1 3

Oecologia (2014) 174:893–907DOI 10.1007/s00442-013-2817-2

Plant-mIcrObe-anImal InteractIOns - OrIgInal research

Bottom‑up and top‑down herbivore regulation mediated by glucosinolates in Brassica oleracea var. acephala

Serena Santolamazza‑Carbone · Pablo Velasco · Pilar Soengas · María Elena Cartea

received: 17 July 2013 / accepted: 21 October 2013 / Published online: 19 December 2013 © springer-Verlag berlin heidelberg 2013

(coleoptera, chrysomelidae) were not affected by the plant genotype. Parasitism rates of M. brassicae larvae and E. ornatum eggs were affected by plant genotype. the ratio of generalist/specialist herbivores was positively correlated with parasitism rate. although both top-down and bottom-up forces were seen to be contributing, the key factor in shaping both herbivore performance and parasitism rate was the glucosinolate concentration, which highlights the impact of bottom-up forces on the trophic cascades in crop habitats.

Keywords brassicaceae · herbivory · glucosinolates · Parasitoids · tritrophic interactions

Introduction

Plant defence strategies against insect herbivores may involve the synthesis of a plethora of biologically active compounds (allelochemicals) which are phylogeneti-cally conserved in specific plant families or genera (mith-öfer and boland 2012). many compounds act directly on the herbivores (bottom-up control), whereas others act indirectly, via the attraction of organisms from other trophic levels (i.e. parasitoids and predators) which, in turn, protect the plants (plant-mediated top-down con-trol) (Ode 2006). the genus Brassica (brassicaceae) has a sophisticated two-part defence system involving glu-cosinolate compounds and a myrosinase protein complex. the enzyme myrosinase breaks down glucosinolates into toxins (isothiocyanates, nitriles) upon leaf tissue dam-age (hopkins et al. 2009). glucosinolates derived from phenylalanine or tyrosine (aromatic), and those derived from alanine, valine, leucine and isoleucine (aliphatic) are typical of the brassicales, although they have also been

Abstract Quantitative differences in plant defence metabolites, such as glucosinolates, may directly affect herbivore preference and performance, and indirectly affect natural enemy pressure. by assessing insect abun-dance and leaf damage rate, we studied the responses of insect herbivores to six genotypes of Brassica oleracea var. acephala, selected from the same cultivar for having high or low foliar content of sinigrin, glucoiberin and glucobras-sicin. We also investigated whether the natural parasitism rate was affected by glucosinolates. Finally, we assessed the relative importance of plant chemistry (bottom-up con-trol) and natural enemy performance (top-down control) in shaping insect abundance, the ratio of generalist/specialist herbivores and levels of leaf damage. We found that high sinigrin content decreased the abundance of the generalist Mamestra brassicae (lepidoptera, noctuidae) and the spe-cialist Plutella xylostella (lepidoptera, Yponomeutidae), but increased the load of the specialist Eurydema ornatum (hemiptera, Pentatomidae). Plants with high sinigrin con-tent suffered less leaf injury. the specialist Brevicoryne brassicae (hemiptera, aphididae) increased in plants with low glucobrassicin content, whereas the specialists Pieris rapae (lepidoptera, Pieridae), Aleyrodes brassi-cae (hemiptera, aleyrodidae) and Phyllotreta cruciferae

communicated by merijn Kant.

Electronic supplementary material the online version of this article (doi:10.1007/s00442-013-2817-2) contains supplementary material, which is available to authorized users.

s. santolamazza-carbone (*) · P. Velasco · P. soengas · m. e. cartea misión biológica de galicia, consejo superior de Investigaciones científicas (mbg-csIc), P.O. box 28, 36080 Pontevedra, spaine-mail: anaphes@gmail.com

894 Oecologia (2014) 174:893–907

1 3

found outside the order, among non-cruciferous dicotyle-donous angiosperms (Fahey et al. 2001). Indole (synthe-sized from tryptophan) and methionine-derived aliphatic glucosinolates are unique to the brassicales, the latter being a group of metabolites characteristic of the fam-ily brassicaceae (bekaert et al. 2012). the synthesis of indole glucosinolates, such as glucobrassicin and neoglu-cobrassicin, tends to be induced by herbivory and fungal infection through jasmonate or other signalling pathways, whereas aliphatic glucosinolates, such as sinigrin and glu-coiberin, tend to be constitutively expressed (harvey et al. 2011). however, it has also been reported that aliphatic compounds may decrease after herbivore attacks (Velasco et al. 2007), and increase after jasmonate induction (Fritz et al. 2010) or by below-ground herbivory (soler et al. 2005). thus, patterns and relative concentrations of these chemicals are consistently subject to variation depending on genetic and environmental factors (Poelman et al. 2008; lankau and Kliebenstein 2009).

glucosinolates may act as a potent feeding deter-rent for generalist insect species, as their toxicity causes developmental and fitness damage. For insects special-ized in brassicaceous plants, however, they may act as oviposition and feeding stimulants (kairomonal role). the toxic effect on the specialist herbivores are circumvented by excretion, detoxification, sequestration and behav-ioural responses (hopkins et al. 2009). nonetheless, the existence of qualitative and quantitative variation of phy-tochemicals among plant genotypes, tissues and ontoge-netic stages still challenges insect survival (Ode 2006). Furthermore, glucosinolate breakdown products are also used by the natural enemies of herbivorous insects, such as parasitoids, as cues for host location. there is evi-dence that levels of attraction vary between parasitoids, however, and consequently the nature of these secondary metabolites may significantly affect their performance (turlings and benrey 1998; gols and harvey 2009). the net effect of glucosinolates on the host-parasite interac-tion is complex. On the one hand, glucosinolates may reduce the immune responses of the hosts, producing a positive effect on parasitoid survival (bukovinszky et al. 2009), but on the other hand plant metabolites may also negatively affect parasitoid survival, through the direct ingestion of harmful phytochemicals from the herbivore and from reducing host quality (gols and harvey 2009). consequently, plants may face a conflict between produc-ing high or low levels of glucosinolates: higher levels can enhance resistance against generalist insect herbivores, but at the same time may attract co-evolved specialists, and also be harmful to the natural enemies of these herbi-vores. lower levels, on the other hand, may increase the abundance of generalist herbivores (lankau 2007; Kos et al. 2011a).

In Brassica oleracea var. acephala l. (kale), two ali-phatic compounds (sinigrin, glucoiberin) and one indole compound (glucobrassicin) dominate the glucosinolate pattern (Velasco et al. 2007; cartea et al. 2008). Previous works have focused on the impact of glucosinolates in insect-plant interactions in Brassica ecosystems, by study-ing: (1) different populations of wild B. oleracea with qual-itative differences in glucosinolate patterns (specifically the presence/absence of aliphatic glucosinolates) (newton et al. 2009a, b, 2010); (2) a wild population of Brassica nigra and cultivated varieties of B. oleracea with high or low total glucosinolate concentration (Poelman et al. 2008); (3) different cultivars of B. oleracea with quantitative dif-ferences in glucosinolate profiles (Poelman et al. 2009); and (4) wild species and cultivated varieties of B. oleracea, with quantitative differences in glucosinolate profiles (i.e. high vs. low levels) (gols et al. 2008; harvey et al. 2011). It is known that cultivars of Brassica species have low lev-els of glucosinolates in leaf tissue compared with wild pop-ulations, which justifies the comparison between wild and cultivated species (gols and harvey 2009). however, it is recognized that other plant traits, such as morphology, phe-nology, primary chemistry and physiology, related to their different origins, life histories and genetic backgrounds, could also play a role in insect responses (carmona et al. 2011). at present, relatively few studies have used artifi-cial selection to create lines of brassicaceous plants with different glucosinolate profiles, and those that did usually only found quantitative variation of sinigrin (lankau 2007; lankau and strauss 2008; lankau and Kliebenstein 2009). In the present study, we performed a divergent selection from a local variety of kale, obtaining six plant genotypes which shared the same phenotype but which differed in glu-cosinolate patterns, having high or low content of aliphatic (sinigrin and glucoiberin) and indole (glucobrassicin) glu-cosinolates. We focused on insect responses to quantitative variation in glucosinolates (bottom-up forces), by sampling a wide range of naturally occurring specialist and generalist insect herbivores. In particular, generalists were expected to be most sensitive to high sinigrin concentration, as indi-cated in the literature (Ode 2006). On the other hand, the role of plant chemistry in attracting natural enemies (top-down forces), such as parasitoids, was assessed through the evaluation of parasitism rate. We also determined the extent to which the cost of the synthesis of secondary defence metabolites is translated into benefits (in terms of reduced herbivory), by assessing the variability of leaf damage rates among plant genotypes. Finally, we investigated whether the differences between plant groups in terms of herbi-vore abundance, the ratio between generalist and special-ist herbivores, and leaf damage rates, could principally be explained by parasitism rate (top-down forces) or by plant constitutive defences (bottom-up forces).

