Fitness assays: quantifying male harm

To study whether male harm is affected by temperature, we established a factorial design to measure survival and lifetime reproduction success (LRS) of female flies under monogamy (i.e., one male and one female per vial) vs. polyandry (i.e., three males and one female per vial), across three stable temperature treatments typical of this population during their reproductively active period in the wild: 20, 24, and 28°C. Comparison of female fitness at monogamy vs. polyandry is a common way to gauge male harm in Drosophila and other organisms (Yun, Agrawal, & Rundle, 2021), and our treatments reflect the low and high-end of the spectrum of sex ratios that are typical of D. melanogaster at mating patches in the wild (Dukas, 2020).

We first randomly divided virgin flies into three groups that we allocated to the three different stable temperature treatments 48 hours before starting the experiment. Flies remained at those temperatures until the end of the experiment. To estimate LRS, we transferred flies to fresh vials twice a week using mild CO2 exposure anaesthesia. We incubated the vials containing female eggs at 24 ± 4°C for 15–20 days (~15 days for 28°C, ~17 days for 24°C and ~20 days for 20°C) to allow F1 offspring emergence, after which we froze them at -21°C for later counting. The differences in incubation time are due to differences in developmental time because of temperature differences in the first 3-4 days of each vial (i.e., before flipping females to new fresh vials). We discarded and replaced males by young (2 to 4 days old) virgin males three weeks after starting (at the same time for all treatments). We kept a stock of replacement males maintained at each of the three temperatures (20, 24 and 28°C) to replace dead male flies if needed. We kept flies under these conditions during six weeks, after which we discarded males and followed females until they died. We started the experiment with 468 females (78 per each temperature x mating system treatment) and 936 males (234 per each temperature x polyandry and 78 per each temperature x monogamy). Final sample sizes were: a) at 20°C: 74 females for polyandry and 76 females for monogamy; b) at 24°C: 72 females for polyandry and 77 females for monogamy; c) at 28°C: 70 females for polyandry and 75 females for monogamy. Differences between start and final sample sizes across treatments are due to discarded/escaped flies since the start of the experiment. We estimated the overall degree of male harm by calculating the relative harm (H) following Yun et al. (2021):

H = (W_monogamy - W_polyandry) / W_monogamy

Finally, and using the data collected above, we partitioned overall effects on LRS into effects of temperature x mating system (i.e., male-male competition level) on early reproductive rate (i.e., offspring produced during the first two weeks of age) and reproductive (i.e., offspring produced over weeks 1–2 vs. 3–4) and actuarial senescence (i.e., survival).

Behavioural assays: quantifying male-male aggression and harassment of females

Immediately after the fitness experiment started, we conducted behavioural observations on the first day of the experiment across all treatments, to investigate the behavioural mechanisms that might underlie the potential fitness effects evaluated above. Due to logistic limitations, we conducted behavioural observations in the same temperature control room, so we had to conduct trials at 20, 24 and 28°C in three consecutive days (with both monogamy and polyandry treatments evaluated at the same time for each temperature), in randomized order (i.e., 20, 28 and 24°C). However, all flies were 5 days old at the start of the experiment. We measured the following behaviours: (a) courtship intensity (number of courtships experienced by a female per hour), (b) male-male aggression behaviours and (c) female rejection behaviours (Bastock & Manning, 1955; Connolly & Cook, 1973). We also recorded the number of total matings during the observation period.

Observations started at lights‐on (10 a.m.) and lasted for 8 hr, during which time we continuously recorded reproductive behaviours using scan sampling of vials. Each complete scan lasted approximately 8 min, so that we always conducted one complete scan every 10 min to ensure recording of all matings (see below). Scans consisted in observing all vials in succession for ca. 3s each and recording all occurrences of the behaviours listed above. We interspersed these behavioural scans with very quick (<1 min) mating scans where we rapidly swept all vials for copulas at the beginning, in the middle and at the end of each behavioural scan. This strategy ensured that we recorded all successful matings (>10min), which typically last between 15 and 25min in our population of D. melanogaster, during our 8‐hr observation. We obtained a total of 49 scans per vial. Behavioural observations were conducted only once, on day 1 of the fitness experiment, as prior experiments have shown that courtship, aggressive and female rejection behaviours are stable over time in D. melanogaster, so that our behavioural indexes are representative of long-term treatment differences (e.g., Carazo, Perry, Johnson, Pizzari, & Wigby, 2015; Carazo, Tan, Allen, Wigby, & Pizzari, 2014). In contrast to courtship and aggression indexes, note that total mating frequency over the first day cannot be taken as a reliable measure of mating rate (Wolfner, 1997), and thus our rationale in recording this variable was just to ensure that early mating ensued normally across treatments (which was the case, see SI2).

