Herbivore-induced plant volatiles mediate host selection by a root herbivore


  • Christelle A. M. Robert,

    1. Laboratory for Fundamental and Applied Research in Chemical Ecology (FARCE), University of Neuchâtel, Rue Emile-Argand 11, 2000 Neuchâtel, Switzerland
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Matthias Erb,

    1. Laboratory for Fundamental and Applied Research in Chemical Ecology (FARCE), University of Neuchâtel, Rue Emile-Argand 11, 2000 Neuchâtel, Switzerland
    2. Root–Herbivore Interactions Group, Max Planck Institute for Chemical Ecology, Beutenberg Campus, Hans-Knöll-Str. 8, 07745 Jena, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Marianne Duployer,

    1. Laboratory for Fundamental and Applied Research in Chemical Ecology (FARCE), University of Neuchâtel, Rue Emile-Argand 11, 2000 Neuchâtel, Switzerland
    Search for more papers by this author
  • Claudia Zwahlen,

    1. Laboratory for Fundamental and Applied Research in Chemical Ecology (FARCE), University of Neuchâtel, Rue Emile-Argand 11, 2000 Neuchâtel, Switzerland
    Search for more papers by this author
  • Gwladys R. Doyen,

    1. Laboratory for Fundamental and Applied Research in Chemical Ecology (FARCE), University of Neuchâtel, Rue Emile-Argand 11, 2000 Neuchâtel, Switzerland
    Search for more papers by this author
  • Ted C. J. Turlings

    1. Laboratory for Fundamental and Applied Research in Chemical Ecology (FARCE), University of Neuchâtel, Rue Emile-Argand 11, 2000 Neuchâtel, Switzerland
    Search for more papers by this author

Author for correspondence:
Ted C. J. Turlings
Tel: +41 32 71 83158
Email: ted.turlings@unine.ch


  • In response to herbivore attack, plants mobilize chemical defenses and release distinct bouquets of volatiles. Aboveground herbivores are known to use changes in leaf volatile patterns to make foraging decisions, but it remains unclear whether belowground herbivores also use volatiles to select suitable host plants.
  • We therefore investigated how above- and belowground infestation affects the performance of the root feeder Diabrotica virgifera virgifera, and whether the larvae of this specialized beetle are able to use volatile cues to assess from a distance whether a potential host plant is already under herbivore attack.
  • Diabrotica virgifera larvae showed stronger growth on roots previously attacked by conspecific larvae, but performed more poorly on roots of plants whose leaves had been attacked by larvae of the moth Spodoptera littoralis. Fittingly, D. virgifera larvae were attracted to plants that were infested with conspecifics, whereas they avoided plants that were attacked by S. littoralis. We identified (E)-β-caryophyllene, which is induced by D. virgifera, and ethylene, which is suppressed by S. littoralis, as two signals used by D. virgifera larvae to locate plants that are most suitable for their development.
  • Our study demonstrates that soil-dwelling insects can use herbivore-induced changes in root volatile emissions to identify suitable host plants.


Different herbivores can interact through physiological changes in shared host plants (van Dam et al., 2003; Erb, 2009; Erb et al., 2009; Gray et al., 2009; Poelman et al., 2010; Pierre et al., 2011). The outcome of these plant-mediated interactions depends on the herbivore species (Wurst & Van der Putten, 2007) and their sequence of arrival (Erb et al., 2011b). Herbivore-induced changes in plant volatile patterns, in particular, have been found to influence oviposition and larval choice aboveground (Carroll et al., 2006, 2008; Soler et al., 2009, 2010). For instance, female moths can use herbivore-induced volatiles to avoid plants that are already infested, probably to avoid competition and/or plants that have otherwise upregulated their defenses (De Moraes et al., 2001; Anderson et al., 2011). For soil-dwelling herbivores, the effects of herbivore-induced changes in plant volatiles on their foraging behavior have not yet been studied, despite the fact that the performance of soil herbivores is affected by the presence of other insects on the same plant (Hausmann & Miller, 1989; Erb et al., 2011b).

