Non-pathogenic rhizobacteria interfere with the attraction of parasitoids to aphid-induced plant volatiles via jasmonic acid signalling



Beneficial soil-borne microbes, such as mycorrhizal fungi or rhizobacteria, can affect the interactions of plants with aboveground insects at several trophic levels. While the mechanisms of interactions with herbivorous insects, that is, the second trophic level, are starting to be understood, it remains unknown how plants mediate the interactions between soil microbes and carnivorous insects, that is, the third trophic level. Using Arabidopsis thaliana Col-0 and the aphid Myzus persicae, we evaluate here the underlying mechanisms involved in the plant-mediated interaction between the non-pathogenic rhizobacterium Pseudomonas fluorescens and the parasitoid Diaeretiella rapae, by combining ecological, chemical and molecular approaches. Rhizobacterial colonization modifies the composition of the blend of herbivore-induced plant volatiles. The volatile blend from rhizobacteria-treated aphid-infested plants is less attractive to an aphid parasitoid, in terms of both olfactory preference behaviour and oviposition, than the volatile blend from aphid-infested plants without rhizobacteria. Importantly, the effect of rhizobacteria on both the emission of herbivore-induced volatiles and parasitoid response to aphid-infested plants is lost in an Arabidopsis mutant (aos/dde2-2) that is impaired in jasmonic acid production. By modifying the blend of herbivore-induced plant volatiles that depend on the jasmonic acid-signalling pathway, root-colonizing microbes interfere with the attraction of parasitoids of leaf herbivores.


Plants have a central position in terrestrial ecosystems by mediating interactions among herbivores and other community members, such as pathogens (Stout, Thaler & Thomma 2006), herbivores of different species (Rodriguez-Saona et al. 2005; Kaplan & Denno 2007; Soler et al. 2012) and carnivores (Dicke 2009; Dicke & Baldwin 2010), even when these community members are located above- and belowground (van der Putten et al. 2001; Rasmann & Turlings 2007; Soler et al. 2007). Recent evidence shows that plants also mediate interactions among beneficial organisms aboveground such as pollinators (Kessler & Baldwin 2011) and carnivores (Poelman et al. 2011). However, how plants mediate responses between beneficial microbes belowground and beneficial insects aboveground is not yet known. In response to herbivory, plants release volatiles that are exploited as host- and prey-location cues by such carnivores (Turlings, Tumlinson & Lewis 1990; Vet & Dicke 1992; Dicke 2009; Dicke & Baldwin 2010). Belowground, plants interact with beneficial microbes, such as mycorrhizal fungi and plant growth-promoting rhizobacteria. These microbes have been extensively shown to not only improve plant nutrition and tolerance to abiotic stress (Dimkpa, Weinand & Asch 2009), but also to induce resistance in systemic tissues against microbial pathogens and against aboveground herbivorous insects (Pozo & Azcon-Aguilar 2007; van Oosten et al. 2008; Pozo et al. 2008; Gehring & Bennett 2009; Koricheva, Gange & Jones 2009; Pineda et al. 2010). Our knowledge on the interactions between herbivores and soil microbes other than mycorrhizal fungi is scarce, but recent studies suggest that different groups of soil-borne microbes affect herbivores aboveground via similar mechanisms (van Oosten et al. 2008; Pineda et al. 2010, 2012).

There are remarkably few studies on the effects of beneficial microbes on the third trophic level, that is, parasitoids and predators, despite their relevance for understanding the functioning of plants in their multitrophic environment (van der Putten et al. 2001; Gange, Brown & Aplin 2003). This understanding is especially relevant for agricultural ecosystems, in which carnivorous insects are released for insect pest control simultaneously to the use of beneficial microbes as biofertilizers or to increase resistance to pathogens. Increasing evidence shows that mycorrhizal fungi generally enhance the attraction of carnivorous insects to plants (Gange et al. 2003; Guerrieri et al. 2004; Hempel et al. 2009; Wooley & Paine 2011), although opposite effects can be observed depending on the mycorrhizal strain (Gange et al. 2003). However, the mechanisms underlying these effects are yet unknown. Exciting discoveries have shown that root herbivores modify the blend of herbivore-induced plant volatiles (HIPVs) and that this alters interactions of the plant with aboveground organisms at several trophic levels (Rasmann & Turlings 2007; Soler et al. 2007). Similarly, recent evidence suggests that mycorrhizal fungi also modify the blend of HIPVs (Fontana et al. 2009; Leitner et al. 2010), but whether such microbial modification of HIPVs affects the behaviour of carnivorous insects, that is, the enemies of aboveground herbivores, remains to be investigated.