Female reproduction and survival assays: quantifying male ejaculate “toxicity”

To examine post-mating mechanisms that might underlie the fitness effects observed in our first experiment, we conducted four additional assays to test whether temperature modulates the previously well-documented effects of male ejaculates on female reproduction and survival in D. melanogaster. Briefly, males of this species transfer seminal fluid proteins (SFPs) produced by their accessory glands that increase short-term female fecundity, as well as decrease female receptivity and survival (Chapman, Liddle, Kalb, Wolfner, & Partridge, 1995; Wigby & Chapman, 2005). In addition, prior studies have shown that males are able to tailor investment into SFPs according to the expected sperm competition risk and intensity (Hopkins et al., 2019). Thus, we set up a factorial design where we manipulated the temperature (i.e., 20, 24 and 28ºC) and perceived sperm competition risk levels (i.e., males kept alone vs. with 7 more males in a vial) at which adult males were kept prior to mating, for two temperature treatment durations (i.e., 48 hours or 13 days), and then measured how reception of their ejaculate in a common garden environment (i.e., 24ºC) affected female fecundity, survival and reproduction.

Receptivity assays

We first collected experimental males as virgins (i.e., within 6 h of eclosion) under ice anesthesia and randomly placed them either individually or in a same-sex group of 8 in medium containing plastic vials. Next, we randomly divided them into three groups that we allocated to the different stable temperature treatments for either 48 hours (experiment 1) or 13 days (experiment 2) immediately before the beginning of each experiment. In experiment 2 (treatment duration of 13 days), we emptied the seminal fluid of experimental males before we allocated them to the different temperature / competition treatments. To do so, we housed individually experimental males with four standard virgin females for 24 hours. This strategy ensured that spermatogenesis took place over the 13 days under the temperature / competition treatments.

We collected all females and competitor males used in receptivity assays as virgins and held them in groups of 15 to 20 flies at 24 ± 4°C. Experiments started by exposing all virgin females to single experimental males for 2.5 h at 24°C. After a successful copulation, we separated the mated females from the males and kept them individually in medium-containing vials until the next mating trial. We discarded unmated females and experimental males. 72 h after the first mating, we individually exposed females to single virgin competitor males for 12 h. After each trial, we transferred unmated females into a new medium containing vial, until the next mating assay 24 h later. We repeated remating trials every 24 h for three consecutive days. We calculated the cumulative percentage of remated females for each of the three days of each experiment. We conducted the experiments in two blocks each: with n = 390 females for each batch in experiment 1 (n = 436 rematings) and n = 420 females for each batch in experiment 2 (n = 676 rematings). We also recorded mating duration for the first mating and mating latency (i.e., time between males being introduced into the female-containing vial and copulation) and mating duration for re-matings. Females that did not remate within those three days were right-censored. Females and all experimental males were 4 days old at the start of experiment 1. In experiment 2, females and “re-mating” competitor males were 4 days old, while experimental males were 18 days old.

