Root herbivores and detritivores shape above-ground multitrophic assemblage through plant-mediated effects

Authors


Correspondence author. E-mail: adelagm@ugr.es

Summary

1. Indirect effects mediated by changes in plant traits are the main mechanism by which above- and below-ground herbivores affect each other and their enemies. Only recently the role of decomposers in the regulation of such plant-based systems has been considered. We hypothesized that: (i) below-ground organisms, both herbivores (negative effect on plants) and detritivores (positive effect on plants), will have a profound effect on the interactions among above-ground arthropods; (ii) floral herbivores will negatively affect other above-ground herbivores associated with the plant; and (iii) not only above- and below-ground herbivores, but also detritivores will affect the production of secondary metabolites, i.e. glucosinolates, in the plants.

2. We manipulated the presence of above-ground herbivores, below-ground herbivores and below-ground detritivores on the Brassicaceae Moricandia moricandioides in the field to disentangle their individual and combined effects on other organism groups. We also investigated their effects on the plant’s chemical defence to evaluate potential mechanisms.

3. Our results show that not only above- and below-ground herbivores, but also detritivores affected other herbivores and parasitoids associated with the host plant. Most effects were not additive because their strength changed when other organisms belonging to different functional groups or food web compartments were present. Moreover, below-ground herbivore and detritivore effects on above-ground fauna were related to changes in glucosinolate concentrations and in quantity of resources.

4. This study indicates that multitrophic interactions in plant-based food webs can dramatically change by the action of below-ground organisms. One of the most important and novel results is that detritivores induced changes in plant metabolites, modifying the quality and attractiveness of plants to herbivores and parasitoids under field conditions.

Introduction

Ecological terrestrial communities are characterized by their complex structures, which result from the variation in organism sizes, trophic groups and type of interactions in which organisms are involved (Tscharntke & Hawkins 2002). Recently, the importance of indirect interactions, and more specifically of trait-mediated indirect interactions (TMII), as determinants of the structure of ecological communities has been highlighted (Abrams et al. 1996; Ogushi 2005; Ogushi, Craig & Price 2007). Trait-mediated indirect interactions involve effects transmitted from one species to another through one or more intermediate species, involving changes in the phenotype of the interacting species (Werner & Peacor 2003; Schmitz, Krivan & Ovadia 2004; Ogushi 2005).

In terrestrial ecosystems, plants act as a conduit between above- and below-ground levels, allowing above-ground herbivores to affect below-ground herbivores indirectly and vice versa (Scheu 2001; Bardgett & Wardle 2003; Porazinska et al. 2003; Schröter et al. 2004; Wardle et al.2004; Kaplan et al. 2008). Therefore, the importance of indirect plant-mediated interactions between these organism groups has been highlighted (van Dam et al. 2003; Bezemer & van Dam 2005; Ogushi et al. 2007). The mechanisms behind plant-mediated interactions between herbivores are diverse, whereas the induction of plant chemical defences plays a crucial role (Agrawal 1998; Bezemer & van Dam 2005; Denno & Kaplan 2007; Ogushi et al. 2007; Kaplan et al. 2008; Hopkins, van Dam & van Loon 2009). Changes in the quality and quantity of metabolites can affect both generalist and specialist herbivores by influencing not only their host-plant acceptance and detoxification ability but also the foraging behaviour of their predators and parasitoids (Scheu 2001; Bezemer et al. 2005; Soler et al. 2005; Wolfe, Husband & Klironomos 2005; Rasmann & Turlings 2007; Hopkins et al. 2009).