The emission of HIPVs is a form of indirect plant defence, whereas the production of toxic compounds against herbivores is a form of direct defence. Both direct and indirect defences are under the control of a complex network of signal-transduction pathways that are regulated by different phytohormones, of which jasmonic acid (JA) is a central regulator (Thaler et al. 2002; Kessler, Halitschke & Baldwin 2004; Shiojiri et al. 2006; Girling et al. 2008; Kusnierczyk et al. 2008; Snoeren, Van Poecke & Dicke 2009; Kusnierczyk et al. 2011). JA has classically been considered the key regulator of defences against leaf chewers, but there is increasing evidence for its crucial role in direct and indirect defences against phloem feeders, too (Thompson & Goggin 2006; Girling et al. 2008; Kusnierczyk et al. 2008, 2011; Walling 2008). Interestingly, induced systemic resistance (ISR) that is mediated by plant growth promoting rhizobacteria is also regulated by the JA signal-transduction pathway and is a form of priming for enhanced defence rather than direct activation of defence (van Oosten et al. 2008; Pozo et al. 2008; Pieterse et al. 2009; Pineda et al. 2012). While most studies on priming by non-pathogenic rhizobacteria have focused on direct defences against herbivores, it has not been investigated if these microbes can also prime plants for an enhanced emission of plant volatiles, as other priming elicitors do (Engelberth et al. 2004; Ton et al. 2006; Peng et al. 2011).

In a previous study with the same rhizobacteria-plant-herbivore system as used in this study, we observed that the rhizobacterium Pseudomonas fluorescens strain WCS417r had a positive effect on the performance of the aphid Myzus persicae feeding on the phloem of Arabidopsis thaliana (Pineda et al. 2012). These results were congruent with the general pattern that has been proposed for mycorrhizal fungi (Gehring & Bennett 2009; Koricheva et al. 2009), which indicates a negative effect on the development of generalist leaf-chewing insects, and neutral or positive effects on phloem-feeding herbivores as well as on specialist leaf-chewing insects. However, to establish predictable patterns in the effects of beneficial microbes belowground on plant defence against herbivores aboveground, more studies including the third trophic level are needed, and especially the underlying mechanisms involved in the plant-mediated interaction between soil microbes and the third trophic level, need to be unravelled. By combining ecological, chemical and molecular approaches, we here examine these mechanisms. Based on the knowledge about below–aboveground interactions (Gange et al. 2003; Guerrieri et al. 2004; Rasmann & Turlings 2007; Soler et al. 2007), on the key role of JA signalling in HIPV emission (Dicke et al. 1999; Thaler et al. 2002; Kessler et al. 2004; Girling et al. 2008; Snoeren et al. 2009), and on our previous results (Pineda et al. 2012), we hypothesize that: (1) rhizobacteria prime the plant for an enhanced emission of herbivore-induced plant volatiles; (2) the aphid parasitoid Diaeretiella rapae will be more attracted, will lay more eggs and will perform better on aphid-infested plants that are colonized by rhizobacteria than on non-colonized aphid-infested plants; (3) the JA signalling pathway regulates the rhizobacteria-induced effects on the emission of aphid-induced plant volatiles. Our results reveal that rhizobacteria modify the blend of HIPV via a JA-dependent mechanism, with unexpected consequences for the behaviour of an aphid parasitoid.


Rhizobacteria, plant and insect material

P. fluorescens strain WCS417r bacteria were grown for 48 h at 28 °C on King's medium B (King, Ward & Raney 1954) agar plates containing 25 mg L−1 of rifampicin (Pieterse et al. 1996). Bacteria were collected and re-suspended in 10 mm MgSO4 to a density of 109 colony-forming units (cfu) mL−1[optical density (OD660) = 1.0] before being mixed through autoclaved soil (see below). In a previous experiment, we confirmed that aphid infestation does not affect rhizobacterial colonization (t-test, P = 0.742), supporting the results of a recent study (Lee, Lee & Ryu 2012). We found that the mean (±SE, based on 10 replicates per plants) was in plants with aphids 1.07 ± 0.12 × 108 cfu g−1 rhizosphere soil, versus 1.13 ± 0.12 in control plants. Both numbers are above the required threshold to induce systemic resistance of 105 cfu g−1 rhizosphere soil (Raaijmakers et al. 1995).

Seeds of A. thaliana (L.) Heynh. [ecotype Columbia (Col-0) and the mutant aos/dde2-2 in the Col-0 background] were sown in a soil mixture for Arabidopsis (Horticoop, Bleiswijk, the Netherlands) that was autoclaved at 121 °C twice for 20 min with a 24 h interval. One-week-old seedlings were transferred to pots containing the same soil mixture and allowed to grow for 5–6 weeks. Before transfer of the seedlings, a suspension of bacteria (109 cfu mL−1) was mixed through the soil to a final density of 5 × 107 cfu g−1, as described previously (Pieterse et al. 1996). Control soil was supplemented with an equal volume of 10 mm MgSO4. Plants were cultivated in a growth chamber with an 8 h day (200 µmol m−2 s−1 at 21 °C) and a 16 h night (21 °C) cycle at 50–70% relative humidity (RH). The plants were watered three times a week, adding a total of 50 mL per pot weekly. During the experiment, all plants remained in the vegetative stage.

