Local population density and group composition influence the signal-preference relationship in Enchenopa treehoppers (Hemiptera: Membracidae)

Introduction

Social and sexual selection often produce rapid evolution and divergence – as well as extravagant elaboration of traits like sexual ornaments, weapons and mate preferences – in part because social and sexual competition are highly dynamic (West-Eberhard, 1983, 2014; Wolf et al., 2007; Lyon & Montgomerie, 2012). Selection due to competition for mates, for instance, favours ornament variants that provide an advantage in a given social environment. However, as sexual ornaments and preferences evolve, the selective (social) environment also changes, so that different ornament and preference variants may be favoured, altering the selective context yet again (West-Eberhard, 1983, 2014). These dynamics create ongoing evolutionary feedback loops as the cause of selection – the social environment – coevolves with the targets of selection – the ornaments and preferences (West-Eberhard, 1983, 2014).

Another reason for the high dynamism of evolution by social and sexual selection is that it is influenced by the outcome of behavioural interactions, and behaviours are highly adjustable and plastic (West-Eberhard, 2003; Foster, 2013; Snell-Rood, 2013; Zuk et al., 2014). For example, the outcome of competition for a mate will depend upon interactions among competing males, as well as between males and the female. The nature of these interactions depends upon not only ornaments and preferences, but also how the individual behaviours dictating the interactions have been shaped by prior social interactions. This phenomenon – behavioural plasticity that arises from variation in experience with the social environment (henceforth, social plasticity) – generates another level of feedback loops between the causes and targets of selection. Furthermore, plasticity-generated feedback loops interact with the above evolutionary feedbacks to influence the evolution of ornaments and preferences (Hebets & Sullivan-Beckers, 2010; Fowler-Finn & Rodríguez, 2012a, b; Verzijden et al., 2012; Rodríguez et al., 2013c; Rebar & Rodríguez, 2016).

Social plasticity in signals and preferences is widespread, and the patterns of plasticity expressed are diverse – in terms of both the social factors and cues involved, and how individuals respond to those cues (Hebets & Sullivan-Beckers, 2010; Rodríguez et al., 2013c). Signalling behaviour has been shown to vary in numerous ways in response to factors as diverse as experience with and/or feedback from females (Dewinter & Rollenhagen, 1993; Patricelli et al., 2002; Dukas, 2004, 2008; Wong & Svensson, 2009; Svensson et al., 2010; Lehtonen et al., 2011; Sullivan-Beckers & Hebets, 2011; Rodríguez et al., 2012; Mulrey et al., 2015); neighbours and competitors (Slater, 1989; Brenowitz & Beecher, 2005; Rebar et al., 2015; Rebar & Rodríguez, 2016); and population-level factors like density and sex ratio (French & Cade, 1989; Jirotkul, 1999; de Jong et al., 2009; Wong & Svensson, 2009; Kasumovic et al., 2011). Mate preferences, too, have been shown to vary in a diversity of ways with experience with potential mates or mating signals (Miller & Fincke, 1999; Wagner et al., 2001; Hebets, 2003; Dukas, 2005; Hebets & Vink, 2007; Bailey & Zuk, 2008, 2009; Fowler-Finn & Rodríguez, 2012a, b; Kozak et al., 2013; Bailey & Macleod, 2014; Rebar et al., 2015; Stoffer & Uetz, 2015b); observations of mating interactions (Godin et al., 2005; Dugatkin & Druen, 2007; Vukomanovic & Rodd, 2007; Tramm & Servedio, 2008; Frommen et al., 2009; Fowler-Finn et al., 2015b; Whitte et al., 2015); the composition of and interactions with neighbours (Albert, 2005; Verzijden & ten Cate, 2007; Verzijden et al., 2007; Tramm & Servedio, 2008; Kozak & Boughman, 2009; Servedio et al., 2009; Kozak et al., 2011), including the genetic identity of neighbours (Indirect Genetic Effects: Rebar & Rodríguez, 2013); and perceived or realized local density (Berglund, 1995; Tinghitella et al., 2013, 2015; Atwell & Wagner, 2014; Tinghitella, 2014).