Insect larvae can disperse in the soil and locate plants using semiochemical cues (Johnson & Gregory, 2006). Carbon dioxide, for instance, which is released by roots and diffuses rapidly in the soil, is known to be a common attractant for soil insects (Johnson & Gregory, 2006). Because emissions of carbon dioxide by roots are ubiquitous and nonspecific, it is not surprising that several studies have identified additional, nonvolatile chemical signals that enable specialized root herbivores to recognize their host plant (Johnson & Gregory, 2006; Bernklau et al., 2009) and host species of high quality (Johnson et al., 2005). However, it remains unclear whether root herbivores are able to use induced changes in plant volatiles to distinguish host plants. Given the considerable physiological variation of plant quality as a result of genetic and environmental factors, including systemic resistance induced by other plant feeders (Moran & Whitham, 1990; Masters, 1995; Erb et al., 2011b; Pierre et al., 2011), root herbivores should be able to assess host quality even among closely related plants within a population. Because the movement of insects through the soil matrix is costly, we hypothesize that root herbivores may make use of long-range signals to assess the suitability of host plants from a distance.

To test this hypothesis, we explored the interaction between the specialist root feeder Diabrotica virgifera virgifera and its main host plant Zea mays. Diabrotica virgifera females oviposit at the end of the vegetation period, and their eggs diapause in the bare soil during winter, waiting for a new generation of maize plants to germinate in the spring. Therefore, it is impossible for females to assess the quality of the host plants that their offspring will eventually encounter. To assess whether, instead, D. virgifera larvae are able to select and orient towards maize plants that are best suited for their development, we performed a series of performance and preference experiments. Diabrotica virgifera larvae showed a stronger performance on plants that were infested with conspecifics than on healthy plants, whereas they performed more poorly when feeding on plants infested by the leaf herbivore Spodoptera littoralis. By analyzing the changes in the volatile bouquets emitted by roots from plants attacked by below- or aboveground herbivores and performing choice experiments with pure compounds and different maize varieties, we identified ethylene, a gaseous phytohormone (Yang & Hoffman, 1984), and (E)-β-caryophyllene, a sesquiterpene emitted by maize on root infestation (Rasmann et al., 2005) as two distinct signals that are used by D. virgifera larvae to evaluate plant quality from a distance.

Materials and Methods

Plants and insects

Maize seeds (Zea mays L.; varieties ‘Delprim’, ‘Pactol’, ‘Ronaldinho’ and ‘C’; Delley DSP, Delley, Switzerland) were sown in plastic pots (height, 11 cm; diameter, 4 cm) by placing them on a layer of moist washed sand (0–4 mm; Jumbo, Marin-Epagnier, Switzerland). The seeds were then covered with 2 cm of commercial soil (Aussaaterde; Ricoter, Aarberg, Switzerland). Seedlings were grown in a climate chamber (23 ± 2°C, 60% relative humidity, 16 h : 8 h light : dark and 50 000 mmol m−2), and MioPlant Vegetable and Herbal Fertilizer (Migros, Neuchâtel, Switzerland) was added every 2 d after plant emergence. Twelve-day-old plants were used for the experiments. Larvae of Diabrotica virgifera virgifera LeConte (Coleoptera: Chrysomelidae) were reared on freshly germinated maize seedlings until use. Spodoptera littoralis larvae were reared on artificial diet.

Diabrotica virgifera performance

To determine whether infestation by other herbivores influences host plant quality for D. virgifera, we conducted two performance experiments. First, to measure D. virgifera larval performance on plants that had been infested previously with conspecifics, maize root systems were infested with five second-instar D. virgifera larvae. Control plants remained uninfested. After 2 d, all the larvae were carefully removed by gently washing the roots with tap water. Control roots were washed under the same conditions. Maize roots were then repotted in 10% (v/v) moist white sand and infested with five new second-instar D. virgifera larvae for 6 h. The induction of root defenses by D. virgifera occurs within 4–5 h after the onset of feeding (Hiltpold et al., 2011), and the chosen time window was thus deemed optimal to compare the performance of the larvae on induced and uninduced plants. The weight of the larvae was recorded before and after infestation, and average individual relative weight gain was calculated.

In the second experiment, D. virgifera larval performance on plants previously infested with S. littoralis was measured by first adding 20 second-instar S. littoralis larvae on maize leaves, a density comparable to field infestations (Martins, 2000). Control plants were left uninfested. Transparent 1.5-l polyethylene terephthalate (PET) bottles were placed upside down over the aboveground part of the plants to confine leaf herbivores as described elsewhere (Erb et al., 2011b). Forty-eight hours after leaf infestation, single preweighed second-instar D. virgifera larvae were placed on the soil of infested and uninfested plants and left to feed on the roots for 5 d, after which they were recovered and weighed again.