The generalist aphid M. persicae (Sulzer) (Hemiptera: Aphididae) was reared in a greenhouse at 22–24 °C, 50–70% RH and a 16L:8D photoperiod, on radish plants, Raphanus sativus L. The aphid parasitoid D. rapae (McIntosh) (Hymenoptera: Braconidae) was reared on M. persicae feeding on radish plants, in a growth chamber at 24 ± 1 °C, 50–70% RH and a 16L:8D photoperiod. Wasp mummies were carefully detached from the radish plants and adults were allowed to hatch in a cage supplemented with water and honey, without contact with plant or aphid material. All mated female wasps that were used in the experiments were 2–3 d old.

Behavioural assays

Dual-choice tests were performed using a closed-system Y-tube olfactometer, which was illuminated from above (Snoeren et al. 2009). The set-up was placed on a table surrounded by white cardboard to reduce any visual bias for the parasitoids. All experiments were conducted in an acclimatized room without natural daylight (21 ± 2 °C). Both parasitoids and volatile sources were placed under the lights for at least 1 h before an experiment was conducted to allow acclimatization to the light level and temperature. Then, 15 min prior to the bioassay, plants were introduced in the jars, and a single female parasitoid was released in the Y-tube. Wasps were given 10 min to make a choice, as pilot experiments and previous studies (Girling et al. 2006, 2008) have shown that this period is sufficient for a response to be elicited. The olfactometer was divided into several sections. A choice was recorded when a female wasp reached a line marked at 1 cm from the end of each olfactometer arm, and did not return to the junction for at least 30 s. Wasps that did not choose within 10 min were excluded from the statistical analysis. After 5–6 females were tested, the position of the jars containing the two odour sources was interchanged, to compensate for any unforeseen asymmetry in the set-up. Experiments were repeated on several days, with ca. 20 female wasps tested per pairwise comparison per day. In total, four sets of plants and 71–96 females were evaluated per pairwise comparison.

As odour sources, four treatments were arranged based on the presence/absence of rhizobacteria and aphids: (1) control uninfested (C); (2) control aphid-infested (CA); (3) rhizobacteria-treated uninfested (R); (4) rhizobacteria-treated aphid-infested plants (RA). In the aphid treatments, Arabidopsis plants (6–7 weeks old) were infested with 60 M. persicae of first and second nymphal stages during 3 d before the experiments. Each odour source consisted of four plants of any of the treatments 1–4. Because D. rapae females require post-emergence experience to find aphid-infested plants (Girling et al. 2006), prior to the bioassays, female wasps were provided with an Arabidopsis plant heavily infested with M. persicae, on which they were allowed to gain post-emergence oviposition experience for 1 h before the trial. To account for a potential effect of associative learning on the choices made, half of the parasitoids tested were provided with oviposition on one of the two odour sources, and the other half on the other odour source.

To confirm the results of previous studies showing that D. rapae can use volatiles from infested A. thaliana Col-0 as host-location cues (Girling et al. 2008), the following dual-choice experiment was conducted as a control: (1a) control uninfested (C) versus control aphid-infested plants (CA). To assess whether D. rapae also used volatiles from rhizobacteria-treated aphid-infested plants as host location cues, the following experiment was conducted: (1b) rhizobacteria-treated undamaged (R) versus rhizobacteria-treated aphid-infested plants (RA). To test the second hypothesis, that the effect of rhizobacteria is a result of the plant's interaction with both aphids and rhizobacteria and not simply a rhizobacteria-plant interaction, the following experiments were conducted: (1c) control aphid-infested (CA) versus rhizobacteria-treated aphid-infested plants (RA); (1d) control uninfested (C) versus rhizobacteria-treated uninfested (R). To test the third hypothesis, that the intact JA signal-transduction pathway in A. thaliana is required for the plant-mediated effect of rhizobacteria on D. rapae response to aphid-infested plants, the following experiment was conducted: (5a) control aphid-infested aos/dde2-2 (CAaos) versus rhizobacteria-treated aphid-infested aos/dde2-2 (RAaos). As a control, to confirm that D. rapae can use volatiles from infested A. thaliana mutant plants aos/dde2-2 as host-location cues, we conducted the following experiment: (5b) control uninfested aos/dde2-2 (Caos) versus control aphid-infested aos/dde2-2 (CAaos).

Oviposition preference assay

Naïve mated female wasps were used to investigate the effects of rhizobacteria on the oviposition preference of the parasitoid. Plants were infested with seven adult aphids for 3 d, which yielded 60 ± 5 aphids of different life stages at the time of the choice assay; this infestation level was similar to that in the behaviour assays. In a greenhouse compartment with the same conditions as described for aphid rearing, four CA and RA were arranged in two groups, 35 cm apart, in individual tents (75 × 75 × 115 cm). One mated naïve female wasp was released per tent at a distance of 40 cm from the plants, and allowed to forage freely and oviposit for 24 h. A total of 11 female wasps were simultaneously evaluated; the experiment was repeated during three consecutive days with new wasps and plants, and in two independent experiments. Subsequently, the wasps were collected and the plants were kept individually in closed ventilated containers, in a climate room under the same conditions as described for parasitoid rearing. The proportion of plants with mummies (selected plants) and the number of mummies per plant were assessed 9 d after the 24 h experiment, as in a previous experiment we observed that all parasitized aphids mummified after this period. The number of mummies observed on in total 400 plants was then analysed.

