Growth and reproductive costs of larval defence in the aposematic lepidopteran Pieris brassicae

Authors

  • Andrew D. Higginson,

    Corresponding author
    1. Institute of Biodiversity, Animal Health & Comparative Medicine, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow G12 8QQ, UK
      Correspondence author. E-mail: a.higginson@bio.gla.ac.uk
    Search for more papers by this author
  • Jon Delf,

    1. Science and Social Sciences, Liverpool Hope University, Hope Park, Liverpool L16 9JD, UK
    Search for more papers by this author
  • Graeme D. Ruxton,

    1. Institute of Biodiversity, Animal Health & Comparative Medicine, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow G12 8QQ, UK
    Search for more papers by this author
  • Michael P. Speed

    1. School of Biological Sciences, Biosciences Building, University of Liverpool, Liverpool L69 3BX, UK
    Search for more papers by this author

Correspondence author. E-mail: a.higginson@bio.gla.ac.uk

Summary

1. Utilization of plant secondary compounds for antipredator defence is common in immature herbivorous insects. Such defences may incur a cost to the animal, either in terms of survival, growth rate or in the reproductive success.

2. A common defence in lepidopterans is the regurgitation of semi-digested material containing the defensive compounds of the food plant, a defence which has led to gut specialization in this order. Regurgitation is often swift in response to cuticular stimulation and deters predators from consuming or parasitizing the larva. The loss of food and other gut material seems likely to impact on fitness, but evidence is lacking.

3. Here, we raised larvae of the common crop pest Pieris brassicae on commercial cabbage leaves, simulated predator attacks throughout the larval period, and measured life-history responses.

4. We found that the probability of survival to pupation decreased with increasing frequency of attacks, but this was because of regurgitation rather than the stimulation itself. There was a growth cost to the defence such that the more regurgitant that individuals produced over the growth period, the smaller they were at pupation.

5. The number of mature eggs in adult females was positively related to pupal mass, but this relationship was only found when individuals were not subjected to a high frequency of predator simulation. This suggests that there might be cryptic fitness costs to common defensive responses that are paid despite apparent growth rate being maintained.

6. Our results demonstrate a clear life-history cost of an antipredator defence in a model pest species and show that under certain conditions, such as high predation threat, the expected relationship between female body size and potential fecundity can be disrupted.

Introduction

Chemical defences against predation are widespread in the animal kingdom, having been found in many terrestrial arthropods (Blum 1981; Pasteels, Grégoire & Rowell-Rahier 1983), marine invertebrates (Hay 1996; Lindquist & Hay 1996), fish (Mebs 2001) and amphibians (Hanifin et al. 1999; Summers & Clough 2001). Chemical defences are especially common in herbivorous insects (Whitman, Blum & Alsop 1990), especially in the Lepidoptera, Coleoptera and Hymenoptera (Nishida 2002; Ruxton, Sherratt & Speed 2004). Lepidopteran larvae often biosynthesize antipredator defences (Whitman et al. 1990), but may also obtain noxious chemicals from their diet (Nishida 2002; Opitz & Muller 2009) and may sequester the chemicals into their haemolymph and/or cuticles (Whitman et al. 1990; Müller et al. 2001). Such chemicals are usually made by the food plant to deter herbivory, and so use of these compounds represents a subversion of plant defences by the insects for their own survival (Leather & Walsh 1993; Fordyce 2001; Karban & Agrawal 2002; Muller et al. 2006).

The creation and/or maintenance of chemical defences is expected in many circumstances to carry some energetic cost measurable in terms of growth rate and adult size (Cohen 1985; Holloway, de Jong & Ottenheim 1993; Grill & Moore 1998; Zalucki, Brower & Alonso 2001; Coley, Bateman & Kursar 2006), which in turn might impact negatively on adult survival and reproductive success (e.g. Blakley 1981; Peters 1983; Davidowitz, D’Amico & Nijhout 2004; Etilé & Despland 2008). Demonstrations of defence costs often involve raising larvae on two or more food plants and assuming that any resulting differences in growth or survival are because of differences among food plants in concentrations of secondary compounds (e.g. Larsson, Björkman & Gref 1986; Dobler & Rowell-Rahier 1994; Hay-Roe & Nation 2007). However, it is difficult to separate the effects of secondary compounds from the effects of nutritional differences among food plants, especially where clear effects are not observed (e.g. Poelman et al. 2008).

