Introduction

Anti-predator behavior is an important determinant of fitness in many animal species (Godin 1997; Ruxton et al. 2004). The existing evidence shows that animals can adjust their reproductive effort (Godin 1995; Candolin 1998) and their foraging effort to the risk of predation (Lima and Dill 1990; Magnhagen 1990; Ylonen 1994). Trade-offs between reproductive investment and predation risk and between starvation and predation risk are central to many decisions individuals make regarding habitat choice (Gilliam and Fraser 1987; Butler et al. 2005), foraging and mating, and diurnal activity (Dall and Witter 1998; Krause and Ruxton 2002; Macleod et al. 2005; Hedrick and Kortet 2006; Luttbeg and Sih 2010).

Differences among prey individuals in responses either to the mere presence of a predator or to real attacks by the predator are typically viewed as adaptive. Anti-predator responses do not represent just some variation around a mean value but reveal genuine differences in behavioral strategies (Dall et al. 2004; Dingemanse and Réale 2005; Bell and Sih 2007; Réale et al. 2007, 2010; Smith and Blumstein 2008, 2010; Bell et al. 2009; Dingemanse and Wolf 2010). In many species, individual differences in behavior and physiology are consistent across contexts and are stable over time. Such consistent individual differences are referred to as temperament or personality (van Oers et al. 2004; Groothuis and Carere 2005; Sih and Bell 2008; Bell et al. 2009; Dingemanse and Wolf 2010; Gyuris et al. 2011), and behavioral correlations that arise as a consequence of individual variation in two or more behaviors are termed behavioral syndromes. Predation has been proposed as one of the most important mechanisms underlying the maintenance of behavioral syndromes by selecting against certain combinations of traits (Stamps 2007; Luttbeg and Sih 2010). However, there are few studies investigating the relationship between personality traits and anti-predator behavior (Quinn and Cresswell 2005; Bell and Sih 2007; Kortet and Hedrick 2007; Réale et al. 2007; Hollander et al. 2008; Jones and Godin 2010; Kortet et al. 2010; Smith and Blumstein 2010), most likely because predation events are infrequently observed in nature compared with other behavioral phenomena.

Careau et al. (2008) proposed a “performance model” to explain the proximate mechanisms that underlie consistent individual differences in behavior. It was suggested that consistent individual differences in energy metabolism, as reflected by resting metabolic rate (RMR), may be correlated positively with individual differences in metabolically costly anti-predator responses (Biro and Stamps 2010). Several studies have shown relationships between RMR and total energy output in animals (Daan et al. 1990; Hammond and Wunder 1995; Hammond et al. 1996; Hammond and Diamond 1997; Niewiarowski et al. 2000). Although empirical studies of the relationships between resting metabolic rates and total energy output across individuals of the same age, sex, and reproductive state are quite rare (Earle and Lavigne 1990; Hayes et al. 1992; McLean and Speakman 2000; Mueller and Diamond 2001; Zera and Zhao 2006; Castañeda et al. 2010), most of them report positive correlations between resting metabolic rate and sustained energy output (Speakman et al. 2003, 2004; Rezende et al. 2005, 2009; Nespolo et al. 2008).

