Development time

Of 65 freshly hatched P. cochleariae larvae, which were kept individually in Petri dishes on cabbage, 52 successfully developed into adults. The development time from freshly hatched larvae to adult hatching ranged from 19-23 days (20.68 ± 0.85 days, mean + SD) and did not differ between males and females (Mann Whitney-U test, U = 306.5, p = 0.712, N_females = 21, N_males = 31).

Analysis of behavioural traits

Behavioural traits were measured five times throughout the lifetime, testing up to eight traits in up to four different contexts. Six behavioural traits were determined for both larvae and adults. The traits covered distance, amount of movements, turning angles < 90° and variance of turning angles were received from video analyses of the locomotion of individuals in Petri dishes within 30 min (larvae) or 1 h (adults), respectively. The duration of thanatosis (tonic immobility; thanatosis 1) and the subsequent activity time (thanatosis 2) were measured after imitating a predator attack. Furthermore, for larvae the latency of reaching a safe refuge in an unprotected environment (light-dark test) was measured. Alternatively, for adults the latency of leaving a safe refuge (hiding context, dark-light test) and of reaching a safe wall in an unprotected environment context (wall time test) were determined. Applying an agglomerative cluster analysis on the measured behaviours combined with a Silhouette plot to determine the likely number of behavioural groups for each life stage separately, two dimensions of behavioural traits could be identified for L2 and L3 larvae as well as young adults (10-17 d old, A1) at the population level (Fig. 1). In the larval stage L2 as well as in the A1 adults, the behavioural traits covered distance, amount of movements, turning angles < 90°, and variance of turning angles were clustered together and defined as the activity dimension (Fig. 1). Furthermore, for the larvae the traits light-dark, thanatosis 1, and thanatosis 2 were clustered together and formed a second dimension termed boldness (Fig. 1, left). This dimension could be considered as equivalent to the traits dark-light, wall time, thanatosis 1, and thanatosis 2 which clustered in A1 adults (Fig. 1, right). The behavioural traits for L3 larvae clustered in the same dimensions as traits of L2 larvae (L3, agglomerative coefficient = 0.55). Thus, at the population level, the behavioural clustering showed similar patterns. The behavioural traits of older adults (24-31 d old, A2; 38-45 d old, A3) clustered almost identically as in A1 adults except that the trait turning angles < 90° was grouped in the boldness dimension (A2, agglomerative coefficient = 0.53; A3, agglomerative coefficient = 0.52). The agglomerative coefficients were similar or higher to those found in comparable studies and indicate a reasonable clustering [3, 17].

Consistency across behavioural traits and contexts

Consistency over time and age-dependent behavioural differences

The amount of movements, chosen as typical activity trait, was significantly lower in the larval compared to the adult stages in individuals measured throughout life (one way repeated measures ANOVA, life stage F_4,153= 33.619, p < 0.001, followed by pairwise multiple comparisons, Fig. 2a). The proportion of individuals showing a thanatosis 1 duration of ≤ 10 s (fast responders) compared to > 10 s (slow responders), chosen as a typical boldness trait, differed significantly between L2 and L3 larvae (Fisher's exact test, p = 0.043; N_L2 = 35, N_L3 = 51), L2 and A1 (p = 0.039; N_A1 = 48) and L2 and A3 (p = 0.004; N_A3 = 41), respectively. After correction for false discovery rate, only the difference between the L2 larvae and the A3 adults remained significant. Most L2 larvae showed only very short thanatosis 1 durations, whereas latencies increased with age, with A3 adults showing more often durations > 10 s (Fig. 2b).

Sex-dependent behavioural differences

Adult females showed a significantly higher activity (composite variable) than males throughout adulthood, but the activity did not differ between adults of different age (two way repeated measures ANOVA; sex, F_1,36 = 13.355, p< 0.001; age, F_2,72 = 0.045, p = 0.956; sex x age, F_2,72= 1.020, p = 0.366; followed by pairwise multiple comparisons; Fig. 3a). In contrast, the composite variable boldness did neither differ between sexes nor between beetles of different age (two way repeated measures ANOVA; sex, F_1,36 = 0.767, p = 0.387; age, F_2,72= 0.062, p = 0.940; sex x age, F_2,72 = 1.398, p = 0.254; Fig. 3b).

Correlation between development time and behavioural dimensions

The development time (from hatching of larvae until adult emergence) and the composite variable activity of L2 larvae were significantly correlated (Spearman rank correlation, r = -0.358, p = 0.048, N = 31) (Fig. 4). No correlations were found between the developmental time and activity of L3 larvae (r = 0.018, p = 0.903, N = 46) and between the developmental time and boldness of L2 (r = -0.118, p = 0.525) and L3 larvae (r = -0.260, p = 0.080), respectively.