Defences which are only manifest to a predator once a prey has been ingested are likely to provide relatively little protection to individuals (Fisher 1930). A potentially less risky strategy is to have an externally operating chemical defence such as a secretion which is triggered shortly before or during the early stages of an attack. The externalization of a defence may help to prevent injury to the prey individual, because the predator does not have to handle the prey to be aware of its noxious qualities (Endler 1986, 1991). Such defences must be renewed after being triggered unless the prey is able to re-ingest the fluids. However, such externalized fluids are often likely to be sampled by the predator and therefore be not completely recoverable for subsequent use (Bowers 1993; Eisner, Eisner & Siegler 2005).

Externalized fluids provide a way for researchers to assess the cost of antipredator defences. Defence secretions of toxin-containing haemolymph have been shown to reduce adult body size in several Coleopteran species (Rowell-Rahier & Pasteels 1986; Grill & Moore 1998). Another mechanism of defence deployment, regurgitation, involves the externalization of enteric fluids that are aversive to predators (Larsson et al. 1986; Peterson, Johnson & LeGuyader 1987; Bowers 1992, 1993; Stamp & Wilkens 1993). Regurgitation is common in the Lepidoptera (Bowers 1993; Grant 2006), where it is associated with a greatly enlarged foregut (Grant 2006) and specialized gut diverticula that act as stores for a larger amount of regurgitant (Whitman et al. 1990). Such anatomical features are costly for growing animals, but costs have been difficult to demonstrate (Ruxton et al. 2004), although there is some evidence that food loss can have costs. Twice-daily stimulation of regurgitation and the resulting loss of semi-digested food, fluids and gut material have been demonstrated to reduce growth rate (Bowers 2003). Reduced growth rate might result in a longer development time and so increased exposure to predators or might result in smaller size at maturation and so lower fitness. Hence, prey animals exposed to a high level of predation are faced with a complex trade-off between mortality and adult fitness.

If we are to understand the evolution and diversity of chemical defences, then greater understanding of their costs as well as their benefits is essential. Previous demonstrations of impacts on life history have typically been limited to single traits, but costs may be shown in a diversity of traits simultaneously that all relate to fitness and can be adaptively traded-off by the animal (Roff 1992). In many species, the timing of metamorphosis is plastic in response to predation risk during the larval stage (Roff 1992; Benard 2004; Relyea 2007). In some cases, prey will metamorphose sooner if they detect predators to reduce premetamorphic mortality risk (Benard 2004; Relyea 2007), as predicted by theory (Werner & Gilliam 1984; Abrams & Rowe 1996). However, if there are costs to defences that are incurred by prey, then metamorphosis might be delayed, thereby increasing mortality (Werner & Gilliam 1984; Roff 1992; Abrams & Rowe 1996). Thus, the cost of defences, the duration of exposure to predators and the importance of reaching a large body size in adulthood (Taylor, Anderson & Peckarsky 1998; Hirschberger 1999; Sokolovska, Rowe & Johansson 2000; Sibley, Ankley & Benoit 2001; Tammaru, Esperk & Castellanos 2002; Thurston & MacGregor 2003; Evenden, Lopez & Keddie 2006; Bauerfeind & Fischer 2008; Togashi et al. 2009) will interact in determining adaptive changes in the timing of metamorphosis (Higginson & Ruxton 2010).