895Oecologia (2014) 174:893–907

1 3

Materials and methods

Plant source

Divergent selection was started in 2006 by using seeds of the kale population mbg-brs0062 (cycle 0), kept at the Brassica germplasm bank at misión biológica de galicia (galicia, northwest spain). this population is a local vari-ety which represents the kale germplasm grown in north-west spain. the objective was to obtain six plant groups which had high or low concentrations of sinigrin, high or low glucoiberin, and high or low glucobrassicin content. In 2006, approximately 750 plants (cycle 0) were transplanted outside into six cages (125 plants each), and fenced with fine-mesh walls to ensure isolation conditions. the leaf glucosinolate content (see details below) of all the plants was assessed 120 days after sowing. In each cage, 20 plants with an extreme content of the relevant glucosinolate (i.e. the highest or the lowest concentration) were selected (20 % selection intensity), and all remaining plants were destroyed before flowering. because kale is an allogamous crop, cross-pollination among the selected plants in each cage was obtained using bumblebees. In 2007, an equal number of seeds were taken from the selected plants of the cycle 0, for each divergent selection, to create the cycle 1 generation (125 plants per cage). according to the proto-col adopted for cycle 0, only those plants which showed an extreme leaf glucosinolate content were selected (20 plants per cage). From 2008 to 2009, this process was repeated for two successive generation cycles. a recent investiga-tion, still unpublished, has recorded the absence of signifi-cant differences in biomass and phenology between the six plant groups. thus, we can reasonably conclude that the main differences among genotypes were due to differences in the glucosinolates subjected to selection, although the possibility exists that other plant traits may also combine with the selected glucosinolates to further influence insect performance.

Insect herbivores

the study was focused on insect herbivores feeding on kale leaves. In northwest spain, the most common lepidopteran herbivores are the generalists Mamestra brassicae l. (noc-tuidae), Autographa gamma l. (noctuidae) and Evergestis forficalis l. (Pyralidae), and the specialists Plutella xylos-tella l. (Yponomeutidae), Pieris rapae l. (Pieridae) and Pieris brassicae l. (Pieridae) (cartea et al. 2009). among hemipterans, the cabbage aphid Brevicoryne brassicae l. (hemiptera, aphididae), the whitefly Aleyrodes bras-sicae Walter (hemiptera, aleyrodidae) and the pentato-mid Eurydema ornatum l. (hemiptera, Pentatomidae), are also specialist herbivores of brassicaceous plants. the

abundance of the adults of the cabbage flea beetle Phyl-lotreta cruciferae goeze (coleoptera, chrysomelidae) was also assessed, because although the larvae feed only on roots and stems, adults feed on the foliage, producing small round holes.

experimental design

the study was conducted during 2011 and 2012 at mis-ión biológica de galicia. Plants of the six genotypes were grown in multi-pot trays in a greenhouse at 20 °c for 40 days and then transplanted into the field (salcedo, spain; 42°24′n, 8°38′W), at the five to six true leaf stage, on 15 april 2011 and a second batch on 26 march 2012. Plant varieties were evaluated in a randomized complete block design with six replications. each experimental block consisted of six rows of 25 plants each (one genotype per row, randomly assigned). rows were spaced 0.8 m apart and plants within rows were spaced 0.5 m apart. Field sam-plings were performed on 23 may, 23 June and 26 July 2011. a total of 1,080 observations were obtained (corre-sponding to 60 plants/genotype, three sampling dates and six genotypes). sampling was interrupted in august 2011 due to the critical conditions of the plants, which were severely affected by whiteflies, aphids and fungal diseases. In 2012, the samplings were carried out on 9 June, 27 July, 27 august, 27 september and 30 October. In 2012 a total of 1,800 observations were obtained (corresponding to 60 plants/genotype, five sampling dates and six genotypes). In november, no more insects were found in the field.

On each sampling date, ten plants per row were thor-oughly inspected for the presence of insect herbivores on leaves. however, the eggs of A. gamma, E. forficalis and P. xylostella and the mining first-instar larvae of P. xylos-tella, were not sampled due to their small size. also, M. brassicae pupae were not sampled because they develop while buried in the soil. Plants inspected on one sampling date were always left untouched on the following date. all the lepidopteran species studied are, in spain, bi- or mul-tivoltine, and we therefore assumed that sampling would not significantly reduce their abundance in the plots. Field-collected material (i.e. lepidopteran eggs, larvae and pupae, and E. ornatum eggs) was transported to the laboratory, identified, counted and placed in plastic 10-cm-diameter Petri dishes, labelled with the collection date, host plant, block and insect identity. larvae were reared individually in Petri dishes and fed with fresh kale leaves. lepidop-teran immature stages were reared until adulthood or until the emergence of parasitoids. eggs were maintained until larvae or parasitoids emerged. all rearing was carried out at room temperature (20 ± 2 °c) and under a natural pho-toperiod. because of the magnitude of the colonies of B. brassicae and A. brassicae, and of adults of E. ornatum

896 Oecologia (2014) 174:893–907

1 3

and P. cruciferae, their abundance was estimated by using a subjective 0–4 rating scale (0, absence; 1, up to five indi-viduals; 2, up to ten; 3, up to 50; 4, more than 50). abun-dance was expressed as the average rating scale measured on ten plants. the parasitism rate of B. brassicae and A. brassicae was not assessed because these hemipterans are more subject to predators (i.e. syrphid fly maggots, green lacewing larvae, anthocorid bugs and ladybird beetles) than to parasitoid control in the system studied (s. santol-amazza-carbone, personal observation). Furthermore, any study of the parasitism rate of these colonies would require several leaves to be removed, thus provoking an altera-tion of the plant architecture. E. ornatum and P. cruciferae adults do not suffer parasitoid attacks at this life stage. a 1–5 rating scale was used to evaluate the damage level of the plants (i.e. the overall amount of injuries caused by the whole herbivore complex), where level 1 represents a healthy plant, without any damage, and level 5 represents a completely damaged plant, with 90–100 % of the leaves attacked by herbivores. the damage level was calculated as the average rating measured on 15 plants.

the impact of parasitoids on insect host populations was calculated as the proportion of available hosts that had been parasitized per sampling date, block and host plant genotype. Unhatched host eggs, host pupae and unhatched parasitoid cocoons were dissected under the microscope in order to take into account any unemerged adult parasi-toids. Parasitoid taxonomical identity was ascertained by s. santolamazza-carbone, and voucher specimens have been conserved at misión biológica de galicia.

glucosinolate analysis

In every experimental plot, two lots of ten fresh leaves per plant genotype were collected. collecting dates were 7 June and 7 July 2011, and 10 July and 30 October 2012. samples were stored at −80 °c until prepared for analy-sis. glucosinolate profiles were determined by ultra high-performance liquid chromatography (UhPlc). sample extraction and desulphation were performed according to Kliebenstein et al. (2001) with minor modifications. Five microlitres of the desulpho-glucosinolate extract from leaves was used to identify and quantify glucosinolates. chromatographic analyses were carried out on an UhPlc (nexera lc-30aD; shimadzu) equipped with a nexera sIl-30ac injector and one sPD-m20a ultraviolet (UV)/visible photodiode array detector. the UhPlc column was an acquity UPlc hss t3 (1.8-μm particle size, 2.1 × 100 mm internal diameter; Waters, ma) protected with a Van guard UhPlc pre-column. the oven tempera-ture was set at 30 °c. compounds were detected at 229 nm and were separated by using the following method in aque-ous acetonitrile (b), with a flow of 0.4 ml min−1: 1.5 min

at 90 % ultrapure water (a), a 3.5-min gradient from 10 to 25 % (v/v) b, a 4-min gradient from 25 to 50 % (v/v) b, a 4.5-min gradient from 50 to 100 % (v/v) b, a 1-min gradient from 100 to 0 % (v/v) b, and a final 3 min at 90 % a. Data were recorded on a computer with the labsolu-tions software (shimadzu). specific glucosinolates were identified by comparing retention times with standards and by UV absorption spectra. glucosinolates were quantified at 229 nm by using sinigrin (sinigrin monohydrate; Phyto-plan, Diehm & neuberger, heidelberg) and glucobrassicin (glucobrassicin potassium salt monohydrate; Phytoplan, Diehm & neuberger) as an external standard and expressed in μmol g−1 dry weight. the regression lines were made with at least five data points, from 0.34 to 1.7 nmol for sinigrin and from 0.28 to 1.4 nmol for glucobrassicin. the average regression equations for sinigrin and glucobras-sicin were y = 148,818x (R2 = 0.99) and y = 263,822x (R2 = 0.99), respectively.

statistical analysis

sampling dates were analysed as independent events, irre-spective of the year, because of the different number of samplings performed in 2011 and 2012. the impacts of plant genotype and sampling date (fixed factors) and the interaction between them on the abundance of insect her-bivores and on leaf damage level, were investigated using a two-way anOVa. blocks were used as a random factor. Pairwise comparisons between two plant genotypes (i.e. high vs. low sinigrin, high vs. low glucoiberin and high vs. low glucobrassicin) and general comparisons (i.e. between the six genotypes), were also carried out. Insect count data were log10(x + 1) transformed prior to the analyses, whereas the insect abundance estimated by rating scales and the leaf damage level were arcsin square root trans-formed. Differences between means were assessed by a least significant difference test. the larvae of P. brassicae, E. forficalis and A. gamma, and the eggs of P. rapae were excluded from the statistical analyses because they were only sporadically found.