Performance assay

To investigate the effects of rhizobacterial colonization of the roots on the development of D. rapae, we observed parasitoid development on aphids feeding on control or rhizobacteria-treated plants. Naïve mated female wasps were used to parasitize 3-day-old aphid nymphs. Parasitism was achieved by offering six nymphs to an individual female, and removing them once parasitization was observed. Three aphids were then placed on a control plant, and the other three on a rhizobacteria-treated plant. Plants were kept individually in a climate room at the same experimental conditions as the parasitoid rearing. Parasitized aphids were observed daily and once mummies were formed, they were kept individually in glass vials closed with cotton wool until adult parasitoids emerged. Freshly emerged wasps were immediately frozen, then dried at 80 °C for 24 h, and weighed on a microbalance. The following performance parameters were assessed: number of mummies relative to the observed aphids (egg plus larval survival of the parasitoids); number of adults relative to the number of mummies (pupal survival of the parasitoids); time from parasitization to mummy formation (egg plus larval developmental time); time from mummy formation to adult emergence (pupal developmental time); dry weight of male and female adult parasitoids. A total of 222 aphids were parasitized, and their development was monitored on 37 control plants and 37 rhizobacteria-treated plants.

Headspace collection and analysis of volatiles

During the olfactometer bioassays, dynamic headspace collection of volatiles from the tested plants was performed. Volatiles were trapped during 4 h by sucking air out of the jars at a rate of 200 mL min−1 through a stainless steel cartridge (Markes, Llantrisant, UK) filled with 200 mg Tenax TA (20/35 mesh; CAMSCO, Houston, TX, USA). The plastic pots were wrapped in aluminium foil to prevent the release of plastic-related volatiles. Additionally, volatiles were collected from empty plastic pots wrapped in aluminium foil as a control; compounds detected in these samples were excluded from the data obtained for the plant samples. Immediately after collection, the Tenax cartridges were dry purged with nitrogen at 20 psi for 10 min at ambient temperature to remove moisture, and then they were stored at −20 °C until analysis.

Headspace samples were analysed with a Thermo Trace GC Ultra (Thermo Fisher Scientific, Waltham, MA, USA) connected to a Thermo Trace DSQ (Thermo Fisher Scientific) quadruple mass spectrometer. Volatiles were desorbed from the cartridges using a thermal desorption system at 250 °C for 10 min (Model Ultra Markes) with a helium flow of 20 mL min−1. Analytes were focused at 0 °C on an electronically cooled sorbent trap (Unity, Markes) and were then transferred in splitless mode to the analytical column (ZB-5, 30 m, 0.25 mm i.d., 1.0 µm film thickness, Phenomenex, Torrence, CA, USA) by rapid heating of the cold trap to 250 °C. The GC was held at an initial temperature of 40 °C for 3.5 min followed by a linear thermal gradient of 5 °C min−1 to 280 °C and held for 4 min under a column flow of 1 mL min−1. The column effluent was ionized by electron impact ionisation at 70 eV. Mass spectra were acquired by scanning from 35–350 m/z with a scan rate of 5.38 scans s−1. Tentative identifications were made by comparison of spectra with the mass spectral databases NIST 2005, Wiley 7th edition spectral library and the Wageningen Mass Spectral Libraries Database of Natural Products. Experimentally calculated linear retention indices (LRI) were also used as additional criterion to identify the compounds. Relative quantification (peak areas of individual compounds) was obtained using a single (target) ion, in selected ion monitoring (SIM) mode. The individual peak areas of each compound were further used in the statistical analysis.

Statistical analysis

A binomial test was used to determine whether plant preferences of the female wasps differed significantly from a 50:50 distribution (p = q = 0.5, two-tailed, α = 0.05) (Genstat 13). Generalized linear models (GLMs) with logit link function and binomial distribution were applied to analyse the effect of previous oviposition experience on parasitoid preference, the proportion of plants with and without rhizobacteria that was selected for oviposition by the wasps, as well as the performance parameters of larval and pupal parasitoid survival. The number of mummies on plants with and without rhizobacteria from the oviposition preference assay, as well as larval and pupal developmental time from the performance assay, was analysed using GLM with Poisson distribution and logarithm link function. Adult dry weight was normally distributed and the variances were homogeneous, therefore the data were analysed with a two-way analysis of variance (anova), with sex and treatment as main factors.

The volatile blends of differently treated plants were log transformed and analysed with the multivariate projection to latent structures-discriminant analysis (PLS-DA module of SIMCA-P 12.0.1, Umetrics, Umeå, Sweden). The number of significant PLS components was determined by cross-validation (Eriksson et al. 2006). The results of the analysis are visualized in score plots, which reveal the sample structure according to the model components. In addition, this method aims to identify which compounds are important for the differences between the complex volatile blends emitted by plants subjected to the aforementioned treatments. The analysis shows the variable importance in the projection (VIP) of each variable, that is, a volatile compound. Variables with VIP values larger than 1 are most influential for the model (Eriksson et al. 2006). To pre-process data, the integrated peak areas were normalized [log(X+1)], mean-centred and scaled to unit variance. Additionally, emitted quantities of individual compounds were tested for significant differences between plant treatments using a t-test.


