Introduction

Many animal taxa respond behaviorally to predators. Consistent with theoretical predictions, prey species generally reduce their activity when presented with predation threats (Werner and Anholt 1993; Lima 1998). This is often not a simple on–off phenomenon. Prey may also show lower activity levels in the absence of predators as a fixed antipredator strategy (Sih 1987). According to the threat-sensitivity hypothesis (Sih 1986), a prey will adjust its activity in accordance with the magnitude of the threat imposed by a predator. Several studies have shown how intrinsic characteristics like size, which affect sensitivity to predators, shape the magnitude of the prey’s antipredator behavior (e.g. Puttlitz et al. 1999). Most of these studies manipulated predation risk by manipulating exposure to visual cues from a predator. Many prey, however, rely on chemical cues to assess predation risk. Fewer studies have evaluated threat-sensitivity of prey behavior to chemical cues released by predators (kairomones) (but see e.g. Kusch et al. 2004).

A largely overlooked characteristic of a prey that will affect its sensitivity to predators is its history of autotomy, a widespread behavior where a prey sacrifices a body part that has been grasped by a predator. Autotomy has been shown to reduce escape speed, and is accompanied by a reduced tendency to flee (Robinson et al. 1991a, 1991b; Stoks 1999). Therefore, prey that have undergone autotomy are often more vulnerable to predators (Stoks 1998a, 1999; Fox and McCoy 2000). Accordingly, after autotomy, prey have been shown to reduce activity. No studies so far have studied threat-sensitive behavior to kairomones in prey that have undergone autotomy. Autotomy is, however, widespread, and especially in aquatic environments, communication about predation risk occurs primarily through kairomones (Bronmark and Hansson 2000).

In this study, we test whether the threat-sensitivity of antipredator behavior in larvae of the damselfly Ischnura elegans to fish kairomones is affected by previous autotomy. Damselfly larvae have three caudal lamellae, which may be sacrificed when attacked by fish or invertebrate predators (McPeek 1990; Stoks and De Block 2000). Lamellae loss in natural larval damselfly populations can be high, with up to 20% missing all three lamellae (Stoks 1998b). Caudal lamellae play a relatively small role in oxygen uptake (Eriksen 1986), and are used to generate thrust when swimming. As a result, escape speed is drastically reduced after lamellae autotomy (Robinson et al. 1991a; Stoks 1998c). We hypothesize that larvae will perceive a higher threat after autotomy and in the presence of predator kairomones, and will act accordingly by decreasing their activity. We evaluated the effects of autotomy and fish kairomones in larvae from a fishpond and in larvae from a fishless pond. As fish are very efficient predators and impose high mortality on damselfly larvae (McPeek 1990; Stoks and McPeek 2003a), it is plausible to expect that larvae from a fishpond perceive a higher threat than larvae from a fishless pond, especially if the latter do not recognize fish kairomones.

Materials and methods

Collection and housing

Larvae of the damselfly I. elegans were collected throughout the winter of 2001–2002 and immediately brought to the laboratory. This continuous collection procedure reduced the amount of time between capturing and testing the larvae and kept the time lag similar for all larvae (about 3 days). Half of the larvae were collected in a fishpond dominated by the top predator pumpkinseed sunfish (Lepomis gibbosus) in the nature reserve De Maten in Genk (Belgium). The other damselfly larvae were collected in a small fishless pond dominated by Anax dragonfly larvae in Overijse (Belgium). In such fishless water bodies, Anax are the top predators (e.g. McPeek 1990; Stoks and McPeek 2003a). Dragonfly larvae likely pose a smaller threat to damselfly larvae than fish as damselfly larvae have a much higher chance of escaping a dragonfly attack than a fish attack (Stoks and De Block 2000). We only collected final instar larvae with three complete, unregenerated lamellae.

Larvae were kept individually in plastic holding containers (10 cm long×8 cm wide×4 cm high) in a walk-in climate room with a constant temperature of 21°C and a photoperiod of L/D 16:8 h. Damselfly larvae were fed ad libitum with Daphnia and mayfly larvae. During the acclimation period of the larvae we manipulated their lamellae status. We randomly assigned larvae of each population to one of two groups: no lamellae or three lamellae. All three caudal lamellae were removed by gently pulling them with two fingers until the animals autotomized these appendages at the specialized breaking joints (as in Stoks 1998a, 1999). The other half of the larvae underwent a sham operation. They were manipulated in the same way, held in the hand but instead of pulling the lamellae, we gently touched these without causing their autotomy. All larvae underwent autotomy or the sham operation at least 1 day before their use in experiments.