This diversity of patterns and causes of social plasticity means that the selective dynamics that arise from social plasticity-generated feedback loops can influence evolution in a variety of ways. In some cases, social neighbourhood effects contribute to surprisingly concordant changes in male signals and mate preferences (Rebar & Rodríguez, 2015). In other cases, a single type of social variable (e.g. experience of mate availability and variability) can result in one kind of change in advertisement signals and a mismatched change in preference (Fowler-Finn & Rodríguez, 2012a, b; Rebar & Rodríguez, 2016). Depending on the emergent patterns, social plasticity can impact a range of important evolutionary processes – the maintenance of variation, divergence among populations or even potentially the colonization of novel environments (Bailey & Moore, 2012; Fowler-Finn & Rodríguez, 2012a; Rodríguez et al., 2013c; Rebar & Rodríguez, 2016). Furthermore, unless male signals and female preferences change in the same way and degree to the same social factors, the nature of sexual selection will necessarily vary across social environments. The result can be marked changes in selection on male signals among or even within populations with social substructure.

Testing hypotheses about the influence of social plasticity on selection requires testing the relationship of male signal distributions relative to female preferences across environments. We use the signal–preference relationship – the comparison of the distribution of signal traits with the shape of mate preference functions – as a powerful way of predicting the action of sexual selection due to mate choice (Ritchie, 1996; Rodríguez et al., 2006, 2013a; Sullivan-Beckers & Cocroft, 2010).

Here we test the hypothesis that social plasticity arising from variation in local population density will influence the form of sexual selection. We focus on population density because it is a key determinant of mating systems and the strength of competition for mates (Emlen & Oring, 1977; Andersson, 1994; Shuster & Wade, 2003). Encounter rates with conspecifics can influence the benefits of signalling and mate acquisition behaviours (Kokko & Rankin, 2006), and density can change the intensity of selection on traits (McLain, 1992).

We assess changes in the signal–preference relationship across local densities in a member of the Enchenopa binotata species complex of treehoppers (Hemiptera: Membracidae) — plant-feeding insects in which males produce signals that vary in frequency (Hz), and females favour some frequencies over others. The hypothesis that social plasticity in response to density alters sexual selection makes the following predictions: (i) density will influence the expression of signal frequency and/or female mate preference frequency; and (ii) the signal–preference relationship will changes across densities because density influences signal frequency and preference frequency in different ways (Fig. 1a–c). Support for these predictions could take various forms: social plasticity might be absent in signals and present in preferences; present in signals but absent in preferences; or present in both but in different directions (Fig. 1a–c). Conversely, if social plasticity in response to density does not alter sexual selection, there should be no change in the signal–preference relationship across densities. The latter outcome could arise if the patterns of plasticity in signals and preferences match one another across densities, or if neither signals nor preferences vary across densities (Fig. 1d–e).

Enchenopa provide excellent opportunities to study the influence of social plasticity on sexual selection. As plant-feeding insects, they spend their entire lifecycle on their host plant, and form aggregations of varying sizes that put them in close contact with conspecifics during juvenile and adult life stages (Cocroft et al., 2008). Groups of nymphs range from a few individuals up to hundreds of individuals (Fig. 2; Cocroft et al., 2008). Adults sometimes fly among stems or plants, but they also aggregate to varying degrees during the mating and egg-laying seasons (Cocroft et al., 2008). Local group size and composition can dramatically vary spatially and temporally within and across populations, as well as across generations (Cocroft et al., 2008). Thus, plasticity in response to social groupings could generate spatial and temporal variation in selection within and across populations. Furthermore, there is now a large body of research demonstrating that Enchenopa binotata are sensitive to various social factors. Females vary mate preferences across different social groupings (Rebar & Rodríguez, 2013, 2015), and in response to experience with different types of male signals as adults (Fowler-Finn & Rodríguez, 2012a, b); males adjust signalling effort with experience with competitors and competitor male–female duets (Rebar & Rodríguez, 2016), as well as in response to immediate feedback from female response cues (Rodríguez et al., 2012).