Rhizobacterial colonization decreases the attraction of parasitoids to aphid-infested plants

In dual-choice olfactometer tests, females of D. rapae showed a strong attraction towards volatiles from wild-type A. thaliana infested by its host, the aphid M. persicae (Fig. 1a; binomial test, P < 0.001). Similarly, in a situation where aphid-infested and control plants had been treated with rhizobacteria, the parasitoids were more attracted towards aphid-infested plants (RA) than to uninfested ones (R) (Fig. 1b; P = 0.003). To evaluate whether rhizobacterial colonization modified the attraction towards aphid-infested plants, females were offered a choice between odours from CA versus RA. The wasps preferred the CA to the RA plants (Fig. 1c; P < 0.001). However, as we hypothesized, when plants were not infested by aphids, parasitoids did not discriminate between the volatiles from C and R plants (Fig. 1d; P = 0.19). Parasitoid experience did not affect preference behaviour, neither in the choice of C versus R nor in the choice of CA versus RA (Supporting Information Fig. S1).

Figure 1.

Behavioural responses in a Y-tube olfactometer of Diaeretiella rapae female parasitoids to volatiles from differently treated Arabidopsis thaliana Col-0 plants. Plants were either treated with rhizobacteria (R) or left untreated (C), and either uninfested (C, R) or infested with 60 Myzus persicae aphids (CA, RA), 3 d before the experiment. Data represent the total number of parasitoids that chose for either of the two odour sources as determined in four replicated experiments with different sets of plants. The number of tested and responsive (in parentheses) parasitoid females is given for each test. Asterisks indicate a significant difference within a choice test: **P < 0.01; ***P < 0.001; ns not significantly different (binomial test).

The effect of rhizobacterial colonization on parasitoid odour preference is also reflected in oviposition behaviour but not in parasitoid performance

To evaluate whether rhizobacteria colonization affects D. rapae performance, the development of parasitized aphids on control and rhizobacteria-treated plants was observed until adult parasitoids emerged. Rhizobacterial treatment did not affect the performance parameters of developmental time (Fig 2a; GLM; egg plus larvae: deviance ratio1,125 = 2.84; P = 0.095; pupae: deviance ratio1,115 = 2.81; P = 0.096), survival (Fig. 2b; GLM; eggs plus larvae: deviance ratio1,70 = 1.01; P = 0.319; pupae: deviance ratio1,54 = 0.94; P = 0.338) and dry weight of adult females and males (Fig. 2c; anova, F1,112 = 0.38; P = 0.536). Since parasitoid survival is not affected by rhizobacterial treatment, the number of mummies can be used as an indicator of oviposition frequency. Female wasps showed a lower oviposition rate in aphids on plants treated with rhizobacteria than in aphids on control plants (Fig. 3a; GLM, deviance ratio1,399 = 21.66; P < 0.001). Similarly, the proportion of plants with mummies was lower when plants were colonized by rhizobacteria than when they were untreated (Fig. 3b; deviance ratio1,399 = 9.48; P = 0.002).

Figure 2.

Performance of Diaeretiella rapae (back transformed means ± SEM) when developing in aphids feeding on control or rhizobacteria-treated plants. Effect of rhizobacterial treatment on (a) developmental time of eggs plus larvae and pupae; (b) survival (%) of eggs plus larvae and pupae; (c) dry weight (µg) of adult females and males. ns not significantly different (P > 0.05).

Figure 3.

Oviposition preference of Diaeretiella rapae for aphids growing on rhizobacteria-treated or untreated control Arabidopsis thaliana Col-0 plants. Back-transformed means (± SEM) of (a) the number of mummies per plant and of (b) the proportion of plants with mummies (i.e. selected for oviposition) (GLM; **P < 0.01, ***P < 0.001). GLM, general linear model.

Aphid herbivory suppresses the emission of several volatiles and induces the emission of others

A total of 56 compounds were detected in the headspace of Arabidopsis plants attacked by the aphid M. persicae (Supporting Information Tables S1 & S2). A multivariate analysis including volatiles emitted by aphid-infested and uninfested control plants (CA and C, respectively) resulted in a model with three significant components (Fig. 4a; PLS-DA, R2X = 0.63, R2Y = 0.99, Q2 = 0.88). In this model, a total of 21 compounds had VIP values higher than 1 (Fig. 5a), which means that these were the most important compounds in terms of differentiating these two odour blends (Fig. 4a). The four most influential compounds were in order of importance (Fig. 5a): (Z)-3-hexenyl acetate (C6 ester), β-caryophyllene, γ-caryophyllene and β-gurjunene (three sesquiterpenes). It is remarkable that these four compounds were almost exclusively detected in the blends of aphid-infested plants. Interestingly, the majority of the compounds that characterized the volatile blend of aphid-infested plants, were emitted in smaller amounts by CA plants than by C plants (Fig. 5a), with significant differences (Supporting Information Table S1; t-test, P < 0.05) for six compounds (longifolene, 2-phenylpropan-2-ol, 5,5-dimethyl-2(5H)-furanone, 2-nonenal, limonene, 1-pentanol).