An appropriate model species for studying the life-history consequences of antipredator defences is the large cabbage white butterfly Pieris brassicae, which is a common pest of Brassica crops worldwide (Feltwell 1982), and so its interactions with its natural enemies are relatively well studied (e.g. Kristensen 1994; Brodeur, Geervliet & Vet 1996; Lyytinen et al. 1999; Mattiacci et al. 2001; Tanaka & Ohsaki 2006; Tanaka 2009). All Brassica plants produce glucosinolates as defensive secondary compounds (Halkier & Du 1997) that are commonly exploited by herbivorous insects (Nishida 2002; Opitz & Muller 2009). Pieris brassicae does not sequester glucosinolates (i.e. transport them through the gut wall and store them in their body; Muller, Agerbirk & Olsen 2003) but does respond to cuticular stimulation by producing regurgitant that is aversive to predators (Muller et al. 2003). Regurgitation causes the caterpillar to lose body fluids and the nutrients and toxins that they contain (one bout of 10 simulated attacks causes third instar larvae to lose c. 6% of their body mass; A. D. Higginson, unpublished data). Repeated stimulation is therefore likely to involve significant losses and impose measurable fitness costs seen as slowed growth rates and/or low pupal and adult mass.

In the wild, P. brassicae is prey to a diversity of predators including parasitoid wasps such as Cotesia glomerata, which may infect up to 80% of larvae (Baker 1970), and many bird species. For example, in Britain, sparrows (Passer domesticus) and blue tits (Parus caeruleus) are effective predators on first and second instar larvae, whereas great tits (Parus major) and the song thrush (Turdus ericetorum) prey on the larger third and fourth instars (Baker 1970). Predation by foraging birds can be intense, for example great tits have been observed to consume half of an identified set of caterpillars within about 10 days (Baker 1970). Nonetheless, there is some evidence that P. brassicae larvae are not entirely edible to birds, so it is likely that their defensive responses can provide some protection from predation (see Wiklund & Järvi 1982).

It seems reasonable to assume that there will be a range of demands on individual defences in natural conditions. For example, P. brassicae is a gregarious species in which females oviposit in batches ranging from 5 to 140 eggs (Feltwell 1982). Individuals found in large groups may experience fewer attacks from predators than those found in solitary conditions or in very small groups. Furthermore, it seems reasonable to assume that for animals in some habitats, the presence of various predators, including ants, birds, parasitoids and social wasps, could lead to a high frequency of attacks, whereas animals in large groups or in areas denuded of predator diversity by anthropogenic changes may suffer far fewer predation events. Given that variations in defensive responses are likely, a key issue which now presents itself is what, if any, are the fitness costs associated with different levels of defensive responses. In this paper, we therefore report the results of an assessment of the cost of regurgitation via subjecting larvae to simulated predator attacks and show that this defensive response reduces several measures of reproductive potential in this species.

Materials and methods

Pieris brassicae larvae and eggs were collected from gardens and nature reserves in North-West England. Pupae were also sourced from another study culture (University of Bangor, UK). Pupae were placed in a flight room and allowed to eclose and mate with other adults in the same climate controlled room (18L : 6D, 22 °C, 40–50% humidity), but sibling mating was avoided by artificial diapause of one sex of each sibling group to prevent potential effects of inbreeding. Cotton wool soaked in 10% sugar solution was provided along with potted cabbage plants on which females could oviposit. Leaves containing groups of eggs were placed in plastic catering cartons (with a gauze covered hole in the lid) in a controlled-environment cabinet (Binder KBW-E5·1) under a 20 °C/10 °C and 18 h L/6 h D regime, and larvae were raised in families for 6–13 days (mean 8·5). Food (commercial organic cabbage from Abel & Cole, London, UK) was provided ad libitum (waste removed and food replaced as necessary). Larvae were fed one of two varieties: Savoy cabbage (Brassica oleracea bullata major) or Spring cabbage (Brassica oleracea capitata), and all individuals received the same cabbage throughout life. These two cabbage varieties were used because they are among the most common commercially grown varieties and are well studied (Hamilton et al. 2005; Bellostas et al. 2007). When most larvae of the family were heavier than 30mg, larvae were weighed, placed individually in boxes and entered into the experimental treatment (range of initial mass 25–130 mg, mean 60 mg). Solitary larvae were weighed every second day until they entered the prepupal stage when silk attaching them to the box prevents removal. For the first few families, most larvae were kept in sibling pairs, but we changed to solitary conditions when statistical tests showed that there was no difference in growth rates (or regurgitation behaviour) between solitary and paired larvae, as had been previously alluded to (Lewis & Thomas 2001). Thus, we used the mean values for the pairs in all analyses except the survival analysis, where paired individuals could not be treated as independent and taking means would be invalid.

