Martine Kos, Laboratory of Entomology, Wageningen University, P.O. Box 8031, 6700 EH, Wageningen, The Netherlands. E-mail: email@example.com
1. Plant resistance against herbivores can act directly (e.g. by producing toxins) and indirectly (e.g. by attracting natural enemies of herbivores). If plant secondary metabolites that cause direct resistance against herbivores, such as glucosinolates, negatively influence natural enemies, this may result in a conflict between direct and indirect plant resistance.
2. Our objectives were (i) to test herbivore-mediated effects of glucosinolates on the performance of two generalist predators, the marmalade hoverfly (Episyrphus balteatus) and the common green lacewing (Chrysoperla carnea) and (ii) to test whether intraspecific plant variation affects predator performance.
3. Predators were fed either Brevicoryne brassicae, a glucosinolate-sequestering specialist aphid that contains aphid-specific myrosinases, or Myzus persicae, a non-sequestering generalist aphid that excretes glucosinolates in the honeydew, reared on four different white cabbage cultivars. Predator performance and glucosinolate concentrations and profiles in B. brassicae and host-plant phloem were measured, a novel approach as previous studies often measured glucosinolate concentrations only in total leaf material.
4. Interestingly, the specialist aphid B. brassicae selectively sequestered glucosinolates from its host plant. The performance of predators fed this aphid species was lower than when fed M. persicae. When fed B. brassicae reared on different cultivars, differences in predator performance matched differences in glucosinolate profiles among the aphids.
5. We show that not only the prey species, but also the plant cultivar can have an effect on the performance of predators. Our results suggest that in the tritrophic system tested, there might be a conflict between direct and indirect plant resistance.
Plants have two different resistance mechanisms against herbivorous insects, namely direct and indirect resistance. Direct resistance affects herbivores through physical (e.g. thorns) or chemical (e.g. toxins or digestibility reducers) plant traits. Indirect resistance influences the effectiveness of natural enemies of herbivores through, for example, the emission of volatile herbivore-induced secondary plant metabolites that attract the natural enemies (Karban & Baldwin, 1997; Dicke & Baldwin, 2010). Secondary plant metabolites that mediate direct resistance, however, may affect not only herbivores. They can also negatively influence natural enemies of herbivores, either directly through feeding on the herbivore that contains the secondary metabolites, or indirectly through reduced host or prey quality (Francis et al., 2001a; Harvey, 2005; Ode, 2006). This may result in a conflict between direct and indirect plant resistance (Sznajder & Harvey, 2003; Gols & Harvey, 2009). With some exceptions (see Sznajder & Harvey, 2003; Gols & Harvey, 2009 and references therein), direct and indirect resistance strategies are mostly studied independently, disregarding the potential evolutionary conflict between them.
The objectives of this study were (i) to test herbivore-mediated effects of glucosinolates on the performance of two generalist predators, the marmalade hoverfly (Episyrphus balteatus de Geer; Diptera: Syrphidae) and the common green lacewing (Chrysoperla carnea Stephens; Neuroptera: Chrysopidae) and (ii) to test whether intraspecific plant variation affects predator performance. The predators were fed either B. brassicae or Myzus persicae Sulzer (Homoptera: Aphididae) reared on four white cabbage (Brassica oleracea L. convar. capitata var. alba) cultivars that have previously been shown to differ in their glucosinolate profiles and resistance to herbivores (Broekgaarden et al., 2008; Poelman et al., 2008; Kabouw et al., 2010). Brevicoryne brassicae and M. persicae differ in the concentration of glucosinolates, as well as in the presence of toxic hydrolysis products of these glucosinolates. Brevicoryne brassicae sequesters glucosinolates from the phloem of its host plant and contains aphid-produced myrosinases (Jones et al., 2001; Francis et al., 2002). Upon tissue damage by carnivores, the sequestered glucosinolates in B. brassicae come into contact with aphid myrosinases, causing the formation of toxic breakdown products (Bridges et al., 2002; Kazana et al., 2007). Myzus persicae does not sequester glucosinolates, but excretes them in the honeydew, and does not contain myrosinases that could break down the glucosinolates that are present in the gut into toxic breakdown products (Francis et al., 2001b). Aphid predators will therefore encounter high concentrations of toxic glucosinolate breakdown products when feeding on B. brassicae, but not when feeding on M. persicae. We hypothesised that, based on the difference in concentrations of glucosinolates and their breakdown products, the performance of C. carnea and E. balteatus will be lower when fed B. brassicae than when fed M. persicae. Furthermore, we expect that variation in glucosinolate composition among the host-plant cultivars would affect the glucosinolate composition of the aphids feeding on these cultivars, as well as the possible formation of breakdown products (which depends on the side chain identity, see above), and thereby the performance of the natural enemies feeding on these aphids.