Plant-centred indirect interaction webs not only involve herbivores and their predators, but also other organisms such as ants or mycorrhizal species can have an important role in shaping the structure of ecological communities (Ogushi et al. 2007). Furthermore, decomposers are prominent members in most food webs, interacting with plants and potentially affecting other organisms of the community (Scheu 2001; Bardgett & Wardle 2003; Porazinska et al. 2003; Schröter et al. 2004; Poveda et al. 2005). Only recently the role of decomposers in the regulation of these types of systems has been considered (Bonkowski et al. 2001; Bezemer & van Dam 2005; Blouin et al. 2005; Poveda et al. 2005; Wolfe et al. 2005). It is well-known that their activity enhances microbial turnover, nutrient recycling and the breakdown of organic matter, favouring plants in many ways (Bardgett & Wardle 2003; Bardgett, Usher & Hopkins 2005). However, only a few studies have explored their effects on other above- or below-ground members of the community, which are mediated by their influences on plant performance, phenotype and chemical defence (Wurst, Dugassa-Gobena & Scheu 2004a; Wurst et al. 2004b, 2006; Blouin et al. 2005; Poveda et al. 2005; Lohmann, Scheu & Müller 2009).

Brassicaceae are well-characterized for their specific defence system, the glucosinolate (GS)–myrosinase system (Halkier & Gershenzon 2006). Root- as well as shoot-feeding herbivores can alter the GS concentrations in both above- and below-ground tissue, locally and systemically, and thereby influence other members of the multitrophic system, including predators and parasitoids (Hopkins et al. 2009). Even detritivores can influence the GS and myrosinase concentrations of above-ground tissue (Lohmann et al. 2009). Therefore, species of this plant family offer useful models to study plant-mediated interactions between organisms, which may be related to induce changes in plant chemistry.

In this study, we examine the role of above- and below-ground multitrophic interactions in the regulation of a plant-based food web. We hypothesized that: (i) below-ground organisms, both herbivores (negative effect on plants) and detritivores (positive effect on plants), will have a profound effect on the interactions among above-ground arthropods; (ii) floral herbivores (FH) will negatively affect other above-ground herbivores associated with the plant; and (iii) not only above- and below-ground herbivores, but also detritivores will affect the production of GS in the plants. To test our hypotheses, we manipulated the presence of above-ground herbivores, below-ground herbivores and below-ground detritivores in the field. We tested their isolated and combined effects on other co-occurring mutualistic and antagonistic organisms using Moricandia moricandioides (Boiss.) Heywood (Brassicaceae) as a host. We also evaluated the isolated and combined effects of the three manipulated organism groups on the plant’s chemical defence, trying to discern potential mechanisms underlying animal–plant–animal interactions under natural conditions, and to understand the importance of TMII in shaping the food web structure.

Materials and methods

Study system

The study was conducted in 2007 at Barranco del Espartal, a seasonal watercourse located in the arid Guadix-Baza Basin (Granada, south-eastern Spain).The climate is continental Mediterranean with strong temperature fluctuations (ranging from −14 to 40 °C) and high seasonality (hot summers and cold winters). The vegetation is an arid open shrub-steppe dominated by Artemisia herba-alba Asso, A. barrelieri Bess., Salsola oppositifolia Desf., Stipa tenacissima L., Lygeum spartum L. and Retama sphaerocarpa L. Furthermore, the annual Brassicaceae species M. moricandioides is highly abundant in this habitat and was used as a model system.

The insect community associated with this plant includes FH, mainly Pontia daplidice L. and Euchloe crameri Batler (Pieridae) and the diamondback moth (Plutella xylostella L., Plutellidae). Many lepidopteran species also feed on the leaves, such as Pieris rapae L., Pieris brassicae L. (Pieridae) and P. xylostella (Gómez 1996).

Experimental set-up

During the winter of 2006–2007, undamaged seedlings of M. moricandioides were grubbed out from the same population the field, taking care of preserving the root systems. These seedlings were potted in plastic pots (7 × 10 cm). Seedlings with no more than three young developed leaves were selected to ensure similar age of all plants. Plants were kept in a common garden until the beginning of spring (end of March 2007) when they were moved back to the field. Plants did not differ at that time point in number of leaves (8·24 ± 0·27 leaves, mean ± SE) or height (3·94 ± 0·40 cm) and none of the plants had developed a flowering stalk. Once in the field, 120 plants were distributed at random into three blocks approximately 100 m apart (40 plants were assigned to each block). Plants were located in five rows of eight plants, each plant 50 cm apart. Plants were re-potted using mixed soil (free of macroarthropods) from the study sites. The new pots consisted of fibre–glass mesh cylinders (10 × 15 cm) of 1-mm mesh size to inhibit the entrance or escape of macroinvertebrates. These pots were then buried with the upper surface even with the ground.