The experimental treatment involved inducing larvae to defensively regurgitate by stroking the dorsal cuticle just posterior to the head with a mounted needle, and the volume of the resulting regurgitant was measured using a standard laboratory pipette with 10-μL tips. The number of attacks needed to provoke regurgitation was noted and stimulation of larvae ceased after one regurgitation or 10 attempts, whichever came sooner. A standard laboratory pipette was used to measure regurgitant volume accurate to 0·01 μL. Regurgitant volume was in the range 0·02–12 μL per regurgitation. Larvae were subjected to one of five treatment groups: larvae were induced to regurgitate once per day (treatment 1); larvae were induced to regurgitate every second day (treatment 2); larvae were induced to regurgitate every third day (treatment 3); larvae were induced to regurgitate every 5 days (treatment 5); larvae were taken from the culture cabinet, the boxes were opened, but larvae were not otherwise disturbed (control). Treatment groups were balanced within family groups, and there were 1–4 individuals per group from each family, depending on family size. The treatment was continued until larvae died or entered the prepupal stage. The date of pupation or the date of death if individuals failed to achieve pupation was noted, and pupating individuals were weighed 4 days after pupation. Pupae were allowed to eclose in the individual cartons. Adult females were kept for 4 days with 10% sugar water and no males in a flight cage (Lepidoptera Breeders Association, UK) before being dissected to count mature eggs, because egg maturation mostly occurs during the first 4 days of adulthood and later generation of eggs is negligible (Jervis, Boggs & Ferns 2005).

Statistical analysis was carried out in R (R Core Development Team 2009). Survival was analysed via Cox regression to assess whether the treatment affected the length of time that larvae survived with treatment groups were ranked in order of frequency of stimulation (to enter as a covariate). The analyses that involved regurgitant production (square root transformed to normalize), pupal mass and number of eggs were restricted to individuals that achieved pupation and used general linear mixed models (with family as the random factor) and nonlinear least-squares regression. We simplified models by combining treatment groups where appropriate (Crawley 2007), ascertained by using likelihood ratio tests and/or the Akaike Information Criterion. Individual sex did not affect pupal mass nor the tendency to regurgitate so it was dropped from all models, which enabled inclusion of the individuals of unknown sex and the mixed sex pairs in all analyses. We took means of paired individuals, but omitted those boxes in which one of the pair died. Potential fecundity (number of mature eggs in ovaries 4 days after eclosion) was assessed for 80 females; some females had not been assayed owing to space restrictions.

Results

Survival

Survivorship differed between families (Wald W20 = 125·403, < 0·001) and decreased with the increasing frequency of stimulation (W1 = 8·36, = 0·004, Fig. 1). There was no effect of the mass of larvae at the start of the treatment (W1 = 0·55, = 0·458). A separate analysis (because cabbage was confounded by family) showed that cabbage type had no effect on mortality rate (W1 = 1·325, = 0·25). Across families, there was no correlation between the probability of survival and either the mean volume of regurgitant produced (Pearson product-moment correlation coefficient = −0·104, = 24, = 0·628) or the mean number of stimulations per occasion (= 0·097, = 24, = 0·652), the latter indicating that the treatment itself did not directly cause the difference between groups.

Figure 1.