Stable behavioural structure through out the lifetime

Throughout the ontogeny of P.cochleariae, two behavioural dimensions, activity and boldness, were concordantly found at the population level. In two earlier studies on adult behavioural phenotypes of this beetle species and another chrysomelid, the traits, which clustered in the activity dimension in the present study, were split in two behavioural dimensions, with the activity dimension being characterised by the covered distance and the amount of movements, whereas an additional non-targeted explorativeness cluster comprised the traits turning angles < 90° and variance of turning angles [3, 34]. The different assignment of traits to behavioural dimensions in the earlier studies [3, 34]compared to the present study may be explained by minor differences in some behavioural tests or slight differences in food quality offered to the insects. For example, thanatosis 2 was exclusively measured in the present study, whereas the movement in the inner area of a Petri dish was only recorded in the earlier studies on chrysomelid beetles [3, 34]. Furthermore, beetles were fed either young or old leaves in the earlier studies [3, 34], whereas food was provided in alternating order in the present study.

A stable correlation structure between various behavioural traits has also been found earlier for some hemimetabolous insects [28, 30]. Our results show that even for insects which are holometabolous the behavioural phenotype can be stable throughout the lifetime. Such persistence in behavioural structure is not necessarily found throughout life in invertebrate species (e.g., not in spiders, see introduction [29]). More studies are needed that investigate the stability of behavioural phenotypes of various invertebrate species to determine whether pattern are taxon-specific or differ in general between animals with contrasting life-histories such as spiders, hemimetabolous, and holometabolous insects.

Age- and sex-dependent behavioural differences

Although the correlation structure of the two behavioural dimensions remained stable over the lifetime of P. cochleariae, the rank order in different behavioural traits was not consistent over time at the individual level, neither among the two tested larval stages nor between the larvae and the adults. In particular, age-dependent differences were found in activity (represented by amount of movements) as well as in boldness (represented by thanatosis 1). Whereas adult beetles were generally more active than larvae, L2 larvae were boldest. Such differences between life stages could be explained either by the stage-specific life-history (differences in anatomy, morphology, and physiology), by special challenges or needs at each specific life stage, and/or by ecological niche shifts between life stages. Thus, different levels of the same behaviour could be advantageous [35], leading to distinct responses of juvenile instars and adults in the same situation [28]. Since juveniles should mainly be adapted to growth and survival and the last larval instar additionally needs to find a site for pupation, whereas adults should maximise their reproductive output [36, 37], from a life-history perspective they are expected to respond differently.

Larvae of P. cochleariae may be less active than adults simply because they are less mobile. In contrast, adults should be more active since they need to find mating partners and oviposition sites, and since they may disperse. Although adult beetles have elytra, we never observed P. cochleariae flying. Nevertheless, in this experiment beetles moved distances with up to 13.5 m/h (personal observation). The bolder behaviour of L2 larvae in comparison to adult beetles may be explained by the asset protection principle [35, 38]. Adult beetles have more to lose if they become victims of predators because they are in the reproductive stage [35, 39]. Furthermore, in the L2 stage, larvae are much lighter than L3 larvae and adult beetles of P. cochleariae. As a consequence, it is most important for young larval instars to gain weight and they cannot afford to show long thanatosis after a predator attack, since they need to forage intensely. Heavier L3 larvae and adult beetles are probably able to cope with longer periods of time without feeding. Moreover, thanatosis may be inefficient in young larvae, since they probably face a larger range of different predators than adults. Similar age- and stage-dependent differences in several characteristic behavioural traits have been found in some hemimetabolous insect species [19, 26, 28], even though hemimetabolous insect larvae change their overall phenotype much less than holometabolous insects.

Within the adult stage of P. cochleariae, a high rank order consistency was found across time in almost all investigated behavioural variables. Thus, results of an earlier study carried out with adult beetles of this species over a period of about three weeks [3] could be confirmed for an even longer period of almost five weeks in the present study, representing almost the entire adult life span. In conclusion, selection might uncouple the behavioural phenotypes over the lifetime if environmental conditions and functional needs of juveniles differ from those of adults [7], but within the adult phase they may be stable, if the environment does not change.

