Male-male lethal combat in the quasi-gregarious parasitoid Anastatus disparis (Hymenoptera: Eupelmidae)

Most animals employ aggressive behaviours to acquire resources such as food, territory and mates. Although mating is important for males, which typically exhibit competitive behaviours to gain mating opportunities, they generally tend to avoid conflict escalation; while extreme combat also occurs in some species and results in death. In this study, male-male lethal combat behaviour in Anastatus disparis was examined (Hymenoptera: Eupelmidae) by investigating the characteristics of fighting and the factors that influence fighting intensity in this species. Male fight intensity in A. disparis increased with both competitor density and female presence, while it was not influenced by the relatedness among male competitors. By comparing the frequency of received attacks between injured and non-injured males, we found that the former were more vulnerable to attack. In contrast to death due to lethal attack, death that occurs as a result of A. disparis combat may be the cumulative effect of injuries sustained over repeated competitive encounters. Combined with the biological characters of A. disparis, we discuss potential factors contributing to the evolution of fatal conflict in this species.

Injured male. In general, there were no significant differences in the frequency of attacks between the uninjured males in four treatments (p > 0.05). In the treatment with three uninjured males, the frequency of attack was 0.72 ± 0.18 over a 10-min period ( Fig. 4. Treatment A). In the treatment consisting of two uninjured males and one slightly injured male, the slightly injured male was attacked at a higher frequency than the two uninjured males were attacked by each other (Fig. 4. Treatment B; 1-tailed sign test; p < 0.05). However, the frequency at which the uninjured males were attacked by the slightly injured male was lower than the attack frequency between each other (1-tailed sign test; p < 0.05). In the treatment consisting of two uninjured and one severely injured male, the frequency of attacks between the uninjured males was significantly lower than the frequency at which the severely injured male was attacked ( Fig. 4. Treatment C; 1-tailed sign test; p < 0.05). Severely injured males did not attack other individuals. The uninjured males sometimes also attacked dead individuals; while the frequency of attacks between uninjured males did not significantly differ from the frequency at which dead males were attacked by uninjured males (Fig. 4. Treatment D; 1-tailed sign test; p > 0.05).

