Predator encounter assay We performed predator encounter assays using 4th instar nymphs of the flower mantid Astyliasula basinigra (n = 26) as the model and 4th instar nymphs of the large-sized mantid Rhombodera longa (n = 9) as the predator.Results showed that: (1) Foreleg stretching behaviour enhanced the survivor possibility of the flower mantids under predation risk. (2) Foreleg stretching can significantly delay and reduce predator attacks, likely by intimidating the predators or confusing predatory mantids regarding the target location. (3) Foreleg stretching can deflect predators' attacks from the more vulnerable body trunk to the less vulnerable forelegs, thereby facilitating escape after the attack. Characterising the types of foreleg stretching behaviours We compared the stretching behaviour among three flower mantid species: A. basinigra, Pseudocreobotra wahlbergii, and Acromantis hesione. Type I is characterised by the alternative extension of both forelegs to opposite sides of the body axis, with each leg extending at a large angle relative to the body's central axis (approximately 56.90°). This mode is often accompanied by the shivering of the tibia and tarsus. Type II is the simplest of the three types, involving simultaneous forward stretching of both forelegs at a much smaller angle (approximately 2.83°) to the body's central axis, making the inner surfaces of the forelegs concealed and resembling the predatory strike behaviour during hunting. Type III involves the alternative extension and withdrawal of the forelegs in a circling motion, with the angle of stretching falling between that of type I and type II (approximately 21.99°). Morphometrics of foreleg femurs We selected 16 adult A. basinigra specimens (n = 8 for each sex) to measure the morphology of foreleg femurs. The length, width and thickness of the femurs were measured using a digital calliper. The right foreleg femur of each specimen was measured. Additionally, to reflect the variation in foreleg femur extension among flower mantids, we selected 93 specimens from 7 species which exhibiting 4 distinct foreleg stretching patterns (Type I, Type II, Type I+II, and Type III). The relative width (width/length) of the femur was measured for these specimens. Result revealed that at the early stage of the evolution of stretching behaviour, with the emergence of the type II behaviour, lineages exhibiting stretching behaviour showed almost no femur expansion. Greater femur expansion appeared following the evolution of the more derived stretching behaviour types, types I and III. Frequency measurement of behaviour To determine whether foreleg stretching behaviour occurs more frequently during moving in A. basinigra flower mantids, we selected seven 3rd instar individuals with similar moulting times. To account for the diurnal activity of mantids, video recording began at 8:00 AM and lasted for 12 hours. The experiment lasted for seven days. Result showed that foreleg stretching behaviour is frequently exhibited even when flower mantids are actively moving, a state in which they are more likely to be detected by ambushing predators. In contrast, flower mantids barely stretch their forelegs when at rest. Speed comparison between control and treatment group To assess whether our manipulation affected the movement speed of mantids, potentially influencing their survival under predation, we compared the mean speeds of individuals from the control (normal) group and the treatment (forelegs-disabled) group. Result showed that there is no significant difference between the two groups. Taxon sampling and phylogenetic analyses We utilized mitogenomes and nuclear DNA from a curated taxon set of 50 mantid species to reconstruct their phylogenetic relationships. By employing advanced mitochondrial phylogenomic workflow and ancestral state reconstruction, our results obtained the phylogenetic tree with high support value, with its topology aligning with widely accepted hypotheses of evolutionary relationships of Hymenopodidae. Divergence time estimation and ancestral state reconstruction To better understand the evolutionary trajectory and temporal rates of change in foreleg stretching behaviors, we conducted ancestral state reconstructions and rate estimations of character transitions using the phylogenetic relationships and foreleg stretching behavioral datasets we established for various mantodean species. Our analysis, facilitated by functions in the phytools package, suggested that the ancestral function of foreleg stretching was associated with prey capture. This function later evolved to incorporate anti-predator mechanisms within the flower mantid clade, aligning with the diversification of large-sized mantids, which are likely the primary invertebrate predators of these flower mantids.

Evolutionary co-option, in which existing traits acquire novel adaptive functions, is a key strategy by which organisms adapt to new environmental challenges. Although such co-option has been widely documented at the genetic and morphological levels, its incidence at the behavioural level remains largely unknown. Mantids stretch their forelegs to capture prey; however, some flower mantids also perform foreleg stretches in the absence of prey. The current study tested whether this behaviour represents a novel function of the foreleg stretch, thus representing a case of behavioural co-option. Predator encounter behaviour assays revealed that foreleg stretching facilitates the escape of flower mantids from large predatory mantids by delaying predator approach or deflecting their attack towards less vulnerable body parts. Phylogenetic analysis suggested that the ancestral function of foreleg stretching involves prey capture, with the anti-predator function subsequently acquired in the flower mantid clade, coinciding with the diversification of large-sized mantids, the most likely invertebrate predators of flower mantids. This study provides a case of behavioural co-option, where a predator uses its predatory organ as a defensive implement to cope with its own predators. These findings further suggest that behavioural co-option may be common in nature, meriting more comprehensive studies.

Funding provided by: National Natural Science Foundation of ChinaROR ID: https://ror.org/01h0zpd94Award Number: Funding provided by: China Agricultural UniversityROR ID: https://ror.org/04v3ywz14Award Number: