Rearing design
Female E. binotata that live on Ptelea trifoliata (Rutaceae) mate throughout the summer until late August (Sullivan-Beckers & Cocroft, 2010). Females mate once (Sullivan-Beckers & Cocroft, 2010) and oviposit eggs beneath their hostplant’s bark for overwintering (Wood, 1980; Wood & Guttman, 1983). Nymphs emerge in the spring when water flow starts in the hostplants and go through five stages before their final molt to adulthood. In May 2019, we collected 1st -2nd instar nymphs from Capen Park (Lat: 38.924925, Long: -92.318294) in Columbia, Missouri, taking care to collect from multiple trees and branches to account for genetic variation in the population. We brought the nymphs to Saint Louis University (St. Louis, MO) and randomly assigned them to one of two developmental temperatures: 26°C or 21°C. These temperatures correspond to daily mean high and low temperatures at the collection locale during the insects’ developmental period (source: ncdc.noaa.gov/cdo-web).
We placed 40–50 nymphs on each of three replicate rearing plants per developmental treatment. We housed the plants in individual incubators set at the respective developmental temperatures, and watered plants weekly. We monitored temperature and relative humidity with data loggers (ECOWITT DS102). We mounted a 100W full spectrum LED plant grow light outside of each incubator set to a 14:10 L:D photoperiod (corresponding with the summer photoperiod); plants received ~ 3000lux at their closest point to the light – sufficient to promote plant growth. A 2.3-liter ultrasonic humidifier placed in each 26°C incubator allowed for us to adjust for differences in relative humidity across temperature treatments.
To control for sexual experience, we transferred newly molted adult males to sex-specific plants daily (Fowler-Finn et al., 2017). Once all individuals on a rearing plant molted to adulthood, we transferred all adult females onto a new plant of the same size as the male plant and changed the incubator temperature to 23.5°C. We gave every individual a unique color-coded ID using non-toxic acrylic paint (Apple Barrel®, multi surface, satin finish) applied to their pronotum. We started assaying thermal sensitivity of adult courtship behavior after the insects reached sexual maturity (one to two weeks after adult emergence for males, and two weeks later for females).
I. Thermal sensitivity of courtship behavior
To test how developmental temperature shaped the thermal sensitivity of courtship activity, we generated thermal courtship activity curves by taking advantage of the duetting system in E. binotata (Fig. 1). During pre-copulatory courtship, males travel from branch to branch producing advertisement signals to attract females; if a female prefers the male’s signal, they may respond with their own sex-specific courtship signal, which initiates a duet that facilitates pair formation (Cocroft et al., 2008; Rodríguez & Cocroft, 2006). For each individual male and female, we tested for courtship activity across a range of temperatures extending somewhat beyond the typical range of thermal variation in the field (Jocson et al., 2019) (18–45°C at 3°C intervals). To do so, we first acclimated an individual at the testing temperature for at least 20 minutes (Greenfield & Medlock, 2007; Jocson et al., 2019),then placed the individual on a potted P. trifoliata plant within the testing incubator and allowed them two minutes to settle. Next, we played back sex-specific primers (see below) and assayed whether or not an individual responded with their own courtship response signal (for detailed methodology on vibrational playback and recording methods see supplemental and Fowler-Finn et al., 2017; Jocson et al., 2019; Macchiano et al., 2019).
For males, we played a primer of a male-female duet every two minutes until ten minutes had passed, or the male produced two separate signal bouts (Fowler-Finn et al., 2017; Jocson et al., 2019; Macchiano et al., 2019). Males that produced any signal during testing were marked as actively courting (Macchiano et al., 2019). For females, we played three primers: the first of a male signaling at the specific testing temperature, the second of a male signaling at three degrees lower, and third a male signaling at three degrees higher (exceptions for 21°C—no 18°C primer, and 45°C—no 48°C primer). These signals varied in dominant signal frequency, so this approach allows us to remove any confound of individual differences in preference in testing for courtship (e.g., Fowler-Finn et al., 2017; Jocson et al., 2019; Macchiano et al., 2019). If females did not respond to the first set of primers, we waited one minute and played the primers in randomized order a second time. If a female responded to any primer, they were marked as actively courting.
Our goal was to test each individual at all 10 temperatures, without repeating a temperature for any individual. Individuals that were tested at five or more temperatures were included in the subsequent analysis, as five data points is sufficient for reconstructing the unimodal curve describing changes in courtship activity across temperatures (Sasson et al., JEB. in press). For males, we ran 394 trials across 46 males reared at 26°C and 224 trials across 24 males reared at 21°C; including males that were tested at five or more temperatures amounted to an n = 40 for 26°C males, and n = 24 for 21°C males. For females, we ran 298 trials across 35 females reared at 26°C and 259 across 32 females reared at 21°C; including females that were tested at five or more temperatures amounted to an n = 30 for 21°C females, and n = 30 for 26°C females.