Host plant selection by D. virgifera

To investigate whether D. virgifera can locate good hosts from a distance using volatile cues, the attraction of larvae to infested and uninfested plants was tested in dual-choice olfactometers. Maize seedlings were potted in glass pots (diameter, 5 cm; depth, 11 cm) with a horizontal connector (29/32 mm) at a height of 0.5 cm and filled with moist (10% water) white sand (Migros). Pots without plants were filled with moist white sand only. Pots were wrapped in aluminum foil to keep the root systems in the dark and to avoid visual cues for the larvae. After 48 h, the pot connectors were linked using a glass tube (24/29 mm; length, 8 cm) with a vertically connected access port in the middle, and one Teflon connector at both sides of the glass tube (24/29 mm to 29/32 mm). The Teflon connectors contained a fine metal screen (2300 mesh; Small Parts Inc., Miami Lakes, FL, USA), which prevented the larvae from reaching the roots. The system was left connected for 30 min before introducing six second- or third-instar D. virgifera larvae via the vertical port of the central connector. As soon as the six D. virgifera larvae had moved from the glass connector into one of the Teflon connectors, the system was disassembled and the position of the larvae was recorded. Larvae that had not entered a Teflon connector after 10 min were scored as ‘no choice’. In preliminary control experiments, larval choice assays with healthy plants vs pots with sand only and healthy plants vs plants with infested roots were conducted using central connectors filled with moist sand. As the use of empty connectors gave similar results, but enabled a more efficient larval recovery (Supporting Information Fig. S1), all subsequent experiments were performed using empty central connectors. This type of set-up was used to test D. virgifera attraction to healthy vs D. virgifera-infested seedlings and healthy vs S. littoralis-infested plants. For this, maize plants were infested with five second-instar D. virgifera larvae, 20 second-instar S. littoralis larvae or left uninfested. Aboveground feeders were confined to the leaves by placing transparent 1.5-l PET bottles over the leaves as described previously (Erb et al., 2011b). To confirm that the roots were the source of the volatiles used by the root herbivore to distinguish between healthy and leaf-infested plants, leaves, leaf herbivore larvae, frass and soil were removed from the system and isolated roots from healthy and leaf-infested plants were offered to D. virgifera.

Volatile collection and analysis

To find potential signals that can be used by D. virgifera to distinguish between infested and uninfested plants, volatiles emitted by the roots of healthy, D. virgifera- and S. littoralis-infested plants were measured in several independent experiments. Twelve-day-old maize plants were infested with either five second-instar D. virgifera larvae, 20 second-instar S. littoralis caterpillars or left uninfested. Transparent 1.5-l PET bottles were placed upside down over the aboveground parts of all the plants as described above. After 48 h, roots were washed with tap water and frozen in liquid nitrogen, and root volatile production was determined using solid-phase micro-extraction-gas chromatography-mass spectrometry (SPME-GC-MS) analysis following a previously described protocol (Erb et al., 2011a). CO2 emission was evaluated in a second experiment by connecting the belowground glass pots to an additional glass vessel (length, 28 cm; diameter, 5 cm) via the female connector and a glass male joint. The glass vessel was closed using parafilm and left to stabilize for 2 h. A CO2 gas meter (Voltcraft, CM-100; Conrad Electronics, Dietlikon, Switzerland) was then introduced into the connected vessel for 3 min, and CO2 levels were recorded. Ethylene measurements were performed by removing the roots from the pots and gently washing them with tap water. The entire root systems were then placed in 20-ml gas-tight vials (Gerstel GmbH, Mülheim, Germany) and incubated at room temperature for 12 h. One milliliter headspace samples were withdrawn from the vials with a 2.5-ml gas-tight syringe and injected into a gas chromatograph equipped with a flame ionization detector (GC-FID; Hewlett Packard HP 6890 GC, Agilent Technologies Inc., Palo Alto, Ca, USA). The GC-FID was operated in split mode (2 : 1) with a liner temperature of 60°C, a column temperature of 50°C and a detector temperature of 300°C. For separation, a GS-alumina column was used at a constant flow rate of 4.8 ml min−1. Ethylene was identified by comparison of the retention time with that of the pure compound. Absolute quantification was based on a standard curve obtained by injecting different concentrations of pure ethylene.