A full factorial design was used to test the effects of three factors with two levels per factor, presence (+) or absence (−), on plant–insect interactions. The three factors were root herbivores (RH hereafter), FH and detritivores (D). FH are defined here as any species that feeds on buds, flowers and fruits regardless of whether they also feed on other parts of the plant. A total of 15 plants were assigned to each of the treatments (five plants per block). One third-instar larva of the detritivore Morica hybrida Charpentier (Coleoptera: Tenebrionidae) was added to the soil of each plant assigned to the D+ treatments. One third-instar larva of the root-feeding herbivore Cebrio gypsicola Graells (Coleoptera: Cebrionidae) was added to the soil of each plant of the RH+ treatments. M. hybrida and C. gypsicola are among the most abundant generalist below-ground detritivores and herbivores, respectively, to be found in the study area (see Doblas-Miranda, González-Megías & Sánchez-Piñero 2007 for more details on below-ground fauna in the study area). The experiment was ended at the end of June (2007). The recovery rate of detritivores and RH at the end of the experiment was very high (94·16% and 91·67% respectively).

Naturally occurring FH were allowed to lay eggs on all FH+ treatment plants, when these plants started to produce flowering stalks. Three species of FH fed on the experimental plants during the experiment, P. daplidice, E. crameri and P. xylostella. All plants belonging to the FH+ treatments had at least one pierid larva (2·46 ± 0·43 pierid eggs/plant) during the experiment, and 33·9% of the plants had at least one P. xylostella larva (0·5 ± 0·11 Plutella eggs/plant). For the FH− treatments, the experiment was checked every 3 days using magnifying glasses and all eggs or larvae of FH were removed by hand from plants.

Data collection

Attack rate and abundance of above-ground herbivores and parasitoids

To score free-living herbivores, the number of naturally occurring sap-suckers (aphids and leafhoppers) and leaf herbivores was recorded on each experimental plant every 3 days. Total abundance of sap-suckers and leaf herbivores was calculated by summing the number of individuals recorded during all the surveys until the end of the experiment (end of June). To avoid problems of summing individuals counted in a previous census, we subtracted at each census the number of individuals of the same instar/type (winged vs. not winged) counted in the previous census. The attack rate was calculated as the probability of a plant being attacked by sap-suckers and leaf herbivores (presence/absence) respectively.

To score seed predators, fruits were collected after complete maturation of seeds but before seed dispersal (28 June 2007). Seed predators were not considered FH because the feed only on mature seeds not altering any other reproductive part of the plant. The abundance and attack rate (proportion of fruits infested by seed predators) were quantified in the lab by opening 10 fruits per plant of larger plants and all fruits of plants which had produced <10 fruits. The abundance of seed predators was quantified by counting the total number of fruits with at least one predator. Because it is possible to find up to three seed predators per fruit, this is a conservative estimate of seed-predator abundance (see Gómez & González-Megías 2002, for a similar procedure).

Cotesia kazak Telenga (Braconidae, Hymenoptera) was the only parasitoid species that attacked P. daplidice and E. crameri in the study area during the study period. Infected larvae of both species are easily identified because of the change in colour and appearance. The solitary parasitoid builds its cocoon on the stems of M. moricandioides. We recorded the proportion of larvae attacked by parasitoids per plant (attack rate), and the abundance of parasitoids per plant as the total number of cocoons found until the end of the experiment.

