the social environment in sexual selection.

antagonistic and is thought to have a shared genetic basis in males and females (Long & Rice, 2007). It is advantageous for males of this species to move more and for females to move less, as higher activity males mate more and higher activity in females' acts as a stimulus for male courtship (Bateman, 1948;Maklakov & Arnqvist, 2009;Olson-Manning, Wagner, & Mitchell-Olds, 2012;Partridge, Green, & Fowler, 1987;Tompkins, Gross, Hall, Gailey, & Siegel, 1982). In D. melanogaster, male courtship results in a general reduction in female lifetime reproductive success, likely due to higher rates of remating and interference during oviposition (Bateman, 1948;Fiumera & Dumont, 2006;Kuijper, Stewart, & Rice, 2006;Long & Rice, 2007;Partridge et al., 1987;Tompkins et al., 1982). Sexually antagonistic interactions, and IGEs expressed in those interactions, could potentially have a large impact on the evolutionary trajectory of a given trait (Marie-Orleach et al., 2017;Moore & Pizzari, 2005).
While the effect of locomotion on fitness has not been explicitly tested in D. melanogaster's sister species D. simulans, there is evidence to suggest that locomotion is not sexually antagonistic in this species. As aforementioned, higher activity level in D. melanogaster results in increased mating, and multiple matings are detrimental to females and beneficial to males (Fiumera & Dumont, 2006;Kuijper et al., 2006;Long & Rice, 2007;Maklakov & Arnqvist, 2009;Partridge et al., 1987;Tompkins et al., 1982). The relationship between locomotion and mating in D. simulans is not known, but there is evidence that undergoing multiple matings is beneficial in D. simulans (Bateman, 1948;Fiumera & Dumont, 2006;Kuijper et al., 2006;Maklakov & Arnqvist, 2009;Taylor, Wigmore, Hodgson, Wedell, & Hosken, 2008;Tompkins et al., 1982). While we cannot assume that the relationship between movement and mating is the same, if it were in D. simulans, it would suggest that locomotion is not sexually antagonistic. Overall, the data indicate that sexual selection and trade-offs vary considerably between Drosophila species. Indeed, sexual dimorphism of activity level is evolutionarily labile-for example, in D. suzukii females are 4× more active than males, while in D. melanogaster males are 3× more active than females (Ferguson et al., 2015;Long & Rice, 2007;Signor et al., 2017).
In a previous study, we used measures of Ψ for the effect of male locomotion on female activity to show that there is variation in Ψ between abiotic environments in D. melanogaster and found that Ψ was positive. This means that male and female locomotory phenotypes will covary, and females will resemble their male social partners without direct genetic effects (Signor et al., 2017). The two environments were ethanol-and nonethanol-exposed, and while Ψ was higher without ethanol, when Ψ was measured in individual genotypes it varied differently in different genotypes between environments. Here, we will compare Ψ for locomotion in D. simulans and D. melanogaster in ethanol-and nonethanol-exposed environments to determine whether there has been evolution of social effects between the two species.
To construct a comprehensive account of Ψ in different environments, we measure movement in D. simulans in sets of one female and two males. Across all experiments, female genotype is the same and the male genotype varies (but are identical within a replicate), allowing for genetic replication. We compare these results to those for D. melanogaster to investigate how Ψ, and differences in Ψ between environments, has evolved between these two closely related species (Signor et al., 2017). We find that locomotion is remarkably different in D. melanogaster and D. simulans, with D. simulans exhibiting little dimorphism in its level of activity. However, we find that Ψ is almost the same in the two species, indicating that for locomotion sexual dimorphism has evolved, but in this case Ψ did not. This has two implications that Ψ may have a partially separate genetic basis from locomotion in these species and that the differences in Ψ between environments are not due to differences in ethanol tolerance. Lastly, we find that while there is a Ψ × environment interaction in D. melanogaster, this is not present in D. simulans, which has interesting implications for the maintenance of variation in populations.

| Drosophila simulans genotypes
Male genotypes were collected from the Zuma organic orchard in Zuma beach, CA in the spring of 2012. They were inbred by 15 generations of full-sib crosses. A subset of six genotypes were used for behavioral assays. Females used in the behavioral assays were an inbred laboratory strain y 1 w 1 (San Diego Species stock number 14021-0251.013).

| Drosophila melanogaster genotypes
Males were collected from lines originating from an orchard in Winters, California in 1998 that were made isogenic by at least 40 generations of full-sibling inbreeding (Campo et al., 2013;Yang & Nuzhdin, 2003).
A subset of six genotypes were used for behavioral analysis. Female flies were an inbred laboratory strain y 1 w 1 (Bloomington stock number 1495).

