Direct and indirect defences induced by piercing-sucking and chewing herbivores in Medicago truncatula


  • Margit Leitner,

    1. Max-Planck-Institut für Chemische Ökologie, Abteilung Bioorganische Chemie, Hans-Knöll-Str. 8, D-07745 Jena, Germany
    Search for more papers by this author
  • Wilhelm Boland,

    1. Max-Planck-Institut für Chemische Ökologie, Abteilung Bioorganische Chemie, Hans-Knöll-Str. 8, D-07745 Jena, Germany
    Search for more papers by this author
  • Axel Mithöfer

    Corresponding author
    1. Max-Planck-Institut für Chemische Ökologie, Abteilung Bioorganische Chemie, Hans-Knöll-Str. 8, D-07745 Jena, Germany
      Author for correspondence:Axel MithöferTel:+49 (0)3641 571263Fax:+49 (0)3641 571256Email:
    Search for more papers by this author

Author for correspondence:Axel MithöferTel:+49 (0)3641 571263Fax:+49 (0)3641 571256Email:


  • • Direct and indirect defences against feeding induced by chewing (Spodoptera littoralis) and piercing-sucking (Tetranychus urticae) herbivores, as well as components of signal transduction, were investigated in the model legume Medicago truncatula.
  • • Emitted volatiles, representing a mechanism of indirect defence, were measured and identified by gas chromatography/mass spectrometry (GC-MS). As elements of direct defence, the accumulation of phenolic compounds and of reactive oxygen species (ROS) was assessed using microscopic techniques. Jasmonic acid (JA) and salicylic acid (SA) concentrations were assessed as putative components of signal transduction.
  • • Volatile profiles revealed a sizeable number of different substances emitted, particularly sesquiterpenoids. The qualitative composition clearly differed depending on the type of herbivory. The same held true for JA and SA concentrations. Also, deposition of phenolic compounds and the production of ROS around the wounding sites could be detected.
  • • Conspicuous differences were found in indirect defence and signalling for different types of herbivory. In contrast, no divergence in direct defences was observed; furthermore, the traits investigated exhibited striking similarities to reactions known to occur upon pathogen attack.


Plants have developed and optimized a considerable diversity of defence mechanisms against adverse environmental conditions caused by either biotic or abiotic factors. These defences are commonly divided into constitutive and induced defences. The former include morphological and structural features as well as constitutively produced defensive compounds. In contrast, induced defences are only activated in situations of actual threat, for example in reaction to herbivore or pathogen attack. Induced defences can further be classified into direct and indirect modes of self-protection. The fortification of cell walls and the accumulation of pathogenesis-related (PR) proteins and various toxic substances constitute means of direct defence, while the emission of certain volatile organic compounds, for example in the case of herbivory, additionally provides indirect defence involving a third trophic level by attracting natural enemies of the attacker of the plant (Takabayashi & Dicke, 1996; Gatehouse, 2002; Kessler & Baldwin, 2002).

Volatile emission seems to be at least in part mediated by changing concentrations of the phytohormones jasmonic acid (JA) and salicylic acid (SA) (reviewed by Van Poecke & Dicke, 2004). Previous studies on tomato (Lycopersicon esculentum Mill.) suggest that a functional JA biosynthetic pathway is required for the induction of volatile release upon spider mite infestation (Ament et al., 2004). However, it has also been reported that, in lima bean (Phaseolus lunatus L.), both salicylate- and jasmonate-related signal transduction pathways are essential for mounting an indirect defence against Tetranychus urticae Koch, while volatile release upon caterpillar feeding seems to be mainly controlled by JA concentrations (Ozawa et al., 2000).

Regarding the possible involvement of SA in defence against piercing-sucking insects (Ozawa et al., 2000; Walling, 2000; Arimura et al., 2002) and the relevance of SA in the resistance of plants to microbial pathogens (Delaney et al., 1994; Dempsey et al., 1999), it is tempting to speculate that there might be some other parallels in the defence mechanisms against pathogens and herbivores. For example, the oxidative burst (the rapid production of reactive oxygen species (ROS)) is a well-described phenomenon occurring in reaction to pathogen attack, at the onset of the hypersensitive response (HR) (Lamb & Dixon, 1997). Hydrogen peroxide, the most stable of the radicals produced, has been proposed to fulfil several roles in defence against pathogens. It might act via direct antimicrobial activity, contribute to structural defence by oxidative cross-linking of the cell wall, or act as a component of intra- and intercellular signal transduction pathways (reviewed by Lamb & Dixon, 1997). However, little attention has been paid to the question of whether oxidative responses also play a role in defence against herbivores. There is one report by Bi & Felton (1995) demonstrating significant increases in lipid peroxidation and OH . − formation, elevated activity of oxidative enzymes and depletion of cellular antioxidants in soybean (Glycine max (L.) Merr.) in reaction to caterpillar feeding. Additionally, the general relevance of ROS as an important defensive factor involved in various forms of stress responses has been suggested recently (Mithöfer et al., 2004).