In this experimental study, we tested whether survival of yellow mealworm beetles exposed to the great tit, Parus major, a diurnal predator, is related to their anti-predatory responses, and whether anti-predator behavior can be related to RMR. It has been recently shown that more aggressive males of the field cricket, Gryllus integer, have shorter latencies to emerge from safe refuge, are more active in general, and are less cautious towards predation risk compared to less aggressive male crickets (Hedrick and Kortet 2006; Kortet and Hedrick 2007; Niemelä et al. 2012, 2013). To study anti-predator responses, we investigated behavioral variation among yellow mealworm beetles, Tenebrio molitor, in their individual response latency time to become immobile and the duration of immobility to treatments mimicking an approach of predator. To assess the role of metabolism, we measured the RMR of each individual beetle and also tested whether RMR was consistently repeatable across individuals, as well as relationships among RMR, survival, and anti-predator responses (Nespolo and Franco 2007). Although yellow mealworms are a popular study object in evolutionary biology (e.g., Moret 2006; Sadd et al. 2006; Vainikka et al. 2006; Kivleniece et al. 2010; Daukste et al. 2012; Krams et al. 2013a, b), repeatabilities in mealworm anti-predator responses and RMR have not been studied so far. It is important to note that repeatability can be considered as the approximate estimate of the upper limit of heritability (Lynch and Walsh 1998; Dohm 2002). When repeatability is low, the measured trait is unlikely to respond to any kind of selection, and it is also unlikely to correlate with the measurements of other traits (Hõrak et al. 2002; Szafranska et al. 2007; Bell et al. 2009).

Materials and methods

Beetles

The progenitors of the beetles used in the experiment originated from a natural population. More than 700 beetles were collected from 12 barns in southeastern Latvia in 2007–2010. The stock culture was maintained at Daugavpils University on bran mixed with wheat flour, fresh carrots, and apples in a controlled environment (at 24 ± 2 °C) on a 14:10-h L/D photoperiod. We removed pupae from the culture on the day of pupation, weighed them, and determined their sex by examining the genitalia on the eighth abdominal segment (Bhattacharya et al. 1970). We then kept the pupae and newly emerged adults individually in numbered 200-ml plastic containers filled with a mixture of bran and wheat flour and with fresh carrot/apple pieces offered ad libitum at least twice a week. We excluded all individuals with visible abnormalities from the experiment. All of the experimental trials were conducted in December 2009 and January 2010 at 23 °C. Beetles were weighed to the nearest 0.001 g with an electronic balance (RADWAG 26–600 Radom, Poland), and this measure was highly repeatable between two subsequent days (r = 0.98, N = 300, P < 0.0001). The beetles used in this study were of similar body weight, varying between 0.95 and 1.05 g.

Flow-through CO2 respirometry to measure RMR

The LI-7000 differential CO2/H2O analyzer (LiCor, Lincoln, NE, USA) was calibrated at different flow rates by means of calibration gases (Trägergase, VEB, Saxon Junkalor GmbH, Dessau) (Quinlan and Lighton 1999; Lighton 2008) with gas injection (see also Mänd et al. 2005, 2006; Muljar et al. 2012). The insect chamber (Eppendorf test tube, volume 3 ml) was perfused with dry (5–7 % RH) CO2-free air, produced by passing air over Drierite (W. A. Hammond Drierite Co. Ltd, Xenia, OH, USA) and soda–lime granules at a flow rate of 60 ml min−1. Baseline drift of the analyzer was corrected during analysis from the measurements at the beginning and end of each trial, with the respirometer chamber empty (Gray and Bradley 2006; Duncan 2003; Duncan and Byrne 2005).

The first measure of RMR was collected when the beetles were 15 days old. Each beetle was kept in the insect chambers until it became motionless and reached the lowest rates of CO2 production. We then waited for an additional 10–15 min and then began recording RMR for 20 min. The mean metabolic rate of each individual was calculated by averaging data obtained over the final 20-min period for each beetle (e.g., Tartes et al. 1999). Since the beetles were either immobile or active just for short periods, the lowest values of recorded CO2 emissions roughly corresponded to the resting metabolic rate of each insect. As soon as the measurements were over, we returned the beetles to their plastic containers.

The RMR trials were repeated 6 days later (age = 21 days). Nespolo and Franco (2007) showed that repeatabilities are linearly reduced with time between measurements, suggesting habituation to experimental manipulations. However, the time interval of 6 days seems to be sufficient to avoid habituation in the case of such short-lived organisms as mealworm beetles. The life span of T. molitor usually reaches 4 months in our laboratory (Daukste et al. 2012; Krams et al. 2013a, b).