Induction of chemical defences

To measure GS concentrations of the above-ground plant tissue, the youngest leaf of one stem of each of the experimental plants was collected at the end of the experiment (28 June 2007). Leaf samples were immediately stored in 100% methanol to avoid any degradation by myrosinase activity. Samples were freeze-dried, and the dried material was ground and extracted three times in 80% methanol after the addition of p-hydroxybenzyl GS as an internal standard. Supernatants of extracts were applied to DEAE Sephadex A-25 (Sigma-Aldrich, St. Louis, MO, USA) columns (0·1 g in 2 mL of 0·5 m acetic acid buffer, pH 5, per column, incubated overnight). Columns were washed several times with water followed by two washes with 0·02 m acetic acid buffer, pH 5. Purified sulfatase of Helix pomatia (EC 3.1.6.1, type H-1; Sigma, Taufkirchen, Germany), dissolved in 0·02 m acetic acid buffer, was applied on each column and GS converted to desulfoglucosinolate overnight. Desulfoglucosinolate were eluted from the columns with 6 × 1 mL of water, and the samples were analysed by HPLC (1200 Series; Agilent Technologies, Inc., Santa Clara, CA, USA) coupled with a diode-array detector on a Supelcosil LC-18 column (250 × 4·6 mm, 5 μm; Supelco, Bellefonte, PA, USA) using a gradient from water to methanol. Desulfoglucosinolates were identified by comparison of UV-spectra and retention-times to those identified in earlier studies (Müller & Sieling 2006). Response factors of 0·5 for p-hydroxybenzyl GS, 1 for aliphatic and 0·26 for indolic GS were considered.

Statistical analyses

Generalized linear models were used to test the effects of each factor (FH, RH and D) and their interactions on herbivores (attack rate and herbivore abundance) and on GS concentrations. Block was included in all the analyses to control for the potential effect of the location. Herbivore attack rates were fitted to a binomial distribution with logit as link function, GS concentrations to a normal distribution with identity as link function, and herbivore abundance to a Poisson distribution with log as link function.

Similarly, we used generalized linear models to test the effects of RH and D factors on the parasitoid attack rate and abundance. Parasitoid attack rate was fitted to a binomial distribution with logit as link function, whereas parasitoid abundance was fitted to a Poisson distribution with log as link function.

Generalized linear models were also used to determine the relationship between GS (the explanatory variable) and insect abundances (the response variable). Herbivore abundances were fitted to a Poisson distribution with log as link function in all analyses. The FH effect on the induction of some GS was modified in some of the analysis by D+; therefore, we tested the relationship between those specific GS and herbivore abundance for: (i) all data; (ii) in the presence of D (D+); and (iii) in the absence of D (D−). Mean ± SE is shown throughout the manuscript.

Results

Insect community associated with M. moricandioides

More than 62% of the plants were infected during the study period with different species of aphids. The generalist Myzus persicae Sulzer was the most abundant aphid species (92·8% of total abundance), followed by the specialist Lipaphis erysimi Kaltenbach (5·3%) and other unidentified species (1·9%). In addition, 80% of the plants were also infected by the generalist planthopper species Agalmatium bilobum Fieber (Hemiptera, Issidae).

Foliar herbivores also attacked the experimental plants, colonizing around 20% of them. The most abundant species found were: larvae of P. rapae (5%; specialist on Brassicaceae), the moth Spodoptera sp. (Lepidoptera, Noctuidae; 48·3%) and the specialist sawfly Tenthredo sebastiani Lacourt (Hymenoptera, Tenthredinidae; 46·7%).

More than 50% of the plants had at least one fruit attacked by a seed predator. The only species found on M. moricandioides fruits until now in the study area was an unidentified moth species of the family Blastobasidae (Lepidoptera).

Attack rate and abundance of above-ground herbivores

Above-ground herbivore attack rates were not affected by our experiment (P > 0·05, in all analyses). However, there was a significant effect on aphid, planthopper, foliar herbivore and seed-predator abundances (Table 1).

Table 1.   Results of the generalized linear models (chi-square values) showing the effect of the different factors (FH, RH and D; see Materials and methods) on insect abundance (aphids, planthoppers, foliar herbivores and seed predators), and three glucosinolate (GS) concentrations
 d.f.PlanthoppersFoliar herbivoreSeed predatorsAphids2-S-2-Hydroxy-3-butenyl GS4-Hydroxyindol-3-ylmethyl GSIndol-3-ylmethyl GS
  1. FH, floral herbivores; RH, root herbivores; D, detritivores.