 Proportion of individuals that survived to pupation for the five treatment groups. As the frequency of stimulation increased, the probability of surviving to pupation decreased.

Regurgitation

The total amount of regurgitant produced by surviving larvae (control group excluded) was affected by treatment (F3,159 = 10·021, P < 0·001), because of a greater number of stimulation occasions leading to more regurgitant produced. However, the mean amount of regurgitant per occasion did not differ among treatment groups (F3,159 = 0·264, P = 0·851) and was not affected by cabbage type (F3,159 = 1·478, P = 0·226). There were significant differences between families in the mean amount of regurgitant per attack occasion (F21,89 = 3·994, P < 0·001), indicating a familial component to regurgitation tendency, but there was no interaction between family and treatment, indicating that all families responded similarly to the treatment (F56,89 = 1·268, P = 0·157). The mean number of simulated attacks per occasion also did not differ among treatment groups (F3,161 = 1·175, P = 0·321), demonstrating that there was no experimenter bias in handling the different groups. Interestingly, the mean number of attacks per occasion was significantly negatively related to the mean amount of regurgitant per occasion (F1,161 = 53·195, P < 0·001), because individuals that took longer to respond to stimulation produced less regurgitant.

Pupation

Pupal mass differed among families (F21,201 = 9·15, < 0·001) so mixed models were used with family as a random factor. Treatment group affected pupal mass (Likelihood ratio test comparing models: L ratio = 16·72, P = 0·0022): pupal mass was smaller when larvae were stimulated every day or every 5 days. Crucially, pupal mass decreased with total volume of regurgitation (t194 = −2·817, = 0·005, Fig. 2), while also increasing with initial mass (t194 = 3·245, = 0·0014) and decreasing with the mean number of simulated attacks per occasion (t194 = −2·865, = 0·005). Cabbage types had no effect on pupal mass (t194 = −0·784, = 0·442). Nonlinear least-squares regression showed that the decrease in pupal mass with increasing regurgitation was nonlinear (controlling for initial mass and the mean number of simulated attacks per occasion): Negative exponential model: Pupal mass = 0·3 − 0·004 × exp(0·19 × Total regurgitant) + 0·52 × Initial mass. The negative exponential model provided a better fit to the data than the linear model (F1 = 6·75, = 0·01, Fig. 2) suggesting that the marginal costs of regurgitation accelerate as greater quantities are produced. Total development time (egg to pupa) did not differ among treatment groups (L ratio = 4·943, P = 0·293). Initial mass was negatively related to the length of larval period (t194 = −4·233, < 0·001): individuals that grew quickly initially pupated sooner. There was no effect of the mean number of attacks per occasion (t194 = 0·45, = 0·65) but there was a difference between cabbage types (t194 = 3·833, = 0·001), with larvae that were fed Savoy cabbage pupating later.

Figure 2.

 Relationship between total regurgitant produced and pupal mass, corrected for the effect of initial mass. We show the different treatment groups for illustrative purposes: treatment 1 (closed squares), treatment 2 (closed circles), treatment 3 (triangles), treatment 5 (open squares), control (open circles). The line represents the fitted nonlinear model (see text). As the total amount of regurgitant produced increased, pupal mass decreased at an accelerating rate.