Figure 4.

Separation of volatile blends emitted by Arabidopsis thaliana Col-0 plants in response to rhizobacterial treatment and/or aphid infestation (PLS-DA; n = 5 samples). The first two components are represented and the percentage of variation explained in parentheses. Discriminant plots for volatile blends emitted by (a) control (C) and aphid-infested plants (CA); (b) rhizobacteria-treated plants that were either uninfested (R) or infested by Myzus persicae (RA); (c) aphid-infested plants that were treated with rhizobacteria (RA) or left untreated (CA). PLS-DA, projection to latent structures-discriminant analysis.

Figure 5.

Contribution of each of the volatile compounds (triangles) to the discrimination of the blends emitted by plants subjected to different treatments (squares). Loading plots of the first two components of the PLS-DA (Fig. 4) of (a) control (C) and aphid-infested plants (CA); (b) rhizobacteria-treated plants that were either uninfested (R) or infested by Myzus persicae (RA); (c) aphid-infested plants that were treated with rhizobacteria (RA) or left untreated (CA). Only the compounds that are most influential for each model are shown (VIP values > 1; Supporting Information, dataset S1). Numbers refer to the volatile compounds listed in Table S1. PLS-DA, projection to latent structures-discriminant analysis; VIP, variable importance in the projection.

The volatile blends of rhizobacteria-treated plants that were uninfested (R) or aphid-infested (RA) were also significantly different (Fig. 4b; PLS-DA model with two significant components: R2X = 0.28, R2Y = 0.98, Q2 = 0.42), and the multivariate analysis shows 16 compounds with a VIP value higher than 1 (Fig. 5b). The four compounds characteristic of aphid-infested plants have also high VIP values in the R versus RA comparison (Fig. 5b). The dry weight of the leaf rosettes was not affected by aphid infestation or by rhizobacterial treatment (two-way anova, aphid: F1,19 = 0.003; P = 0.96; rhizobacteria: F1,19 = 0.198, P = 0.662). Therefore, volatile emissions were not corrected for plant biomass differences.

Rhizobacterial colonization modifies the blend of plant volatiles emitted during aphid infestation

The volatile blends emitted by aphid-infested plants that were treated with rhizobacteria (RA) or that remained untreated (CA) were significantly different (Fig. 4c; PLS-DA resulted in a model with four significant components: R2X = 0.70, R2Y = 0.99, Q2 = 0.60). The model showed 30 compounds with VIP values higher than 1 (Fig. 5c). Most of the 30 compounds were emitted in larger amounts by RA than by CA plants (Fig. 5c), with significant differences (Fig. 6; Supporting Information Table S1; t-test, P < 0.05) for eight compounds [2-nonenal, isovaleric acid, dimethyl sulfoxide, 2-cyclopenten-1-one, (R)-verbenone, (E)-2-heptenal, 1-pentanol, 5,5-dimethyl-2(5H)-furanone]. It is remarkable that all the compounds emitted in significantly larger amounts by RA than CA, were emitted in lower amounts by CA plants compared with C plants (although these differences were significant only for 2-nonenal, 1-pentanol and 5,5-dimethyl-2(5H)-furanone). In support of our initial hypothesis, none of the compounds that were emitted in larger amounts by RA- than by CA-plants differed in emission rate between R plants and C plants (Fig. 6; t-test, P > 0.05). Additionally, the volatile blends emitted by uninfested plants that were either untreated (C) or treated with rhizobacteria (R) were similar, and did not result in a significant model in the PLS-DA analysis (Fig. 7).

Figure 6.

Emission rate (mean ± SE) of the compounds whose presence in the blends of rhizobacteria-treated (RA) or untreated (CA) aphid-infested plants, was significantly different. Arabidopsis thaliana Col-0 plants were either treated with rhizobacteria (R) or left untreated (C), and either uninfested (C, R) or infested with 60 Myzus persicae aphids (CA, RA), 3 d before the experiment. Asterisks indicate a significant difference between compared pairs: *P < 0.05; t-test.

Figure 7.

Score plot of principal component analysis (PCA) of the volatile blends of Col-0 uninfested plants that were untreated (C) or treated with rhizobacteria (R). These two blends overlapped in the multivariate space (the PLS-DA analyses did not result in a significant model; n = 5 samples).