  2. Chi-square values: msP = 0·07; *P < 0·05; **P < 0·01; ***P < 0·001.

Whole model9.97212·88***19·65*40·02***1254·50***22·94**18·05*23·17**
FH134·65***4·88*4·01*115·96***0·013·62ms1·21
RH11·850·941·510·376·03*0·648·27**
D12·90**6·24*8·03**157·72***1·074·78*0·57
FH × RH11·731·414·13*185·14***2·500·110·0001
FH × D11·960·863·890·184·99*6·75**2·63
RH × D18·07**4·75*0·064·05*1·310·460·43
FH × RH × D10·033·310·46124·37***1·090·411·17
Block2121·49***5·5720·19***707·99***8·35*4·7310·88**

Planthopper abundance was negatively affected by FH (Table 1; Fig. 1a). The two-way interaction between RH and D was significant (Table 1). Planthoppers were negatively affected by D but only in RH− treatments (Fig. 1a).

Figure 1.

 Effects of flower herbivores (FH), root herbivores (RH) and detritivores (D) on : (a) planthopper abundance (N planthoppers), (b) foliar herbivore abundance (N folivores) and (d) seed predator abundance (N seed predators). *Significantly different (P < 0·05). Treatments with different letters are significantly different (post hoc log-likelihood ratio test, P < 0·05).

Foliar herbivore abundance was positively affected by FH (Fig. 1b). The significant interaction term between RH and D indicated that foliar herbivores abundance was negatively affected by D and RH, and only in RH−FH− treatment their abundance was significantly higher than in the other treatments (Fig. 1b).

Seed-predator abundance was negatively affected by D (Table 1; Fig. 1c). There was also a significant interaction between RH and FH (Table 1). Only in RH− treatments there was a clear negative effect of FH on seed predators (Fig. 1c).

There was a significant treatment effect on aphid abundance (Table 1), however, the significant three-way interaction term indicated no main effect of any of the factors (Table 1). FH-and D-affected aphid abundance on plants but, in both cases, their effects were modified by the presence of RH. Aphid abundance was negatively affected by FH and D but only when RH was present (Fig. 2).

Figure 2.

 Effects of flower herbivores (FH), root herbivores (RH) and detritivores (D) on aphid abundance (N aphids). Treatments with different letters are significantly different (post hoc log-likelihood ratio test, P < 0·05).

Attack rate and abundance of parasitoids

Parasitoid attack rate differed between treatments because of D (P < 0·05; Appendix S1). Attack rate was higher in D+ treatments (0·11 ± 0·05 D− vs. 0·35 ± 0·08 D+; Appendix S1).

Although the model for the abundance of parasitoids was not significant (Appendix S1), there was a significantly positive effect of D presence on parasitoid abundance (0·23 ± 0·01 parasitoids/plants on D− vs. 0. 46 ± 0·11 on D+).

Glucosinolate concentrations

Seven GS were found in M. moricandioides leaves, four aliphatic and three indolic GS. Total GS concentrations were on average 7·31 ± 0·46 μmol g−1 dry weight. The variation of all GS was rather high but only the concentration of 2-S-2-hydroxy-3-butenyl GS, 4-hydroxyindol-3-ylmethyl GS and 4-hydroxyindol-3-ylmethyl GS on M. moricandioides varied significantly between treatments (Table 1).

RH+ negatively affected the concentration of 2-S-2-hydroxy-3-butenyl GS in plants (Table 1; Fig. 3). Plants from FH+ treatments also had lower concentrations of 2-S-2-hydroxy-3-butenyl GS, but only in the presence of D (Fig. 3) as indicated by the significant interaction term (Table 1).

Figure 3.

 Effects of flower herbivores (FH), root herbivores (RH) and decomposers (D) on concentrations (μmol g−1 dry weight) of 2-S-2-hydroxy-3-butenyl GS (2-S-2-hydroxy GS), 4-hydroxyindol-3-ylmethyl GS (4-hydroxyindol GS) and indol-3-ylmethyl GS. *Significantly different (P < 0·05). Treatments with different letters are significantly different (post hoc log-likelihood ratio test, P < 0·05).

The concentration of 4-hydroxyindol-3-ylmethyl GS differed significantly between treatments (Table 1). As shown by the interaction term, FH+ plants had lower 4-hydroxyindol-3-ylmethyl GS concentrations but only in the absence of D (Fig. 3).