Statistical analyses for the thermal sensitivity of courtship behavior – To test for changes in the thermal of courtship activity across developmental temperatures, we constructed generalized linear mixed effects models with a binomial response of courtship yes/no. We constructed a global model including the following terms: sex, developmental temperature (21°C or 26°C), testing temperature (describes a linear response to testing temperature), temperature × temperature (this quadratic term describes whether there is a parabolic response to testing temperature, and is included because many thermally-sensitive traits are unimodal; Huey & Stevenson, 1979), days-since-temperature change (this term—the time between changing the incubator temperature to 23.5C and testing individuals—captures variation due to both age and possible acclimation to the intermediate temperature they were held as adults), as well as two- and three-way interactions between the terms. The interaction terms allowed us to test changes in thermal sensitivity due to developmental plasticity and sex-specific responses. For example, a significant developmental temperature × quadratic temperature would indicate differences in the shape of the thermal courtship activity curve across developmental treatments. Linear and quadratic temperature terms, and days-since-temperature change values were Z-transformed to make parameter estimates more comparable and improve model performance (Schielzeth, 2010). We included rearing plant replicate number and individual ID (nested within plant replicate ID) as random effects. We tested the significance of terms by performing likelihood ratio tests of models with versus without each effect, analyzed in R using the package lme4 (function: glmer(); Bates et al., 2015).
We detected a significant interaction between developmental temperature, testing temperature, and sex in the global model (x2 = 23.545, p = 1.2e− 6; Supplemental Table 1), which indicates that the interaction between developmental and ambient temperature differs between the sexes. To identify the specific patterns of sex-specificity, we constructed models split by sex and then split by developmental temperature (Supplemental Tables 2–3). For the models split by sex, a significant developmental temperature × testing temperature term would indicate that developmental temperature alters the thermal sensitivity of courtship signaling. We also tested whether time since emergence to adulthood affected courtship activity within each developmental treatment and for each sex by testing the significance the days-since-temperature change term and its interactions with developmental temperature and linear and quadratic temperature terms. A significant days-since-temperature change × quadratic temperature indicates changes in the shape of the courtship activity curve over the number of days we tested them. This may be interpreted as the effect of developmental temperature diminishing over time, or age affecting signaling behavior. Finally, in the models split by developmental temperature, we can identify if thermal sensitivities differ between sexes within developmental treatments by including the sex × quadratic temperature term, which would indicate differences in the shapes of the sex’s thermal courtship activity curves.
We visualized the fitted response from the global model for each sex in each developmental treatments using the r package visreg (2.7.0.1) (Breheny & Burchett, 2017) (Fig. 2).
II. Thermal sensitivity of mating
Following the conclusion of the courtship activity trials, we conducted mating trials between males and females from the same treatments. We used males and females up to four times, but retired females after a successful mating. We prioritized pairing males and females from different replicate plants when possible (all except two trials per developmental treatment). Our final sample size was 35 mating trials from the 26°C treatment, and 25 from the 21°C treatment. Following Leith et al. (2020), we acclimated individuals to the testing temperature for at least 20 minutes before first placing the female on a testing plant standardized for size, giving her five minutes to settle and then introducing the male ~ 10cm below the female on the plant. We monitored each pair for four hours or until copulation ended, recording the start and end time of copulation; it is unlikely for pairs to initiate mating in the lab after four hours (Leith et al., 2020). If either individual flew off the plant, they were placed back on. We ran mating trials at a subset of the temperatures for the courtship activity trials. These temperatures were selected to cover most of the range (i.e. 18, 24, 30, 36, 42°C) while prioritizing testing a higher number of pairs at each temperature.
Statistical analyses for the thermal sensitivity of mating – To analyze the effects of developmental and ambient temperature on mating rates, we constructed a generalized linear mixed effects model with the binomial response of mating yes/no. The fixed effects were developmental temperature, linear and quadratic testing temperature, and their two-way interactions; we included male and female ID as random effects. The number of times the male and female in each pair had previously been tested were included as covariates, but neither effect was significant and these terms were removed from the final model. We tested the significance of terms by performing likelihood ratio tests of models with, versus without each effect in R using the package lme4 (function: glmer(); Bates et al., 2015). We visualized the fitted response of this model using visreg (Breheny & Burchett, 2017) (Fig. 3, Supplemental Table 4).
To assess whether thermal mating curves differed in shape from male or female thermal courtship activity curves, we performed pairwise Tukey post hoc tests comparing the quadratic trends between male, female, and courtship pair activity across developmental treatments (Supplemental Table 5).