However, sex-dependent differences were found in adult P. cochleariae, with females being more active (composite variable) than males. The activity differences between the sexes might be due to different challenges or functional needs of females compared to males. Females are generally heavier than males and hence might have higher food requirements. Additionally, females need to find appropriate oviposition sites and accordingly may benefit from a higher searching activity effort. In contrast, we could not detect any sex-dependent differences in boldness of mustard leaf beetles. So far, most behavioural differences between the sexes of insects have been found for boldness traits [17, 19]. For example, females of the long-winged morph of firebugs are bolder than males [17]. Boldness behaviour of female field crickets is repeatable over the lifetime, whereas males become less bold with maturation, which might be explained by the risky courtship songs of the males which could attract predators [19].

Correlations between development time and behaviour in the larval stage

The activity (composite variable) of L2 larvae was significantly correlated with the developmental time from the first larval stage until hatching of the adult beetles. In other words, larvae which developed faster were more active. This result supports the POLS hypothesis, which states that a shorter development time and consequently a faster growth rate is positively correlated with activity and hence potentially a higher dispersal ability [32]. The fact that only development time and activity of L2, but not that of L3 larvae, were correlated might be explained by stage-specific features or challenges. At the end of the third instar, larvae become quite inactive and do not feed any more before pupation. In contrast, since small larvae are most vulnerable to different predators, they should develop quickly to outgrow predation risk from smaller predators. Thus, high activity during the L2 stage might be crucial to increase the probability to survive this stage. Additionally, a more active behaviour may result in higher energy gains if more food is foraged and consequently may result in a shorter development time [40].

Interestingly, in P. cochleariae activity and boldness are not necessarily related, since we did not find a correlation between boldness and the developmental time. Environmental conditions can even lead to opposite responses in activity and boldness in this beetle [3]. Until now, only few empirical studies exist about possible correlations between different behavioural dimensions and life-history parameters in invertebrates. Faster growing crickets mature earlier and invest less in immune defence in comparison to slower growing individuals [26]. Red morphs of the pea aphid (Acyrthosiphon pisum) show a strong trade-off between early reproduction and their life span which, moreover, correlates with their boldness behaviour [41]. In contrast to invertebrates, in many vertebrates such relationships, mostly between boldness and reproduction or growth, have been well described [42–44].

Study organism and rearing conditions

The chrysomelid beetle P. cochleariae is a specialist on species of the Brassicaceae family and can reach pest status on some crops [46]. Half of the adults survive for up to 60 days under our laboratory conditions in this experimental set-up. Beetles were collected in different parts of Germany and reared for several generations under constant conditions in a climate cabinet (20°C, L16:D8, 65% r.h.). The beetles were kept in ventilated plastic boxes (20 x 20 x 6.5 cm) and provided with leaves of Brassica rapa L. ssp. pekinensis var. Michihili (Brassicaceae) ad libitum. Food plants were grown from seeds (Kiepenkerl, Bruno Nebelung GmbH, Konken, Germany) in pots (12 cm diameter) filled with composted soil in a greenhouse (L16:D8, 60% r.h.). Leaves of non-flowering, 8-10 week old plants were used as food.

Experimental set-up

To investigate the development time and the behavioural phenotype of P. cochleariae, 65 neonate larvae were kept individually in Petri dishes (5.5 cm diameter) covered with plastic cups (4.5 cm diameter, 5 cm height). Cabbage leaf disks (2.5 cm diameter) supported with moistened sponge rubber to prevent desiccation were offered. The insects received young and mature cabbage leaves in an alternating order to supply them with food of mixed quality. For larvae, food was switched every day, for adult beetles every other day. The development time from the first larval stage until hatching of the adult beetles was recorded. After metamorphosis, adult beetles were sexed and kept in pairs of one female and one male.