Effects of sequestration of glucosinolates by B. brassicae on aphid predators (mainly lady bird beetles and hoverflies) have been tested before (see, for example, Francis et al., 2001b; Kazana et al., 2007; Pratt, 2008). Previous studies often (i) only reported the glucosinolate concentrations in the host plant, and not in the prey insect itself, or (ii) linked glucosinolate concentrations in total leaf material to glucosinolate concentrations in the aphid, whereas aphids do not chew leaf tissue, but feed on phloem sap exclusively. Our study presents novel data because we have analysed glucosinolate concentrations as well as profiles in the phloem sap and the aphids feeding on this phloem sap. This resulted in a detailed comparative analysis of the effects of glucosinolate sequestration of a specialist aphid on two of its main predator species, of which one, the green lacewing, has not yet been the subject of testing effects of glucosinolates on its performance. We discuss the effects of intraspecific plant variation in glucosinolate composition on aphid–predator interactions in the context of a possible conflict between direct and indirect plant resistance.
Materials and methods
Plants and insects
We used four white cabbage (Brassica oleracea L. convar. capitata var. alba) cultivars: Christmas Drumhead and Badger Shipper (Centre for Genetic Resources, CGN, Wageningen, The Netherlands), representing open pollinated cultivars, and Lennox and Rivera (Bejo Zaden BV, Warmenhuizen, The Netherlands), representing more recently cultivated F1 hybrids. Plants were cultivated in a greenhouse compartment at 20 ± 2 °C, 60–70% RH and a LD 16 : 8 h photoperiod. When the light dropped below 200 W m−2 during the photoperiod, supplementary illumination was provided by sodium lamps (SON-T Philips, Eindhoven, The Netherlands). Seeds were germinated on peat soil (Lentse potgrond, no. 4, Lent, The Netherlands) and after 8 days, individual seedlings were transferred to the same peat soil in 1.45-l pots. All plants were watered daily and were fertilised by applying Kristalon Blauw (Hydro Agri, Rotterdam, The Netherlands) (N-P-K-MgO) 19–6–20–3 (2.5 mg l−1) to the soil twice a week from the age of 4 weeks onwards. Six-week-old plants were used in the experiments.
Brevicoryne brassicae and M. persicae were reared on the four white cabbage cultivars and Brussels sprouts (Brassica oleracea L. convar. gemmifera cv. Cyrus) in greenhouse compartments at 22 ± 2 °C, 60–70% RH and a LD 16 : 8 h photoperiod. Fresh plants were provided on a weekly basis and plants were watered every other day. The M. persicae and B. brassicae cultures were originally collected from B. oleracea in the vicinity of Wageningen (The Netherlands) in 2004 and 2008 respectively.
Episyrphus balteatus pupae and C. carnea eggs were provided by Koppert Biological Systems (Berkel en Rodenrijs, The Netherlands) and kept in a greenhouse compartment at 22 ± 2 °C, 60–70% RH and a LD 16 : 8 h photoperiod. Episyrphus balteatus pupae were kept in gauze cages (67 × 50 × 67 cm) and adults emerging from the pupae were provided with water, sugar, and bee-collected pollen provided by Koppert Biological Systems. Females were allowed to lay eggs on Brussels sprouts plants infested with either B. brassicae or M. persicae. After egg hatch, neonate larvae of both species were used in the experiments. (Note: it has been recently discovered that C. carnea is actually a complex of many cryptic sibling species (Henry et al., 2002), and the specific C. carnea we used was identified by Koppert Biological Systems as the sibling species C. affinis).
