Motion-triggered defensive display in a Tephritid fly

1,3Samuel Aguilar Argüello, 1Francisco Díaz-Fleischer, 1,2Dinesh Rao

1Inbioteca

Universidad Veracruzana,

Av. CulturasVeracruzanas No.101,

Col. E. Zapata, CP 91090,

Xalapa, Veracruz, México

3Current address, Instituto de Ecología, A.C,

Apartado Postal 63, CP 91000

Xalapa, Veracruz, México

2Author for Correspondence;

ABSTRACT

Interactions between prey and predators are often mediated by signals sent by the prey. Passive signals such as aposematic colouration, and active signals such as pursuit deterrence signals are thought to prevent attack from predators. In true fruit flies (Diptera: Tephritidae), the defensive wing display is called supination, and studies have shown that supination effectively reduces the chance of being attacked by salticid predators. In this study, we investigated the proximal causes of supination in staged interactions in an arena. We asked whether the movement of the display target influences the likelihood of triggering supination in the Mexican fruit fly Anastrephaludens. We tested the effect of motion on fly display in three different ways using 1. amanually moveddead spider or beetle, 2.live bouts with a spider and a katydid, and 3. video playback experiments where movement of the display targetwas controlled. Our results show that flies are more likely to perform supination when the display target moves. The identity of the display target did not influence display propensity suggesting that the supination of flies is a generalizeddisplay behaviour against any possible threat.

Keywords: supination, jumping spiders, aggressive behaviour, Anastrephaludens

INTRODUCTION

Interactions between prey and predators are often mediated by signals sent by the prey. These signals can be passive (e.g., aposematic colouration, (Bond 2007)) or active (e.g., honest signals, (ZahaviandZahavi, 1997)). In active signalling, there may also be an element of deception, wherein a prey can send misleading information to the predator. Prey can also signal to predators to interrupt an attack (e.g., startle displays, (Ruxton et al. 2004)) or to prevent the launch of an attack (e.g., pursuit deterrence, (Caro, 1995)). By signalling the relative difficulty in prey capture, the prey benefit by thwarting an attack, and the predator can benefit by conserving energy and redirecting its attack towards a facile prey. Active signalling to avoid predation has been documented in a taxonomically wide variety of animals such as tail-wagging in mot mots (Murphy 2006), tail flagging in ground squirrels (Rundus et al. 2007), inspection behaviour in guppies (Godin and Davis 1995) and the shimmering behaviour of the Asian hive bee (Tan et al. 2012).

Interactions between jumping spiders (Araneae: Salticidae) and the true fruit flies (Diptera: Tephritidae) are an example of a system where a prey signals to the predator and successfully avoids attack. In this system, flies perform a wing display called supination, where the fly brings the wings forward, perpendicular to the long axis of the body, while the ventral surface of the wing is turned to face anterior of the fly such that the costal margin of the wing is dorsal (HeadrickandGoeden 1994, Supplementary Video S1). Supination can be asynchronous or synchronous, i.e., it can occur with both wings simultaneously or sequentially (HeadrickandGoeden 1994). This display is common in both male and female flies and has been observed during conspecific interactions (Briceno et al. 1999; Headrickand Goeden 1994, Benelli2013, 2015, Benelli et al. 2014). Supination has been found effective in preventing an attack in 4 species of flies against up to 25 species of salticids (Greene et al. 1987; Hasson 1995; Mather andRoitberg 1987; RaoandDíaz-Fleischer 2012). The signalling is thought to be deceptive in function, since the flies have bands on their wings which, when viewed from a certain angle, mimics the leg patterns of the spiders (Eisner 1985). The display of the flies may mimic the courtship or aggressive displays of jumping spiders. This hypothesis, termed the predator mimicry hypothesis because the flies purportedly mimic their predators, has been invoked to explain the functional significance of these displays (Greene et al. 1987; Mather andRoitberg 1987). However, there are unresolved questions with respect to the hypothesis. Firstly, salticids are known to kill other salticids (Jackson 1977), and the mere spider-like appearance of a fly is not enough to grant it complete immunity from attack. However, the spider-like appearance may contribute to confusion on the part of the spider and thus allow sufficient time for the fly to escape, akin to evasive mimicry or imperfect mimicry (Ruxton et al. 2004). Secondly, the display is successful against a range of salticid species, even though salticid displays are highly species specific (Elias, Land, Mason, and Hoy 2006). Thirdly, though most true fruit flies have banded wings (Sivinskiand Pereira 2005), the display is seen even in species that are lightly banded. In previous experiments, we established that the appearance of the bands was not as important as the display itself in order to prevent an attack (RaoandDíaz-Fleischer 2012).