The concentration of indol-3-ylmethyl GS also varied between treatments (Table 1; Fig. 3). In this case, RH induced the production of indol-3-ylmethyl GS in leaf tissue (Fig 3).

Relationships between glucosinolates and insect attack rate and abundance

Aphid abundance was positively related to 2-S-2-hydroxy-3-butenyl GS (β = 0·72 ± 0·32; Appendix S2), but the interaction was significant only in D+ (β = 4·06 ± 0·56; Appendix S2). Foliar herbivore abundance was also positively related to 2-S-2-hydroxy-3-butenyl GS (β = 11·09 ± 1·61; Appendix S2), the relationship was not affected by D (P > 0·05). The relationship was significantly negative for seed predators (β = −2·96 ± 1·03), but it was modified by D (Appendix S2). There was no significant relationship for planthoppers. Furthermore, there was a significantly negative relationship between parasitoid attack rate and 2-S-2-hydroxy-3-butenyl GS (β = −26·02 ± 12·03,) but only in D+ (Appendix S2). This interaction was also marginally significant for parasitoid abundance in the presence of D (β = −9·91 ± 6·90; * Appendix S2).

Furthermore, there was a significantly negative relationship between 4-hydroxyindol-3-ylmethyl GS and aphid abundance (β = −8·56 ± 1·29), and seed predator abundance (β = −24·75 ± 7·41; Appendix S2). Planthopper abundance was positively correlated with this GS (β = 3·01 ± 0·37; Appendix S2). The relationship was not significant for foliar herbivores or parasitoids. D did not modify any of the interactions between insect abundance and this GS.

With regard to indolic GS, the abundances of aphids and planthoppers (β = −1·90 ± 0·26 and −1·94 ± 0·43 respectively), as well as seed predators (β = −2·84 ± 0·79; Appendix S2) were negatively related to indol-3-ylmethyl GS concentrations. There was no significant relationship between GS concentrations and other insects.

Discussion

Above- and below-ground herbivore effects on above-ground organisms

Our experiment demonstrates that both above- and below-ground herbivores significantly affect other herbivores co-occurring on M. moricandioides (Fig. 4). Antagonistic relationships between herbivores are usually expected but also positive effects between herbivores feeding on the same plant have been reported in previous studies (Agrawal & Sherriffs 2001; Hopkins et al. 2009). In most cases, the underlying mechanism seems to be related to the induction of plant metabolites. Feeding damage can induce the production of secondary compounds by the plant, attracting specialist herbivores, which in turn trigger a boost in the plant’s chemical defences (but see Hopkins et al. 2009 for a review). In our system, FH favoured the abundance of foliar herbivores on M. moricandioides. Although FH had a strong effect on the GS content in M. moricandioides (Fig. 5), individual GS responded in different ways, being either reduced or increased. Thus no general pattern could be derived from our experiment. Because FH also had a negative effect on above-ground plant biomass and no effect on the quality of the above-ground tissue in terms of N content (A. González-Megías, unpublished) none of those factors seem to clarify the positive effect of FH on foliar herbivores.

Figure 4.

 Effects of floral herbivores and decomposers on other organisms associated with Moricandia moricandioides (a) in absence of root herbivores and (b) in presence of root herbivores.

Figure 5.

 Effects of floral herbivores and root herbivores on three different glucosinolates (GS) (indol-3, indol-3-ylmethyl GS; 4-hydroxy, 4-hydroxyindol-3-ylmethyl GS; 2-S2-hydroxy, 2-S-2-hydroxy-3-butenyl GS), and the relationship between GS and the other insects associated to the host-plant (a) in presence of detritivores and (b) in absence of detritivores. Dashed line, effects on GS; solid line, responses to GS (or other unidentified correlated traits).