The behavioural traits of individuals were investigated in a test series five times, two times in the larval stages [second (L2) and third larval instar (L3)] and three times during adulthood [between day 10-17 (A1), day 24-31 (A2) and day 38-45 (A3) of adult life]. Since the L2 larval stage only lasts for three to four days, 35 L2 larvae could be tested for their behaviour within this short time period. In the subsequent live stages, behavioural tests were performed with 41-51 individuals. For activity measurements of larvae and behavioural tests of adults, we used a similar test series as described earlier [3]. To test behavioural traits in a locomotion context, the movements of individual larvae in a Petri dish (9 cm diameter) were observed for 30 min and that of adults for 60 min with a camera (Life Cam VX-2000, Microsoft, Seattle, WA, USA). The observation time differed since larvae cannot starve as long as adults. Using the motion detection software Cam Alert (2.9.23, Max Christian Pohle) and the programs Power batch (Power batch 6.1.2.2, UniDream Marketing Technologies Inc) and Virtual Dub 1.9.11 (Avery Lee) a video sequence was created. Applying the software Bug Tracer Program (version C, Robert Winkler, based on a MATLAB motion detection script of Lokesh Peddireddi) the traits covered distance [cm], relative amount of movements (percentage of time an individual is moving), number of turning angles < 90°, and variance of turning angles were revealed. After motion detection, individuals had a 30 min resting period for feeding. Afterwards, three additional behavioural traits were measured in larvae and four in adults, using tests which are appropriate to the respective life stage [30]. All following tests were ended after 300 s irrespective of whether the individual had finished the tasks or not. In a light-dark test, only performed with larvae, the latency (light-dark) needed to reach a safe refuge within a Petri dish (9 cm diameter) was measured. The safe refuge consisted of a round black cardboard disk (2 cm diameter), which was accessible from two sites 1 cm off the ground and in a distance of 1 cm from the edge of the Petri dish. The larva was inserted at the opposite site from the refuge in the Petri dish. In contrast, adults were tested in a dark-light test. In this test, beetles were put in a brown glass vial (1 cm diameter, 6 cm height) for one minute. Thereafter, the vial was shortly shaken, then placed horizontally and the latency the beetle needed to completely exit the vial was recorded (dark-light). The light-dark and dark-light test served to investigate the behaviour in an unprotected environment (larvae) or hiding context (adults), respectively. The dark-light test was followed by a wall time test to investigate the adult behaviour in the context of an unprotected environment. In the wall time test, a beetle was placed in the middle of an arena (17.2 cm diameter) surrounded by a wall (0.8 cm height) and the latency the beetle needed to reach the wall was determined (wall time). Finally, a thanatosis test was performed with larvae and adults after a 15 min break to recover from the previous tests. To test the behaviour in a predator response context, the individuals were gently squeezed with spring steel forceps inducing a thanatosis reaction (tonic immobility). The duration of thanatosis reaction (thanatosis 1) and the duration until the beetle stopped again after moving (thanatosis 2) were recorded. We interpret a longer lasting thanatosis 2 duration as a bolder behaviour, since the individual does not seem to be long-term affected by a predator attack under these circumstances.

Statistical analyses

To test for variance-homogeneity and normal distribution of data a Shapiro-Wilk test as well as a Levene's test were used. Since development time was not normally distributed, differences between males and females were analysed with a Mann Whitney-U test.

The behavioural traits were analysed according to [3, 17]. Kendall's W coefficient of concordance was calculated to analyse if larvae (L2 or L3) or adults (A1, A2 or A3), respectively, were similarly ranked across all tested behavioural traits [47]. To investigate the relation between behavioural traits within one life stage, an agglomerative cluster analysis using agnes function with Ward's clustering method within the R-package irr was conducted [48]. The likely number of behavioural dimensions was identified by a Silhouette plot [49] with the R-package cluster, but was not further formally tested. Kendall's W coefficient of concordance was analysed [47] both to determine the consistency across the behavioural traits measured in three contexts for larvae and four for adults within a test series and to test for consistency of each behavioural trait over time between the different life stages. To test for the consistency over time between the larval and the adult stage only the behavioural traits which were tested for both stages were used. Consistency over time was computed for each behavioural trait separately as well as for each behavioural dimension by using a composite variable, adding the ranks of each individual over all behavioural traits included in this dimension per life stage. All behavioural traits were ascendingly ranked, except the behavioural traits dark-light and thanatosis 1 which were descendingly ranked. In this way, high ranks imply bolder or more active behaviour, respectively. Because of multiple testing, the p-values within a set of tests were adjusted by using the false discovery rate [50].

To test for changes within behavioural traits over lifetime, the amount of movements was analysed as typical activity trait using a one way repeated measures ANOVA, followed by pairwise multiple comparisons (Holm-Sidak method), since data were normally distributed. To compare the duration of thanatosis 1 as typical boldness trait between the life stages we divided the individuals in two categories, those responding within 0-10 s (fast responders and thus considered as bold) and those responding within 11-300 s (slow responders and thus considered as shy). The proportion of individuals belonging to these two categories within each test phase was compared between the larval stages and between the larval and the adult stages using Fisher's exact tests, followed by a correction for false discovery rate.

To test for differences in adult behavioural dimensions between sexes and depending on age, a two way repeated measures ANOVA (sex and age as factors) followed by pairwise multiple comparisons (Holm-Sidak method) was used based on the composite variables.

To test for relationships between life-history and behavioural traits we used a Spearman's rank correlation analysis correlating the larval development time and the composite variables of the activity and the boldness dimension of L2 and L3 larvae, and A1 adults, respectively. The statistical tests were conducted by using the program R (3.0.3), STATISTICA 10 (Statsoft, Germany), and SigmaPlot 11.0 (Systat Software, USA).