Escape and surveillance asymmetries in locusts exposed to a Guinea fowl-mimicking robot predator

Escape and surveillance responses to predators are lateralized in several vertebrate species. However, little is known on the laterality of escapes and predator surveillance in arthropods. In this study, we investigated the lateralization of escape and surveillance responses in young instars and adults of Locusta migratoria during biomimetic interactions with a robot-predator inspired to the Guinea fowl, Numida meleagris. Results showed individual-level lateralization in the jumping escape of locusts exposed to the robot-predator attack. The laterality of this response was higher in L. migratoria adults over young instars. Furthermore, population-level lateralization of predator surveillance was found testing both L. migratoria adults and young instars; locusts used the right compound eye to oversee the robot-predator. Right-biased individuals were more stationary over left-biased ones during surveillance of the robot-predator. Individual-level lateralization could avoid predictability during the jumping escape. Population-level lateralization may improve coordination in the swarm during specific group tasks such as predator surveillance. To the best of our knowledge, this is the first report of lateralized predator-prey interactions in insects. Our findings outline the possibility of using biomimetic robots to study predator-prey interaction, avoiding the use of real predators, thus achieving standardized experimental conditions to investigate complex and flexible behaviours.

rolling and yawing of their body, thus they exhibit their trajectory decision after their hind legs have been cocked for jumping 32 . Interestingly, the desert locust Schistocerca gregaria (Forskål) (Orthoptera: Acrididae) was found lateralized at individual level in its forelimb use 37 . Strongly biased locusts made fewer errors with their preferred forelimb, suggesting that stronger lateralization provides an advantage in terms of boosted motor control in an invertebrate with individual-level lateralization 37 . These inferences suggested us to investigate the presence of lateral bias in Acrididae during predator-prey interactions.
In this research, lateralized escape and predator surveillance behaviours in neanids, nymphs and adults of Locusta migratoria Linnaeus (Orthoptera: Acrididae) at gregarious phase 67 , were investigated at individual and at population-level during interactions with a Guinea fowl-mimicking robotic predator. It has been proposed that lateralization at population-level is more likely to evolve in social/gregarious species 9,68 , although lateralization at population-level has been reported in several solitary species 21,24,27 . We performed two different experiments to evaluate: (i) if L. migratoria shows any lateral bias when -startled by an approaching predator -it jumps off, and (ii) if there is an eye preference used for overseeing a potentially threatening animal. In addition, we investigated if these lateralized responses to a predator change varied among the different developmental stages, since in insects post-embryonic development affects numerous morphological, physiological and behavioural features [69][70][71][72] .

Experiment 1: laterality of escape responses in locusts.
The jumping escape response to a Guinea fowl-mimicking robotic predator was lateralized at the individual level, while the same was not true at the population level (F 2,89 = 0.312; P < 0.7330). Right-biased adult locusts were not significantly more abundant than left-biased adult locusts, the same was true for IV and II instars.
The presence of a jumping escape response at the individual level (regardless the left or the right direction), was significantly affected by the insect instar (F 2,89 = 95.151; P < 0.0001). It was higher in adults and IV instar locusts over II instar ones (Fig. 1a).
The absolute values of the laterality index (ABLI) were significantly influenced by the locust instar (F 2,89 = 207.404; P < 0.0001). ABLI in adults was significantly stronger over ABLI in IV instar locusts; ABLI in IV instars was stronger than the one calculated for II instars (Fig. 1b). Experiment 2: laterality of predator surveillance in locusts. Locusta migratoria adults and young instars showed lateralized eye use during the surveillance of the Guinea fowl-mimicking robotic predator. The laterality index (LI) value significantly differed between left-and right-biased individuals (F 1,84 = 1319.947; P < 0.0001). No significant differences were detected among adult, IV instar and II instar locusts showing the same lateral bias (Fig. 2a). The absolute value of ABLI concerning the use of a compound eye for predator surveillance did not differ among tested instars (F 2,87 = 1.561; P = 0.216). A right-biased lateral dominance was observed in adult, IV instar and II instar locusts (Fig. 2b).