Enchenopa binotata communicate with vibrational signals that travel through plant stems in the form bending waves (Cocroft & Rodriguez, 2005; Cocroft et al., 2008). Males fly from stem to stem producing advertisement signals when they land; females respond to signals they find attractive, establishing male–female duets that facilitate pair formation (Rodríguez et al., 2004, 2006; Rodríguez & Cocroft, 2006; Cocroft et al., 2008). Female responses to male signals provide a good indication of whether they find specific male signals attractive, and are easily analysed to extract underlying mate preference functions (Rodríguez et al., 2004, 2006, 2012; Rodríguez & Cocroft, 2006; Cocroft et al., 2008, 2010; Fowler-Finn & Rodríguez, 2013). Females show the strongest preference for male signal frequency, and this is the trait that varies most among species in the complex (Rodríguez et al., 2006; Cocroft et al., 2008, 2010).

In our experiment, we manipulated local population density by rearing Enchenopa treehoppers from early instar nymphs through adulthood on potted exemplars of their host plant. After sexual maturity, we described the resulting variation in male advertisement signals and female mate preferences. We compared patterns of plasticity in signals and mate preferences – with a focus on signal frequency and female preference for signal frequency – to test for changes in the signal–preference relationship across social environments.

Materials and methods

Quantifying variation in male signals

Male E. binotata produce advertisement signals when placed on a plant in the laboratory, much like they do when they land on a plant in the field (Cocroft et al., 2008). At ~6 weeks post-adult moult, we placed each male individually on a recording plant, played a primer consisting of a recording of a live male and female duet and allowed 10 min for him to signal. All individuals received the same primer, and only heard it once, to minimize any effect of exposure to the primer signals on male behaviour. After recording, males that signalled were not returned to their rearing plant in order to reduce the likelihood that females on the plant would mate and thus become sexually unreceptive and unresponsive to playbacks (Cocroft et al., 2008). If a male did not signal by the end of ten minutes on a given testing day, he was individually marked with nontoxic paint and placed back on his rearing plant for later testing (once per week for up to three attempts; males that did not signal by the third attempt, ~2 weeks after the first one, were removed from the rearing plant). We tested males in randomized fashion such that removal of males was similar across treatments and replicate plants; we removed all males from all plants within 15–19 days of when recording started. The numbers of males per replicate plant that signalled were (mean ± SE) 3.2 ± 0.6 in low density, 6.8 ± 1.0 in medium density and 11.3 ± 2.5 in high density.

We recorded signals using a laser Doppler vibrometer (Polytec CLV 2534; Polytec Inc., Auburn, MA, USA), which allows recording substrate-borne acoustic signals without contacting the substrate and thus avoids distorting the signal. The laser output was sent to an Apple desktop computer and recorded and analysed using the sound analysis program AUDACITY (v. 1.2.4; http://audacity.sourceforge.net/). We isolated our recording set up from building vibrations by placing the recording plant on a rubber pad on top of a large ~135 kg iron plank isolated from building vibrations by floating the plank on top of a table using partially inflated inner tubes. The table was further isolated from building vibrations using rubber pads under the table legs.

Males produce advertisement signals in bouts, and the features of the signals vary slightly along those bouts (Cocroft et al., 2010). To select the signal from which we took measurements, we used a landmark signal position along bouts: we measured the 3rd signal of the 2nd signalling bout, or the closest to this landmark as possible. From these landmark signals, we obtained the following measurements: dominant frequency, whine length, number of pulses, pulse rate and intersignal interval. Other signalling behaviours we measured included the time elapsed between the primer and first signal produced, the number of signal bouts and the number of signals produced in the first bout (Fig. 3).