All experiments were performed in a greenhouse compartment at 22 ± 2 °C, 60–70% RH and a LD 16 : 8 h photoperiod. Aphid performance on the four white cabbage cultivars was assessed, because it is known that the performance of a natural enemy is often positively correlated with the performance of its host or prey (Benrey et al., 1998; Sznajder & Harvey, 2003). Data on performance of B. brassicae on the four white cabbage cultivars were derived from Broekgaarden et al. (2008). We assessed performance of M. persicae according to the protocol of Broekgaarden et al. (2008) to allow for direct comparisons of performance of both aphid species on the selected cultivars. Of each cultivar, 15 6-week-old plants were infested with M. persicae by placing one neonate nymph on each of five older leaves. Plants were placed individually in gauze nets and distributed randomly over the greenhouse. Plants were watered every other day and fertilised weekly in the same way as described under Plants and insects. Nymphs were monitored daily to estimate their development time (number of days between birth and reproduction) and nymphal survival was scored on day 11, the day by which most of the individuals had reproduced. From day 11 onwards, the number of aphids on each plant was recorded twice a week until day 34 of the experiment. Aphid multiplication factor was calculated by setting the number of aphids at day 11 of the experiment to 1.
In order to examine glucosinolate concentrations in B. brassicae feeding on different cultivars, and to link this to glucosinolate concentrations in the phloem of these cultivars, we infested 10 6-week-old plants of each cultivar with hundreds of neonate nymphs by allowing adults to reproduce on the plant for 24 h, after which the adults were removed from the plants. When nymphs reached the third instar, half of the aphids on each plant were collected for glucosinolate analysis, resulting in 10 replicates of several hundred nymphs per cultivar. At the same time, we collected phloem exudate from the third youngest fully expanded leaf of each of these 10 plants per cultivar. For phloem collection we followed the procedure of Bezemer et al. (2005) with minor adaptations: we used 2 ml 8 mM EDTA solution, initially placed the petiole of the leaf for 5 min in the EDTA solution to remove any plant chemicals from the incision, and afterwards placed the petiole for 2 h in a new vial with 2 ml EDTA solution. Our method inherently sampled a small amount of mesophyll fluids mixed with the phloem sap. Therefore, if we refer to phloem sap we mean phloem sap plus these potential contaminants from the mesophyll. Subsequently, the leaf was dried at 80 °C for 3 days and its dry weight was measured on a balance (Mettler-Toledo PM200, Tiel, The Netherlands). When the aphids reached adulthood, the remaining aphids were collected for glucosinolate analysis, resulting in 10 replicates of several hundred adult aphids per cultivar. At the same time, additional phloem samples were taken from the fourth youngest fully expanded leaf of each of these 10 plants per cultivar. Glucosinolate concentrations were analysed separately for third-instar nymphs and adult aphids to allow for investigation of differences in glucosinolate sequestration between nymphs and adult aphids.
All samples were frozen at −20°C immediately after collection. Aphid samples were freeze-dried, weighed and ground to a fine powder. Approximately 50 mg of the ground material of third instar nymphs and 100 mg of adult aphids was weighed into a micro-centrifuge tube. Glucosinolates were extracted and purified by using the methods of van Dam et al. (2004) and Kabouw et al. (2010) and glucosinolate content was assessed by high-performance liquid chromatography (HPLC). Glucosinolate detection was performed with a photodiode array detector with 229 nm as the integration wavelength. Different concentrations of sinigrin (Acros, New Jersey) were used as an internal standard. The correction factors at 229 nm from Buchner (1987) and the European Community (1990) were used to calculate the concentrations of the glucosinolates. We identified desulphoglucosinolate peaks by comparison of HPLC retention times and ultraviolet spectra with standards provided by M. Reichelt (Max Planck Institute for Chemical Ecology, Jena, Germany) and a certified rapeseed standard (Community Bureau of Reference, Brussels, Belgium, code BCR-367 R).
To extract glucosinolates from the phloem, we used a modified protocol. One millilitre from the collected 2 ml of EDTA solution was used for glucosinolate extraction. The solution was boiled in a water bath at 70 °C, subjected to an ultrasonic bath for 15 min to inactivate myrosinase activity, and afterwards transferred directly onto a Sephadex column. To concentrate the samples, after elution the freeze-dried eluate was resuspended in 100 µl instead of 1000 µl water.