Why do the flies display to predators? The display was found to be ineffective against non-salticid predators such as preying mantids, lizards and assassin bugs (Greene et al. 1987). The display is also used in aggressive interactions with conspecifics. Therefore this display may be a reaction to the presence of any potential aggressor, and not necessarily directed to a specific predator such as a spider. In this study, we sought to investigate the proximate causes of supination in the tephritid fly Anastrephaludens. In particular, we hypothesised that the display is triggered by the motion of the display-target rather than its identity. If motion is the proximate trigger for the supination display, then the fly should display regardless of the source (spider or control) or type of the motion (manually moving objects, live animals or video playback).

Methods

In this study we used the large jumping spider, Phidippusaudax (Araneae: Salticidae), which is distributed all across North America (Edwards 2004), and frequently found in citrus orchards, where it is likely to encounter tephritid fruit flies. Female spiders (mean body length: 10.93 mm) are bigger than males(mean body length: 8.39 mm)(Edwards 2004). The abdomen is generally black with a white spot, though there is some variation in colour in this species (Edwards 2004). Spiders were collected from an abandoned maize plantation on the outskirts of Xalapa, Veracruz, Mexico. They were brought to the laboratory of the Inbioteca campus of the Universidad Veracruzana in Xalapa and housed individually in small plastic containers (7 cm diameter x 5 cm height). Spiders were fed grasshoppers weekly and watered every three days. Mass-reared Anastrephaludensflies were obtained from the MoscaFrut plant in Metapa de Dominguez, Chiapas. Flies were acquired as pupae and were allowed to emerge in wooden cages (30 x 30 x 30 cm) covered in mesh cloth within the laboratory. Flies were fed yeast hydrolysate and sugar (proportion 1:3) ad libitum. There is sexual dimorphism in Anastrephaludens flies since female flies (total body length; mean ± std. dev.: 9.99 ± 0.46 mm) are largerthan male flies (total body length; mean ± std. dev.: 7.40 ± 0.49 mm), and can be distinguished by the presence of an ovipositor (Tejeda et al. 2014).

Experimental Design

All experiments were carried out in the laboratory under natural light conditions from 10 am to 4 pm. Flies were chosen randomly from a holding cage for each experiment and each fly was used only once. Flies were not mated. In all experiments, flies were introduced into the test arena first and allowed to acclimatise for 1 min. All experiments were recorded with a Sony HDR-XR260 video camera from above.

Experiment 1: Manual movement

In this experiment, we tested whether male and female flies (n = 60) would respond to movement. One spider and one beetle were used for this experiment.The animals were anaesthetized with CO2 and subsequently frozen. We used a dead female spider and mounted it with wax on a small circular disc on a wooden platform. The disc was moved by means of a lever placed at the base of the platform. The arena consisted of a petri dish (9 cm diameter and 1.3 cm height). As a control, we used a dead beetle (Calosoma sp.; Coleoptera: Carabidae)of similar size and colour as the spider. Each interaction trial lasted for 3 min. There were two treatments: Moving and Still Treatments. For the Moving treatments, the disc was rapidly rotated when the fly was facing the spider or the beetle. During the Still treatment, the disc was not moved. The models were rotated only when the fly was facing it, extending to an angle approximately 45 degrees on either side. We imported the video clips to an Apple iMac computer and recorded the following variables: presence or absence of displays, rate of display, bout duration (time from initiation of interaction to outcome) and proportion of flies that displayed. We analysed the data with a full factorial Generalized Linear Models, with link functions according to the distribution of the response variable.The p values of the whole model correspond to the comparison between the model to the model that contains only the intercept parameter. Analysis was carried out in JMP v9.