Differences in the ratio of generalist (i.e. M. brassicae, A. gamma, E. forficalis) to specialist (i.e. P. xylostella, P. rapae, P. brassicae, E. ornatum) herbivores among plant genotypes and sampling dates (fixed factors) were assessed by using a generalized linear model with binomial propor-tion (logistic regression) and logit link function.

the influence of plant genotype, sampling date and the interactions between them on the parasitism rate was assessed by means of generalized linear model (logis-tic regression) with binomial proportion and logit link function. the binomial proportion (i.e. number of para-sitized hosts/number of available hosts) was treated as the response variable, whereas plant genotype and sampling

897Oecologia (2014) 174:893–907

1 3

date were the independent variables. Pairwise and general comparisons among genotypes were assessed. Parasitism rate was only assessed for immature stages of the lepidop-teran species and E. ornatum eggs.

In order to assess whether overall herbivore abundance, the ratio of generalist to specialist herbivores and the leaf damage level were correlated with glucosinolate concentra-tions or with the parasitism rate (fixed factors), a multiple linear regression was adopted. glucosinolate concentrations were expressed as the mean values for each sample date (June and July 2011, and July and October 2012), with six replications per genotype. For each glucosinolate, data from plants selected for having high and low concentrations were pooled. Parasitism rate was arcsin square root transformed and herbivore abundance was log10(x + 1) transformed prior to the analysis. significance was declared at P < 0.05. sta-tistical tests were carried out by using the genstat 12.1 soft-ware package (Vsn International, hemel hempstead, UK).

Results

glucosinolates

glucosinolate analyses confirmed that in both years the mean concentrations of sinigrin, glucoiberin and glucobras-sicin varied significantly between genotypes (high vs. low concentration), according to the divergent selection pre-viously performed [online resource, table s1, electronic supplementary material (esm)]. In some cases, significant variation between genotypes in the concentration of other glucosinolates, not subjected to divergent selection, was also detected. In fact, this outcome was to be expected owing to the existence of links between the biosynthetic pathways of the different glucosinolates (Fahey et al. 2001). however, this factor is unlikely to have affected the reliability of the study because these variations were of considerably lower magnitude than those recorded in the selected glucosinolates.

herbivore responses to glucosinolate patterns

Variation in the herbivore numbers recorded between the two years was due to natural population fluctuations and to the different number of sampling events (table 1). We found that M. brassicae (84 %) dominated the lepidopteran community, followed by P. rapae (8 %) and P. xylostella (5 %). Pairwise comparison of plant genotypes showed that the leaf damage level was significantly lower in plants with high sinigrin content (table 2; Fig. 1a). general compari-sons across the six plant genotypes also showed the high sinigrin genotype to have significantly lower leaf damage, as well as a significant variation in the extent of leaf dam-age depending on the sampling date (table 3; Fig. 1a).

among the lepidopterans, M. brassicae larvae were significantly less abundant on the high sinigrin content genotype (pairwise comparisons) (table 2; Fig. 1b). by performing general comparisons, we found that plant gen-otype, sampling date and the interactions between them exerted a significant effect on the larval stage (table 3), which avoided the plants selected for having high glucoi-berin content (Fig. 1b). the abundance of M. brassicae eggs significantly decreased on the plants with high sini-grin concentration, but also on the genotype with low glu-coiberin content (pairwise comparisons) (table 2; Fig. 1c). however, general comparisons only detected significant effects of the sampling date (table 3).

Pairwise comparisons between genotypes showed P. xylostella larvae and pupae to be less abundant in the plant genotype with high sinigrin content (table 2; Fig. 1d); however, no genotype effect was observed on P. rapae lar-vae or pupae abundance (table 2), although populations of this species did fluctuate depending on sampling date (Fig. 1e). When general comparisons were performed, only the sampling date was significant for both P. xylostella and P. rapae (table 3).

among the hemipterans, the abundance of E. ornatum eggs significantly increased in plants with high sinigrin content, and also varied depending on the sampling date (pairwise comparisons) (table 2; Fig. 1f). general compar-isons confirmed the attraction exerted by plants with high sinigrin concentration (table 3; Fig. 1f). numbers of adult E. ornatum, however, were only significantly affected by the sampling date, as indicated by both pairwise (table 2; Fig. 1g) and general comparisons (table 3).

the size of the colonies of B. brassicae consistently increased in plants with low glucobrassicin concentration and depended on the sampling date (pairwise comparisons) (table 2; Fig. 1h). When considering general comparisons, only the sampling date and the interaction between sam-pling date and plant genotype were found to be significant for this species (table 3).

the magnitude of the colonies of the whitefly A. bras-sicae (Fig. 1i) and adults of the coleopteran P. cruciferae (Fig. 1l) varied depending on the sampling date in both pairwise (table 3) and general comparisons (table 3).

the ratio of generalist to specialist herbivores signifi-cantly differed among plant genotypes (Wald test = 80, P < 0.001) and sampling dates (Wald test = 431.7, P < 0.001; genotype × sampling date, Wald test = 313.54, P < 0.001). In particular, when both years are considered together, plants selected for having high sinigrin and low glucobrassicin content were seen to be significantly more visited by specialists than generalists (online resource, Fig. s1a, esm). considering the 2 years separately, specialists dominated all plant groups in 2011, but preferred the geno-type with high sinigrin concentration (online resource, Fig.

898 Oecologia (2014) 174:893–907

1 3

s1b, esm); in 2012, however, generalists were more abun-dant, especially in plants with low sinigrin, and high or low glucoiberin content (online resource, Fig. s1c, esm).

Parasitoid responses to glucosinolate patterns

lepidopterans were parasitized by hymenopteran para-sitoids belonging to the braconidae, Ichneumonidae, encyrtidae, Pteromalidae, eulophidae, scelionidae and trichogrammatidae, and by tachinid dipterans, whereas

the eggs of the hemipteran E. ornatum were attacked by Trissolcus sp. (hymenoptera, scelionidae) [see santol-amazza-carbone et al. (2013) for details of the parasitoid complex of B. oleracea herbivores]. In 2011, the parasit-ism rate experienced by the generalist herbivores was lower (0.12 ± 0.01) than that experienced by the specialists (0.26 ± 0.03). a similar trend was found in 2012 for gener-alists (0.15 ± 0.05) and specialists (0.41 ± 0.01).

the parasitism of M. brassicae larvae (table 1), mainly exerted by the braconid wasps Microplitis

Table 1 mean (±se) parasitism rate (parasitized hosts/total hosts) sorted per insect species, developmental stage, year and plant genotype

absence of numerical data in columns means that the insect was not found on the plant genotype. sixty plants per genotype were sampled on each sampling date. Differences in insect abundance between years also depended on the number of sampling events (may–July 2011, n = 1,080 observations; and June–October 2012, n = 1,800 observations)

H-SIN high sinigrin, L-SIN low sinigrin, H-GIB high glucoiberin, L-GIB low glucoiberin, H-GBS high glucobrassicin, L-GBS low glucobras-sicin

n h-sIn l-sIn h-gIb l-gIb h-gbs l-gbs

Parasitism rate 2011

Mamestra brassicae (larvae) 22 0.20 (±0.10) 0.01 (±0.01) 0.02 (±0.01) 0.00 0.00 0.33 (±0.31)

M. brassicae (eggs) 231 0.00 0.00 0.00 0.00 0.00 0.00

Pieris rapae (larvae and pupae) 8 0.00 0.00 0.50 (±0.02) 0.00 0.00 0.00

P. rapae (eggs) 10 0.00 0.00 0.00 0.00 0.00 0.00

Pieris brassicae (larvae) 30 0.00

P. brassicae (eggs) 5 0.00 0.00

Plutella xylostella (larvae and pupae) 66 0.33 (±0.10) 0.47 (±0.13) 0.17 (±0.01) 0.38 (±0.03) 0.75 (±0.12) 0.44 (±0.14)

Autographa gamma (larvae and pupae) 2 0.00 1.00 (± 0.00)

Evergestis forficalis (larvae) 1 0.00

Eurydema ornatum (eggs) 2,203 0.05 (±0.01) 0.27 (±0.07) 0.26 (±0.16) 0.07 (±0.05) 0.38 (±0.12) 0.05 (±0.01)

Parasitism rate 2012

Mamestra brassicae (larvae) 729 0.10 (±0.03) 0.14 (±0.02) 0.18 (±0.05) 0.06 (±0.01) 0.14 (±0.04) 0.04 (±0.05)