Performance of predators was tested in no-choice situations. We assessed the performance of both E. balteatus and C. carnea fed either M. persicae or B. brassicae reared on each of the four white cabbage cultivars, thus creating eight prey species × plant cultivar combinations. For each predator species 40 individual predator larvae were monitored for each of the eight combinations, set up in a randomised design. Neonate larvae were transferred individually to a Petri dish (diameter 9 cm) with filter paper on the bottom by using a fine paintbrush. Larvae were fed ad libitum with aphids of mixed instars (representing the natural situation in which aphids occur in colonies of mixed ages) from the selected prey treatment that were provided together with a leaf fragment of the corresponding plant cultivar. Each day the prey items and leaf fragments were replaced. Survival rate, larva-to-adult development time, pupal fresh weight, sex, adult dry weight, head width, and wing length were measured. Adult dry weight was obtained by weighing adults that had been dried to constant weight at 80 °C for 3 days on a microbalance (Sartorius CP2P, Göttingen, Germany). Head width and wing length were measured using a stereo microscope (Olympus SZX12), attached to a digital camera (Euromex CMEX-1) and the program Image Focus (version 1.0).
Statistical analyses were performed using SPSS for Windows (15th edition, Chicago, Illinois), unless indicated otherwise. Nymphal survival of M. persicae was analysed by logistic regression in Gen Stat (12th edition, VSN International, Hemel Hempstead, UK). Development time was log-transformed to obtain normality, and analysed by anova and post hoc LSD test for cultivar comparisons. Aphid multiplication factor was log-transformed and repeated measures anova and post hoc LSD test for cultivar comparisons were used to assess the impact of different cultivars on the multiplication factor over time. Time was considered a within-subjects factor and cultivar a between-subjects factor. Differences in aphid multiplication factors over time between B. brassicae and M. persicae were analysed by repeated measures anova.
Differences in indole, aliphatic, and total glucosinolate concentrations in phloem and aphids among the cultivars were analysed by Kruskal–Wallis H-tests, as assumptions on normality were violated. Mann–Whitney U-tests with a Bonferroni correction for the number of comparisons (six) were used to compare the mean differences between the groups. Differences in indole, aliphatic, and total glucosinolate concentrations between both aphid developmental stages and between both phloem sampling dates were analysed by Mann–Whitney U-tests. Correlations between indole, aliphatic, and total glucosinolate concentrations in the phloem and the concentrations of these compounds in the aphids feeding on those plants were tested with Spearman's correlation test.
To analyse glucosinolate profiles of phloem and aphids, we used projection to latent structures-discriminant analysis (PLS-DA) and partial least squares projections to latent structures (PLS), in SIMCA-P (12th edition, Umetrics, Umeå, Sweden) (Eriksson et al., 2006). PLS-DA is a multivariate discriminant analysis that we used to test if glucosinolate profiles in the phloem of the different cultivars differed significantly and if glucosinolate profiles in the aphids feeding on these cultivars also differed significantly. PLS is a multivariate method for regression analysis that we used to test the relationship between glucosinolate profiles in phloem and glucosinolate profiles in aphids feeding on those plants. To pre-process data, glucosinolate concentrations were log-normalised, mean-centred, and scaled to unit variance.
Survival and sex ratio of predators were analysed with logistic regression in Gen Stat, including the factors prey species and plant cultivar. T-probabilities were calculated to test pairwise differences between means. Development time, adult dry weight, wing length, and head width were analysed using a three-way multivariate analysis of variance (manova), including the factors prey species, plant cultivar and sex of the predator. manova is used to test difference among groups for multiple dependent variables simultaneously. Besides this ‘overall effect', manova also provides results from the univariate analysis for each individual performance parameter. Pupal weight was not included in the manova, because sex of the pupae could not be determined, and was therefore analysed only by a two-way anova on the factors prey species and plant cultivar. If necessary to obtain normally distributed data, log-transformation was applied. Significant differences amongst prey treatments were further analysed with a post hoc Tukey test.