JA signalling is required for the plant-mediated effect of rhizobacteria on parasitoids

To investigate whether the JA signalling pathway is involved in the plant-mediated effects of rhizobacteria on the odour-mediated host-preference behaviour of aphid parasitoids, we evaluated the responses of D. rapae towards the herbivore-induced volatiles emitted by the Col-0 mutant dde2-2, which is defective in allene oxide synthase (AOS), one of the key enzymes of the JA biosynthesis pathway (von Malek et al. 2002). The mutant aos/dde2-2 is blocked in the production of jasmonates and, consequently, in the induction of JA-responsive genes upon herbivory (Snoeren et al. 2009; Leon-Reyes et al. 2010). In contrast to the results for the wild-type Col-0 accession, the parasitoids did not discriminate between the volatile blends emitted by aphid-infested dde2-2 mutant plants that were either treated with rhizobacteria (RAaos) or untreated (CAaos) (Fig. 8a; binomial test, P = 0.12). This lack of preference as observed in Col-0 plants is not due to a loss of the recognition by D. rapae of aphid-infested plants, because more females were attracted by volatiles from CAaos plants than by volatiles from Caos plants (Fig. 8b; P < 0.001). The blends of volatiles emitted by CAaos and RAaos overlap in the multivariate space (Fig. 9). Additionally, emission rates of all detected compounds were similar for CAaos and RAaos plants when compared pairwise per compound (Supporting Information Table S1).

Figure 8.

Behavioural responses of Diaeretiella rapae females to volatiles of the JA-deficient mutant aos/dde2-2 plants, as assessed in a Y-tube olfactometer. Plants were either treated with rhizobacteria or left untreated, and either uninfested (Caos) or infested with 60 Myzus persicae aphids (CAaos, RAaos), 3 d before the experiment. Data represent total number of parasitoids that chose for either of the two odour sources as determined in four replicated experiments with different sets of plants. Asterisks indicate a significant difference within a choice test: ***P < 0.001; ns not significantly different (binomial test). The number of tested and responsive (in parentheses) females is given for each test. JA, jasmonic acid.

Figure 9.

Score plot of principal component analysis (PCA) of the volatile blends of aos/dde2-2 aphid-infested plants that were untreated (CAaos) or treated with rhizobacteria (RAaos). These two blends overlapped in the multivariate space (the PLS-DA analyses did not result in a significant model; n = 6 samples).


This study provides the first evidence that non-pathogenic rhizosphere bacteria can affect the interactions of a plant with carnivorous aboveground insects by modifying the blend of herbivore-induced volatiles. However, in contrast to our initial hypothesis, the aphid parasitoid D. rapae had a lower preference for volatiles from aphid-infested plants that were colonized by rhizobacteria than for volatiles from aphid-infested plants without rhizobacteria. Carnivorous insects can recognize a herbivore-plant complex by the presence in the blend of specific HIPVs or by specific ratios of ubiquitous compounds (Vet & Dicke 1992). Parasitoids have evolved highly efficient mechanisms to perceive small differences in the blends of HIPV that are associated with high quality hosts that promote performance of their offspring (Soler et al. 2005, 2007; Gols et al. 2009). In this study, the HIPV blends emitted by aphid-infested rhizobacteria-treated and aphid-infested control plants contained the same compounds but in different ratios, and the wasps' preference for the latter was strong (80% of the wasps showed such a preference), consistent across replicates and independent of parasitoid experience. However, the function of this preference behaviour was not explained by avoidance of lower host quality, because parasitoid performance was not affected by rhizobacterial colonization. Therefore, the modified HIPV blend may have misinformed D. rapae that a less suitable host is feeding on rhizobacteria-colonized plants compared with plants without rhizobacteria.

Aphid herbivory induces the emission of plant volatiles that are recognized by their predators and parasitoids (Du et al. 1998; Girling et al. 2006). To our knowledge, this is the first study unravelling the volatile blend of A. thaliana after aphid attack, but some of the aphid-induced volatiles that we recorded [e.g. caryophyllene, (E)-β-farnesene, (Z)-3-hexenyl acetate] have also been detected in the headspace of other plant species infested by aphids (Pareja et al. 2007; Gosset et al. 2009; Staudt et al. 2010). Moreover, positive responses of aphid parasitoids have been reported to those compounds (Du et al. 1998; Pareja et al. 2007; Sasso et al. 2009). In general, the volatile blend emitted by aphid-infested Arabidopsis plants is characterized by an increased emission of a group of compounds (some of them exclusively present in the blend of aphid-infested plants), whereas the emission of another group of compounds is suppressed. A recent study has also shown suppression of the emission of volatiles by Vicia fabae after Aphis fabae attack (Schwartzberg, Böröczky & Tumlinson 2011). Such an effect is highly intriguing, especially because phloem feeders have been proposed to suppress the JA signalling pathway via cross-talk, by activating the SA pathway (Thompson & Goggin 2006; Walling 2008). Indeed, the phloem-feeding whitefly, Bemisia tabaci, has recently been observed to suppress the emission of the monoterpene (E)-β-ocimene by suppressing the JA signal-transduction pathway (Zhang et al. 2009). Thus, the suppression of HIPV emission by M. persicae that we have observed here may also be mediated by cross-talk of the JA and SA signalling pathways.