Discussion
Predator-prey interactions are crucial in shaping the fitness of animal species worldwide [1][2][3][4] . In several vertebrate species, lateralized processes have been reported for predator-prey interactions 5-10 , while little is known about them in arthropods. In our experiments, we investigated the lateralization of escape and surveillance responses in young instars and adults of L. migratoria during interactions with a biomimetic robot-predator inspired by the helmeted Guinea fowl. Results showed an individual-level lateralization in the jumping escape of L. migratoria exposed to a simulated predator attack. The laterality of this response increased in L. migratoria adults if compared to young instars.
We noted that the L. migratoria escape response was lateralized at individual-level in adults and IV instar individuals, although lateralization at population level was not detected. Previous studies also reported lateralization at individual level for forelimb use of S. gregaria 37 , showing that stronger lateralization provides an advantage in terms of boosted motor control. Furthermore, II instar locusts did not show lateralized jumping escape responses. One of the main disadvantage of lateralization is predictability, which may be used for predation by other species or cannibalization by individuals of the same species 9,73 . The presence of individual-level lateralization in locusts could be a strategy to avoid the cost of predictability, everyone has a different lateral bias when approached by the predator. Moreover, individual-level lateralization in the escape behaviour can contribute to the jumping performance thanks to doubled reinforcement by experience (e.g., improvement of the jumping escape performance over time centred on one of the two forelegs), as reported for other insect species 21,26,37 . The role of experience in producing stronger lateralized locusts can contribute to explain our data reporting a significant increase of lateralization strength in jumping escape from II instar to adult locusts. Indeed, locusts can exhibit motor learning at single ganglia-level 37,74 . In addition, it has been proposed that the specialization of forelimb movements control can be related to motor circuits and mechano-sensory reflexes within the prothoracic ganglion of locusts 75,76 . Lastly, we cannot exclude the effect that the post-embryonic development of the neural system can play, leading to insects with a stronger lateralization as they grow 69-72 . Furthermore, population-level lateralization of predator surveillance was found testing both adults and young instars, showing that L. migratoria used the right compound eye to oversee the Guinea fowl-mimicking predator (Fig. 2c). Static cryptic behaviour of locusts overseeing the predator 30,31 was efficiently performed by right-biased insects over left-biased ones. The latter was "more visible" to a potential predator by exhibiting a higher number of jumps and a longer walking activity. Interestingly, it has been reported that right-biased honeybees (Apis mellifera Linnaeus) learn better a colour stimulus compared to left-biased bees 77 , suggesting an analogy with our findings.
Locusts showed several physiological and morphological changes during the gregarious phase, reflecting a modulation of the individual's metabolism to favour greater mobility 78 . It has been argued that lateralization at population-level is more likely to evolve in social/gregarious species 9,68 . However, concerning gregarious Orthoptera, previous studies reported only individual-level lateralization 37,79 . It should be noted that, in terms of jumping, population level laterality would not evolve in this gregarious species, since this would make the group movements more predictable than if only the single individuals had lateralized jumping performances. The population-level lateralization in the eye use in L. migratoria could be partially linked to the need to perform specific group tasks, such as swarm coordination 9,16 , while predictability can be avoided by individual-level lateralization of escape responses. In addition, while the surveillance is mainly accomplished by visual sense, more stimuli are involved in the escape behaviour, such as mechano-receptive hairs that sense air displacement around insects and alert them when a predator is attacking 80 . This can be related to different reactions of locusts to predators, concerning lateralized responses during surveillance and escape. Further research is needed to investigate how these lateralized traits can be connected each other in L. migratoria, and how they change during different social phases.
To the best of our knowledge this is the first report of lateralized predator-prey interactions in insects. Our findings outline the possibility of relying to biomimetic robotic predators to study predator-prey interactions in arthropods, avoiding the use of real predators, thus achieving highly standardized experimental conditions to investigate complex and flexible behaviours.

Materials and Methods
Ethic statements. This research adheres to the guidelines for the treatment of animals in behavioural research and teaching 81  were used; light intensity around the test arena was ca. 1000 lx, estimated over the 300-1100 nm waveband with an LI-1800 spectroradiometer (LICOR Inc., Lincoln, NE, U.S.A.), equipped with a remote cosine receptor. Directional light cues were avoided by using diffuse laboratory lighting to reduce reflection and phototaxis. For each experiment, the behaviour of locusts was directly recorded by an observer 21,27,83 . A white wall of filter paper (Whatman) surrounded both the arena and the robot, the observer was dressed in a white coat, to minimize his impact on L. migratoria behaviour 27,84 and was placed symmetrically behind the robot at a reasonably distance from the arena.

Robotic predator model.
To simulate a predator of locusts, a N. meleagris head in acrylonitrile butadiene styrene (ABS) (Fig. 4a) [38][39][40] was designed in SolidWorks (Dassault Systemes, Vélizy-Villacoublay, France) and fabricated by rapid prototyping. The N. meleagris head-replica had a diameter of 40 mm a thickness of 30 mm and a total length, including the beak, of 75 mm. The bird head (except the beak and helmet), was covered by a thin layer of silicone rubber (Dragon Skin), by turning molding and then coloured, reproducing the colour pattern of real N. meleagris birds. A 300-mm steel rod, connected the bird head to a DC motor (Precision Microdrives: 225-202), producing a simple robotic arm. The DC motor was placed in a Plexiglas pipe section (height 150 mm, diameter 120 mm), partially filled with iron weights, as a support body. A grey-black sheet with white polka dots covered both the support body and partially the rod, to improve the resemblance with the N. meleagris plumage (Fig. 4b). A microcontroller (Arduino, Mega 2560) was used to control the movement of the robotic arm, capped with the plastic bird head. Depending on the experiment, the robotic stimulus can be moved in an upright way or horizontally by changing the support base of the pipe.