Experimental setup

To test the context-dependency of the different behaviors of the larvae of I. elegans we set up a full factorial 2×2×2 experiment with two levels of kairomone type (control and fish medium), two levels of autotomy (three lamellae and no lamellae) and two populations (fishless pond and fishpond). The behaviors of ten individual larvae per treatment combination (total of 80 larvae) were observed for 20 min.

Before a trial we added 40 Daphnia to a plastic container (16 cm long×9 cm wide×11 cm high) filled to a height of 10 cm with control medium or fish kairomone medium. Then, we carefully introduced a single damselfly larva in the container with a spoon. Each trial started when the larva hit the bottom of the holding container. A computer was used to record the behaviors of the larvae. We registered the following behaviors: not moving, walking (a change of position on the substrate by moving the legs), swimming (leaving the bottom substrate and moving through the container by swinging the abdomen and caudal lamellae), making rigid abdomen bends, orienting (turning its head towards a Daphnia without changing position), advancing (walking to chase a Daphnia), striking (striking at Daphnia with labium) and catching a Daphnia (see Table 1). For the behaviors walking, swimming, advancing and making rigid abdomen bends we determined not only their frequency but also total, mean and maximum duration. As on average only 4.09 (standard error: 0.24) Daphnia were caught during a trial we did not replace them to avoid any disturbance. Medium with kairomones was made by keeping two pumpkinseeds (standard length 6 cm) in a 16-l bucket filled with 8 l of aged tap water. Fish were fed with Daphnia and mayfly larvae.

Table 1 Correlations between the rotated principal components and the original ln-transformed behavioral variables
Full size table

Statistical analysis

Prior to analysis, frequency variables were ln(x+1)-transformed, all other variables were ln(x)-transformed. To reduce the dimensionality of the data set and obtain a smaller set of uncorrelated variables we first extracted four principal components from the correlation matrix of the behavioral variables (see e.g. McPeek 1990). Principal component axes were rotated using the varimax-normalized procedure to aid their interpretation. Afterwards, we performed univariate analyses of variance (ANOVA) testing for effects of fish kairomone presence, lamellae autotomy and population origin on the principal component scores. Although each effect was tested in four separate ANOVAs, one for each of the extracted principal components, we chose not to perform a correction for multiple testing for two reasons. First, although the threat-sensitivity hypothesis generates clearly directional predictions about how behaviors should change, the ANOVA approach did not allow for one-sided testing, making our results conservative. Moreover, significant results of main effects were in the direction as expected under the threat-sensitivity hypothesis and not random as would be expected if they resulted from chance effects due to multiple testing.

Results

The first four principal components summarized 82% of the variation in the original behavioral data set. PC1 explained 24% of the variation and was highly positively correlated with variables associated with swimming and negatively correlated with the duration of the longest period spent motionless (Table 1). PC2, which explained 21% of the variation, was strongly positively correlated with variables describing foraging effort. PC3 explained 20% of the variation and was strongly positively associated with the number and duration of rigid abdomen bends. Finally, PC4, which explained 17% of the variation in the behavioral data set, was especially linked with the duration of walks: larvae with higher loadings on this PC showed longer walks. As distances covered by walks were typically 2–3 cm, larvae with higher loadings on this PC walked more slowly.

Overall, there was a strong decrease in swimming activity in the presence of fish kairomone (PC1, Table 2, Fig. 1a). This reduction was, however, absent in larvae with lamellae from the fishless pond, generating a significant three-way interaction between fish kairomone presence, autotomy and population. Foraging was lower in larvae from the fishpond compared to larvae from the fishless pond (PC2, Table 2, Fig. 1b). Fish kairomone presence, autotomy and their interactions did not affect foraging (all P>0.23). Larvae from the fishpond made fewer rigid abdomen bends than larvae from the fishless pond (PC3, Table 2, Fig. 1c). Larvae without lamellae also showed this behavior less often than larvae with lamellae. Fish kairomone presence did not affect rigid-abdomen-bend behavior. Finally, larvae from the fishpond walked more slowly than larvae from the fishless pond (PC4, Table 2, Fig. 1d). Fish kairomone presence, lamellae status and their interactions did not affect foraging (all P>0.08).