Fecundity

There were differences among families in the number of eggs (F21,58 = 3·464, < 0·001), so generalized mixed models were fitted with family as a random factor and with a Poisson error structure. Model simplification lead to categorizing treatment group as infrequent (control, treatment 5, and treatment 3) and frequent (treatment 1 and treatment 2) and categorizing volume of regurgitant as small amount (<2 μL) or large amount (>2 μL). There were very few individuals (four) in the small amount-frequent stimulation group (potentially leading to spurious three-way interactions) and they did not appear different to the small amount-infrequent stimulation group. We therefore treated all small amount individuals as one group, giving us three groups: small amount, large amount-infrequent stimulation and large amount-frequent stimulation. In the most parsimonious model, there was an interaction between pupal mass and the treatment/regurgitant category in determining the number of eggs (inline image = 16·92, < 0·001, Fig. 3), because the effect of pupal mass on number of eggs was more positive for large-infrequent than large-frequent (z48 = 2·655, P = 0·008) and in turn more positive for small amount than large-infrequent (z48 = 4·044, P < 0·001). There was also a main effect of treatment/regurgitant category (inline image = 98·5, < 0·001, Fig. 4), because individuals that produced a small amount of regurgitant had more eggs than others (z48 = −3·867, P = 0·004). There were no main effects of pupal mass (z48 = 0·617, P = 0·537), initial mass (z48 = −0·912, = 0·362) or cabbage type (z48 = 0·188, = 0·851). However, the number of eggs declined with the mean number of stimulations per occasion (z48 = −9·274, < 0·001). Nonlinear regression (controlling for pupal mass) revealed that the number of eggs decreased with total volume of regurgitant (Linear model: Number of eggs = 193 × Pupal Mass − 2·657 × Total regurgitant, F3,77 = 9·675, < 0·001) but that a negative exponential model did not provide a better fit to the data (F1,76 = 2·17, = 0·145). Hence, potential fecundity decreased at a constant rate with increasing total volume of regurgitation (Fig. 5).

Figure 3.

 Effect of stimulation frequency and amount of regurgitant on the relationship between pupal mass and the number of eggs. There was a strong positive relationship between pupal mass and number of eggs only when individuals produced a small amount of regurgitant (small). For those individuals that produced a large amount of regurgitant, the relationship between pupal mass and number of eggs was weaker when they were stimulated daily or every other day (large-freq) than when they stimulated less frequently (large-infreq).

Figure 4.

 Difference between treatment/regurgitant groups in the number of eggs. Individuals that produced a small amount of regurgitant produced significantly more eggs than others.

Figure 5.

 Relationship between total amount of regurgitant and number of eggs, corrected for pupal mass. Symbols for the five treatment groups are the same as in Fig. 2. As individual females produced more regurgitant, they contained fewer eggs and this decrease was linear.

Discussion

Our results show that defensive regurgitation by Pieris brassicae larvae reduces the probability of survival to pupation, reduces body size at pupation and reduces the number of eggs in adult females. Additionally, a high frequency of regurgitation disrupts the positive relationship between female body size and egg number, such that the potential fecundity of large individuals is no greater than that of small individuals. Thus, our study provides a demonstration of life-history costs of an antipredator defence in a specialist herbivore and shows that some expression of these costs may be cryptic, in that body size at pupation is partially maintained if there are defensive responses but it is no longer an indicator of the number of eggs produced. Some studies have demonstrated later pupation as a cost of defence but no effects on adult size (Larsson et al. 1986; Gols et al. 2008b). Our results suggest that such studies may have missed the cryptic cost of reduced potential fecundity, which is also likely to be present even when a decrease in adult body size has been found (Rowell-Rahier & Pasteels 1986; Grill & Moore 1998; Harvey et al. 2007; Gols et al. 2008a).

We found that the costs of producing regurgitant in terms of pupal mass accelerated with increasing quantities of regurgitant produced, while the cost in terms of potential fecundity increased at constant rate. Hence, costs are primarily paid in egg production initially, and only when predator attack frequency is high is a growth cost incurred. It is not immediately obvious why individuals subject to high predation risk should pay this cost cryptically. An alternative strategy would be to instead reach a smaller size and produce more eggs. It is possible that the rate of growth is maintained at the expense of investment in reproduction, in order that if predation risk diminishes close to pupation, then the female can stop growth and invest in eggs. This hypothesis depends on the assumption that females that suddenly find themselves not being attacked can invest in eggs faster than they can invest in growth and eggs simultaneously. A decrease in risk might be the case if prey reach a size-refuge where predators are less likely to attack them. A size-refuge is unlikely against predators that tend to prefer larger prey (Berger, Walters & Gotthard 2006; Remmel & Tammaru 2009), but is the case for parasitoids that preferentially attack first instar larvae, and some birds (Baker 1970; Brodeur et al. 1996). Hence, a strategy of paying costs cryptically, at least initially, is likely to benefit adult females in at least some circumstances.