Identification of attractants

The attractiveness of (E)-β-caryophyllene and ethylene to D. virgifera larvae was evaluated in three different assays using the dual-choice set-up as described above. For all of these assays, healthy vs D. virgifera-infested plants or healthy vs S. littoralis-infested plants were included as positive controls. In a first experiment, D. virgifera larvae were given the choice between a D. virgifera-infested plant and a healthy plant whose rhizospace was complemented with synthetic (E)-β-caryophyllene (Sigma Aldrich Chemie GmbH, Buchs SG, Switzerland). (E)-β-Caryophyllene was added using slow-release capillary dispensers as described previously (Mérey et al., 2011). To verify that the dispensers release (E)-β-caryophyllene at a similar rate to infested maize roots, we performed a series of quantification experiments. Using SPME-GC-MS as described above, we first established a calibration curve with different doses of pure (E)-β-caryophyllene (0, 12, 25 and 50 ng) dissolved in 50 μl of 0.1% ethanol (v/v). Second, we measured (E)-β-caryophyllene emissions from dispensers with a 1-μl capillary every hour over a period of 8 h, and calculated the release rate in nanograms per hour. To compare the release of dispensers with real maize plants, 12-d-old plants were infested with six D. virgifera larvae for 48 h. After this period, the root system was gently washed with tap water, excised from the stem and placed in an SPME vial. To minimize the effects of removing the leaves, (E)-β-caryophyllene emissions were measured immediately after cutting by SPME.

For the behavioral experiments, the dispensers with a 1-μl capillary were placed upside down into a small cavity in the sand surface for 24 h. Dispensers continuously released up to 40 ng h−1 of (E)-β-caryophyllene, which is well within the physiological range of infested maize roots (Fig. S4). Empty dispensers were added to D. virgifera-infested plants. In a second experiment, D. virgifera selection between healthy and D. virgifera-infested plants was tested with plants of the variety C, which do not emit (E)-β-caryophyllene (Erb et al., 2011a). In a third experiment, larvae were offered uninfested plants with empty control dispensers and uninfested plants with (E)-β-caryophyllene-filled dispensers. Finally, to test the effect of previous (E)-β-caryophyllene exposure during the rearing of the larvae, D. virgifera larvae were reared on either maize seedlings that emit (E)-β-caryophyllene (variety ‘Ronaldinho’; Landi Lyss, Lyss, Switzerland) or do not (variety ‘Pactol’; Delley Semences DSP SA, Delley, Switzerland) on wounding. Choice experiments were then performed for ‘naive’ and ‘experienced’ larvae as described above. The role of ethylene was investigated using similar complementation experiments as above. Diabrotica virgifera larvae were given a choice between plants whose rhizospace was enriched with 2 ppm of ethylene and plants who received ambient air. This increased ethylene concentrations by approximately one-half the amount of ethylene released by maize roots over 12 h (see the Results section). To achieve the enrichment, 10 nl of ethylene in 10 μl of ambient air or ambient air only were injected into the soil with a gas-tight syringe 10 min before the experiment started.

Statistical analysis

All analyses were performed using the software package R, version 2.8.1. Data were first analyzed using Levene and Kolmogorov–Smirnov tests to determine the heteroscedasticity of error variance and normality. Diabrotica virgifera performance was compared using Student’s t-tests. Host selection was analyzed using a general log-linear model (glm) as described previously (D’Alessandro & Turlings, 2006) and the proportions of choosing larvae were compared with control experiments using chi-squared tests. When volatile emission data passed the Levene and Kolmogorov–Smirnov tests, root volatiles were compared using Student’s tests (t-test) and one-way ANOVAs. Pairwise comparisons following ANOVAs were conducted using Tukey’s honestly significant difference tests. If the data did not pass the Levene and Kolmogorov–Smirnov tests, nonparametric Mann–Whitney U-tests or Kruskal–Wallis ANOVA on ranks (H-tests) were carried out. Pairwise comparisons were realized by performing Dunn’s tests. The correlation between CO2 emission and larval preference was tested using Pearson’s correlation coefficients.


Performance of D. virgifera larvae on infested plants

Diabrotica virgifera larvae gained over 30% more weight on plants that had been infested with conspecifics for 2 d compared with healthy plants (n = 7; Student’s t-test, t = −2.675, df = 12, = 0.020; Fig. 1a). However, the weight gain of D. virgifera on plants infested with S. littoralis larvae was only one-quarter of that of healthy plants (control plants, n = 14; infested plants, n = 9; Student’s t-test, t = 2.515, df = 21, = 0.020; Fig. 1b).

Figure 1.