Nymphal survival of M. persicae was on average 63 ± 3% (mean ± SE) and did not differ among cultivars (logistic regression, d.f. = 3,deviance ratio = 1.39,P = 0.256). Development time differed among the cultivars (anova,F3,181 = 7.71,P < 0.001). Nymphs on cultivar Lennox developed slower than on the other three cultivars [post hoc LSD tests; development time on Lennox 13.0 ± 0.4 days (mean ± SE), on the other three cultivars on average 11.1 ± 0.2 days]. Aphid multiplication factor increased over time (repeated measures anova,F6,336 = 337.20,P < 0.001) and was different between the cultivars (repeated measuresanova,F3,56 = 5.55;P = 0.002). At the end of the experiment (day 34), the multiplication factor of M. persicae was more than twice as high on Christmas Drumhead and Badger Shipper (107 ± 30 and 100 ± 17 per plant respectively) than on Rivera and Lennox (39 ± 10 and 47 ± 9 per plant respectively).
Brevicoryne brassicae multiplication factors increased faster than M. persicae multiplication factors (repeated measures anova,F1,127 = 152.16;P < 0.001) (Fig. 1). The ranking of the four white cabbage cultivars in terms of aphid multiplication factors of B. brassicae on these cultivars was similar to that of M. persicae (Fig. 1).
Glucosinolates in phloem of white cabbage plants. At both phloem sampling times, phloem samples contained higher concentrations of indole glucosinolates than aliphatic glucosinolates, as analysed for all four cultivars combined (Mann–Whitney U,first sampling : U = 156.00,P < 0.001; second sampling: U = 309.00,P < 0.001; Table 1). Aromatic glucosinolates were not detected in the phloem. There was no difference in total glucosinolate concentrations in phloem between both sampling times (Mann–Whitney U,U = 551.00,P = 0.053; Table 1).
Table 1. Mean (±SE) concentration (n = 40) of glucosinolate compounds detected in the phloem of white cabbage (averaged over four cultivars) at two sampling dates and in Brevicoryne brassicae nymphs and adults.
There were no differences in concentrations of total, indole and aliphatic glucosinolates in the phloem among the different cultivars (Kruskal–Wallis H-test, d.f. = 3, n = 10 per cultivar, P > 0.05 for all analyses), except for the second sampling time when the phloem of Badger Shipper plants contained higher total glucosinolate concentrations than Lennox (Mann–Whitney U,U = 11.00, P = 0.004). Glucosinolate profiles in the phloem were not different among the cultivars (no significant PLS-DA components could be extracted).
Glucosinolates in Brevicoryne brassicae. In third instar and adult aphids, concentrations of aliphatic glucosinolates were higher than concentrations of indole glucosinolates (Mann–Whitney U,L3 aphids : U = 255.00, P < 0.001; adult aphids: U = 14.00,P < 0.001; Table 1; Fig. 2). This contrasts to phloem samples, in which indole glucosinolate concentrations were higher than those of aliphatic glucosinolates. Aromatic glucosinolates were not detected in B. brassicae. Adult aphids contained on average two times higher total concentrations of glucosinolates than third instar nymphs when averaged over all cultivars (Mann–Whitney U,U = 326.00,P < 0.001; Table 1; Fig. 2), although adult aphids on Christmas Drumhead contained only about 70% more glucosinolates than third instar nymphs (Fig. 2). Compared with third instar nymphs, adult aphids contained higher concentrations of aliphatic and lower concentrations of indole glucosinolates when averaged over all cultivars (Mann–Whitney U, aliphatic: U = 269.00, P < 0.001;indole : U = 520.00,P = 0.024; Table 1; Fig. 2).
In both aphid developmental stages, there were no differences in concentrations of indole, aliphatic and total glucosinolates among aphids reared on the different cultivars (Kruskal–Wallis H-test, d.f. = 3, P > 0.05 for all analyses; Fig. 2). In contrast to the total glucosinolate concentration, the glucosinolate profiles of third instar nymphs did differ among nymphs reared on the different cultivars (3PLS − DA principal components,R2Xcum = 0.811, R2Ycum = 0.363, Q2cum = 0.201; Fig. 3a). PLS-DA mostly separated glucosinolate profiles of aphids reared on Christmas Drumhead and Badger Shipper from profiles of aphids reared on Rivera and Lennox (Fig. 3a). Figure 3b shows the contribution of the glucosinolate compounds to the discrimination among the aphid groups, based on the first two principal components. Glucosinolate profiles in adult aphids did not differ among aphids feeding on different cultivars (no significant PLS-DA components could be extracted).