Signal traits covary within and among species of E. binotata (Cocroft et al., 2010), and may therefore not offer independent data for our analyses. Using a PCA, we generated noncorrelated traits for analysis: the first three male signal PCs had Eigen vectors greater than 1.0, with signal frequency weighted heavily on male signal PC1 (Table 2).

Table 2. Variation in male Enchenopa signals in our experiment, summarized by the principal components with eigenvalues > 1.0 and the contributions of each signal trait to the PC. PC1, PC2 and PC3 explained 26.5%, 18.0% and 13.3% respectively

Signal trait	PC1 Eigenvalue = 2.12	PC2 Eigenvalue = 1.44	PC3 Eigenvalue = 1.06
Dominant signal frequency	−0.45	−0.09	−0.44
Whine length	0.46	−0.02	0.49
No. pulses/signal	−0.24	0.54	0.28
Pulse rate	−0.41	0.37	0.22
Intersignal interval	0.27	−0.16	−0.40
First signal	0.38	0.41	−0.40
No. bouts	−0.14	−0.58	0.32
Signals/first bout	0.34	0.17	0.13

Quantifying variation in female preferences

We assayed female mate preferences when females became sexually receptive ~7 weeks post-adult moult. We focused on female preferences for male signal frequency because they are the strongest and make the clearest contribution to sexual selection and signal evolution in the E. binotata complex (Rodríguez et al., 2006; Cocroft et al., 2008, 2010, Sullivan-Beckers & Cocroft, 2010).

Quantifying variation in mate preferences requires recognizing their nature as function-valued traits (Stinchcombe & Kirkpatrick, 2012; Fowler-Finn & Rodríguez, 2013; Rodríguez et al., 2013b, c). Mate preference functions describe variation in attractiveness over a range of signal traits (Ritchie, 1996; Wagner, 1998). To generate individual preference functions, we conducted vibrational playback experiments that took advantage of the duetting system that facilitates pair formation in E. binotata treehoppers (Rodríguez et al., 2006; Cocroft et al., 2008). We placed each female individually on the plant stem of a potted host plant and allowed her 2 min to acclimate. We then played back a series of 19 signal bouts in random order, each bout consisting of four stimulus signals (which corresponds to the mean number of signals/bout in our study species). Stimuli were presented to a female by imparting them to a plant stem at an amplitude of 0.15 mm s⁻¹ using a piezoelectric controller and actuator (Thorlabs, Newton, NJ, USA). Each bout varied from others only in dominant signal frequency, with all other signal values set to the mean of the population (Fowler-Finn et al., 2015a). The playback stimuli had signal frequencies of 165 Hz, and steps of ± 2, 4, 6, 8, 10, 15, 20, 30, 40 Hz above and below this value. To generate and play stimuli, we used custom scripts written in MATLAB v. 7.5.0 (The Mathworks, Inc., Natick, MA, USA; scripts available upon request). We recorded female responses using a laser vibrometer, and determined the number of responses for each of the 19 stimulus frequencies for each female. The numbers of females per replicate plant that responded and thus from which we were able to obtain full preference functions were (mean ± SE) 4.4 ± 0.7 in low density, 6.7 ± 0.7 in medium density and 7.0 ± 1.5 in high density.

We used cubic spline regressions to generate individual preference functions from the raw response data; this method makes no assumptions about the shapes of functions other than that they are smooth (Schluter, 1988). We optimized the smoothing parameter for each function using the mgcv package, gam functions, and custom-written script in R v. 2.14.1 (R Development Core Team, 2011; script available upon request) (Fowler-Finn & Rodríguez, 2012a, 2013; Rodríguez et al., 2013b). We then extracted two important values from these individual preference functions that describe the shape of the function in independent ways. Peak preference is the stimulus frequency estimated to elicit the greatest female response (Fowler-Finn & Rodríguez, 2012a, b, 2013; Rodríguez et al., 2013b; Fig. 4). Selectivity summarizes variation in the shape of the preference other than peak, and describes how strongly females favour the preferred signal frequency (Fig. 4). Selectivity is the PC summarizing variation in the aspects of the shape of the mate preference that are independent from the peak (Fowler-Finn & Rodríguez, 2012a; Fowler-Finn et al., 2015a). They include responsiveness (the mean number of responses across all stimuli, which corresponds to the overall elevation of the preference functions), tolerance (how quickly the female response drops as the signal frequency deviates from the preferred signal values) and strength (the level of variation in responses between more and less preferred signal frequencies). Our PCA generated a single selectivity PC with an eigenvalue = 2.0 explaining 71% of the variation, with the following contributions from each measured selectivity trait: tolerance = 0.47, strength = −0.61, responsiveness = 0.63.