Experiment 2: Live bouts

In this experiment, male (n = 62) and female (n = 58) flies were placed into a plastic petri dish (14 cm diameter, 2.5 cm height). The petri dish was divided into 2 partitions, separated by an opaque divider. Flies were introduced first into one of the partitions and the order ofintroduction (i.e. left or right) was randomised. Spiders or a control (a katydid)were introduced into the opposite partition (Supplementary Video S1,2). The katydid was chosen as a control due to its easily quantifiable sudden movements (pers. obs, Rao. D.). Both animals were allowed to acclimatize for a minute. Once the trial started, the divider was removed, and the interaction was filmed from above. The following variables were recorded: presence or absence of displays, number of displays, presence or absence of movement of the animals (we considered only movements that involved a change in position, i.e. we did not include movement of body parts). From the video clips we recorded: bout duration, the distance at which the fly initiated display (Display Initiation Distance) and the distance at which the fly escaped or walked away (Flight Initiation Distance). Distances were measured using the software GraphClickVer 3.0. We analysed the data with full factorial Generalized Linear Models, with link functions according to the distribution of the response variable. The p values of the whole model correspond to the comparison between the model to the model that contains only the intercept parameter. Distance data were analysed using a standard least squares fit model.

We recorded the onset of spider or katydid movement (after approximately 5 seconds of inactivity), fly attention (defined as the moment when the fly detected the presence of the spider or katydid and moved to face the spider by positioning its body directly at the animal), display (when the fly finished a display cycle, defined as from one outstretched wing pose to another outstretched wing pose), end of fly attention and end of movement. These variables were analysed using the event recording software JWatcherVer 1.0 (Blumstein and Daniel, 2007). We performed an analysis of non-repeating sequences to test whether a given behaviour was more likely to be followed by another. We computed transitional probabilities for two key behaviour sequences: movement followed by fly attention, and fly attention followed by display. To test whether these probabilities were significant, we obtained the Z-scores (adjusted residuals) of the sequence analysis. These Z-scores were calculated using a log-linear analysis for data with structural zeros using the software iLogver 4.0 (Bakemanand Robinson 1994). Z-scores greater than 1.96 are considered significant at the 0.05 level (BakemanandGottman 1986).

Experiment 3: Video playback

We recorded short clips of spider and katydid movement and configured these clips to play on an Apple iPod Touch (Supplementary Video S3). As a control we used a still image of a spider or a katydid. The iPod was integrated into one wall of an arena(15 x 15 x 7 cm). Male and female flies (n = 60)were introduced into the arena and the interactions were filmed from above. The trials lasted 3 min. We recorded the number of displays, bout duration, fly attention (while facing the screen) and the onset of these behaviours.

Data Analysis

We analysed the data with a full factorial Generalized Linear Models, with link functions according to the distribution of the response variable(Crawley, 1993; Agresti, 2007).Count data were analysed using ageneralised linear model (GLM) with Poisson errors, a log-link function and type IIIsignificance tests.For binary data, we used binomial errors and logit-link function and type III significance tests. The p values of the whole model correspond to the comparison between the whole model and the model that contains only the intercept parameter.Contrasts were used to test for differences in levels within variables.

RESULTS

Experiment 1: Manual movement

Fly display (presence or absence) was significantly triggered by the factors (i.e., movement, sex of the fly or the display target) in the model (GLM, binomial distribution, Logit Link: 2= 20.32, df = 7, p = 0.005; Fig 1A). Of these factors, sex of the fly, movement and the interaction between display target and movement were significant (Table 1). Female flies (probability 0.90) displayed more than male flies (probability 0.79). Flies were more likely to display to a still beetle (mean probability of display: 0.85 ± 0.05) than to a moving beetle (mean probability of display: 0.59 ± 0.14), but this was not seen in the case of the spider (still: 0.66 ± 0.09; moving: 0.63 ± 0.13). Post hoc contrasts revealed that there were significant differences between movement and still treatments (2= 6.6, p< 0.001) and no significant differences between beetle and spider (2= 2.4, p = 0.12).

There was a significant effect of the factors on bout duration (GLM, Poisson distribution, Log link, 2= 35.73, df = 7, p < 0.0001). Of these factors, movementand sex of the fly were significant (Table 1). There was also a significant effect of the interaction between the display target and the sex of the fly, as well as between movement and display target (Table 1). Female flies displayed for longer against beetles (Mean ± S.D.; 6.1 ± 3.5 s;) than against spiders (Mean ± S.D.; 4.15 ± 1.08 s), Male fliesdisplayed for longer against spiders (Mean ± S.D.; 7.01 ± 0.08 s) than against beetles (Mean ± S.D.; 6.45 ± 2.25 s). However, neither was significant in post-hoc contrast analysis.