M. brassicae (eggs) 1,061 0.05 (±0.11) 0.33 (±0.12) 0.08 (±0.11) 0.13 (±0.14) 0.33 (±0.20) 0.10 (±0.22)

Pieris rapae (larvae and pupae) 155 0.38 (±0.23) 0.54 (±0.22) 0.35 (±0.28) 0.44 (±0.26) 0.52 (±0.17) 0.44 (±0.14)

P. rapae (eggs) 15 0.00 0.00 0.00 0.00 0.00

Pieris brassicae (larvae) 8 1.00 (±0.00) 0.83 (±0.01)

P. brassicae (eggs) 0

Plutella xylostella (larvae and pupae) 54 0.67 (±0.23) 0.79 (±0.15) 0.54 (±0.27) 0.68 (±0.26) 0.42 (±0.12) 0.58 (±0.23)

Autographa gamma (larvae and pupae) 19 0.40 (±0.25) 0 1.00 (±0.00) 0.50 (±0.13)

Evergestis forficalis (larvae) 6 0.00 0.00 0.00 0.00

Eurydema ornatum (eggs) 348 0.00 0.00 0.00 0.00 0.00 0.00

total parasitism rate (2011–2012)

Mamestra brassicae (larvae) 751 0.15 (±0.07) 0.14 (±0.05) 0.17 (±0.05) 0.06 (±0.03) 0.14 (±0.04) 0.07 (±0.06)

M. brassicae (eggs) 1,292 0.00 0.25 (±0.16) 0.10 (±0.10) 0.00 0.20 (±0.20) 0.00

Pieris rapae (larvae and pupae) 163 0.36 (±0.16) 0.50 (±0.13) 0.37 (±0.11) 0.00 0.49 (±0.11) 0.38 (±0.10)

P. rapae (eggs) 25 0.00 0.00 0.00 0.00 0.00 0.00

Pieris brassicae (larvae) 38 1.00 (±0.00) 0.83 (0.01) 0.00

P. brassicae (eggs) 5 0.00 0.00

Plutella xylostella (larvae and pupae) 120 0.40 (±0.25) 0.53 (±0.13) 0.50 (±0.02) 0.25 (±0.17) 0.60 (±0.24) 0.40 (±0.15)

Autographa gamma (larvae and pupae) 21 0.40 (±0.25) 0 1.00 (±0.00) 0.00 0.63 (±0.24)

Evergestis forficalis (larvae) 7 0.00 0.00 0.00 0.00

Eurydema ornatum (eggs) 2,551 0.04 (±0.02) 0.20 (±0.09) 0.05 (±0.03) 0.19 (±0.10) 0.22 (±0.07) 0.03 (±0.02)

899Oecologia (2014) 174:893–907

1 3

mediator l. and, occasionally, by Cotesia rubecula mar-shall, was significantly greater in those plants selected for high glucoiberin and high glucobrassicin content (pairwise comparisons) (table 4). When considering gen-eral comparisons (table 5), it was found that the parasit-ism rate was higher in genotypes with high glucoiberin and high glucobrassicin content (table 1), and also

affected by sampling date, although the interaction was not significant.

When performing pairwise comparisons, we found that plant genotypes did not influence the parasitism rate of M. brassicae eggs, or of P. rapae and P. xylostella larvae (tables 1, 4). similar results were obtained from general comparisons of plant genotypes for the parasitism rate of

Table 2 effects of plant genotype (pairwise comparisons: high vs. low concentration of sinigrin, high vs. low concentration of glucoiberin and high vs. low concentration of glucobrassicin), sampling date, and interactions, on leaf damage and insect abundance, investigated by anOVa

blocks were used as random factor

P-values in italics indicate significant effect (P < 0.05)

sinigrin glucoiberin glucobrassicin

F P df F P df F P df

leaf damage

genotype 14.86 <0.001 1, 95 0.33 0.57 1, 95 0.14 0.712 1, 95

sampling date 11.31 <0.001 7, 95 13.27 <0.001 7, 95 18.52 <0.001 7, 95

sampling date × genotype 1.12 0.362 7, 95 0.35 0.926 7, 95 1.33 0.249 7, 95

Mamestra brassicae (lepidoptera) larvae

genotype 4.96 0.029 1, 95 0.03 0.872 1, 95 1.97 0.165 1, 95

sampling date 12.14 <0.001 7, 95 19.12 <0.001 7, 95 9.35 <0.001 7, 95

sampling date × genotype 3.2 0.005 7, 95 0.52 0.814 7, 95 1.25 0.29 7, 95

M. brassicae eggs

genotype 30.24 0.012 1, 95 15.17 0.022 1.47 0.36 0.55 1, 95

sampling date 13.01 0.005 7, 95 0.68 0.433 3.47 1.21 0.315 7, 95

sampling date × genotype 45.1 0.016 7, 95 0.55 0.789 3.47 0.7 0.628 7, 95

Plutella xylostella (lepidoptera) larvae and pupae

genotype 4.24 0.039 1, 95 3.52 0.126 1, 95 0.45 0.534 1, 95

sampling date 18.65 0.001 7, 95 12.43 0.009 7, 95 24.56 <0.001 7, 95

sampling date × genotype 3.67 0.134 7, 95 0.15 0.978 7, 95 3.56 0.134 7, 95

Pieris rapae (lepidoptera) larvae and pupae

genotype 0.77 0.383 1, 95 0 0.991 1, 95 0 0.964 1, 95

sampling date 2.91 0.032 7, 95 3.33 0.018 7, 95 2.94 0.031 7, 95

sampling date × genotype 0.55 0.701 7, 95 0.82 0.519 7, 95 0.42 0.79 7, 95

Eurydema ornatum (hemiptera) eggs

genotype 15.04 0.03 1, 95 0.05 0.638 1, 95 1.08 0.11 1, 95

sampling date 44.23 <0.001 7, 95 33.81 <0.001 7, 95 44.87 <0.001 7, 95

sampling date × genotype 0.67 0.333 7, 95 0.01 0.945 7, 95 0.5 0.437 7, 95

E. ornatum adults

genotype 0.03 0.877 1, 95 0.02 0.912 1, 95 0.12 0.489 1, 95

sampling date 18.23 <0.001 7, 95 23.67 <0.001 7, 95 44.98 <0.001 7, 95

sampling date × genotype 0.07 0.965 7, 95 0.12 0.344 7, 95 0.23 0.652 7, 95

Aleyrodes brassicae (hemiptera) colony

genotype 3.23 0.076 1, 95 2.25 0.138 1, 95 0.02 0.9 1, 95

sampling date 110.7 <0.001 7, 95 120.39 <0.001 7, 95 114.52 <0.001 7, 95

sampling date × genotype 1.11 0.365 7, 95 1.38 0.224 7, 95 0.85 0.552 7, 95

Brevicoryne brassicae (hemiptera) colony

genotype 0.28 0.601 1, 95 1.01 0.318 1, 95 10.86 0.001 1, 95

sampling date 2.23 0.041 7, 95 5.22 <0.001 7, 95 13.29 <0.001 7, 95

sampling date × genotype 1.25 0.285 7, 95 1.02 0.422 7, 95 3.73 0.002 7, 95

Phyllotreta cruciferae (coleoptera) adults

genotype 1.36 0.248 1, 95 1.45 0.301 1, 95 0.07 0.79 1, 95

sampling date 11.13 <0.001 7, 95 13.11 <0.001 7, 95 6.44 <0.001 7, 95

sampling date × genotype 2.53 0.022 7, 95 0.63 0.726 7, 95 0.58 0.767 7, 95

900 Oecologia (2014) 174:893–907

1 3

M. brassicae eggs (table 5). sampling date did signifi-cantly affect the parasitism rate of P. rapae larvae (general comparisons), but the interaction between genotype and sampling date did not (table 5). the parasitism rate of P.

xylostella larvae was not affected by the plant genotype, the sampling date or their interactions when considering both pairwise (table 4) and general comparisons (table 5).

In 2011, a proportion of 13.5 % of E. ornatum eggs were parasitized, and parasitism rate was highest in plants with low sinigrin and high glucobrassicin content (tables 1, 4). In 2012 we did not detect any parasitism (table 1). Fur-ther analysis (general comparisons) confirmed the positive effect of high glucobrassicin plants on the egg parasitism rate of this hemipteran, as well as the importance of the sampling date (table 5).

top-down vs. bottom-up effects

In 2011, herbivore abundance was significantly and posi-tively correlated with sinigrin and glucobrassicin content, while the parasitism rate calculated in these plant geno-types was not (table 6). the leaf damage rate was signifi-cantly and negatively correlated with sinigrin concentra-tion, while the parasitism rate did not produce any effect on it (table 6). the ratio of generalist to specialist herbivores calculated on the genotypes with high and low glucobras-sicin content did not show any significant relationship with glucosinolate concentration, but it had a positive relation-ship with the parasitism rate (table 6).