Imitation of predator approach

There is a wide variety of experimental methods to measure anti-predatory behaviors (here, latency time to become immobile and time spent immobile) in response to imitated predator approach. Many insects, especially beetles, respond to a variety of mechanical or optical stimuli by abruptly reaching a state of more or less complete tonic immobility lasting from a few seconds to several hours. The true tonic immobility in beetles usually is characterized by the lack of regular pumping movements involving both abdominal and thoracic segments, while abdominal movements may be present during freezing behavior (Metspalu et al. 2002). When predators attack, mealworm beetles often feign death by entering a state of tonic immobility. In our studies, as soon as insects reached the age of 25 days, we handled individual beetles for 2 min and then gave them the opportunity to escape into a conical Eppendorf test tube (1.5 ml), which was used as insect chamber. In this tube, the beetles either became less active or completely immobile, as if they were hiding in a dark and safe crevice, whereas they never adopted an immobile pose during handling. The insect chamber was combined with an infrared optical–actographic system using IR-emitting diodes (TSA6203) and IR-sensor diodes (BP104), which were placed on the sides of the insect chamber. The light from the IR-diode was modulated by the abdominal contractions, leg and body movements, and heart activity, and we could discriminate among these activities because the level of the output voltage reflected the vigor of the muscular contractions of the insect (see Hetz (1994); Mänd et al. 2005; Karise et al. 2010). Thus, we could detect whether the beetle was immobile or active. The beetles were left in the insect chamber, and we waited until they became active again as indicated by movements of legs, head, and antennae. When beetles just resumed locomotion by moving the legs, head, and antennae, we exposed them to a sudden and brief mechanical stimulus by flicking the tube from above with a stick to apply a mass of about 20 g dropped from a height of 10 cm. This mimicked a light jump of a landing great tit (P. major) onto the insect chamber. We recorded individual response latency time to the mechanical stimulus (time to become immobile) and the time beetles spent immobile. We repeated the trials 5 days later. The repeatabilities of the behavioral responses obtained by I.K. and I.K. were high (r = 0.97 and 0.982 for response latency time and the time beetles spent immobile; all P < 0.0001). In total, we obtained behavioral data from 300 individuals (150 males and 150 females).

Predators and survival experiment

Major predators of wild mealworm beetles living in barns include birds and rodents. Since great tits often visit barns and have been seen attacking mealworms under twilight conditions of barns (Krams et al., numerous personal observations), we used these birds as predators in this study. Fifteen great tits (14 males and one female) were captured a day before the beetle survival trials. Prior to the trials, birds were kept in individual cages with water and food ad libitum. As soon as the trials were over, the birds were immediately released at the place of capture. The longest time a great tit was held in captivity for this study was 18 h.

To provide a measure of the effect of individual anti-predator responses on survival, we placed beetles in bird cages, which served as predation exposure arenas (30 × 40 × 40 cm, L × W × H) and exposed them to great tits deprived of food for 2 h before the onset of the trials. The predation trial arena was set up to be as natural as possible, given the constraints of the small enclosure. The floor of each arena was covered with a 2-cm layer of bran, which provided beetles their usual food supply. The bran layer was thick enough to allow the beetles to walk and run without slipping and to burrow out of sight. All of the beetles were individually marked by drawing small dots on the dorsal side of the abdomen with a correction pen. During a preliminary study, the birds consumed the marked beetles in 72 of 145 cases (two-tailed sign-test: Z = 0.03, P = 0.98), which shows that markings did not significantly influence the survival of mealworm beetles.

An hour before each predation trial, beetles were released into the predation cage to familiarize themselves with the environment. This gave them ample time to search out and find cover. In each trial, we used ten randomly chosen individuals. We carried out survival experiments separately for male and female mealworm beetles. Before the predator was released into the predation exposure arena, we ensured that no beetles were left lying on their backs. The birds started to attack the beetles within the first minutes of the survival trial. Each predation trial lasted for 15 min. We observed the behavior of birds and beetles from a hide. As soon as the trial was over, we removed the birds and counted the survivors.