Discussion
In this study, male-male combat behaviour in A. disparis was observed to lead to male injury and death. In general, fight intensity was influenced by competitor density. Prior predictions have suggested either a negative 23 or domed 13 relationship between fight intensity and competitor density, depending on whether the benefit of winning or the cost of fighting is more influential; an increased linear relationship has also been proposed 8 . Our results showed that the intensity of fighting increased with competitor density (Fig. 1), which may be the result of an increased rate of male encounters 8,13,31 . It also might suggest that each male increased its fighting level in response to increased competitor density.
If males can discriminate kin from non-kin, Hamilton's theory of kin selection suggests that fight intensity should be lower between closely related competitors than between less related ones and that more altruism should be evident in the former 32,33 . Additional studies have also indicated that limited dispersal would increase local competition among relatives 23,[34][35][36] . However, when A. disparis males were allowed to interact with both nonrelatives and relatives in an arena, we found no evidence that male competitors adjust their fighting behaviour in response to relatedness (Fig. 3. see also studies in Melittobia species 8,37 ). This result may be consistent with a comparative study on certain non-pollinating fig wasps, whereby relatedness between males did not influence fighting level 38 . Moreover, as discussed and reviewed in West et al. 36,38 , if males recognize relatives, they are expected to adjust their fighting level depending on relatedness. However, if a male cannot recognize relatives, fighting levels are not expected to be influenced by relatedness. Therefore, our results suggest that A. disparis may be unable to discriminate kin-an ability that is rarely observed in non-social insect species 8,[39][40][41][42][43] . Another possibility is that the degree of relatedness may show little variation 44 . Theory predicts that genetic kin discrimination based on genetic polymorphism will be rare because kin discrimination would reduce genetic diversity if common alleles are frequently recognized 45,46 . In the present study, the use of third-generation lines from a potentially inbred or genetically uniform colony might have led to little variation in relatedness among treatments. Molecular methods could be used to quantify relatedness in future studies, and distinct strains, ideally collected from the wild, would be useful.
As mating is directly related to male fitness, access to female mates is important to males [9][10][11] , and the occurrence and intensity of combat is influenced by the value of the resource, such as mating opportunity 6,16 . Our results showed that the intensity of male fighting was greater under female presence (Fig. 3). This finding suggests that males may escalate conflict to gain the valuable resource of mating opportunities. It is likely that in environments with few mating opportunities, the potential benefits of winning can exceed the costs of fatal combat 6,13,47 . However, as the number of females present in the arena increased (as a proxy for the number of mating opportunities, Fig. 3), fighting intensity (proportion of dead and injured) was still at a high level and did not change much. As observed in some other parasitoid species (e.g., Nasonia vitripennis 48 ), individual A. dispar males can mate with more than 10 females in 24 hours (Liu & Hao's observation). And mating is directly related to male fitness 9-11 , therefore, each male may be not expected to adjust their fighting level in response to increased females in an effort to acquire more mating opportunities. Another possible reason is that males lack the ability to estimate the number of females in the environment.
Inconsistent with the lethal attack (e.g., decapitation) that occurs in Melittobia species 31,49-51 , a single attack cannot lead to death of an individual in A. disparis. We found that the frequency of attack by others was higher for injured males than for uninjured males; in addition, even dead males were sometimes attacked. These results suggest that injured males may be vulnerable to groups and to frequent and repeated attacks by competitors. One possible reason for these findings is that the capacity of injured males (and dead males) to evade attack is decreased due to reduced mobility. Alternatively, weak individuals may generally receive more aggression within a range of contexts; further work is needed to test this possibility. Death from combat in A. disparis appears to have been the result of the cumulative effects of injuries sustained from frequent and repeated attacks.
Because of damage to the organism, most species avoid conflict escalation 1 , but this condition was not observed in A. disparis. First, the mating system perhaps contributes to the evolution of extreme conflict. Although there is only one A. disparis parasitoid per host, the hosts tend to be spatially clumped 24,25 , which leads to mating 24,28,29 and male competition for mates nearby the emergence site. When males compete for mates, the females are limited in space, time, or both, and extreme competition and fatal combat behaviours can evolve 8,12,13,23 . Second, the short lifetime of these males (5-8 day) in field 24,26,29 may lead to low expected-lifetime male mating success 6 . Thus, each successful mating may represent a considerably larger proportion of lifetime reproduction, and in turn, the benefits of winning far outweigh the potential costs of conflict and may lead to extreme forms of fighting that can be fatal 6,7 . Moreover, the vulnerability of injured males may serve as a factor contributing to the evolution of fatal combat behaviours. pernyi are cultivated commercially for their silk; in this study, they were purchased from a farmer in Qinglong Manchu Autonomous County, Qinhuangdao City, Hebei Province, China. Before acquiring the host eggs, the pupae of A. pernyi were maintained at 26-28 °C prior to adult emergence. Upon emergence of the females, they were sacrificed and dissected, and the eggs were removed from their abdomens. The eggs were maintained at 0 °C prior to use for no longer than 60 days 52 .
To reduce male relatedness in each experiment, third-generation males were obtained by separating the grandmothers during oviposition. Because sex determination in A. disparis is haplodiploid, virgin females lay unfertilized eggs that develop into males 27,28 . To obtain males, we housed virgin females with 20-30 A. pernyi eggs for oviposition and then incubated the hosts at 26-28 °C. After 24 hours, we isolated the hosts individually in polyethylene tubes (height: 7.5 cm; diameter: 1 cm) whose openings were covered by a cotton ball to prevent any aggressive male-male behaviour before the experiment began. After each virgin female's androgenesis, they were mated and again supplied with 20-30 hosts and allowed to oviposit for 24 hours. Then, the hosts were individually isolated in polyethylene tubes to prevent mating behaviour. After approximately 18 days 27 , females and males were checked daily and grouped by eclosion date. For the subsequent experimental replicates, we selected insects that had emerged within the same 24-hour period.