Most interactions between FH and RH with the other co-occurring herbivores were negative and were mediated by the plant (Figs 4 and 5). FH had a negative effect on aphid and plant-hopper abundance. Similar negative effects of chewing herbivores on sap-suckers have been reported in several studies (Inbar et al. 1999; González-Megías & Gómez 2003; van Zandt & Agrawal 2004; Poveda et al. 2005; Gómez & González-Megías 2007). Two main mechanisms have been proposed for these negative effects on sap-suckers feeding on stems associated with the reproductive parts of the plant such as flower stalks, buds and even flowers: induced resistance by altered plant chemical defences (Faeth 1992; Faeth & Wilson1997; Denno & Kaplan 2007), and reduced resources (see Poveda et al. 2005). In the M. moricandioides system, FH led to reduced concentrations of GS, specifically 2-S-2-hydroxy-3-butenyl GS and 4-hydroxyindol-3-ylmethyl GS. Conversely, aphid abundance was positively correlated with 2-S-2-hydroxy-3-butenyl GS. Whether this is because of a preference of the aphid species for this GS, or whether the aphids induced particularly 2-S-2-hydroxy-3-butenyl GS, or both, can not be discriminated. It has been shown that aphids, at least specialist species, prefer tissue with higher GS concentrations (Gabrys, Tjallingii & van Beek 1997), allowing them to escape from competition (Hopkins et al. 2009). However, the most abundant aphid species on M. moricandioides was a generalist aphid, M. persicae. In addition, the reduction of resources (a density-mediated indirect interaction) can be an alternative mechanism by which FH-affected sap-suckers because FH reduced the production of flower stalks and flower production in M. moricandioides (A. González-Megías, unpublished) and thereby diminished the resources for sap-suckers.

Floral herbivores and RH also negatively affected seed predators (Fig. 4). FH might affect seed predators by exploitative competition, as a consequence of reducing the availability of oviposition sites. FH reduced flower production by nearly 50% in M. moricandioides during the study period (A. González-Megías, unpublished). Similar negative effects of FH on seed-predator abundance have been observed in other systems (Juenger & Bergelson 1998; Gómez & González-Megías 2002, 2007; Freeman, Brody & Neefus 2003; Milbrath & Nechols 2004). An additional mechanism observed in other systems is incidental predation. FH can incidentally predate on seed predator eggs when feeding on flowers (Gómez & González-Megías 2002, 2007). In addition, FH affected seed-predator abundance by inducing changes in GS concentrations (TMII). In this case, the negative interaction between FH and seed predators was somewhat mitigated by the negative effect of FH on the production of aliphatic GS, which were negatively related to seed predator abundance. The RH effect on seed predators was also related to GS induction. There was a clear negative relationship between seed predators and indol-3-ylmethyl GS, which was induced in the study system by RH. By inducing this GS or through changes of plant emitted volatile pattern, RH probably increased host-plant attractiveness to parasitoids of different herbivores, as found in other studies (Masters, Jones & Rogers 2001; van Dam et al. 2003).

Herbivore interactions on M. moricandioides are therefore mainly based on indirect interactions: on density-mediated indirect interactions by reducing resources, and on TMII through altered concentrations of chemical plant compounds, i.e. GS or other traits. The response of the plant to a given herbivore species as well as the response of herbivores to the plant chemical profile is highly species-specific with regard to the induction and effect of individual GS.

Detritivore effects on above-ground organisms

Detritivore effects on plant growth and reproductive success can lead to indirect interactions with above- and below-ground organisms sharing the same host plant, as found in several studies (Bonkowski et al. 2001; Scheu 2001; Bezemer & van Dam 2005; Bezemer et al. 2005; Wolfe et al. 2005; Kessler & Halitschke 2007; Rasmann & Turlings 2007). Some of these studies found a positive effect of detritivores in sap-suckers (Bonkowski et al. 2001; Poveda et al. 2005), which was apparently associated with changes in the content of soluble nitrogen or amino acids in plant phloem (Bonkowski et al. 2001). However, we found a negative effect of detritivores on the proportion of N content on above-ground tissue (A. González-Megías, unpublished). This negative effect of D on plant quality could explain the negative effects of D on aphid, plant-hopper, seed predator and leaf herbivore abundance in the M. moricandioides system.