Correlations between glucosinolates in phloem and Brevicoryne brassicae. No correlations in indole, aliphatic, and total glucosinolate concentrations between the phloem of a plant and aphids feeding on that plant were found for both aphid developmental stages (Spearman's correlation, rs values were between −0.039 and 0.216 and P > 0.05 for all correlations). Furthermore, we did not find a relationship between glucosinolate profiles in the phloem and profiles in the aphids feeding on these plants at both sampling times (no significant PLS components could be extracted).
Survival. Survival of C. carnea until adult emergence was on average 92%. Prey species, plant cultivar or the interaction between both did not affect survival of C. carnea (logistic regression, P > 0.05 for both factors and the interaction). Survival of E. balteatus until adult emergence was on average 60%. Survival of E. balteatus was affected by prey species (logistic regression, d.f. = 1, deviance ratio = 11.79, P < 0.001) as it was lower when fed B. brassicae (49% survival) than when fed M. persicae (70% survival). Plant cultivar or its interaction with prey species had no effect on E. balteatus survival (logistic regression, P > 0.05 for both analyses).
Development time, adult weight, and adult size. Prey species, plant cultivar, and predator sex affected the performance of both predator species in terms of development time and adult weight and size (Table 2). Although for E. balteatus interactions between the factors were observed (Table 2), effects on performance were mostly consistent for both predator species. In general, both predator species developed faster, although into smaller adults, when fed the generalist M. persicae compared with the specialist B. brassicae (Figs 4 and 5). For C. carnea, development was fastest and adults were larger when fed aphids (both the specialist and the generalist aphid) reared on Christmas Drumhead and Badger Shipper, while development was slowest and adults were smaller when fed aphids reared on Rivera and Lennox (Figs 4 and 5). For E. balteatus, the effect of plant cultivar on development time and adult size was less consistent and depended on the prey species (Table 2), but mostly similar cultivar effects were observed as for C. carnea (Figs 4 and 5).
Table 2. Effects of prey species, plant cultivar, predator sex, and their interactions on performance parameters of Chrysoperla carnea and Episyrphus balteatus as analysed by manova.
Effect of predator sex. Sex ratios were 48% females for C. carnea and 52% females for E. balteatus and did not differ among the different prey species × plant cultivar combinations (logistic regression, P > 0.05 for all combinations). The sex of the predatory larvae affected their performance (Table 2). Irrespective of prey species or host plant, C. carnea females developed slower than males [24.3 ± 0.2 days (mean ± SE) and 23.8 ± 0.2 days respectively] and developed into heavier adults than males (2.04 ± 0.02 and 1.72 ± 0.02 mg respectively). Development times of E. balteatus females and males (20.0 ± 0.3 and 20.1 ± 0.3 days respectively) did not differ, but only when fed B. brassicae, E. balteatus females developed into lighter adults than males (3.20 ± 0.10 and 3.91 ± 0.15 mg respectively).
Prey species effect
The two predator species, C. carnea and E. balteatus, exhibited slower development and E. balteatus exhibited lower survival when fed the specialist herbivore B. brassicae, than when fed the generalist herbivore M. persicae. Although the two aphid species probably differ in many ways, and displayed differential population growth on white cabbage, we propose that the observed difference in predator performance can be attributed to an important extent to the difference in concentrations of glucosinolates and their hydrolysis products between the prey species (Gols & Harvey, 2009). Brevicoryne brassicae sequesters glucosinolates in high concentrations [100–150 times higher than in the phloem of its host plant, according to Hopkins et al. (2009)] and contains its own aphid-specific myrosinase that hydrolyses the glucosinolates in its body upon damage by natural enemies (Jones et al., 2001; Francis et al., 2002). Myzus persicae does not sequester glucosinolates, and excretes the ingested glucosinolates in the honeydew (Francis et al., 2001b), but it does contain glucosinolates in the gut that could potentially harm predators. However, M. persicae does not contain myrosinases that could break down the glucosinolates into toxic breakdown products. Therefore, predators face much higher concentrations of glucosinolates and glucosinolate hydrolysis products when feeding on B. brassicae than when feeding on M. persicae, which seems to correlate negatively with predator performance as observed in a separate experiment.