| Heterozygous crosses
Heterozygous crosses were used to generate the male flies used in the behavioral assays. In general, flies collected from the field will be heterozygotes, so in order to more closely replicate these conditions, the genotypes from Zuma Beach and Winters were crossed to a single reference genotype (D. simulans: w 501 San Diego Species stock number 14021-0251.011, D. melanogaster: w 1118 , Bloomington stock number 3605), and the F1 flies from this cross were used in the behavioral assays (Nuzhdin, Friesen, & McIntyre, 2012). This allows both the use of heterozygous flies that are more similar to wild flies and the replication of behavioral observations as the flies are genetically identical (Brakefield, 2003;Wahlsten, 2001).

| Rearing conditions
All flies were reared on a standard medium at 25°C with a 12-h light/12-h dark cycle. Mothers were crossed at 1 day old in groups of 10 males and 10 females. F1 males and females were collected as virgins and maintained separately by sex in vials at a density of 24-30 individuals (males) or 12-15 individuals (females). Males and females were then aged 3-4 days before use in the behavioral assays. To more closely mimic natural situations in which most of the female flies will be mated, 3-5 males of a single standard genotype were added to the virgin females the day before the behavioral observation. There is potential that not all females will have mated, but this effect will be random across genotypes and will not create systematic differences in the data.

| Experiment setup
All assays were performed within 2 hr of dawn to avoid any unnecessary variation introduced by circadian rhythms, and replicates were conducted randomly under standardized conditions (25°C, 70% humidity). Each group of two males and one female was put into a single chamber using a paintbrush, for a total of 12 chambers per assay (each chamber is 2.54 cm, VWR cat. no. 89093-496 see (Signor et al., 2017)). The females were all of the standard y 1 w 1 genotypes, and male genotypes varied across replicates but not within them. Each chamber contained 5 ml of either standard grapefruit medium or medium in which 15% of the water has been replaced with ethanol.
To avoid introducing artifacts from the use of CO 2 for sedation, the flies were sedated at 4°C for 10 min. The flies were allowed to acclimate in the chambers for 10 min before commencing recording with PointGrey Grasshopper digital cameras. This is long enough for any initial startle responses to ethanol to have concluded and is standard for behavioral assays (Cho, Heberlein, & Wolf, 2004;Grosjean et al., 2011;Li, Fink, El-Kholy, & Roeder, 2015). FlyCapture (PointGrey) was used to properly set up the assays for filming, which was automated using VideoGrabber (http://code.google.com/p/ video-grabber). Flies were filmed for 10, 20, or 30 min for three replicates each of the ethanol-and nonethanol-exposed conditions, thus each genotype was filmed in 216 arenas. The different time intervals are intended for an analysis of the transcriptome of flies exposed to ethanol for different time periods, which will be reported in subsequent papers. Thus, for each genotype, three replicates of these 12 chambers were performed for each of three time periods in ethanol-and nonethanol-exposed conditions (12 chambers per replicate, three replicates, three time periods, six genotypes, and two conditions). Arenas with damaged flies or mating events were excluded from the final analysis.

| Automatic tracking
In brief, we used a background subtraction method for each frame of the video and a Gaussian mixture model to determine the exact position of each fly. Matching flies in consecutive frames created a representation of fly movement by recording the number of pixels traversed each second by each fly. Please see (Signor et al., 2017) for a more thorough description of the automatic tracking of fly movement.
The background subtraction method was modified from that in (Signor et al., 2017), and it is described in the Appendix S1.

| Validation of methods
To obtain a measure of error for our tracking algorithms, we randomly selected 360 20-s intervals from the videos and compared it with the recorded tracking data. We evaluated the frequency of errors using two metrics: (1) how often the tracking switched between two different males (i.e., the fly was incorrectly identified) and (2) how often the sex of the fly was incorrectly inferred.

| Movement
The phenotype for each arena was defined as the average movement in pixels per second for each 5-min interval. For each 5-min interval, the male phenotype was averaged between the two individuals, as they are of the same genotype. Thus, for each interval, there are two data points, one for females and one for the average of the two males.

| Analysis of movement rates and sexual dimorphism
Male and female movements were analyzed separately. We used a t-test to test for differences between activity levels between species.
We used measurements from 779 arenas and normalized the data using a logarithmic transformation. We accounted for repeated measures (the 5-min intervals for each arena), and fit a linear mixed-effects model. Random effects included genotype and interactions between genotype, time, and environment. The ID of individual arenas was nested within genotype as a random effect, but it was included with only an intercept term to reflect individual baseline movement but not change over time. Fixed effects were day, environment, time, and the interaction between environment and time. The day of the assay was included to control for any batch effects.
Model fit was evaluated using the lme (nlme) function in R (See Appendix S1 for the R commands). Significance of fixed effect variables was assessed using an F test with the anova (nlme) function.
Using the same function, significance of variables of random effect was assessed using a likelihood ratio test to compare model fits with and without the variable of interest. We used REML (restricted maximum likelihood) to fit the lme model parameters. Our fixed effects are common across models, and thus, REML is an appropriate test.