A strong positive relationship between body size and reproductive success, especially potential fecundity, has been shown in many insect species (Taylor et al. 1998; Hirschberger 1999; Sokolovska et al. 2000; Sibley et al. 2001; Tammaru et al. 2002; Thurston & MacGregor 2003; Evenden et al. 2006; Bauerfeind & Fischer 2008; Togashi et al. 2009), while other studies have failed to find such a relationship in the lepidopterans Pieris napi, Pararge aegeria and Lycaena tityrus and the damselfly Ischnura verticalis (Richardson & Baker 1997; Bauerfeind & Fischer 2008). Egg number has only been shown to relate to actual realized success in a few studies, owing to the difficulties of assessing this in the field (Sopow & Quiring 1998; Togashi et al. 2009), while in others, the potential fecundity has been shown to be not realized (Ohgushi 1996; Bauerfeind & Fischer 2008), in some cases demonstrably because of limitations on reproductive opportunities such an oviposition sites or adult longevity (Leather 1988; Gotthard, Berger & Walters 2007; Togashi 2007; Berger, Walters & Gotthard 2008). Hence, while this is the first study to show that under certain conditions, such as high predation threat, the relationship between female body size and egg number might be disrupted it is not certain that this will necessarily lead to lower realized reproductive success. Such a reduction in this life history parameter might therefore have a relatively small impact on fitness compared to a decrease in body size, especially if larger females have greater flight ability and endurance (Kimura & Tsubaki 1986; Shirai 1995) and even if large size does not promote longevity (Kimura & Tsubaki 1986; Shirai 1995; Calvo & Ma Molina 2004) because female reproductive success is constrained by host plant finding ability more than egg number.

The finding that the costs of the defence are apparently not expressed as a delay in development, but only as smaller size at pupation, could show that, at least in this species, there is adaptation to pay costs as reduced fecundity rather than delayed metamorphosis. This might indicate that in this species, there is great importance to metamorphosing early, perhaps to get access to host plants or because this species is multivoltine (Feltwell 1982), so individuals that produce offspring sooner should have more generations per year. A complimentary explanation is that this shift could result from costs slowing growth rate and there being an opposing adaptive life-history shift to escape a dangerous environment (Abrams & Rowe 1996; Benard 2004; Relyea 2007), forcing costs to be paid in fecundity rather than delayed pupation. Under this perspective, the shift is determined both by the cost of defences and by the avoidance of further predation attempts. This is compatible with the additive effects on pupal mass of frequency of stimulation, number of ‘attacks’ per occasion and volume of regurgitant, indicating that the stimulation per se negatively affected life history. As well as regurgitant costs, it is possible that feeding behaviour instead changes in response to heighted predation risk, which might reduce growth rate. However, because the reduction in pupal mass is not observed until high regurgitant production, the magnitude of the reduction in pupal mass observed here may be unlikely to occur in nature and so no changes in size or age in metamorphosis might be observed, but only paid in fecundity. Furthermore, any costs of potential fecundity might actually be insignificant because ultimate reproductive success might not be reduced, especially when multiple generations per year are still realized.

Although there may be small impacts on adult fitness, depending on the environment, there may be impacts on the probability of reaching adulthood. The levels of survival observed here are in agreement with previous rearing studies of this species (Feltwell 1982; Jindal & Kular 2003). The stimulation per se did not affect survival. However, our results suggested that the regurgitation defence could also impose a cost in terms of the probability of mortality during the larval period. Thus, it appears that regurgitation is intrinsically costly, perhaps because it expends nutrients required by the animal. We found that the amount of regurgitant was negatively associated with the number of stimulations necessary to provoke regurgitation, suggesting that individuals are evolved to trade-off predation risk against loss of fitness. Thus, when fitness has already been severely impacted by the defensive response, then prey may be forced to risk predation by being more reluctant to deploy the costly defence. Interestingly, we found significant differences between families in the propensity to regurgitate, and such variation in defensive responses is also seen in ladybirds (Holloway et al. 1991). This suggests that there exist multiple antipredator strategies and prey either shows vigorous defence over a long developmental period or shows little defence to mature as quickly as possible.