Each great tit was used only twice: during the male survival experiment and during the female survival experiment, which were separated by a 1-h break for each individual bird. The order of trials was chosen randomly for each sex of beetles and for each predator. Thus, each individual test beetle with known RMR (see above) was exposed to two predator approach tests in a hide (the insect chamber) and to the survival test.

Data analysis

To ascertain whether individual test beetles were consistent in their behavior, the repeatabilities (R) of response latency time to the mechanical stimulus and the total time beetles spent immobile were calculated across two trials using Lessells and Boag’s (1987) method. From both measures, mean values were calculated and used thereafter. The existence of a behavioral syndrome was then assessed from the correlation between response latency to mechanical stimulus and the duration of immobility. Statistical tests were two-tailed and all of the analyses were performed in SPSS 17.0 for Windows (SPSS Inc). All dependent variables were non-normally distributed and subjected to square root transformations to comply with normality assumptions of ANCOVA. Since body weight used as a covariate in ANCOVA analyses did not affect behavioral responses, we removed it from the remaining analyses.

Ethical note

The great tits were kept for the needs of the behavioral experiment and were released within 2–3 h after the mealworm survival tests at their place of capture. We performed the experiment using live beetles to obtain really valid dependent measures linking RMR to survival, and it was not possible to “model” predation events in this study (Barnard 2007; Cuthill 2007; Fedigan 2010).

Results

Consistency of behavior over time

The latency to become immobile in response to flicking the insect chamber varied between 2 and 118 s (26.89 ± 20.34, mean ± SD), and the total time spent motionless after flicking varied between 4 and 920 s (287.03 ± 195.98, mean ± SD). Individuals were significantly repeatable both in the response latency after flicking of the insect chamber (R = 0.854, F 299,300 = 26.304, P < 0.001) and in the total time spent immobile (R = 0.593, F 299,300 = 13.095, P < 0.001). The RMR (49.00 ± 40.53 VCO2 μl h−1, mean ± SD) likewise was found to be repeatable among individual beetles (R = 0.574, F 299,300 = 16.217, P < 0.001).

Determinants of anti-predator behavior

The response latency to flicking and the total time spent immobile were negatively correlated (Pearson’s r = −0.83, t 298 = −6.204, P < 0.001; Fig. 1), suggesting that individuals more prone to freezing behavior spent more time motionless under the risk of predation. GLM analysis showed that both the response latency to the mechanical stimulus and the total time spent immobile were dependent on the RMR (a covariate in GLM), while sex and the interaction between sex and the RMR did not affect the behavioral responses under predation risk (Table 1). The response latency to flicking the insect chamber and the RMR were positively correlated (Pearson’s r = 0.489, t 298 = 4.517, P < 0.001; Fig. 2). In contrast, the total time of immobility and the RMR were negatively correlated (Pearson’s r = −0.428, t 298 = −2.78, P < 0.001; Fig. 3).

Fig. 1
figure 1

Negative correlation between the response latency to the mechanical stimulus and the duration of time spent motionless (Pearson’s r = −0.83, t 298 = −6.204, P < 0.001) (all data were square root-transformed)

Full size image
Table 1 The effects of resting metabolic rate and sex on response latency and duration of immobility (GLM analysis) of mealworm beetles
Full size table
Fig. 2
figure 2

Positive correlation between resting metabolic rate and the response latency to predation risk (Pearson’s r = 0.489, t 298 = 4.517, P < 0.001) (all data were square root-transformed)

Full size image
Fig. 3
figure 3

Negative correlation between resting metabolic rate and the total time spent immobile (Pearson’s r = −0.428, t 298 = −2.78, P < 0.001) (all data were square root-transformed)