Post-hoc contrast analysis showed that flies displayed for significantly longer to beetles when there was no movement (Mean ± S.D.; Still: 8.3 ± 0.38 s; Moving: 4.23 ± 0.86s, 2= 18.50, p< 0.0001), while they displayed for similar durations to spiders irrespective of movement (Mean ± S.D.; Still: 5.99 ± 1.52 s; Moving: 5.17 ± 2.52s, 2= 1.97, p = 0.16).

There was no significant effect of the three factors on display rate (GLM, Poisson distribution, Log link, Χ2= 1.1, df = 7, p = 0.99).

Experiment 2: Live bouts

During live bouts, flies did not significantly differ in their propensity to display (GLM, binomial distribution, Logit Link: 2= 2.56, df = 3, p = 0.463); and this pattern was seen whether the display target was a spider or a katydid or whether the fly was male or female (Table 2). There was a significant effect of the factors on bout duration (GLM, Poisson distribution, Identity Link:2= 29.01, df = 3, p < 0.0001). For bout duration, both the display target (Katydids: Mean ± S.D. = 14.14s ± 12.01;Spiders: Mean ± S.D. = 11.68s ± 8.66) and the sex of the fly (Males: Mean ± S.D. = 14.61s ± 11.79; Females: Mean ± S.D. = 11.27s ± 8.79; Table 2) were significant.

There was no significant effect of display target or sex of the fly on the rate of display (GLM, poisson distribution, Log Link: 2= 0.908, df = 3, p = 0.823; Table 2). Display Initiation Distance (Least squares fit, R2 = 0.06, F 3, 76, p = 0.202) and the Flight Initiation Distance (Least squares fit, R2 = 0.06, F 3, 76, p = 0.158) were not significantly different according to whether the flies displayed to a spider or a katydid.

Flies were significantly more likely to display when the spider or katydid moved. The relevant transitions of behaviour along with the Z-scores are presented in Fig. 2.

Experiment 3: Video playback

The likelihood of display was influenced by the factors (GLM, binomial distribution, Logit Link: 2= 54.71, df = 7, p < 0.0001). Of the three factors, i.e. sex of the fly, movement or identity of display target, only movement significantly influenced presence of display (Table 3). Flies were more likely to display to a moving video clip (Fig. 1B). There was no significant effect of sex of the fly or the display targetor the interactions (Table 3).

Bout duration was significantly influenced by the factors (GLM, Poisson distribution, Log link, 2= 359, df = 7, p < 0.0001). Of the factors, both the display targetand movementinfluenced the length of the bouts (Table 3). All interactions were significant(Table 3). Post hoc contrasts revealed that there were significant differences between movement and still treatments (2= 114.56, p< 0.001). Bouts by female flies were longer against katydids and spiders when there was movement (predicted value: 34.27 s). On the other hand, male flies displayed for longer against katydids when there was movement (predicted value: 31.06 s), but not against spiders (predicted value: 15.33 s).

The rate of display was not significantly influenced by the factors (GLM, Gaussian distribution, Identity Link: 2= 1.98, df = 7, p = 0.961; Table 3).

DISCUSSION

In all three experiments, fly displays were triggered by the motion of the display target, irrespective of the type of motion — manual, live or video playback. Our results suggest that rather than being a predator-directed display (Greene et al. 1987), supination in A.ludens is a generalized display against any potential threat and is primarily triggered by motion.

A potential drawback of our methodology lies in the fact that we did not use multiple examples of the target in experiments 1 and 3. There is the possibility that the spiders were responding to the target individual in question and not the movement per se. However, since the treatments are very distinct (i.e. still and movement), we can consider the response of the fly as representative. Further studies should incorporate several exemplars in order to resolve this issue.

Of all the display parameters measured, the likelihood to perform display was influenced by movement across the three experiments. However, other parameters such as bout duration and rate of display were also significant. In terms of interactions between factors, from the manual movement and video playback experiments, it was apparent that male and female flies do not respond similarly to the stimulus, which is understandable since they have different needs and motivations. In A. ludens, there is a lek mating system where courting males guard non-resource territories and wait for the approach of females. Aggression between males is more likely at this stage; whereas between females aggression may be more likely over oviposition locations (Benelli 2015).