Interestingly, we document that the sequestration of glucosinolates by B. brassicae from the host plant's phloem is selective. While indole glucosinolates dominated in the phloem of white cabbage plants, aliphatic glucosinolates dominated in the aphids. (Note that our method of collecting phloem sap inherently sampled a small amount of mesophyll fluids mixed with the phloem sap.) Total glucosinolate concentrations were more than two times higher in adult aphids than in third-instar nymphs, confirming the observation by Kazana et al. (2007) that B. brassicae continues to sequester glucosinolates during its development. Furthermore, from the third instar to the adult stage, aliphatic glucosinolates were sequestered in higher concentrations, while concentrations of indole glucosinolates decreased. The high sequestration of aliphatic glucosinolates, but not of indole glucosinolates, can be explained by the difference in toxicity between both. Formation of toxic hydrolysis products of plant aliphatic glucosinolates is prevented in aphids due to the intercellular path taken by aphid stylets to reach the phloem (Tjallingii & Hogen Esch, 1993), thus allowing aphids to ingest phloem glucosinolates without bringing these compounds into contact with plant myrosinases (Andreasson et al., 2001; de Vos et al., 2007; Kim & Jander, 2007). Indole glucosinolates, in contrast, have been shown to be broken down to toxic hydrolysis products independently of myrosinase activity (Kim & Jander, 2007; Kim et al., 2008). Indole glucosinolates may, therefore, be detrimental for B. brassicae, as was suggested by Cole (1997), providing an explanation for the low sequestration of these glucosinolates. Aliphatic glucosinolates undergo fast enzymatic degradation by purified aphid myrosinase (Francis et al., 2002), and higher sequestration of aliphatic glucosinolates by B. brassicae may therefore lead to higher toxicity to predators, without affecting aphid performance itself.
The mechanism underlying the high sequestration of aliphatic glucosinolates, but low sequestration of indole glucosinolates by B. brassicae could be that transporters for glucosinolate uptake in insects may be rather specific. While the glucosinolate-sequestering sawfly Athalia rosae sequesters mostly aliphatic glucosinolates, and almost no indole glucosinolates (Müller, 2009), another species of this genus, A. liberta, has been shown to be able to sequester indole glucosinolates (Opitz et al., 2010). Unfortunately, nothing is known yet about the exact mechanisms underlying glucosinolate sequestration or the specificity of glucosinolate transporters (Opitz et al., 2010).
Although development of both predator species was slower and survival of E. balteatus was lower when fed B. brassicae, predator size was larger when feeding on this aphid species. Predator size is generally positively correlated with lifetime fecundity, as was shown for E. balteatus (Branquart & Hemptinne, 2000). Fitness of an individual, however, is not necessarily affected equally by the different performance parameters we have measured. Aphid species are so-called r-selected, colonising species, characterised by fast development, fluctuating population densities in response to changing environments and abundant offspring production (Lewontin, 1965; Caswell & Hastings, 1980; Carter & Dixon, 1981; Ankersmit et al., 1986). Under such selection pressures an increase in developmental rate has a more pronounced effect on an individual's fitness than an increase in lifetime fecundity (Caswell & Hastings, 1980). If natural enemies of aphids are under the same selection pressures as their prey, short development time would also be favoured over offspring production, as has been suggested for aphid parasitoids (Sequeira & Mackauer, 1994 and references therein). We assume, therefore, that development time and survival of aphid predators are more important components of their fitness than adult size.