Diabrotica virgifera selects optimal host plants. Average (± SE) individual relative weight gain of D. virgifera larvae after 6 h of feeding on healthy or D. virgifera (D.v.)-infested plants (a), or healthy or Spodoptera littoralis (S.l.)-infested plants (b). Average number (± SE) of larvae that chose volatiles from a healthy or a D. virgifera (D.v.)-infested plant (c), or from a healthy or S. littoralis (S.l.)-infested plant (d). Pie charts show the proportion of larvae that entered an arm. Asterisks indicate significant differences between treatments (*, < 0.05; **, < 0.01).

Diabrotica virgifera detects optimal host plants using volatile signals

In accordance with current literature (Bernklau & Bjostad, 1998), D. virgifera clearly preferred pure CO2 or maize roots over controls (Fig. S2a). When D. virgifera larvae were given a choice between healthy plants and plants infested with conspecifics, they significantly preferred the latter (n = 20; glm, F = 7.418, df = 38, = 0.009; Fig. 1c). Diabrotica virgifera larvae were not attracted by conspecifics alone (n = 9; Fig. S2b). When offered control or leaf-infested plants, D. virgifera larvae preferentially selected healthy plants over S. littoralis-infested plants (n = 15; glm, F = 6.4257, df = 28, = 0.017; Fig. 1d). Removing leaves, larvae, frass and soil from the set-up did not change this preference (n = 20; glm, df = 38, F = 7.7377, = 0.008; Fig. 2).

Figure 2.

Diabrotica virgifera larvae detect leaf herbivore-induced changes in root volatiles. Average number (± SE) of larvae that chose isolated roots of uninfested plants or isolated roots of Spodoptera littoralis (S.l.)-infested plants. The pie chart shows the percentage of larvae that entered an arm. Asterisks indicate significant differences between treatments (**, < 0.01).

Above- and belowground herbivory induces changes in root volatile emission

Maize roots infested with D. virgifera produced a distinct bouquet of volatiles compared with healthy roots. Infested roots released significant amounts of (E)-β-caryophyllene, a compound that was not detected in uninfested roots (Kruskal–Wallis one-way ANOVA on ranks, df = 2, H = 21.083, < 0.001). Furthermore, α-humulene (Kruskal–Wallis one-way ANOVA on ranks, df = 2, H = 7.499, = 0.024), hexadecanal (one-way ANOVA, df = 20, F = 13.655, < 0.001) and tetradecanal (one-way ANOVA, df = 20, F = 8.812, = 0.002) were emitted in greater quantities from infested plants (n = 9; Fig. 3a). Plants infested with D. virgifera were also found to emit less CO2 than healthy plants (n = 8; Student’s t-test, df = 14, t = 2.767, = 0.015; Fig. 3b), an effect that could be attributed to a loss of root biomass following herbivory (Fig. S3a,b). No difference in ethylene emission was noted between D. virgifera-attacked and healthy plants (n = 12; Student’s t-test, df = 22, t = 1.309, = 0.204; Fig. 3d). The GC-MS root volatile profile of plants whose leaves were infested by S. littoralis was not different from that of healthy plants (Fig. 3a). There was also no effect of S. littoralis on root CO2 emission (n = 12; Student’s t-test, df = 22, t = 0.814, = 0.424; Fig. 3c). However, roots of S. littoralis-infested plants emitted 50% less ethylene than root systems of healthy plants (n = 14; Student’s t-test, df = 26, t = 247.5, = 0.043; Fig. 3e).

Figure 3.

Root volatile emission changes after leaf and root attack. (a) Average relative amounts (± SE) of root volatile compounds detected with solid-phase micro-extraction-gas chromatography-mass spectrometry (SPME GC-MS) analysis. Different letters indicate significant differences (< 0.05). White bars, healthy plants; gray bars, Spodoptera littoralis-infested plants; black bars, Diabrotica virgifera-infested plants. Average CO2 emissions (± SE) of healthy and D. virgifera (D.v)-infested plants (b) and healthy and S. littoralis (S.l.)-infested plants (c). Average ethylene emissions (± SE) of healthy and D. virgifera (D.v.)-infested plants (d) and healthy and S. littoralis (S.l.)-infested plants (e). Asterisks indicate significant differences between treatments (*, < 0.05).