Full size image

Predation experiment

On average, great tits preyed upon 7.63 ± 0.77 beetles during each experimental trial, and only 71 out of 300 beetles survived. The number of female (N = 40) and male survivors (N = 31) appeared to be the same (χ 2 = 1.181, df = 1, P = 0.28). Captured individuals had significantly higher RMR (74.42 ± 42.03 VCO2 μl h−1, mean ± SD) than individuals that survived the predation trials (23.58 ± 11.45 VCO2 μl h−1, mean ± SD) (t-test, t = 22.32, N 1 = 229, N 2 = 71, P = 0.01; Fig. 4). All of the survivors (N = 71) became immobile once the predator was put in the arena, while most of the non-survivors (149 out of 229 individuals) remained active (χ 2 = 89.1971, df = 1, P = 0.0001). The response latency to flicking of survivors (20.96 ± 38.78 s, mean ± SD) was significantly shorter than the response latency of non-survivors (69.60 ± 139.13, mean ± SD; t-test, t = −2.685, N 1 = 71, N 2 = 229, P = 0.008). GLM analysis confirmed that higher RMR (Wald χ 2 = 5.00, df = 1, P = 0.03) and longer response latency (Wald χ 2 = 4.43, df = 1, P = 0.038) impaired the survival of mealworm beetles. The RMR by the response latency interaction was non-significant (Wald χ 2 = 2.68, df = 1, P = 0. 092), showing that only beetles with greatest rates of the RMR and longer response latency to flicking were significantly more often attacked by predators. The total time spent motionless after flicking of survivors (277.48 ± 163.90, mean ± SD) was significantly longer than that of non-survivors (186.04 ± 174.15, mean ± SD; t-test, t = 3.783, N 1 = 71, N 2 = 229, P = 0.001).

Fig. 4
figure 4

Resting metabolic rate of survivors and non-survivors. Bars show means ± SE

Full size image

Discussion

The major novel finding of this study is that consistent individual differences in anti-predator behavior of male and female mealworm beetles were significantly correlated with their RMR and with survival. RMR was positively correlated with the latency of immobility response and was negatively correlated with the total duration of immobility, revealing a behavioral syndrome in the anti-predator behavior of T. molitor. These results suggest that individuals with higher levels of resting metabolic rate were bold, explaining their higher exposure to the risk of predation (Artacho and Nespolo 2009). This supports some recent findings showing that RMR may be positively related to mortality as evidenced in land snails, juvenile squirrels, and captive voles (Jackson et al. 2001; Alvarez and Nicieza 2005; Artacho and Nespolo 2009; Boratynski and Koteja 2009; Larivee et al. 2010; Smith and Blumstein 2010; Lantovį et al. 2011). We note that repeatabilities for our three main measures, while all statistically significant, ranged from 0.57 (RMR) to 0.85 (latency to become immobile). Although behavioral responses are typically thought to be more “plastic” than physiological reactions (West-Eberhard 2003), perhaps RMR is affected by some other important factors, which need to be studied in the future.

The results of this study support the hypothesis suggesting that consistent individual differences in energy metabolism might underlie consistent individual differences in behavior (Biro and Stamps 2008; Careau et al. 2008). Since energy is required to fuel all of the metabolic processes of the body, differences in energy metabolism may explain consistent individual differences in many behaviors, particularly those that are energetically costly (Schmidt-Nielsen 1991; Biro et al. 2010). Biro and Stamps (2008) suggested that consistent individual differences in resting metabolic rate might support consistent individual differences in growth and fecundity, in which case fast-growing and highly fecund individuals with higher resting metabolic rate should be more active, aggressive, and bold in contexts related to food acquisition (Burton et al. 2011).

It has been recently suggested that behavioral responses that require maximal aerobic output over brief periods of time, such as predator escape or activities related to mating, may be correlated with RMR (Sadowska et al. 2005; Wone et al. 2009). Some studies have revealed that in animals competing for food or feeding territories, individuals with higher RMR tend to be dominant (Metcalfe et al. 1995; Yamamoto et al. 1998). In contrast, in many laboratory settings, every individual has easy access to food ad libitum. In such cases, it is assumed that intake rate need not depend upon behavior, in which case one might not detect correlations between behavior and RMR or productivity. Indeed the ad libitum food accessibility in our laboratory setting might help explain the low (though significant) correlations between RMR and response latency (0.489) and between RMR and response duration (−0.428).