| Calculation of Ψ for different abiotic environments
We used the following formula to estimate Ψ: See Table 1 for a description of all of the terms. Ψ is the regression coefficient from this model, which we estimate for each abiotic environment with male phenotype as a fixed effect. It is possible that there is feedback between males and females as well; however, we could not evaluate this with our model. The model was fitted with lme (nlme) in R, and REML was used to fit the lme model parameters.

| Ψ j for individual genotypes
To estimate the effect each genotype has on female behavior, we fit a separate regression for each genotype for the effect of male movement on female movement. While Ψ has been reported in the literature for individual genotypes, male and female behaviors can covary for reasons other than Ψ (Bleakley & Brodie, 2009). That is, they can vary due to their shared environment, or due to Ψ, and the two factors will always be confounded. We have used extensive measures to control for environment, so if environmental covariation is affecting our estimates, it should be equivalent across experiments. This does not rule out the possibility that there are genetic differences in the sensitivity to shared environment. We calculate it here with the caveat that measures are confounded with shared environment and refer to Ψ as Ψ j . We used the following formula to estimate Ψ j : See Table 1 for a description of the variables. The model was fitted as previously described, using lme (nlme) with REML to fit the lme model parameters, accounting for day effect and repeated measures. To estimate Ψ j, genotype was included as a random effect, we calculated the slope of the male phenotype effect for each genotype (Appendix S1).

| RESULTS
In the following section, we will include a selection of results from Signor et al. (2017) so that we may explicitly compare between these sister species.

| Analysis of movement rates and sexual dimorphism
Overall male movement level in D. simulans was lower than in D. melanogaster, in both ethanol-and nonethanol-exposed conditions (Figure 1) (Difference in mean movement = 15.23 pixel/s, t 1148 = 41.232, p < 10 −4 ). However, the effect of genotype and ethanol was significant for male activity level in this species (Figure 1, Table 2) (male movement, genotype, χ 2 (df = 1) = 65.88, p < 10 −4 ; male movement, environment, F 1,705 = 11.89, p = .0006)). There was also a much larger decrease in movement in D. simulans in response to ethanol than in D. melanogaster ( Figure 1)   The same results as shown in (b) but for female flies. These females are all the same genotype, and thus, differences in movement will be largely due to male genotype Locomotion in D. melanogaster was sexually dimorphic, and we previously confirmed the results of other studies that males move approximately three times more than females (Long & Rice, 2007) ( Figure 1a-c). In each species, a single female genotype was used, so any overall variation in movement is due to the abiotic environment and to females' interaction partners. Sex and genotype are confounded because females are a different genotype than males. We tested the significance of the difference between males and females in D. simulans using a linear regression model and found that males moved slightly less than females. This is in contrast with D. melanogaster, where males move 2.7× as much as females (0.77× vs 2.7×) (F 1,1642 = 35.29, p < 10 −4 ).

| Calculation of Ψ for different abiotic environments
Here, we estimated Ψ in D. simulans and compared Ψ between D. melanogaster and D. simulans (Figure 2a,b). Male phenotype was defined, as before, as the average movement of a male in either ethanol-or nonethanol-exposed conditions. In D. simulans, we estimated Ψ = 0.13 in nonethanol environments and Ψ = 0.06 in ethanol environments. This is very similar to the estimates of Ψ in D. melanogaster Ψ (0.11 and 0.04, respectively) (Figure 2a,b). The authors note that in both cases a single female genotype was used, and it is possible that the inclusion of additional genotypes would alter this comparison.

| Ψ j for individual genotypes
To determine whether the Ψ is genotype-specific, and whether that varies with environment, we tested for variation in Ψ j (Tables 4 and 5, Appendix S1). With ethanol, we found significant differences between genotypes (genotype × male movement, LRT result for models with and without a genotype × male movement term: χ 2 (df = 1) = 4.95, p =.03). This effect is not significant in environments without ethanol (genotype × male movement, LRT results for models with and without a genotype × male movement term: χ 2 (df = 1) = .01, p = .91) (Tables 2 and 5). What is unique to D. simulans is that they did not vary in unexpected ways between environments, but rather all changed in a consistent manner. That is to say that measures of Ψ j were reduced uniformly across genotypes upon   (Table 4) (Signor et al., 2017).