Table 2 ANOVAs testing for effects of fish kairomone presence, autotomy and population origin on the four behavioral principal components
Full size table
Fig. 1
figure 1

The effect of fish kairomone presence, autotomy and population origin on the behavioral principal components of the damselfly I. elegans. Means are given ±1 SE

Full size image

Discussion

Although damselfly larvae are important intermediate predators in aquatic food webs (Johnson 1991), where chemical detection of predators is widespread (Bronmark and Hansson 2000), few studies have explicitly looked at their response to kairomones (Chivers et al. 1996; Wisenden et al. 1997; Gyssels and Stoks 2005). We showed here that larvae reduced swimming in the presence of fish kairomones. This is adaptive as swimming behavior is very conspicuous to fish predators (Baker et al. 1999; Elkin and Baker 2000). Former studies showed reductions in swimming behavior of damselfly larvae in response to visual cues of fish predators (e.g. Stoks and McPeek 2003b). Interestingly, this behavioral response to fish kairomones was also present in larvae from the fishless pond population that had undergone autotomy (but not in those that still had three caudal lamellae). As larvae from the fishless pond did not have the opportunity to learn to recognize fish kairomones, this may suggest that the ability to recognize fish kairomones is innate (being genetic or due to maternal effects). Alternatively, fish kairomones may be very similar to kairomones of other predators present in fishless ponds. However, studies on Daphnia show that they are able to discriminate between fish kairomones and kairomones from invertebrate predators (De Meester et al. 1999). In contrast to damselfly larvae without lamellae, those with lamellae have a high chance of escaping the invertebrate top predators in fishless ponds (Stoks and De Block 2000). This may explain why we did not find the reduction in swimming activity due to fish kairomones in damselflies with intact lamellae.

The few studies that have looked at behavioral responses of damselfly larvae to kairomones have found mixed support. Chivers et al. (1996) and Wisenden et al. (1997) showed a short-term decrease in foraging activity in Enallagma larvae confronted with fish kairomones. Other studies on I. elegans did not show responses to predator kairomones. Heads (1985) reported a decrease in movements of I. elegans during the night in the presence of Notonecta but this could be caused by a response to kairomones or to mechanical stimuli from the predator. Schaffner and Anholt (1998) did not find a behavioral response under continuous exposure of I. elegans larvae to kairomones of dragonfly predators, and Gyssels and Stoks (2005) found no effects of fish kairomones and dragonfly kairomones on escape responses. This does not mean that in these other studies the larvae were unable to recognize the kairomones. Possibly larvae did not feel threatened enough to react differentially to the chemical stimuli. For example, if we had only studied larvae with lamellae, we would have concluded that larvae of the fishless pond did not show the ability to perceive fish kairomones.

Few studies have looked to see how autotomy affects antipredator behavior in invertebrates. In accordance with their higher vulnerability to predators, damselfly larvae that underwent autotomy showed an adaptive reduction in the number of rigid abdomen bends. These bends make larvae very conspicuous to fish predators, often leading to attacks (R. Stoks, personal observation). In line with this, it has been shown that Enallagma damselfly species that invaded fishless lakes increased rigid abdomen bends compared to the species that continued to occupy the ancestral fish lakes (Stoks et al. 2003). Alternatively, the reduced number of rigid abdomen bends after autotomy may be due to oxygen shortage. However, this seems unlikely as even under low oxygen levels oxygen uptake through the lamellae is limited (Eriksen 1986). Moreover, if oxygen shortage were to occur then reductions in behaviors that demand more oxygen, such as swimming, would also be expected. In another study on damselfly larvae, it was shown that mobility was reduced in Lestes sponsa larvae with no lamellae (Stoks 1999). In the present study we could not detect such an effect of autotomy. One reason for this may be that the latter study used visual cues of Notonecta. Autotomy of the lamellae reduces escape swimming speed (Stoks 1999), and therefore increases vulnerability much more toward invertebrate predators than toward fish predators, for which prey swimming speed is of less importance (McPeek 1990; Stoks and De Block 2000; Gyssels and Stoks 2005).

In line with the assumed higher predation threat in fishponds, larvae from the fishpond population showed behavior that was more risk-sensitive. This was reflected in their fixed lower foraging activity, fewer rigid abdomen bends and slower walks. Fixed reductions in these behaviors have been shown at the species level among Enallagma species occupying fish lakes and fishless lakes (McPeek 1990; Stoks et al. 2003). These fixed lower activities in the larvae of the fishpond may be interpreted as adaptive antipredator behavior. Predators are much more likely to detect and attack moving and foraging larvae (Baker et al. 1999; Stoks and Johansson 2000). Note, however, that we studied populations from only one fishpond and one fishless pond, and we cannot exclude at present that other differences between the populations, such as hunger level and parasite abundance, caused these effects.

To conclude, our results show that the studied larval behaviors were differentially affected by fish kairomones, autotomy and population origin and that these responses were to some extent in agreement with the threat-sensitivity hypothesis. Most of the observed behaviors could be interpreted as adaptive fixed antipredator behaviors in line with the differential sensitivity of the larvae to predation risk. Also a flexible reduction in swimming activity in the presence of fish kairomones was observed. The latter was, however, dependent upon autotomy and population origin. Such context-dependent responses in activity to kairomones should be kept in mind when evaluating the ability of a prey to recognize kairomones.