III. Thermal sensitivity of mating signals and mate preferences
For males that were responsive in the courtship activity trials, we analyzed signals from the recorded WAV files (see Vibrational Playback and Recording Set up). We measured the frequency of a standardized landmark signal: the third signal of the second bout (following: e.g., Cocroft et al., 2010; Fowler-Finn et al., 2014, 2017; Jocson et al., 2019; Rodríguez et al., 2012; Sattman & Cocroft, 2003). Frequency functions in Enchenopa binotata are linear, so we included males that responded at a minimum of three temperatures in our analyses; for males reared at 21°C, this included 90 data points across 21 individuals, and for males reared at 26°C, this included 225 data points across 39 individuals.
Females that were responsive in the courtship activity trials were tested for their preference for fundamental signal frequency—the most important signal trait for mate selection in this system (Rodríguez et al., 2006; Rodríguez & Cocroft, 2006)—using a function-valued approach (Kilmer et al., 2017; Leith et al., 2021; Stinchcombe & Kirkpatrick, 2012). Within a minute after a female exhibited responsiveness in the courtship activity trials, we began playing a series of 16 synthesized male signal bouts (6 signals per bout, reflecting the population mean) that varied in frequency. We played the signal bouts in randomized order using a custom Matlab script (v.8.3, 2014). This playback design allows females to respond to each frequency up to six times and enables us to quantify their preference for frequency at each of the testing temperatures (Fowler-Finn et al., 2015). Using the raw data, we generated cubic splines for each temperature a female was tested at using the program Pfunc (Kilmer et al., 2017). We selected a standardized smoothing value of 0.01 to minimize wobbly and noisy functions (following Kilmer et al., 2017). From these splines, we extracted peak preference (stimulus frequency where females are most responsive), which was our primary interest and our response variable in our models described below. As with males, to assess the effect of developmental temperature on female signal preference we included females that had preferences quantified at three or more testing temperatures; for 21°C females this included 62 data points across 14 individuals, and for 26°C females 81 data points across 17 individuals.
We additionally tested for the effects of developmental and ambient temperatures on female mate preference selectivity, with selectivity indicating how much a female may reduce responses to frequencies that deviate from their preferred signal frequency. We quantified selectivity by performing a principal components analysis (R package 'psych', Revelle, 2022) on the correlation matrix of the following traits also extracted from individual cubic splines: peak height (the maximum elevation of the function on the y-axis), tolerance (the width of the function at a given distance from the peak), strength (how much the preference punishes deviation from the peak), and responsiveness (the mean of the y-axis values along the curve) (Fowler-Finn & Rodríguez, 2012; Kilmer et al., 2017). One principal component had an eigenvalue greater than 1 and explained 74% of the variation in these four traits. We used the scores of this principal component to represent female selectivity, with higher scores indicating increased selectivity (Supplemental Table 9). For these selectivity analyses, we included data for all females that had preferences obtained (not just those tested at 3 or more temperatures) including an n = 23 for 21°C females, and n = 26 for 26°C females.
Statistical Analyses for thermal sensitivity of mating signals and mate preferences – To test for the effects of developmental temperature on the frequency of male signals and female preferences, we constructed linear mixed effects models fit with maximum likelihood with lme4 (Bates et al., 2015). First, we constructed a global model with the response as “Frequency” (for males this was the peak frequency of their signal at the testing temperature; for females, this was peak preference at the given testing temperature). The fixed effects were: sex, developmental temperature, linear and quadratic testing temperature, days since temperature change, and their two-, three-way interactions. We included rearing plant replicate and individual ID (nested within replicate) as random effects, and Z-transformed linear and quadratic temperature and days since temperature change terms to account for data scaling (Schielzeth 2010). We detected significant interactions between developmental and linear testing temperature (x2 = 4.618, p = 0.032, Supplemental Table 6), as well as sex and linear testing temperature (x2 = 4.667, p = 0.031, Supplemental Table 6) in the global model. To determine the source of these significant factors, we constructed models split by sex and developmental temperature as we did for our courtship activity analyses (Supplemental Tables 7–8). In these subsequent models, we were primarily interested in the linear temperature × developmental treatment or linear temperature × sex, respectively, which could describe changes in signals or preferences due to developmental temperature, and potential uncoupling of male signals and female preferences.
To test for differences in female selectivity across developmental and testing temperatures, we constructed linear mixed effects models with the principal component of female selectivity as the response. We first constructed a global model with the fixed effects: developmental temperature, linear and quadratic temperature, days since temperature change, and their two-, three-way interactions, and individual ID as a random effect. Because we found a significant three-way interaction between developmental temperature, days since temperature change, and quadratic temperature, we proceeded to construct two similar models parsing females by developmental treatments with the same fixed and random effects (except developmental temperature and its interactions) (Supplemental Table 10). We conducted analyses in R, using the package lme4 (function: lmer; Bates et al., 2015). We tested the significance of terms in all mate preference and selectivity models by performing likelihood ratio tests of models with versus without each effect. We performed likelihood ratio tests with the rand() function to test for significance of random effects (‘lmerTest’; Kuznetsova et al., 2017)