Diabrotica virgifera can use (E)-β-caryophyllene and ethylene to locate optimal hosts

Following the above results, we carried out a set of behavioral experiments to investigate the role of (E)-β-caryophyllene and ethylene in the attraction of D. virgifera to infested maize plants. We found that the preference of D. virgifera larvae for plants infested with conspecifics could be counterbalanced by adding capillary dispensers releasing synthetic (E)-β-caryophyllene at a rate of c. 40 ng h−1 to healthy roots (n = 32; glm, df = 62, F = 0.0047, = 0.954; Fig. 4a). When a maize variety that does not emit (E)-β-caryophyllene (variety ‘C’) (Erb et al., 2011a) was offered to the larvae, they did not distinguish any longer between plants infested with conspecifics and healthy plants (n = 17; glm, df = 32, F = 0.0383, = 0.846; Fig. 4a). In addition, D. virgifera larvae were found to selectively orient towards healthy plants with (E)-β-caryophyllene-diffusing dispensers rather than to healthy plants with water dispensers (n = 10; glm, df = 18, F = 20.696, < 0.001; Fig. 4a). Interestingly, (E)-β-caryophyllene in the absence of plants did not attract D. virgifera larvae, even in a context of moderate CO2 levels (Fig. S5a,b). The addition of (E)-β-caryophyllene-releasing dispensers did not alter the emission of other root volatiles (Fig. S6).

Figure 4.

Diabrotica virgifera larvae use root-derived cues to locate good quality hosts. (a) Diabrotica virgifera choice between healthy (white bars) and D. virgifera (D.v.)-infested (black bars) plants; speckled bars indicate plants with added (E)-β-caryophyllene (EβC). Var C refers to a maize variety that does not emit (E)-β-caryophyllene. (b) Correlation between larval choice and CO2 emission. Ratios between the numbers of larvae that preferred healthy plants over infested plants plotted against the ratio of CO2 emission between healthy and infested plants. (c) Average number (± SE) of experienced or naive larvae choosing a healthy plant (white bars) or a plant infested with conspecifics (black bars). (d) Diabrotica virgifera host preference between healthy (white bars) and S. littoralis (S.l.)-infested (gray bars) plants, speckled bars indicate plants with ethylene (ET) addition. Pie charts show the percentage of larvae that entered an arm. Asterisks indicate significant differences between treatments (*, < 0.05; **, < 0.01).

Because, in a few individual cases, D. virgifera larvae preferred healthy over infested plants, we explored the relative role of CO2 and (E)-β-caryophyllene in attracting D. virgifera in more detail. We found that the ratio of larval choice was positively correlated with the ratio of CO2 between healthy and infested plants (n = 16; Pearson’s product moment correlation, df = 14, Qobs = 3.376, = 0.004; Fig. 4b). When the two source plants emitted similar amounts of CO2, D. virgifera larvae oriented towards the infested plants (Fig. 4b). This preference was reversed when healthy plants emitted > 1.2 times more CO2 than infested plants (Fig. 4b). In an additional experiment, we found that the attraction of D. virgifera to infested plants was innate and not dependent on previous feeding experience in the presence of (E)-β-caryophyllene, as larvae that were reared on seedlings that did not emit the compound were equally attracted to infested plants as experienced larvae that were reared on an emitting line (n = 20 per treatment; glm, df = 76; plant status, F = 19.2850, < 0.001; larvae experience, F = 0.1245, = 0.7252; Fig. 4c).

To test the impact of ethylene on host selection behavior, we increased ethylene concentrations by 2 ppm by direct injection into the sand surrounding the roots 10 min before the preference assays. In the control experiment, D. virgifera larvae again preferred healthy plants to S. littoralis-infested plants (n = 23; glm, df = 44, F = 8.048, = 0.007; Fig. 4d). The addition of ethylene to the rhizosphere of S. littoralis-infested plants counterbalanced this effect, resulting in similar attractiveness of both odor sources (n = 15; glm, df = 28, F = 0.199, = 0.659; Fig. 4d). Furthermore, the injection of ethylene into the rhizosphere of a healthy plant made it more attractive for D. virgifera (n = 12; glm, df = 32, F = 7.4381, = 0.012; Fig. 4d).


This study demonstrates that soil-dwelling herbivores can use herbivore-induced plant volatiles to select the most suitable host plants. Previous infestation of maize plants by root or leaf herbivores changes the host quality for D. virgifera. Aggregation of the root feeder on the same host plant, for instance, was beneficial for the insect (Fig. 1a). Similar effects have been documented for a number of leaf-feeding beetles (Dickens, 2006; Weed, 2010). Previous experiments have shown that D. virgifera is entirely resistant to benzoxazinoids, the major defensive secondary metabolites in maize roots (Robert et al., 2012), and the results presented here add to the growing evidence that maize roots do not possess any effective defenses against this specialist feeder. Contrary to root herbivory, leaf infestation by S. littoralis reduced the growth of D. virgifera via systemically induced changes in root physiology (Fig. 1b). This result confirms earlier laboratory and field studies showing that leaf feeders have a general negative impact on root herbivores (Erb et al., 2009; Gill et al., 2011; Pierre et al., 2011). Thus, even in a genetically uniform plant population, D. virgifera larvae encounter plants of different suitability, which could have led to the evolution of efficient host location and selection strategies.