Silencing the JA signalling pathway has been shown to have a strong impact on the plant-mediated interaction between parasitoids and their hosts in different plant-herbivore systems (van Poecke & Dicke 2002; Kessler et al. 2004; Girling et al. 2008; Snoeren et al. 2009; Wei et al. 2011). This study provides support for an important role of the JA signalling pathway in the plant-mediated interactions between parasitoids and root-colonizing bacteria, because the effect of rhizobacteria on parasitoid attraction to aphid-infested plants and on the emission of HIPVs disappeared in the JA-deficient genotype aos/dde2-2. Plants that are colonized by P. fluorescens rhizobacteria show a stronger activation of the JA response after attack by phloem-feeding or leaf-chewing insects (van Oosten et al. 2008; Pineda et al. 2012). Stronger activation of the JA response in the plant, for instance by spraying increasing concentrations of JA, is positively correlated with an enhanced attraction of carnivores (Gols et al. 2003; Bruinsma et al. 2009). Based on the latter studies, we expected that rhizobacteria would enhance the attraction of aphid parasitoids. However, the enhanced attraction in those previous studies was found for parasitoids and predators of chewing insects or mites (Gols et al. 2003; Bruinsma et al. 2009), which strongly activate the JA signal-transduction pathway. Phloem feeders, in contrast, induce both the SA and JA pathways (Thompson & Goggin 2006; Walling 2008; Kusnierczyk et al. 2011), and both pathways are needed to produce the full blend of volatiles that is most attractive to the aphid parasitoid D. rapae (Girling et al. 2008). Due to the important role of the SA pathway in indirect defences against aphids, we interpret this such that the stronger activation of the JA signalling pathway after aphid attack via priming by rhizobacteria, has modified the blend to the extent that it becomes less attractive for D. rapae.

It is remarkable that, as we hypothesized, the effect of rhizobacteria on the emission of HIPVs and through these on parasitoid behavior is only significant after herbivore attack. Several volatiles were emitted in larger amounts by rhizobacteria-treated aphid-infested plants than by aphid-infested control plants, but not when they were uninfested. This finding is in line with the priming mechanism by which rhizobacteria induce systemic resistance (Conrath et al. 2006; van Oosten et al. 2008; Pozo et al. 2008). However, the traditional priming concept assumes that the attacker is inducing a defensive response, whereas aphids suppressed the response, that is, the emission of plant volatiles. Priming is generally regarded as an adaptive response, but the final outcome of the response will depend on the effect it has on the community of herbivorous attackers and higher trophic levels.

We have exploited the model plant A. thaliana which has proven to be an excellent model for studies of rhizobacteria–plant interactions (Pineda et al. 2010; Zamioudis & Pieterse 2012), for studies of plant–attacker interactions (van Poecke & Dicke 2004; van Poecke 2007; Ehlting et al. 2008; Snoeren et al. 2010; Huang et al. 2012), for studies on the chemical ecology of plant volatiles (Godard, White & Bohlmann 2008; Snoeren et al. 2010; Chen et al. 2011) and for manipulative studies on phytohormonal signalling (Pieterse et al. 2009; Snoeren et al. 2009). This plant is a valid model that is extensively used as a stepping stone towards interactions among other brassicaceous plants and their community members (van Poecke & Dicke 2004; Bruinsma et al. 2009).

In this study, we show that the effect of rhizobacteria at inducing susceptibility to aphids that we previously observed (Pineda et al. 2012), extends to a multitrophic context in the sense that the attraction of a parasitoid is reduced. In contrast, in the same microbe-plant system, rhizobacteria induce resistance against generalist caterpillars (van Oosten et al. 2008). These results support the emerging idea that beneficial organisms do not benefit the plant in the defence against all attackers, as can be concluded from the observation that plant-mediated microbial effects on phloem feeders and leaf chewers or on generalist and specialist herbivores differ (Gehring & Bennett 2009; Koricheva et al. 2009; Pineda et al. 2010). Plants, herbivores, microbes and members of the third trophic level all exert their own, sometimes opposite, selective pressures on plant traits (Dicke & Baldwin 2010; van Dam & Heil 2011). The rhizobacteria–plant–aphid–parasitoid interactions studied here likely lead to different benefits for the different players. We did not find evidence of a benefit for the parasitoid to avoid rhizobacteria-treated plants, as parasitoids performed equally well in hosts on control or rhizobacteria-treated plants. The logical next step is to evaluate whether beneficial soil-borne microbes benefit plant fitness in a natural community context. Further studies should then evaluate whether the modification of the HIPV blend by rhizobacteria affects the attraction of other parasitoids, predators and pollinators, and whether the induced resistance against pathogens and generalist chewers overrides the effect of inducing susceptibility against aphids. This study contributes to the understanding of the fundamental aspects of microbe–plant–insect interactions required for application of beneficial soil microbes in agriculture.


We thank L. Westerd, F. van Aggelen and A. Gidding for rearing the insects; D. Lucas-Barbosa for advice during chemical identification; and M. Kos for advice on olfactometer experiments. Research activities by A.P. were supported by a post-doctoral EU Marie Curie Individual Fellowship (grant no. 234895). R.S. was financially supported by a VENI grant (nr 863.08.028) from the Netherlands Organization for Scientific Research (NWO).