Although not directly related to roots, detritivores can alter plant defence levels such as GS concentrations (Bezemer & van Dam 2005; Lohmann et al. 2009) and thereby indirectly affect herbivores. The effects appear to be related to changes in nitrogen availability to the plants as well as to changes in gene expression related to plant defence (Bezemer & van Dam 2005). In Sinapis alba L., aromatic GS concentrations significantly increased because of the presence of earthworms (Lohmann et al. 2009). In M. moricandioides, D presence altered the effects of FH in the production of aliphatic GS, supporting the hypothesis that D prompts changes in plant chemistry (Fig. 5).

It bears noting that D presence also affected parasitoid abundance on M. moricandioides, although the mechanism involved is not clear. The use of induced volatile metabolites to attract natural enemies is a mechanism observed in other systems for both specialists and generalist parasitoids (Turlings et al. 2002; Soler et al. 2005). Isothiocyanates and nitriles, the hydrolysis products of GS, are known to be involved in attraction of parasitoid species in other Brassicaceae systems (Bradburn & Mithen 2000; Mumm et al. 2008) and may have been involved also in parasitoid attraction in the M. moricandioides system. However, an increased attraction of plants to parasitoids after detritivore activity is a misleading signal to the parasitoids and should therefore not be evolutionary stable.

Complex interactions among food web components

More than 50% of the pairwise interactions of FH, RH and D with the co-occurring organisms of the community and with the plant GS were not additive but were affected by another manipulated component. These interactions did not affect the manipulated organisms with the same intensity, as most D effects were influenced by RH, some FH effects were influenced by RH or D, but barely any of the RH effects appeared to have been influenced by any of the other organisms (Figs 4 and 5). Indeed, two different scenarios of the multiple trophic relationships in our system arose, depending on the presence of RH (Fig. 4). Most of the negative relationships between FH and D on the other herbivores were evident only in the absence of RH. The impacts of RH on plants are well-known, ranging from effects on plant chemical defences to the modification of the phloem nitrogen concentration, which can increase sap-sucker abundance (Gange & Brown 1989; Masters & Brown 1997; Masters et al. 2001; Poveda et al. 2005). Therefore, RH probably had an indirect negative effect on aphids via induced defences but at the same time a positive interaction with aphids via other modifications of plant traits. This positive interaction offset the negative effect of FH and D on aphids, acting as a buffer. Even more importantly, the influence of FH on GS concentration was clearly altered by detritivores.

Surprisingly, two alternative scenarios also become evident in our system with regard to plant chemical defences (Fig. 5), where plant secondary metabolites and the relationship with other co-occurring organisms varied according to the presence of detritivores (Fig. 5). Detritivores altered the influence of FH on GS, potentially affecting the response of the other insects to the plant defence mechanisms. It is important to highlight that most of these complex interactions occurred between organisms belonging not only to the same trophic guild but also to the same food web compartment, thereby increasing the complexity of the food web functioning and dynamics. In a less complex experiment, where not all of the interacting organisms are considered, such interactions would remain unnoticed. Many of the mechanisms by which a third species altered a pair-wise interaction in our system remain unknown and more specific and mechanistic experiments are necessary to elucidate this issue. Jasmonic acid may be an important signal mediating the modifications of plant metabolites such as GS, as shown earlier (Textor & Gershenzon 2009) but also the release of volatile organic compounds (Ozawa et al. 2000), which attract herbivores and parasitoids from a distance.

Summarizing this study shows that below-ground organisms through plant-mediated indirect effects can shape the interactions among above-ground organisms. One of the most important and novel result is that detritivores induced changes in plant metabolites, varying the quality and attractiveness of plants to herbivores and parasitoids.

Acknowledgements

The authors would like to thanks Francisco Sánchez-Piñero, and José Antonio Hódar for their frequent discussion on food web interactions, and their help in the field. We thank J. M. Nieto Nafría (Aphids), T. Oltra and J. Vicente Falco (Braconidae), Oscar Aguado (Tenthedinidae) and Vladimir Gnezdilov (Issidae) for the identification of the specimens. Jose M. Gómez, Michael Rostás, Martín Pareja and Rosa Menendez kindly revised an early version of this manuscript. This work was partially funded by CICYT grant BOS2001-3806.

Ancillary