Indeed, D. virgifera larvae were able to select host plants from a distance using herbivore-induced volatile signals. The root feeder was more strongly attracted to root-infested plants than to uninfested plants (Fig. 1c). Changes in plant volatiles rather than larval cues were responsible for the observed differential attraction (Fig. S2b). Similarly, D. virgifera oriented towards control plants rather than leaf-infested plants (Fig. 1d). This preference was still present after removing leaves, S. littoralis larvae, frass and soil from the set-up (Fig. 2), demonstrating that D. virgifera is able to detect systemic changes in root volatile emissions to avoid leaf-infested plants. This remarkable capacity to detect changes in root volatile signals to orient towards the most suitable hosts is likely to be adaptive for this highly specialized root feeder, as it optimizes its growth and fitness.

Leaf and root herbivory resulted in distinct volatile patterns. Infested roots released large amounts of (E)-β-caryophyllene, a sesquiterpene that was not detected at all in uninfested roots (Fig. 3a). (E)-β-Caryophyllene is known to be emitted on D. virgifera attack in maize (Rasmann et al., 2005; Hiltpold et al., 2011) and diffuses well through the soil (Hiltpold & Turlings, 2008). Plants infested with D. virgifera larvae also produced more α-humulene, hexadecanal and tetradecanal and less CO2 than control plants (Fig. 3). The reduction in CO2 emission may be explained by a decrease in metabolically active root mass following root herbivore attack (Fig. S3). Leaf herbivory by S. littoralis did not change the abundance of most detected root volatile compounds, with the exception of ethylene, which was emitted in smaller amounts by leaf-infested plants (Fig. 3). Ethylene is a gaseous phytohormone involved in root growth (Yang & Hoffman, 1984), and the reduced emission may reflect leaf herbivore-induced changes in elongation or branching, as they are known to occur in Nicotiana attenuata on wounding (Hummel et al., 2007). The determination of ethylene emissions in the roots in vivo remains a technical challenge, as the sensitivity of GC-FID methods is insufficient to detect ethylene over short sampling intervals. The use of highly sensitive photo-acoustic lasers may eventually make it possible to test the observed effect in real time and to exclude possible artifacts which may arise from the removal of the shoots of maize plants to measure root emissions and the relatively long incubation period.

The choice assays demonstrate that D. virgifera larvae can use (E)-β-caryophyllene as a signal to locate D. virgifera-infested plants (Fig. 4a) and ethylene to distinguish uninfested from S. littoralis-infested plants (Fig. 4d). As (E)-β-caryophyllene in the absence of plants did not attract D. virgifera larvae (Fig. S5a,b), we suggest that the attraction of D. virgifera larvae to plants infested with conspecifics stems from an attractive effect of (E)-β-caryophyllene within a plant volatile background. Our study adds to the growing evidence that semiochemicals, including sesquiterpenes, are only active in the presence of a plant background odor (Mumm & Hilker, 2005; Schroeder & Hilker, 2008). In addition to (E)-β-caryophyllene, CO2 is a well-known attractant for D. virgifera (Strnad et al., 1986; Johnson & Gregory, 2006), which may be used by the larvae to locate a host plant. In our assays, this volatile did not apparently serve by itself to distinguish root-infested plants from healthy plants, as emissions were lower in the more attractive plants (Fig. 3b). Interestingly, however, the preference for root-infested plants was reversed whenever healthy plants emitted large amounts of CO2 (Fig. 4b). As demonstrated previously, high emissions of CO2 can override the attractiveness of other volatile signals (Bernklau & Bjostad, 1998), possibly because D. virgifera larvae will be better off, in some cases, to feed on inferior roots that are close by rather than venturing over longer distances to reach a higher quality plant. Nevertheless, our results demonstrate that (E)-β-caryophyllene is an attractant for D. virgifera. Its attractiveness does not depend on previous feeding experience (Fig. 4c) and is therefore innate.