Plant cultivar effect
The performance of a natural enemy is often positively correlated with the performance of its host or prey (Benrey et al., 1998; Sznajder & Harvey, 2003). In support of this, the performance of C. carnea and E. balteatus when fed aphids reared on the four cultivars reflected the performance of the aphids themselves on these cultivars: aphids and predators performed better on Christmas Drumhead and Badger Shipper than on Rivera and Lennox (Figs 1, 4 and 5). However, in the case of B. brassicae, not only aphid performance itself, but also the glucosinolate content of the aphids has likely influenced predator performance, because glucosinolate concentrations reached high levels in these aphids. These high concentrations of glucosinolates likely led to high concentrations of glucosinolate hydrolysis products after breakdown by the aphid myrosinase, although we did not quantify glucosinolates and their hydrolysis products separately. Predators developed generally faster and into larger adults when fed B. brassicae reared on Christmas Drumhead and Badger Shipper than when fed B. brassicae reared on Rivera and Lennox. These two distinct groupings in predator performance matched the groupings of glucosinolate profiles of the aphids reared on the different plant cultivars, but not the total indole, total aliphatic or overall total glucosinolate concentrations. Although we believe that the differences in glucosinolate profiles, rather than total concentrations of glucosinolates, among aphids developing on the different cultivars influenced predator performance, we can not rule out potential effects of other aspects of aphid quality on predator performance. In a study with A. rosae, haemolymph of the larvae deterred ants and predatory wasps more strongly than the individual major glucosinolate compounds in the haemolymph (Müller et al., 2002; Müller & Brakefield, 2003). It is, however, not clear whether this stronger deterrence was due to other glucosinolate compounds in the haemolymph, in agreement with our hypothesis that glucosinolate profiles rather than total concentrations influence natural enemies, or whether this stronger deterrence was due to completely different compounds.
Our study shows that not only the prey species but also the plant cultivar can have an effect on the performance of predators. When fed B. brassicae populations reared on different plant cultivars, differences in predator performance matched differences in glucosinolate profiles among the aphids, although we did not test other aspects of aphid chemistry. Predator performance, in terms of survival and development time, was lower when predators were fed the specialist aphid B. brassicae that selectively sequestered glucosinolates from its host plant and contains its own aphid-specific myrosinase, than when fed the non-sequestering M. persicae. Breakdown products of glucosinolates are known to confer direct resistance against a wide variety of herbivores, and our results imply that aphid glucosinolates and their hydrolysis products negatively affected aphid predators. This suggests that in the tritrophic system tested, in which glucosinolates are the main secondary metabolites, there might be a conflict between direct and indirect resistance.
The conflict between direct and indirect resistance may not be a general trend in nature and may, for example, not arise for specialist natural enemies. Specialist natural enemies that are adapted to feeding on glucosinolate-containing herbivores, such as the aphid parasitoid Diaeretiella rapae, are probably not affected negatively by higher concentrations of glucosinolates in their hosts (Le Guigo et al., 2011). Parasitoid wasps such as D. rapae have been shown to be attracted to volatile breakdown products of glucosinolates that are indicators of host presence (Bradburne & Mithen, 2000; Blande et al., 2007). For attraction of natural enemies by volatile breakdown products of glucosinolates, not only total concentrations of glucosinolates are important, but also the side-chain of the glucosinolate, as breakdown of aliphatic glucosinolates leads to higher emission of volatiles than breakdown of indole glucosinolates (Hopkins et al., 2009).
Our findings suggest that plants might be subject to a conflict between the consequences of producing higher or lower levels of glucosinolates. Higher levels enhance resistance against herbivores and attract specialist natural enemies; lower levels improve the performance of generalist natural enemies, possibly leading to balancing selection on plant glucosinolate levels. This hypothesis may be tested by varying only levels and profiles of glucosinolate hydrolysis products in the diet of natural enemies, either by using artificial diets containing glucosinolates that are offered in combination with myrosinases, or by rearing host and prey on plants that have been genetically engineered to produce modified glucosinolate levels and profiles.
We thank two anonymous reviewers for constructive comments on an earlier version of the manuscript; Colette Broekgaarden for the permission to use B. brassicae performance data; Rieta Gols, Roland Mumm, and Erik Poelman for help with the statistics; Koppert Biological Systems for the delivery of E. balteatus pupae and C. carnea eggs; CGN and Bejo Zaden for providing seeds of the cabbage cultivars and Unifarm, especially Andre Maassen and Alex Super, for plant rearing. This work was supported by a grant from the Earth and Life Sciences Council of the Netherlands Organization for Scientific Research (NWO-ALW) under the ERGO program (number 838.06.010).