Results

Variation across treatment densities

Realized densities across treatments are shown in Fig. 5a. Treatments varied significantly in density and group size (social PC1; F_2,20 = 28.7, P < 0.0001; Fig. 5b), but not in sex ratio (social PC2; F_2,20 = 1.3, P = 0.3063; Fig. 5c).

Variation in signals

Male signal PC1 varied across replicate plants but not treatment densities, whereas male signal PC2 and PC3 showed no significant variation across replicate plants or densities (Table 3).

Table 3. Variation in Enchenopa male signalling behaviour across the density treatments (low density, medium density, high density) and replicate plants. Replicate plant is included as a random term nested within treatment

Trait	Factor	DF (num,den)	F	P	N
Male signal PC1	Treatment	2,24.5	0.5	0.5942	163
	Replicate	20,141	4.5	< 0.0001
Male signal PC2	Treatment	2,41.1	0.4	0.6511	163
	Replicate	20,141	1.0	0.4102
Male signal PC3	Treatment	2,40.5	0.3	0.7692	163
	Replicate	20,141	1.1	0.3769

• Significant tests in bold.

Male signal PC1 did not correlate with social PC1 (r = −0.01, P = 0.9481, n = 23; Fig. 6a), and showed a negative but statistically nonsignificant correlation with social PC2 (r = −0.39, P = 0.0681, n = 23; Fig. 6b).

Signal–preference relationship across replicate plants

The significant sex term indicates a population-level mismatch between preference and signal frequency; the significant sex × social PC1 term indicates that signals and preferences vary in different ways from one another across the social factors comprising social PC1 (density and group size) (Table 5; Fig. 6). This change in the signal–preference relationship suggests that the form of sexual selection becomes more directional (favouring higher signal frequencies) at higher population densities (Fig. 6, 7). The marginal sex × social PC2 term suggests that the signal–preference relationship may also vary with sex ratio (Table 5), approaching stabilizing selection at more heavily male-biased sex ratios (Fig. 6b).

Table 5. Replicate plant-level variation in signal frequency and peak preference in Enchenopa binotata treehoppers across social factors (social PC1 and PC2)

Factor	DF (num,den)	F	P
Sex	1,36	158.5	< 0.0001
Social PC1	1,36	6.2	0.0173
Social PC2	1,36	0.1	0.7808
Sex × Social PC1	1,36	5.3	0.0271
Sex × Social PC2	1,36	3.7	0.0630

• Significant tests in bold.

Discussion

We used Enchenopa treehoppers to ask whether sexual selection by mate choice is influenced by social plasticity in signals and mate preferences arising from variation in local population density and other social factors. We found that males and females respond to different social factors, a phenomenon broadly observed in other taxa, but rarely tested for both sexes simultaneously (crickets: Tinghitella et al., 2013; fish: Kozak & Boughman, 2009). Male signals varied across replicate plants but not across treatment densities, whereas female preference varied across density, but not across replicate plants within the same treatment densities. These sex-dependent patterns of plasticity generated differences in the relationship between the distribution of male signals and female preference – thereby likely influencing the expected form of sexual selection – across social contexts. Social plasticity could therefore have important consequences for signal evolution across social groupings.