We found no effects of sex on survival, defensive responses or pupal size. A model of the effects of defence costs on pupal mass (Higginson & Ruxton 2009) suggested that there might be a difference in the magnitude of the response to predator attacks between sexes. That such a difference was not found suggests that males and females of this species benefit equally from large body size. P. brassicae appears to be monogamous or only mildly polyandrous (David & Gardiner 1961; Wiklund, Karlsson & Leimar 2001), and previous studies have demonstrated little or no sexual dimorphism (Feltwell 1982; Lewis & Thomas 2001). It is likely that males are selected to be large to compete for females, and that females are large because body size scales with fecundity in the absence of the predation-related effect discussed here. Thus, no difference in pupal mass between the sexes is expected, so larval behaviour does not differ between the sexes.

We found little effect on larvae of the cultivar of cabbage that they were fed. While cabbage type did not affect survivorship, the quantity of regurgitant, pupal mass or potential fecundity, a diet of Savoy cabbage resulted in larvae taking longer to reach pupation. Later pupation on Savoy cabbage has been observed elsewhere (Hamilton et al. 2005), where it was associated with an oviposition preference for Savoy over other green cabbage cultivars. Larvae fed Savoy cabbage might spend more energy in maintaining a high level of defences and so pupate at a later time. Plants that provide better defences and are preferred by herbivores can be associated with reduced survivorship (Larsson et al. 1986; Dobler & Rowell-Rahier 1994; Hay-Roe & Nation 2007). Savoy cabbage has been shown to have a lower concentration of glucosinolates than other cabbage cultivars (Bellostas et al. 2007), but this difference was apparently not sufficient to cause differences in mortality in our study. This is likely to be because of sustained artificial selection on crop cultivars to reduce the content of toxins (Harvey et al. 2007; Gols et al. 2008b; Gols & Harvey 2009), meaning that larvae are unlikely to consume toxic levels of plant defences.

The finding that there are demonstrable costs in terms of reproductive success of a level of stimulation that may not be excessive relative to natural exposure (see data and discussion in Baker 1970) suggests that predators need not kill many individuals to have profound impacts on the population dynamics of small herbivores (Lima & Dill 1990; Lima 1998; Ajie et al. 2007; Orrock et al. 2008). However, the significance of these effects in natural systems can only be assessed if a true estimate of regurgitation frequency is known. Intuitively, we might expect that when there is a high density of naïve predators, larvae might be subject to greater frequency of testing by predators. On the other hand, as P. brassicae is gregarious on host plants (Hunter 2000) and if predators are likely to learn to avoid larvae after a small number of experiences of regurgitation, larvae would have only been selected to tolerate a low frequency of attacks. Here, we have shown that there can be significant life-history costs to defensive regurgitation, so an assessment of the natural rate of utilization of such behaviours would be invaluable to the understanding of multitrophic interactions involving the herbivores that use them.

Acknowledgements

The authors are grateful to Katie Mather for help with animal husbandry and data collection, Tom Heyes for logistical assistance, and Owen Lewis, Ann Pennell, Hannah Rowland, and Ilik Saccheri for husbandry advice. This manuscript was improved by the comments of three anonymous reviewers and our editors. A. D. H. was supported by NERC grants NE/E016626/1 awarded to G. D. R. and NE/E018521/1 awarded to Mike Speed. G. D. R. is also supported by NERC grants NE/F002653/1, NE/D010500/1 and NE/D010772/1.

Ancillary