Experiment 1: laterality of escape responses in locusts.
We evaluated if locusts showed a bias in jumping to the left or to the right when the robot-predator approached them frontally. Individual L. migratoria were gently placed on a cubic platform (100 × 100 × 100 mm) of white cardboard exactly centered with respect to the robotic stimulus, in the centre of a rectangular white arena (800 × 600 × 600 mm) 37 . Insects were placed to an identical distance from the right and left side of the arena. The robotic stimulus was visually isolated from the tested subjects by a wall of the arena (600 × 600 mm) fitted with a white curtain with a vertical slot in the centre, which allowed the robotic arm, capped with the bird head, to enter in the arena when simulating predation (Fig. 5a).
A jumping escape was evoked when the robot predator simulated an attack, striking moving from top to bottom, 100 mm from the locust head, and then returning. The robot was perfectly symmetric in appearance and movement to avoid any lateral bias. The top of the arena was covered with a transparent partition, to allow focal observations of the locust behaviour. For laterality observations, we considered only locusts which were approached by the robot predator when they were perfectly centered with respect to the robot predator. For each insect, the laterality of 30 jumps was recorded. Each jump was evoked by the Guinea fowl-mimicking robot predator after 10 min from the previous one. 30 II instar, 30 IV instar and 30 adult locusts that jumped after a perfectly symmetric robotic stimulus were analyzed in this experiment. Experiment 2: laterality of predator surveillance in locusts. The eye preference used by locusts to scan a Guinea fowl-mimicking robotic predator was investigated. Locusts were placed individually in the centre of a rectangular white arena (800 × 600 × 600 mm), with their body axis orthogonal to a transparent wall (800 × 600 mm, Plexiglas). Insects were placed to an identical distance from the right and left side of the arena. The robot predator was placed outside the arena in correspondence of its middle, with the bird head 250 mm from the transparent wall. The rod capped with the bird head was yawed by the DC motor, 45° to the left and 45° to the right with a frequency of 0.5 Hz, to be perfectly symmetric in appearance and movement. Tests (lasting 30 minutes) started 5 minutes after the locust was introduced in the arena, removing an opaque partition from the transparent wall, that allowed visual contact with the robotic stimulus (Fig. 5b). At the beginning of the experiment, the locust was placed head on to the robotic stimulus and was able to orientate the body, according to its overseeing of the robot predator.
For each locusts, we recorded how long a given side of the insect's body (e.g., steered body axis forming an angle >45°, with the initial orientation of the body axis perfectly centred with the stimulus, to have just one eye able to see the robot 85,86 ), was exposed to the Guinea fowl-mimicking robot predator. Furthermore, since locusts assume a static pose to go unnoticed when a predator is nearby 30,31 , we recorded the number of jumps and the duration of the walking behaviour, to evaluate if the cryptic behaviour of subjects overseeing the predator was affected by lateral bias. The further distance of the robot predator and its slow movement, if compared to the Experiment 1, was predicted to evoke the cryptic behaviour in locusts over the jumping escape 30,31 . 30 II instars, 30 IV instars and 30 adults of L. migratoria were tested.

Data analysis.
A laterality index (LI) was calculated for each insect, to analyse the differences in the direction of jumping escape responses: We calculated a LI for each insect evaluating bias in the use of the right and left eye during predator surveillance: LI = [(duration of surveillance with the right eye − duration of surveillance with the left eye)/(duration of surveillance with the right eye + duration of surveillance with the left eye)] 88 .
Individual asymmetrical dominance was determined by comparing the size of the LI value (ranging from −1 to + 1) 88,89 , with a threshold (LI TH ) set to 0.3 (e.g. LI >0.3 right-biased; LI <−0.3 left-biased) 89,90 . Furthermore, the absolute value of the laterality index (ABLI) was considered, to discriminate individuals with a bilateral dominance from individuals with a lateral dominance, regardless the left or the right direction of the bias 88,89 .
Laterality differences among the numbers of locusts (II young instars, IV young instars and adults) displaying right-or left-biased jumping escapes, as well as right-or left-biased eye use during surveillance were analysed by JMP 9 (SAS) using a weighted generalized linear model (glm): y = Xß + ε where y is the vector of the observations (i.e., escape response or surveillance), X is the incidence matrix linking observations to fixed effects, ß is the vector of fixed effects (i.e., the tested instar and laterality) and ε is the vector of the random residual effects. A probability level of P < 0.05 was used for the significance of differences between means. Furthermore, differences in the (i) mean duration of walking response as well as (ii) the number of jumps during the surveillance of a robotic predator were analysed using the glm described above with normal distribution, fixed effects were the tested instar and laterality of the behavioural response. Averages were separated by Tukey's HSD test. A probability level of P < 0.05 was used for the significance of differences between means.
Within each locust instar, the difference in the number of locusts using left or right eyes during the exposure to the predator was analyzed using a χ 2 test with Yates correction (P < 0.05).