Experiment 1: Competitor density.
To investigate whether the intensity of fighting is influenced by the number of competitors 8,13,23 , we established 7 biologically relevant levels of male density: 1, 2, 4, 6, 10, 15 and 25 males (see above for details of A. disparis biological characteristics) with 15,8,17,26,27,22, and 20 repetitions, respectively. Groups of age-and relatedness-matched males of each density were placed into cylindrical arenas (height: 5 cm, diameter: 10 cm) at 28 °C for 24 hours. For each density, males that emerged within the 24-hour period and from a single female were selected for this experiment. We then isolated all the males (dead, injured and healthy) individually in polyethylene tubes (height: 7.5 cm; diameter: 1 cm). The number of dead and injured males was recorded using a microscope; we also scored each male injury visible with the microscope according to a scale from 0-7 (e.g., the loss of an antennae scored 0.5 points) adapted from Murray 13,[20][21][22] . We then calculated the mean injury per wasp and the proportions of male dead and injured for each arena. Experiment 2: Relatedness. In each arena containing 4 or 10 males, we varied the relatedness of the male competitors among 3 treatments: "high relatedness", in which both males were from a single female (14 and 16 repetitions for the 4-and 10-male density levels, respectively); "mixed relatedness", in which the males were from unrelated females (13 and 15 repetitions for the 4-and 10-male density levels, respectively); and "partial relatedness", in which an equal number of males came from 1 of 2 females, meaning that each male was relatively less related to the males from the unrelated female and more related to the males from the same female (15 and 15 repetitions for the 4-and 10-male density levels, respectively). We also calculated the mean injury per wasp and the proportions of male dead and injured for each arena.

Experiment 4: Characteristics of lethal combat.
In lethal combat, death may result from a single lethal attack, such as decapitation in Melittobia 31,[49][50][51] . Although A. disparis males were frequently observed to die following combat, especially under high competitor density (results of Experiment 1), deaths were not observed to occur due to single lethal attacks. In combat, most competitors collided with each other and severed their opponents' limbs using their mouthparts. Based on our casual observations, injured males appeared to be attacked more often and more readily than did uninjured males. Therefore, we hypothesized that injured males are vulnerable in groups and are frequently and repeatedly attacked by competitors, which eventually results in death. To test this hypothesis, we compared the frequency of attack by uninjured males between injured and uninjured males. We considered the degree of injury, i.e., severe or slight, as well as death. Non-moving males were considered to be severely injured. This level of injury generally involves the loss of two or more legs or more serious injuries. Males that had lost part or all of a single leg or antenna were considered slightly injured. Three males (comprising two uninjured males and one slightly injured, severely injured or dead male or comprising three uninjured males; 18, 13, 18, and 17 repetitions, respectively) were placed into cylindrical arenas (height: 5 cm, diameter: 10 cm) at 28 °C. Then, we recorded the number of times each individual was attacked (e.g., male-male contact) over a 10-min period.

Statistical analysis.
As proportional data often have non-normally distributed error variance and unequal sample sizes, the generalized linear model was performed with binomial errors and using a logit link function. Proportional data can also be overdispersed, leading to overestimation of significance. If the heterogeneity factor (HF = residual deviance/residual degrees of freedom) is <4, the data can be scaled and the significance tested using the F distribution to correct for overdispersion 53 . In Experiments 1 and 3, we used the generalized linear model to test for the effect of male group size and female number on the proportion of males dead or injured after 24 hours. In Experiment 2, the generalized linear mixed model was applied to test for effects of male group size and competitor relatedness on mortality and injury. We used the measures of fight intensity as the response variables for each model, including group size and relatedness as fixed effects. Interactions are presented only where significant at the level of p < 0.01; this criterion for significance is recommended when testing interactions 53 . In addition, the data for scores (mean injury per male) was analysed by the nonparametric method using the Kruskal-Wallis H test. In Experiment 4, male-male attack frequency in the same treatment was analysed using sign tests, and the Kruskal-Wallis H test was also used to analyse the attack frequency between different treatments. All statistical analyses were performed using SPSS software (version 20).