Concomitantly with the HR, the accumulation of phenolic compounds has been described as a mechanism of resistance to pathogens (Dixon & Paiva, 1995; Kuc, 1995; Dixon et al., 2002) and some animal pests (Ollerstam et al., 2002). Reports on the latter are mainly restricted to more sedentary organisms such as nematodes, mites, galling insects, bark beetles, adelgids and siricids (Fernandes, 1990; Ollerstam et al., 2002). Although elevated concentrations of phenolic compounds were detected upon feeding by mobile herbivores (Bi et al., 1997a), so far no evidence has been provided for the localized accumulation of these feeding deterrents.

It is known that different herbivores induce different volatile profiles (for an overview, see Van Poecke & Dicke, 2004), but there are few reports on comparative analyses of volatiles induced by herbivores with distinct feeding behaviours (Turlings et al., 1998; Ozawa et al., 2000). The present study addresses the question of parallels and differences in the modes of defence a plant uses against different types of herbivores. For the first time, the model legume barrel medic (Medicago truncatula Gaertn.) was used to investigate components of signal transduction as well as direct and indirect defences against either chewing or piercing-sucking herbivores, in this case cotton leafworm (Spodoptera littoralis Boisduval) and two-spotted spider mite (T. urticae), respectively.

Thus, as there is limited information available on the occurrence of localized direct defences such as oxidative burst and the deposition of phenylpropanoid metabolites in the context of herbivory, we studied the accumulation of ROS and phenolic compounds at the wounding sites caused by S. littoralis and T. urticae. Moreover, in the light of the growing importance of M. truncatula as a model organism (Cook et al., 1997; Oldroyd & Geurts, 2001) and the increasing insights being gained into the value of indirect defences, we aimed to characterize the volatiles emitted by the vegetative plant parts of this species. Identification of the compounds emitted will also contribute to the metabolic profiling of this model plant.

Materials and Methods

Plant material and growth conditions

The plants used were M. truncatula Gaertn. cv. Jemalong. Seeds were allowed to germinate in the dark for 4 d, then the seedlings were grown in a glasshouse at 18–23°C with a light period from 07:00 to 21:00 hours. Humidity was kept at 60–70%.

Rearing of insects and mites

Larvae of the cotton leafworm (S. littoralis) were kept on an artificial diet (500 g hackled beans, 9 g ascorbic acid, 9 g 4-ethylbenzoic acid, 0.7 g vitamin E and 4 ml formaldehyde per litre of water mixed with approx. 650 ml of a 7.5% agar solution) at 23°C with a light period from 07:00 to 21:00 hours. Two-spotted spider mites (T. urticae) were reared on M. truncatula plants under the same conditions as described for the plant growth.

Volatile collection

Volatiles were collected over a 48-h period using the closed-loop-stripping method as described by Donath & Boland (1995) because of its high sensitivity. Plants were enclosed in exsiccators and connected to a circulation pump containing a charcoal trap. Air circulation allowed emitted volatiles to be continuously collected on charcoal filters. Finally, desorption was achieved using methylene chloride (2 × 20 µl) containing 100 µg ml−1 n-bromodecane as internal standard, and the volatiles were analyzed using gas chromatography/mass spectrometry (GC-MS) (TRACE 2000 series; Finnigan, Manchester, UK). As a control, plants were cut and placed into tap water. For experiments on caterpillar and spider mite infestations, both intact plants and detached plants were used, with no differences detected regarding volatile release.