Diabrotica virgifera larvae are frequently found to be clustered in maize fields (Ellsbury et al., 1999), but this effect has not been attributed to plant-produced volatile signals, as it is known for aboveground coleopterans (Sakuma, 1994; Loughrin et al., 1996; Soroka et al., 2005; Dickens, 2006; Beran et al., 2011). Based on our results, it seems possible that (E)-β-caryophyllene is one of the signals that can be used by D. virgifera to aggregate. As (E)-β-caryophyllene also attracts entomopathogenic nematodes to herbivore-infested plants (Rasmann et al., 2005), it is tempting to speculate on the evolution of its induced emission. In the light of our results, a possible scenario is that (E)-β-caryophyllene initially served to protect wounded sites of maize roots against opportunistic and pathogenic microorganisms in the soil. Several studies have suggested that (E)-β-caryophyllene acts as an antibiotic (Alma et al., 2003; Lourens et al., 2004; Pichette et al., 2006), and we have shown previously that it is released directly from wounded tissue rather than systemically (Hiltpold et al., 2011), which supports the notion that it serves an antimicrobial role at the site of injury. Over evolutionary time, the signal may have been hijacked by D. virgifera, as a host location and aggregation kairomone, and by entomopathogenic nematodes as a cue to locate root herbivores. The attraction of the beetle larvae to this compound could also explain why it is no longer emitted by American maize cultivars (Köllner et al., 2008), as breeders may have unknowingly selected for less attractive maize lines.

The ethylene complementation experiments reveal that this compound is also attractive to D. virgifera (Fig. 4d). Ethylene has been described previously as an attractant for a variety of insects, such as moths (Raina et al., 1992) and beetles (Arita et al., 1988; Gonzalez & Campos, 1996), and the experimental evidence presented here suggests that D. virgifera can use ethylene to locate plants that are leaf herbivore free. It remains to be determined how specific is this signal as an indicator for the presence of leaf feeders. From a physiological perspective, it seems likely that ethylene is a general belowground indicator for plant growth and quality (Pierik et al., 2006), and we hypothesize that D. virgifera integrates this signal as a general cue for healthy and vigorously growing plants rather than using it as a specific signal to detect leaf herbivores.

So far, the dispersal and distribution of D. virgifera in the field have been shown to be determined by soil texture (Ellsbury et al., 1994), moisture (Ellsbury et al., 1994), porosity (Gustin & Schimacher, 1989) and plant density (Toepfer et al., 2007). The distribution of other root herbivores is known to depend on vegetation cover (Toepfer et al., 2007) and nonvolatile plant metabolites (Johnson et al., 2005). Our study reveals an additional important role of plant volatiles in the distribution of soil insects. The maize specialist D. virgifera seems to have evolved recognition mechanisms to detect specific changes in volatile emissions from roots in order to locate plants of superior quality and avoid plant-mediated competition from a distance. Although it has been suggested that D. virgifera has poor sensory capabilities to use volatiles for orientation (Bernklau & Bjostad, 1998), our assays show that the beetle larvae can recognize at least two additional specific signals, apart from CO2, and use them for successful host location.


From an ecological perspective, our study shows that the distribution and abundance of belowground herbivores are influenced by their capacity to locate and evaluate plant quality from a distance, and suggests that both above- and belowground herbivory can influence the structure of soil-dwelling communities via indirect, plant volatile-mediated effects. From an applied perspective, the results may be relevant for the development of push–pull approaches in crop protection, which combine attractive and repellent plants to lure herbivores away from the crops and attract natural enemies into the field (Cook et al., 2007). Our results could help to establish a basis for such an approach against D. virgifera. The identification of plants that have an increased turnover of root mass resulting in high emissions of CO2 and ethylene, combined with either natural or engineered production of (E)-β-caryophyllene (Degenhardt et al., 2009), may yield an ideal trap crop for D. virgifera larvae. The fact that entomopathogenic nematodes are also attracted by (E)-β-caryophyllene would further increase the efficacy of the approach, as this would result in aggregation of D. virgifera larvae on roots together with the biocontrol agent.


We thank Roland Reist from Syngenta (Stein, Switzerland) for providing S. littoralis eggs. Wade French and Chad Nielson (US Department of Agriculture-Agricultural Research Service-North Central Agricultural Research Laboratory (USDA-ARS-NCARL), Brookings, SD, USA) supplied D. virgifera eggs. Research activities by C.A.M.R., M.E., M.D., C.Z., G.R.D., and T.C.J.T. were supported by the Swiss National Science Foundation (FN 31000AO-107974). This project was partially funded by the National Center of Competence in Research (NCCR) ‘Plant Survival’, a research program of the Swiss National Science Foundation.