| DISCUSSION
The role of IGEs and how they vary among environments and species is an important part of understanding the impact of social environments in evolution (Bailey & Zuk, 2012). Few studies have measured IGEs using Ψ for any trait, and this is the only study that compares Ψ There is a different relationship between male and female locomotions in D. melanogaster as compared to D. simulans. In D. melanogaster, males move 2.7× more than females, while in D. simulans, males move only 0.77× as much as females. This difference in sexual dimorphism could have implications for sexual selection in each system given that in D. melanogaster, selection on locomotion is sexually antagonistic (Long & Rice, 2007). Absence of sexual dimorphism does not necessarily indicate a lack of sexual conflict, although this has not been investigated in D. simulans. Locomotion in other members of this species group has not been well characterized, although there is some evidence that D. melanogaster is more active than its close relatives overall (Cobb, Connolly, & Burnet, 1987). D. simulans and D. melanogaster do have more divergent courtship behaviors than other members of their species group, likely because they are both cosmopolitan species that occupy many of the same substrates (Cobb, Burnet, & Connolly, 1986;Cobb, Connolly, & Burnet, 1985).
The evolution of locomotory behavior, and of sexual dimorphism in locomotory behavior, has not been well characterized in many systems. The best-studied systems are lizards and D. melanogaster, and in both, locomotion is an important component of sexual selection and fitness (Colomb & Brembs, 2014;Husak & Fox, 2008;Lailvaux et al., 2003;Perry, 1996;Peterson & Husak, 2006). In general, this manifests as males that move faster, but there are exceptions to this that also demonstrate the evolutionary lability of locomotory behavior.
For example, in D. suzukii, females move 4× more than males, and in one species of lizard movement was not sexually dimorphic despite more active males having higher reproductive success (Ferguson et al., 2015;Peterson & Husak, 2006).
One large difference between selection regimes for females of D. melanogaster and D. simulans is that in D. simulans, multiple matings are beneficial for female fecundity and neutral with respect to longevity, while in D. melanogaster, it is neutral to fecundity and deleterious to longevity (Chapman, Liddle, Kalb, & Wolfner, 1995;Kuijper et al., 2006;Taylor et al., 2008;Wigby & Chapman, 2004).
Furthermore, other metrics of sexual selection, such as the cost of female choice on longevity or fecundity, indicate that they are largely neutral in D. simulans but are costly in D. melanogaster (Friberg & Arnqvist, 2003;Pitnick, 1991;Pitnick & Garcia-Gonzalez, 2002;Taylor, Wedell, & Hosken, 2007;Taylor et al., 2008). While the present work does not measure variation in fitness and therefore cannot make any conclusions about this in relation to locomotion, it may be an interesting question for future research to consider that this may be a manifestation of different selective regimes in the two species.
For example, because multiple matings are beneficial to both females and males in D. simulans, selection toward a shared level of activity may be beneficial. In contrast, D. melanogaster has been selected for higher activity levels because of sexually antagonistic selection (Long & Rice, 2007). This would suggest a lack of sexually antagonistic selection for locomotion in D. simulans.
It is interesting that despite these differences in dimorphism and patterns of context-specific change in locomotion overall, estimated Ψ-values are approximately equal in D. melanogaster and D. simulans.
Attempting to estimate Ψ j requires confounding the effect of shared environment and the effect of different genes in the environment. As such it is an overestimation of the effect of individual genotypes, which is most likely why values of Ψ j are consistently higher than Ψ. However, there is no other way to attempt to calculate this term, and extensive environmental controls were used such that the effect of shared environment should have been constant across assays. This does not preclude the existence of genetic variability in the response to shared environment in males, and thus, it is possible that even the extensive environmental controls do not eliminate this effect. As such Ψ j is a valuable contribution to understanding the patterns in our data, however the authors stress that absolute values are not interpretable and it must be considered with caution.
In conclusion, we have explored differences in Ψ for locomotion between abiotic environments for two closely related species, D. simulans and D. melanogaster. We have found extensive evidence for context-specific Ψ that varies between species, with D. melanogaster having interaction terms with abiotic environment that are lacking in D. simulans. We have found that Ψ is positive in both abiotic environments for both species, and approximately equal within a given abiotic environment between species. We have also found evidence for differences in sexual dimorphism in locomotion despite similarities in Ψ between species. These findings present interesting pathways for future research into the evolution of sexual antagonism and sexual selection, and the role of IGEs in both of these processes.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
SS performed the experiments, wrote the manuscript, and created the figures. MA analyzed the data. PM and SN conceived of the experiments and contributed to their design.