In 2012, variation in sinigrin, glucoiberin and glucobras-sicin concentrations or parasitism rate did not have any sig-nificant relationship with herbivore abundance (table 6). however, sinigrin concentration and the parasitism rate were significantly and negatively related to the leaf damage rate (table 6). the ratio between generalist and specialist herbivores was not related to glucosinolate concentration or to parasitism rate (table 6).

Discussion

the study indicates that quantitative variations in glucosi-nolate profiles, and in particular variations in sinigrin con-tent, influence both the overall abundance of insect herbi-vores and the relative proportion of specialist and generalist species, which is translated into a significant variation in leaf damage levels. sinigrin is known to be the principal glucosinolate found in kale varieties of northwest spain (cartea et al. 2008), and this investigation shows that a high concentration of this major chemical defence can help deter both generalist and specialist lepidopteran herbivores. however, high sinigrin content may allow for an increased load of specialist hemiptera, and this can lead to an over-all positive correlation between sinigrin concentration and herbivore abundance. this means that the role of special-ist hemiptera on altering the net value of plant defence traits could be crucial, as also reported by lankau (2007).

Table 3 effects of plant genotype (general comparisons), sampling date, and interactions, on leaf damage and insect abundance, investi-gated by anOVa

blocks were used as random factor

P-values in italics indicate significant effect (P < 0.05)

F P df

leaf damage

genotype 3.55 0.004 5, 287

sampling date 42.71 <0.001 7, 287

sampling date × genotype 0.81 0.774 35, 287

Mamestra brassicae (lepidoptera) larvae

genotype 2.9 0.015 5, 287

sampling date 35.49 <0.001 7, 287

sampling date × genotype 2.18 <0.001 35, 287

M. brassicae eggs

genotype 0.43 0.829 5, 287

sampling date 4.35 <0.001 7, 287

sampling date × genotype 0.7 0.896 35, 287

Plutella xylostella (lepidoptera) larvae and pupae

genotype 1.97 0.091 5, 107

sampling date 25.18 <0.001 2, 107

sampling date × genotype 1 0.451 10, 107

Pieris rapae (lepidoptera) larvae and pupae

genotype 1.13 0.349 5, 287

sampling date 12.04 <0.001 7, 287

sampling date × genotype 0.64 0.904 35, 287

Eurydema ornatum (hemiptera) eggs

genotype 7.45 <0.001 5, 287

sampling date 1.03 0.098 7, 287

sampling date × genotype 0.13 0.433 35, 287

E. ornatum adults

genotype 1.67 0.089 5, 287

sampling date 57.52 <0.001 7, 287

sampling date × genotype 0.67 0.441 35, 287

Aleyrodes brassicae (hemiptera) colony

genotype 1.37 0.237 5, 287

sampling date 338.88 <0.001 7, 287

sampling date × genotype 1.03 0.43 35, 287

Brevicoryne brassicae (hemiptera) colony

genotype 1.92 0.091 5, 287

sampling date 14.58 <0.001 7, 287

sampling date × genotype 1.77 0.007 35, 287

Phyllotreta cruciferae (coleoptera) adults

genotype 1.35 0.244 5, 287

sampling date 29.06 <0.001 7, 287

sampling date × genotype 1.25 0.173 35, 287

901Oecologia (2014) 174:893–907

1 3

1

2

3

4

5D

amag

e ra

te (

arbi

trar

y un

its) a

0

5

10

15

20

25

30

Abu

ndan

ce (

larv

ae/6

0 pl

ants

)

b Mamestra brassicae

0

5

10

15

20

25

30

35

40

Abu

ndan

ce (

eggs

/60

plan

ts) c Mamestra brassicae

0

1

2

3

4

5

Abu

ndan

ce (

larv

ae+

pupa

e/60

pla

nts)

d Plutella xylostella

0

1

2

3

Abu

ndan

ce (

larv

ae+

pupa

e/60

pla

nts)

0

20

40

60

80

100

120

Abu

ndan

ce (

eggs

/60

plan

ts) e Pieris rapae f Eurydema ornatum

0

0.5

1

Adu

lt ab

unda

nce

(arb

itrar

y un

its)

0

1

2

Col

ony

abun

danc

e (a

rbitr

ary

units

)

g Eurydema ornatum h Brevicoryne brassicae

0

1

2

3

4

Col

ony

abun

danc

e (a

rbitr

ary

units

)

Sampling date

H-SIN

L-SIN

H-GIB

L-GIB

H-GBS

L-GBS

0

1

2

Adu

lt bu

ndan

ce (

arbi

trar

y un

its)

Sampling date

i Aleyrodes brassicae l Phyllotreta cruciferae H-SIN

L-SIN

H-GIB

L-GIB

H-GBS

L-GBS

Fig. 1 average leaf damage rate (mean ± se) a estimated by using a 1−5 subjective rating scale [healthy plant, without any damage (1); completely damaged plant, with 90–100 % of the leaves attacked by herbivores (5)]. abundance (mean ± se) of b Mamestra brassicae larvae and c eggs, d Plutella xylostella larvae and pupae, e Pieris rapae larvae and pupae, and f Eurydema ornatum eggs, obtained by direct counting. mean number of g E. ornatum adults, h Brevicoryne brassicae colony, i Aleyrodes brassicae colony, and l adults of Phyl-lotreta cruciferae, were estimated by subjective 0–4 rating scale

[absence (0), up to five individuals (1), up to ten (2), up to 50 (3), more than 50 (4)]. Data (not transformed) sorted by sampling date and plant genotype. means obtained from 60 plants per genotype in each sampling date. H-SIN high sinigrin (solid black line, black tri-angle), L-SIN low sinigrin (solid black line, black circle), H-GIB high glucoiberin (dotted grey line, grey triangle), L-GIB low glucoiberin (dotted grey line, grey circle), H-GBS high glucobrassicin (dotted black line, black triangle), L-GBS low glucobrassicin (dotted black line, black circle)

902 Oecologia (2014) 174:893–907

1 3

Tabl

e 4

eff

ects

of

plan

t ge

noty

pe (

pair

wis

e co

mpa

riso

ns:

high

vs.

low

con

cent

ratio

n of

sin

igri

n, h

igh

vs.

low

con

cent

ratio

n of

glu

coib

erin

and

hig

h vs

. lo

w c

once

ntra

tion

of g

luco

bras

sici

n),

sam

plin

g da

te, a

nd in

tera

ctio

ns, o

n pa

rasi

tism

rat

e ex

peri

ence

d by

inse

ct h

erbi

vore

s, in

vest

igat

ed b

y lo

gist

ic r

egre

ssio

n

est

imat

e an

d se

err

or o

f sl

opes

are

als

o pr

ovid

ed

P-v

alue

s in

ital

ics

indi

cate

sig

nific

ant e

ffec

t (P

< 0

.05)

Para

sitis

m r

ate

sini

grin

glu

coib

erin

glu

cobr

assi

cin

Wal

dP

dfe

stim

ate

seW

ald

Pdf

est

imat

ese

Wal

dP

dfe

stim

ate

se

Mam

estr

a br

assi

cae

(lep

idop

tera

) la

rvae

gen

otyp

e0.

640.

422

1, 5

9−

0.66

10.

683

8.12

0.00

41,

59

−0.

052

0.03

65.

260.

021

1, 5

9−

1.20

10.

678

sam

plin

g da

te0.

231

0.79

56,

59

−0.

455

0.56

618

.961

0.00

26,

59

−0.

123

0.12

98.

201

0.51

16,

59

−0.

206

0.36

5

sam

plin

g da

te ×

gen

otyp

e1.

540.

818

6 ,5

90.

130.

142

18.3

20.

019

6, 5

9−

0.08

50.

126

2.01

0.97

6, 5

90.

488

0.21

8

M. b

rass

icae

egg

s

gen

otyp

e0.

060.

483

1, 1

2−

0.54

40.

565

01

1, 1

5−

0.80

90.

820.

330.

254

1, 8

0.56

60.

59

sam

plin

g da

te7.

010.

011

3, 1

2−

1.45

0.28

90.

020.

543

3, 1

5−

0.45

50.

467

1.77

0.33

33,

80.

134

0.14

1

sam

plin

g da

te ×

gen

otyp

e0.

310.

566

3, 1

20.

233

0.24

50.

560.

654

3, 1

50.

123

0.15

60

13,

80.

654

0.68

8

Plu

tell

a xy

lost

ella

(l

epid

opte

ra)

larv

ae a

nd p

upae

gen

otyp

e0.

061

0.81

1, 2

1−

1.43

0.60

80.

410.

526

1, 2

80.

234

−0.

256

1.54

0.21

41,

29

0.20

90.

222

sam

plin

g da

te3.

770.

052

1, 2

10.

051

0.45

90.

371

0.69

72,

28

0.35

5−

0.36

73.

970.

137

2, 2

90.

583

0.58

9

sam

plin

g da

te ×

gen

otyp

e0.

130.

723

1, 2

10.

267

0.25

70.

110.

898

2, 2

80.

211

0.22

0.68

0.40

82,

29

0.11

10.