In all social contexts, the signal–preference relationship predicts directional selection on signal frequency because females consistently preferred signal frequencies higher than the population mean. However, because male signal frequency did not vary across densities, the degree of directional selection is likely to vary across social groupings, being strongest at high densities when female peak preference is highest. Demographics directly impact mate availability and competition (Emlen & Oring, 1977; Miller & Svensson, 2014) and often influence the form and strength of sexual selection (Kokko & Rankin, 2006; Ryder et al., 2012; Aronson et al., 2013; Wacker et al., 2013). Whether or not the patterns we observed are adaptive, they can have important consequences for spatial and temporal variation in selection.

In prior research with Enchenopa, we found that female preference selectivity is highest when females are exposed as adults to signals of variable frequency (Fowler-Finn & Rodríguez, 2012a, b). These previous results are consistent with the idea that, with higher availabilities of mates, females are often more discriminating (Gwynne, 1984; Crowley et al., 1991; Palokangas et al., 1992; Berglund, 1995; Wagner et al., 2001; Kokko & Rankin, 2006; Hebets & Vink, 2007; Lehmann et al., 2007; Willis et al., 2011; Atwell & Wagner, 2014; Stoffer & Uetz, 2015a). In our current study, we found no variation in selectivity across densities. However, we minimized adult female exposure to male advertisement signals by removing males once they began singing. Therefore, it seems that plasticity in selectivity may be restricted to social effects occurring during adulthood – this would make sense if the juvenile social environment does not predict the quality and availability of mates in the future. In contrast, we found in our current experiment and a prior experiment on E. binotata (Rebar & Rodríguez, 2013) that juvenile experience is likely to influence peak preference. The effects of social experience at different life stages on preference therefore have the potential to impact selection and evolution in different ways (Tramm & Servedio, 2008).

Patterns of variation in male signals across replicate plants could be due to either neighbourhood effects – when the identity of your close neighbours matter – or effects due to the variation in the host plant. Previous work on related species in the E. binotata complex supports both ideas, with findings that signalling varies with experience of competitors and neighbours (Rodríguez et al., 2012; Rebar & Rodríguez, 2016), female responses to male signals (Rodríguez et al., 2012) and also plant genetics (Rebar & Rodríguez, 2013, 2014, 2015). Social environments may alter signal–preference relationship across spatially small scales, particularly when males and females respond in different ways to the same social cues. The influence on selection may vary across densities, as we expect stronger directional selection on male signals in denser groups, and across social composition, as we expect weaker directional selection with higher proportions of males in the group. Although density and sex ratio are direct outcomes of our random placement of individuals on the different replicate plants, the insects had free choice of the aggregations they formed within a plant, in both size and composition. Therefore, it is important to note that any factors influencing the choice of social environment – genetics or indirect genetic effects (Saltz & Nuzhdin, 2014) – could also contribute to the patterns we see.

In short, we find variation in the signal–preference relationship stemming from social plasticity in both male signals and female preferences. Although the cause of the different effect of density and local populations (i.e. replicate plants) on females and males in this study is unclear, the prevalence of social factors affecting males and females differently (e.g. Rebar & Rodríguez, 2016) suggests that social plasticity, and fluctuations in social variables, could be major causes of fluctuating sexual selection in the wild. Overall, these findings may help explain the great dynamism of sexual selection (West-Eberhard, 1983, 2014; Kingsolver et al., 2001; Wolf et al., 2007), but are also relevant for evolution under social selection more generally, whereby within-population competition for mates or other resources results in rapid and extravagant evolution (West-Eberhard, 1983; Lyon & Montgomerie, 2012). We suggest that the same insight may also apply, at least in part, to any case in which behavioural interactions constitute both the cause and the target of selection, including interactions between species. For example, adjustment of behaviour on the part of both predator and prey can influence predator–prey dynamics (McGhee et al., 2013). Therefore, spatial and temporal variation in social composition, in conjunction with behavioural plasticity, could play a key role in determining the course of evolution across a wide variety of contexts.