Empirical studies of the relationships between RMR and energy output across individuals, and the repeatability of such relationships, are rare. However, the first evidence has revealed positive correlations between metabolic rate and sustained energy output in different species (Chappell et al. 2004; Fu et al. 2005, 2008; Rezende et al. 2005, 2009; Vezina et al. 2006). Our results linked these relationships to fitness outcomes. Individuals with longer latency of immobility response, and with shorter total duration of time spent immobile, were consumed by great tits more often than less active individuals. We cannot tell exactly whether it was the long latency of anti-predator response or the short time spent motionless that increased the probability of being killed by the predator. However, it is clear that immediate digging in the layer of bran and staying immobile for several minutes was the most effective way to avoid predation by great tits. This unanswered question reflects a problem to be solved in the future. Since some beetles immediately stopped their activities under the risk of predation while others spent a few seconds to dig in the layer of bran, this may suggest different survival strategies of survivor beetles. We now need to determine whether anti-predator responses of the survivors represent different survival strategies and whether these strategies are related to metabolism.

High RMR has been shown to have fitness advantages in mating contexts. For example, locomotor activity is often energetically costly (Koteja 2000; Schmitz 2005; Yeates et al. 2007), and in the case of fighting (Kotiaho et al. 1998; Finerty et al. 2009) and courtship, aggressive defense of mates may also be energetically costly (Kotiaho et al. 1998). If increased rates of performance of these behaviors are advantageous, boldness may point to a positive link between reproductive success and RMR. It is generally known that males in most species achieve higher fitness by obtaining copulations with as many females as possible (Alcock 2001). In the T. molitor system, bold males may have more copulations than shy males because males should chase females to solicit copulations (Worden and Parker 2001). This could explain why bold males may have more chances to increase fitness prospects than their shy opponents (Godin and Dugatkin 1996; Smith and Blumstein 2008). Since a female T. molitor often mate repeatedly, and both precopulatory and postcopulatory behaviors are important components of sexual selection in this species (Worden and Parker 2001; Worden et al. 2005), it may mean that mate choice in T. molitor should be time- and energy-consuming also in female individuals. However, our results show a RMR-based trade-off between fitness benefits arising from high RMR and predation risk. It is important to note that artificial selection experiments have shown that there is heritable variation for length of death feigning in beetles, and when a predator is introduced, individuals selected for longer death feigning durations are at a selective advantage relative to those at shorter durations (Miyatake et al. 2004, 2009). A link between longer durations of death feigning and survival may provide some support for gene-based links between RMR and anti-predator behavior in T. molitor.

The results of this paper are consistent with a recently proposed hypothesis suggesting that personality traits may couple with life history traits and form integrative pace-of-life syndromes, where behavior and life history co-vary in a fast–slow lifestyle continuum (Réale et al. 2010). At the slow end of the lifestyle continuum, arising mainly from the general growth–mortality trade-off (Stamps 2007), individuals would have slow metabolism, low aggressiveness, and low boldness compared to individuals on the fast end of the continuum (Réale et al. 2010). In this scenario, the life history-driven consistency in animal behavior is ultimately caused by superior fitness of certain combinations of life history and behavior (Wolf et al. 2007; Dingemanse and Wolf 2010). For example, bold behavior should yield better fitness when combined with high metabolism and early rather than late maturation since higher mortality of bold individuals would make them unsuccessful in ever reaching the age required for late maturation. These predictions need to be tested in future studies since different combinations of life history and behavioral variables may yield equal expected lifetime fitness and thus explain the maintenance of different behavioral types in population (Biro et al. 2006; Dingemanse and Wolf 2010; Réale et al. 2010; Wolf and Weissing 2010).