211

Pie

ris

rapa

e (l

epid

opte

ra)

larv

ae a

nd p

upae

gen

otyp

e0.

060.

801

1, 4

2−

0.49

30.

990.

30.

581,

42

−1.

281.

083.

820.

051

1, 4

2−

0.12

30.

133

sam

plin

g da

te1.

980.

744,

42

−0.

122

0.16

92.

630.

624,

42

−0.

114

0.13

26.

130.

189

4, 4

2−

0.11

10.

123

sam

plin

g da

te ×

gen

otyp

e0.

730.

864

4, 4

20.

140.

189

0.23

0.88

74,

42

0.22

40.

169

0.46

0.99

64,

42

0.26

70.

356

Eur

ydem

a or

natu

m (

hem

ipte

ra)

eggs

gen

otyp

e23

.19

<0.

001

1, 2

2−

1.36

80.

031

2.61

0.10

61,

22

−0.

112

0.13

268

.58

<0.

001

1, 2

20.

134

0.04

5

sam

plin

g da

te0.

930.

628

2, 2

20.

366

0.37

813

.962

<0.

001

2, 2

21.

654

0.59

0.26

0.87

62,

22

0.23

60.

267

sam

plin

g da

te ×

gen

otyp

e0.

084

0.95

92,

22

0.23

40.

239

0.23

20.

892,

22

0.44

40.

456

0.04

0.94

42,

22

0.11

10.

122

903Oecologia (2014) 174:893–907

1 3

Furthermore, the third trophic level represented by hyme-nopteran and dipteran parasitoids also performed differ-ently depending on the plant genotype, which is likely to be due to variation in both the host density and the volatile blends emitted by the host plants (gols and harvey 2009).

herbivore responses to glucosinolate patterns

host plants with high sinigrin and low glucobrassicin con-centrations harboured a significantly higher load of spe-cialists. In particular, lepidopterans were outnumbered by specialist hemipterans in 2011, especially in plant groups with high sinigrin content, although this did not occur in 2012. there are a number of reasons that may explain the wide distribution of the hemipterans across the plant gen-otypes: for example, insects may build up high numbers on preferred host plants, eventually spilling over onto less preferred ones (associational susceptibility) (White and Whitham 2000). Furthermore, variation in the ratio between generalists and specialists also depended on the temporal components of herbivory: generalist species tended to avoid plants previously consumed by specialists, whereas spe-cialists often colonized in large numbers plants that were already being consumed by other phytophagous insects, irrespective of their feeding mode (Poelman et al. 2010).

as expected, egg and larval abundances of M. brassi-cae were lower in plants with high sinigrin concentration.

attraction of the M. brassicae female to patches of plants may be mediated by visual and biochemical cues, while the decision of where to oviposit within a plant population would depend on the specific plant chemistry. the response of M. brassicae larvae to glucoiberin is less obvious. On the one hand, larvae were less abundant in september in plants with a high content of this aliphatic glucosinolate; this agrees with previous studies showing high glucoiberin concentrations to be negatively correlated with herbivore abundance and species richness (Poelman et al. 2009; Kos et al. 2011a). however, it was also found that M. brassicae eggs were more abundant in plants with high glucoiberin concentration. these discrepancies are commonly found in the literature, which reports impacts of the host plant on the M. brassicae fitness components ranging from no effect (newton et al. 2009a, b), to effects similar to those expe-rienced by specialist herbivores (Poelman et al. 2008), to negative effects on larval survival, especially those geno-types with a high concentration of aliphatic glucosinolates (gols et al. 2008; harvey and gols 2011).

In agreement with previous studies on neutral P. rapae responses to change in glucosinolate patterns (newton et al. 2009a, b, 2010; gols et al. 2008), but in contrast with others where the negative effect of indole glucosinolates on oviposition preference has been reported (de Vos et al. 2008), we found that P. rapae did not respond to the range of glucosinolate profiles. It has been reported that this

Table 5 effects of plant genotype (general comparisons), sampling date, and interactions, on parasitism rate investigated by logistic regression

estimate and se of slopes are also provided

P-values in italics indicate significant effect (P < 0.05)

Parasitism rate

Wald P df estimate se

Mamestra brassicae (lepidoptera) larvae

genotype 15.14 0.01 5, 184 −0.325 0.112

sampling date 16.951 0.009 7, 184 −0.425 0.061

sampling date × genotype 0.78 0.731 35, 184 −0.375 0.671

M. brassicae eggs

genotype 1.19 0.366 5, 37 −0.222 0.385

sampling date 1.972 0.169 3, 37 0.128 0.188

sampling date × genotype 0.841 0.608 11, 37 0.396 0.444

Plutella xylostella (lepidoptera) larvae and pupae

genotype 0.77 0.574 5, 80 −0.245 0.13

sampling date 1.96 0.13 3, 80 −0.004 0.107

sampling date × genotype 1.601 0.153 7, 80 0.068 0.042

Pieris rapae (lepidoptera) larvae and pupae

genotype 0.28 0.923 5, 113 −0.338 0.499

sampling date 2.624 0.041 4, 113 −0.079 0.056

sampling date × genotype 0.28 0.923 5, 113 −0.044 0.049

Eurydema ornatum (hemiptera) eggs

genotype 2.73 0.036 5, 66 −0.012 0.006

sampling date 5.912 0.005 2, 66 −0.017 0.003

sampling date × genotype 0.87 0.562 9, 66 0.166 0.195

904 Oecologia (2014) 174:893–907

1 3

Tabl

e 6

mul

tiple

line

ar r

egre

ssio

ns to

inve

stig

ate

the

corr

elat

ion

of s

inig

rin,

glu

coib

erin

and

glu

cobr

assi

cin

conc

entr

atio

ns (

botto

m-u

p fo

rce)

or

para

sitis

m r

ate

(top

-dow

n fo

rce)

with

her

bivo

re

abun

danc

e, le

af d

amag

e ra

te a

nd th

e ra

tio g

ener

alis

t/spe

cial

ist h

erbi

vore

s

ran

ge o

f co

ncen

trat

ion

of e

ach

sele

cted

glu

cosi

nola

te o

btai

ned

by p

oolin

g da

ta f

rom

gen

otyp

es w

ith h

igh

and

low

con

tent

. her

bivo

re a

bund

ance

obt

aine

d by

poo

ling

all i

nsec

t spe

cies

che

cked

by

dir

ect c

ount

ing.

est

imat

e an

d se

of

slop

es a

nd c

oeffi

cien

t of

dete

rmin

atio

n (R

2 ) ar

e al

so p

rovi

ded

P-v

alue

s in

ital

ics

indi

cate

sig

nific

ant e

ffec

t (P

< 0

.05)

sini

grin

glu

coib

erin

glu

cobr

assi

cin

Wal

dP

R2

est

imat

ese

Wal

dP

R2

est

imat

ese

Wal

dP

R2

est

imat

ese

glu

cosi

nola

te c

once

ntra

tions

201

1

her

bivo

re a

bund

ance

4.31

0.03

90.

757

0.01

80.

007

0.88

0.88

70.

009

−2.

771

2.89

9.53

0.00

60.

778

0.06

40.

001

lea

f da

mag

e10

.97

0.00

40.

709

−0.

012

0.00

31.

380.

273

0.07

71.

444

2.74

40.

330.

332

0.23

31.

223

1.87

7

gen

eral

ists

/spe

cial

ists

0.77

0.39

0.12

21.

427

2.48

60.

630.

434

0.03

5.32

25.

90.

410.

526

0.14

53.

234

3.67

5

glu

cosi

nola

te c

once

ntra

tions

201

2

her

bivo

re a

bund

ance

1.25

0.27

50.

088

1.33

33.

678

1.12

0.71

50.

123

1.87

72.

135

1.56

0.94

50.

234

−2.

32.

877

lea

f da

mag

e8.

760.

007

0.87

9−

0.00

20.

001

0.34

0.96

30.

201

2.43

32.

51.

90.

333

0.03

21.

345

1.65

gen

eral

ists

/spe

cial

ists

0.78

0.38

70.

092

2.34

52.

433

2.05

0.16

60.

122.

456

2.68

70.

350.

563

0.23

12.

552.

6

Para

sitis

m r

ate

2011

her

bivo

re a

bund

ance

0.1

0.93

80.

078

−2.

321

2.37

70.

340.

654

0.07

81.

877

2.23

40.

330.

332

0.07

7−

1.23

1.76

6

lea

f da

mag

e1.

550.

064

0.17

52.

129

3.32

81.

120.

30.

242.

983.

101

1.44

0.26

70.

12.

854

2.98

7

gen

eral

ists

/spe

cial

ists

1.24

0.27

70.

211

1.76

51.

887

0.33

0.33

20.

037

1.22

31.

318

.86

<0.

001

0.78

5.03

1.22

2

Para

sitis

m r

ate

2012

her

bivo

re a

bund

ance

2.56

0.11

20.

232.

566

2.99

81

0.91

50.