Fecundity and survival assays

To gauge effects on female short-term fecundity and long-term survival, we performed two experiments (experiments 3 and 4) where we compared the oviposition and egg fertility of females mated with male flies subject to the same factorial design imposed in receptivity experiments. We collected and treated all experimental males as in the receptivity assays described above, and then proceeded to mate ~4-day-old virgin females in single pairs to either 4- (experiment 3, 48h temperature treatment duration) or 18-day-old experimental males (experiment 4, 13d temperature treatment duration) for 2.5h at 24°C. After copulation, we separated mated females from males and kept them individually in medium-containing vials. We discarded unmated females. We then transferred females to fresh vials every 24h for 4 days, and then every 3 days twice. Later, we transferred female flies to fresh vials once a week, and we combined vials to maintain density at ten flies per vial. We kept female flies under these conditions until they died. We removed dead flies at each transfer and recorded female deaths. We counted eggs laid the first 3 days and vials from days 1, 2, 3, 4, 5, and 8 after mating were retained to count progeny to determine egg viability. We tested 545 females at the starting point of the experiment 3, and 480 females at the starting point of the experiment 4.

Statistical analyses

We performed all statistical analyses using R statistical software (version 3.5.2). In all models, we assessed fit by visual inspections of diagnostic plots on raw and residual data (Zuur, Ieno, & Elphick, 2010). To examine temperature effects on male harm, we evaluated the interaction between mating system and temperature on female fitness (LRS), survival and male and female reproductive behaviours. We fitted generalized linear models (GLMs) with temperature, mating system and their interaction as fixed effects. We also used Cox proportional hazards survival model to analyse potential differences in mortality risk across treatments, using the survival and survminer packages, including the females lost during manipulations as “right censored” individuals (i.e., individuals that are taken into account for demographic analysis until the day they disappear, Kleinbaum & Klein, 2012). We also censored any remaining females after the six weeks of the experiment. Graphical inspection of LRS and reproductive behaviours (courtship intensity, intersexual aggression behaviours and female rejection) revealed that the normality assumption was apparently violated, as well as the independence assumption for LRS. Box–Cox transformation (Quinn & Keough, 2002) solved these problems and allowed us to run a GLM with a Gaussian error distribution. In all these afore mentioned cases, fitted models were subsequently validated. In addition, we compared GLMs with their corresponding null GLMs using likelihood ratio test. In all models, we used ANOVA type III to compute p-values. We detected a problem of collinearity between mating system and the interaction in the LRS model. We fitted the model again without the main mating system effect (which was not our main interest) and also ran models separately for each temperature level. As a complementary analysis, we ran a model with temperature as factor with a quadratic contrast table predetermined (note that the relation between LRS and temperature does not seem to be linear, Fig 1), obtaining similar results. For the total matings recorded across the 8h observation period (i.e., mating rate), we initially used a GLM with Poisson errors. However, the dataset contained many zero values. We therefore analysed the data using a Hurdle model, in which the zero values are modelled separately from the nonzero values (Zuur et al., 2010). Additionally, we compared GLM and Hurdle models using an information-theoretic approach based on Akaike’s Information Criterion (AIC). The minimum AICc value indicates the best-supported model given the trade-off between fit to the data and model complexity (Konishi & Kitagawa, 2008). To estimate how male harm impact estimated population growth across varying demographic scenarios (i.e., decreasing, stable and growing populations), based on our LRS measures, we calculated rate-sensitive fitness estimates (Edward, Fricke, Gerrard, & Chapman, 2011) against different population background growth rates of r = -0.1, r = -0.05, r = 0, r = 0.05 and r = 0.1.

For receptivity (mating duration and remating latency) and fecundity (total adults) analyses, we fitted generalized linear models (GLMs) with temperature, sperm competition risk level, treatment duration and their interaction as fixed effects. For egg production data, we fitted a generalized linear mixed model (GLMM) with temperature, sperm competition risk level, treatment duration and their interaction as fixed effects and day number as a random effect, with a zero inflated distribution, in which the zero values are modelled separately from the nonzero values (Zuur et al., 2010). We assessed significance of factors by dropping individual terms from the full model using the “drop1” function, refitting where the interaction was nonsignificant. We also used Cox proportional hazards survival model to analyse potential differences in mortality risk across treatments, using the survival and survminer packages, including the females lost during manipulations as “right censored” individuals. Finally, we also ran models separately for each temperature level for the cases in which a significant interaction between sperm competition risk level and temperature treatment was detected.       

Data files included are .xlsx.

We performed all statistical analyses using R statistical software.