159

2.55

52.

92.

090.

115

0.20

3−

1.33

31.

414

lea

f da

mag

e7.

010.

015

0.68

8−

0.07

30.

002

1.56

0.23

80.

124.

245

5.77

71.

110.

321

0.1

2.45

2.5

gen

eral

ists

/spe

cial

ists

3.45

0.07

70.

321

0.12

30.

277

1.01

0.32

60.

342

2.56

73.

986

0.17

0.68

20.

211

2.13

2.97

7

905Oecologia (2014) 174:893–907

1 3

butterfly can redirect the course of the normal hydrolysis reaction that is provoked by the enzyme myrosinase upon insect feeding, by producing nitriles instead of toxic isothi-ocyanates, which are then excreted by the larvae (hopkins et al. 2009). this detoxification system may allow P. rapae to be relatively insensitive to changes of plant chemical concentrations, as reported by harvey et al. (2007).

several studies on host plant recognition have reported the ability of P. xylostella to employ glucosinolates as olfactory cues for oviposition (hopkins et al. 2009). In particular, the role of aliphatic glucosinolate breakdown products (spencer et al. 1999; renwick et al. 2006) and the impact of intact indole glucosinolates on oviposition behaviour (reed et al. 1989; sun et al. 2009) have been shown. Furthermore, as a crucifer specialist, physiological counter-adaptations to plant defence compounds have been evolved by P. xylostella larvae, which possess a glucosi-nolate sulphatase enzyme in the gut that enables the con-version of glucosinolates to desulphoglucosinolates, rather than toxic nitriles and isothiocyanates (raztka et al. 2002). Interestingly, we found that P. xylostella abundance did not increase in plants with high sinigrin content, which does not agree with the evidence that, under laboratory condi-tions, elevated sinigrin concentration is highly attractive to this moth (spencer et al. 1999).

Piercing–sucking insects, such as hemipterans, are exposed to intact glucosinolate or possibly to the by-prod-uct produced by damage provoked by other herbivores (hopkins et al. 2009). the aphid B. brassicae is a glucosi-nolate-sequestering specialist herbivore, which uses aphid-specific myrosinase enzymes to form toxic hydrolytic prod-ucts against its natural enemies (cole 1997). among the hemipteran species studied, only B. brassicae showed sig-nificant responses to glucobrassicin variation, being more abundant on the low glucobrassicin genotype. this finding is consistent with the evidence that the sequestration of glu-cosinolates by this aphid from plant phloem is selective, with a clear preference for aliphatic instead of indole sec-ondary metabolites (Kos et al. 2011b). It is interesting that this herbivore, which has a passive dispersal mechanism, displayed a clear response to certain glucosinolate profiles.

this is the first time that the performance of the pentato-mid E. ornatum in response to glucosinolate polymorphism of the host plant has been investigated. although the adult bugs did not show a preference for a specific plant geno-type, their egg masses were especially abundant in plants with a high sinigrin content. the role of sinigrin as an ovi-position stimulant for the E. ornatum female has never been reported in the past. a previous study on the responses of this hemipteran to the glucosinolate pattern of different Brassica crops did, however, highlight the importance of having a high content of aliphatic compounds, such as pro-goitrin, epiprogoitrin, gluconapin and glucoraphanin, and

of the indole glucobrassicin, in order to reduce the extent of damage, although the effects also depended on plant age (bohinc et al. 2013).

Parasitoid responses to glucosinolate patterns

Parasitisation of insect hosts is the result of a complex pro-cess which involves the attraction of parasitoids by semio-chemical cues emitted by both the host (kairomones) and the plant (synomones), parasitoid arrestment and host searching, host selection, acceptance and oviposition (god-fray 1994). In the present study, the proportion of herbi-vores that were parasitized was affected by differences in the herbivore-induced volatile emissions of the different plant genotypes. herbivore density was not manipulated because we were also interested in measuring herbivore abundance on different plant genotypes, and investigating its impact on the parasitism rate. however, the host density-dependent effect was not seen to contribute to top-down control. E. ornatum eggs, for example, were more abun-dant in plants with high sinigrin content, but suffered from a higher parasitism rate in plants with low sinigrin content. M. brassicae larvae were more abundant in plants with low glucoiberin content but were not more parasitized there. similarly, M. brassicae larvae and E. ornatum eggs were more parasitized in plants with high glucobrassicin content, even though their populations were not especially high on this genotype. these findings agree with work carried out on other B. oleracea varieties, which has shown volatile indole derivates to be important in the attraction of natu-ral enemies of pentatomid bugs (conti et al. 2008). In fact, egg oviposition by herbivores could induce specific plant responses which are relevant for host location by egg para-sitoids (Fatouros et al. 2012).

top-down vs. bottom-up effects

Whether or not plant resources, natural enemies, or both, determine the abundance of insect herbivores in natu-ral multi-trophic systems has long been a topic of debate (hunter and Price 1992; halaj and Wise 2001) that is yet to reach a general consensus. In comparison with natural hab-itats, managed crop systems have a relatively simple food web structure, characterized by the presence of a homoge-neous plant community, for which the impact of bottom-up and top-down forces appears more predictable. the higher propensity of crop habitats to experience strong trophic cascades was clearly shown through meta-analysis by halaj and Wise (2001). In this type of habitat the reduction of predator abundance, more than plant chemical defences, generally increased herbivory and reduced primary plant production (halaj and Wise 2001). Our results, however, indicate the opposite trend. In our study system, the impact

906 Oecologia (2014) 174:893–907

1 3

of bottom-up forces on insects agrees with the general view about the importance of plant-mediated forces in terres-trial ecosystems (Denno et al. 2002). Furthermore, previ-ous studies focused on brassicaceous plants (newton et al. 2009b; Kos et al. 2011a) have also highlighted that plant chemistry and morphology have more impact than natural enemy activity in shaping herbivore abundance.

the statistical analyses showed that glucosinolate con-centrations have a significant effect on the overall herbi-vore abundance, on the relative abundance of generalist and specialist herbivores, on the parasitism rate and on the leaf damage rate. On the other hand, parasitism rate also contrib-uted to shaping the leaf damage rate in 2011, and the ratio between generalist and specialist insects in 2012, which suggests that in reality many ecological forces combine to determine the patterns observed in the field. Interestingly, in 2011 the increase in sinigrin content was positively corre-lated with herbivore abundance, whereas an opposite trend was found for leaf damage rate. the massive presence of the hemipteran E. ornatum, and in particular the prefer-ence manifested by ovipositing females for plants with high sinigrin content, explained the positive relationship between sinigrin concentration and herbivore abundance. the reduc-tion in leaf damage, on the other hand, can be explained by the fact that although overall herbivore numbers increased with high sinigrin content, the numbers of generalist lepi-dopteran herbivores decreased. this prompted a net decline in overall leaf damage because the damage provoked by chewing insects (including lepidoptera) was dispropor-tionately high and more widely detected than damage from hemipteran sap-sucking species. In 2012, the leaf damage rate in plants with high sinigrin content was also reduced as a result of the increase in parasitism rate among the lepi-doptera (the same pattern was not seen in hemiptera: E. ornatum eggs, for example, were parasitized in 2011 but not in 2012). these findings highlighted the role of sinigrin in plant protection. however, an ecological cost of plant defence through sinigrin synthesis does exist, because spe-cialists were more attracted to a high concentration of this secondary metabolite. Previous field studies on the role of B. oleracea chemicals on insect herbivore biodiversity showed the impact of glucoiberin in shaping insect com-munities (Poelman et al. 2009; Kos et al. 2011a). Data sup-porting the idea that sinigrin influences herbivore choices come from laboratory trials (shields and mitchell 1995; gols et al. 2008) and from field tests performed in con-trolled environments (by manipulating insect presence and abundance) (lankau 2007; lankau and Kliebenstein 2009; lankau and strauss 2008; Kos et al. 2011a). this is the first time that an investigation under natural conditions reported significant responses to sinigrin, glucoiberin and glucobras-sicin expressed by herbivores with different feeding modes and behavioural ecologies, and by their parasitoids.

to conclude, our results illustrate how quantitative vari-ation in aliphatic and indole glucosinolates of kales may influence herbivore abundance and the control exerted by the parasitoid complex. Furthermore, the plant glucosi-nolate pattern contributes to reducing leaf damage rate. although the top-down force indeed acts in concert with bottom-up regulation and with the other environmental fac-tors, parasitoid pressure seems to be a weaker force in our study system.

Acknowledgments this research was supported by the national Plan for research and Development (agl2012-35539). the authors thank rosaura abilleira, Pilar comesaña, césar gonzález, and tamara sotelo for their valuable help in field samplings. We also thank David brown for improving the english. s. santolamazza-car-bone acknowledges a post-doctoral research contract (Jae-Doc) from the csIc.

References

bekaert m, edger PP, hudson cm, Pires Jc, conant gc (2012) met-abolic and evolutionary costs of herbivory defense: systems biol-ogy of glucosinolate synthesis. new Phytol 196:596–605

bohinc t, hrastar r, Kosír IJ, trdan s (2013) association between glucosinolate concentration and injuries caused by cabbage stink bugs Eurydema spp. (heteroptera: Pentatomidae) on different brassicas. acta sci agron 35:1–8

bukovinszky t, Poelman eh, gols r, Prekatsakis g, Vet lem, harvey Ja, Dicke m (2009) consequences of constitutive and induced variation in plant nutritional quality for immune defence of a herbivore against parasitism. Oecologia 160:299–308

carmona D, lajeunesse mJ, Johnson mtJ (2011) Plant traits that pre-dict resistance to herbivores. Funct ecol 25:358–367

cartea me, Velasco P, Obregón s, Padilla g, de haro a (2008) sea-sonal variation in glucosinolate content in Brassica oleracea crops grown in northwestern spain. Phytochemistry 69:403–410

cartea me, Padilla g, Vilar m, Velasco P (2009) Incidence of the major Brassica pests in northwestern spain. J econ entomol 102:767–773

cole ra (1997) the relative importance of glucosinolates and amino acids to the development of two aphid pests Brevicoryne bras-sicae and Myzus persicae on wild and cultivated brassica species. entomol exp appl 85:121–133

conti e, Zandra c, salerno g, leonbruni b, Volpe D, Frati F, m c, bin F (2008) changes in the volatile profile of Brassica oleracea due to feeding and oviposition by Murgantia histrionica (heter-optera: Pentatomidae). eur J entomol 105:839–847

de Vos m, Kriksunov Kl, Jander g (2008) Indole-3-acetonitrile pro-duction from indole glucosinolates deters oviposition by Pieris rapae (white cabbage butterfly). Plant Physiol 146:916–926

Denno rF, gratton c, Peterson ma, langellotto ga, Finke Dl, huberty aF (2002) bottom-up forces mediate natural-enemy impact in a phytophagous insect community. ecology 83:1443–1458

Fahey JW, Zalcmann at, Zalalay P (2001) the chemical diversity and distribution of glucosinolates and isothiocyanates among plants. Phytochemistry 56:5–51

Fatouros ne, lucas-barbosa D, Weldegergis bt, Pashalidou Fg, van loon JJa, Dicke m, harvey Ja, gols r, huigens me (2012) Plant volatiles induced by herbivore egg deposition affect insects of different trophic levels. Plos One 7(8):e43607. doi:10.1371/journal.pone.0043607

907Oecologia (2014) 174:893–907

1 3

Fritz Va, Justen Vl, bode am, schuster t, Wang m (2010) glucosi-nolate enhancement in cabbage induced by jasmonic acid appli-cation. hortscience 45:1188–1191

godfray hcJ (1994) Parasitoids: behavioral and evolutionary ecol-ogy. Princeton University Press, Princeton

gols r, harvey Ja (2009) Plant-mediated effects in the brassicaceae on the performance and behaviour of parasitoids. Phytochem rev 8:187–206

gols r, bukovinszky t, van Dam n, Dicke m, bullock Jm, harvey Ja (2008) Performance of generalist and specialist herbivores and their endoparasitoids differs on cultivated and wild Brassica populations. J chem ecol 34:132–143

halaj J, Wise Dh (2001) terrestrial trophic cascades: how much do they trickle? am nat 157:262–281

harvey Ja, gols r (2011) Population-related variation in plant defense more strongly affects survival of an herbivore than its solitary parasitoid wasp. J chem ecol 37:1081–1090

harvey Ja, gols r, Wagenaar r, bezemer tm (2007) Development of an insect herbivore and its pupal parasitoid reflect differences in direct plant defense. J chem ecol 33:1556–1569

harvey Ja, van Dam n, raajimakers ce, bullock Jm, gols r (2011) tri-trophic effects of inter- and intra-population variation in defence chemistry of wild cabbage (Brassica oleracea). Oecolo-gia 166:421–431

hopkins rJ, van Dam n, van loon JJa (2009) role of glucosinolates in insect–plant relationships and multitrophic interactions. annu rev entomol 54:57–83

hunter mD, Price PW (1992) Playing chutes and ladders: heteroge-neity and the relative role of bottom-up and top-down forces in natural communities. ecology 73:724–732

Kliebenstein D, lambrix V, reichelt m, mitchell-Olds t (2001) gene duplication and the diversification of secondary metabolism: side chain modification of glucosinolates in Arabidopsis thaliana. Plant cell 13:681–693

Kos m, broekgaarden c, Kabouw P, Oude lenferink K, Poelman eh, Vet lem, Dicke m, van loon JJa (2011a) relative importance of plant-mediated bottom-up and top-down forces on herbivore abundance on Brassica oleracea. Funct ecol 25:1113–1124

Kos m, Kabouw P, noordam r, hendriks K, Vet lem, van loon JJa, Dicke m (2011b) Prey-mediated effects of glucosinolates on aphid predators. ecol entomol 36:377–388

lankau ra (2007) specialist and generalist herbivores exert opposing selection on a chemical defense. new Phytol 175:176–184

lankau ra, Kliebenstein DJ (2009) competition, herbivory and genetics interact to determine the accumulation and fitness conse-quences of a defence metabolite. J ecol 97:78–88

lankau ra, strauss sY (2008) community complexity drives pat-terns of natural selection on a chemical defence of Brassica nigra. am nat 171:150–161

mithöfer a, boland W (2012) Plant defense against herbivores: chem-ical aspects. annu rev Plant biol 63:431–450

newton el, bullock Jm, hodgson DJ (2009a) bottom-up effect of glucosinolate variation on aphid colony dynamics in wild cab-bage populations. ecol entomol 34:614–623

newton el, bullock Jm, hodgson DJ (2009b) glucosinolate poly-morphism in wild cabbage (Brassica oleracea) influences the structure of herbivore communities. Oecologia 160:63–76

newton el, bullock Jm, hodgson DJ (2010) temporal consist-ency in herbivore responses to glucosinolate polymorphism in populations of wild cabbage (Brassica oleracea). Oecologia 164:689–699

Ode P (2006) Plant chemistry and natural enemy fitness: effect on herbivore and natural enemy interactions. annu rev entomol 51:163–185

Poelman eh, broekgaarden c, van loon JJa, Dicke m (2008) early season herbivore differentially affects plant defence responses to subsequently colonizing herbivores and their abundance in the field. mol ecol 17:3352–3365

Poelman eh, van Dam n, van loon JJa, Vet lem, Dicke m (2009) chemical diversity in Brassica oleracea affects biodiversity of insect herbivores. ecology 90:1863–1877

Poelman eh, van loon JJa, van Dam n, Vet lem, Dicke m (2010) herbivore-induced plant responses in Brassica oleracea prevail over effects of constitutive resistance and result in enhanced her-bivore attack. ecol entomol 35:240–247

raztka a, Vogel h, Kliebenstein DJ, mitchell-Olds t, Kroymann J (2002) Disarming the mustard oil bomb. Proc natl acad sci Usa 99:11223–11228

reed DW, Pivnick Ka, Underhill eW (1989) Identification of chemi-cal oviposition stimulants for the diamondback moth, Plutella xylostella, present in three species of brassicaceae. entomol exp appl 53:277–286

renwick Ja, haribel m, gouinguené s, städler e (2006) Isothiocy-anates stimulating oviposition by the diamondback moth, Plutella xylostella. J chem ecol 32:755–766

santolamazza-carbone s, Velasco P, selfa J, soengas P, cartea me (2013) Intraspecific variation of host plant and locality influence the lepidopteran-parasitoid system of Brassica oleracea crops. J econ entomol 106:1134–1144

shields VDc, mitchell bK (1995) sinigrin as a feeding deterrent in two crucifer-feeding, polyphagous lepidopterous species and the effects of feeding stimulant mixtures on deterrency. Phil trans r soc lond b 347:439–446

soler r, martijn bezemer t, Van der Putten Wh, Vet lem, harvey Ja (2005) root herbivore effects on above-ground herbivore, parasitoid and hyperparasitoid performance via changes in plant quality. J anim ecol 74:1121–1130

spencer Jl, Pillai s, bernays ea (1999) synergism in the oviposi-tion behavior of Plutella xylostella: sinigrin and wax compounds. J Insect behav 12:483–500

sun JY, sonderby Ie, halkier ba, Jander g, de Vos m (2009) non-volatile intact indole glucosinolates are host recognition cues for ovipositing Plutella xylostella. J chem ecol 35:1427–1436

turlings tcJ, benrey b (1998) effects of plant metabolites on the behavior and development of parasitic wasps. Ècoscience 5:321–333

Velasco P, cartea me, gonzález c, Vilar m, Ordás a (2007) Factors affecting the glucosinolate content of kale (Brassica oleracea acephala group). J agric Food chem 55:955–962

White Ja, Whitham tg (2000) associational susceptibility of cotton-wood to a box elder herbivore. ecology 81:1795–1800