Determination of SA and JA concentrations

Salicylate and jasmonate concentrations were determined according to the protocol for jasmonate quantification of Koch et al. (1999), with minor modifications. Briefly, plants were frozen in liquid nitrogen. After the addition of 30 ml acetone: 50 mm citric acid (7 : 3, volume/volume (v/v)) and 150 ng [9,10-2H2]-9,10-dihydro-JA and 500 ng [3,4,5,6-2H4]-SA as internal standards, plants were homogenized using an ultra-turrax, T25 (IKA, Staufen, Germany). The acetone was allowed to evaporate overnight at room temperature. Samples were cleared by filtration and subsequently extracted with 3 × 10 ml diethyl ether. The extracts were then loaded onto solid-phase extraction cartridges containing 500 mg aminopropyl (Chromabond, Macherey-Nagel, Düren, Germany). After washing with 5 ml chloroform: isopropanol (2 : 1, v/v), bound acids were eluted using 12 ml diethyl ether:formic acid (98 : 2, v/v). After evaporation of the solvents, the residues were methylated using excess diazomethane. The final sample volume was adjusted to 50 µl with dichloromethane and analysis was performed using GC-MS (TRACE 2000 series) in the selective ion mode. The fragment ions were monitored at m/z = 120, 124 and 83 for SA, [3,4,5,6-2H4]-SA and JA and [9,10-2H2]-9,10-dihydro-JA, respectively. The endogenous concentrations of salicylate and jasmonate were calculated from the peak areas of the respective substance and its standard using calibration curves.

Detection of phenolic compounds

For the detection of phenolic compounds, whole leaves were used without further preparation. The fluorescence microscope (Axioskop; Zeiss, Jena, Germany) was operated with a 365-nm band pass excitation filter and a 397-nm long pass emission filter. Documentation was obtained using a digital imaging system (Spot; Visitron Systems, Puchheim, Germany).

Detection of ROS

The detection of ROS was conducted using the method described by Olson & Varner (1993), with slight modifications. The samples were covered with a staining solution containing 4% (weight/volume (w/v)) starch and 0.1 m NaI, and incubated for 15–30 min at room temperature. Then the samples were rinsed with water and bleached overnight in ethanol (100%). Slides were prepared using 70% ethanol and were viewed and documented using a light microscope (Axioskop) equipped with a digital imaging system (Spot).


The collection of volatiles upon herbivory in M. truncatula revealed a considerable variety of emitted compounds. These comprised different classes of hydrocarbons such as alkanes, alkenes, aldehydes, alcohols, esters, and aromatics, although the compounds most abundantly present were terpenoids. Figure 1 shows examples of gas chromatograms depicting volatile blends after caterpillar and spider mite feeding. These chromatograms illustrate the general picture for most substances emitted, although without the separation of all sesquiterpenoids. This was carried out using an alternative gas chromatography method (spectra not shown). Compounds (Table 1) were identified according to their fragmentation pattern (MS) and in addition, as far as standard substances were available, by calculation and comparison of retention indices on two different columns with different polarity (EC 5, Alltech, Unterhaching, Germany; DB 225MS, WiCom, Heppenheim, Germany).

Figure 1.

Gas chromatograms of volatiles emitted by Medicago truncatula. (a) Control. (b) Volatiles induced by Spodoptera littoralis feeding. (c) Volatiles induced by Tetranychus urticae infestation. For identification of the compounds, see Table 1. IS, internal standard (100 µg ml−1 n-bromodecane). Asterisks mark contaminations of abiotic origin (plasticizer).

Table 1.  List of compounds identified in the volatile blends emitted by Medicago truncatula upon Spodoptera littoralis or Tetranychus urticae infestation
CompoundNo. in chromatogram1S. littoralisT. urticae
  1. 1 As indicated in Fig. 1.

  2. 2 As determined by fragmentation pattern (mass spectrometry).

  3. 3 As determined by fragmentation pattern (mass spectrometry) and calculation of Kovats indices (in comparison to authentic standards).

  4. 4 Artefacts generated from (E)-ocimene during adsorption to the charcoal trap (Kaiser, 1993).

  5. +, relative abundance below 25%; ++, relative abundance between 25 and 50%; +++, relative abundance above 50%.

 n-Pentadecane3 + 
 3,5-Dimethyl-1-hexene2 + 
 2,6-Dimethyl-1,3,5,7-octatetraene (E)2,4 + 
 Benzaldehyde3 + 
 n-Decanal3 ++
 2-Ethyl hexanal3 1++
 n-Nonanal3 4++
 2,6-Dimethyl-3,5,7-octatrien-2-ol (E)2,4 7+ 
 6-Methyl-1-heptanol2 ++
 1-Octene-3-ol3 2++++
 cis-3-Hexenyl acetate3 3+++
 Cresol2  +
 3,5-Dimethoxytoluene3 + 
 3,5-Dimethylanisole2  +
 Methyl salicylate3 + 
 Trimethylbenzene2 ++
 3-Carene3 + 
 Limonene3 + 
 α-Pinene3 ++
 α-Bisabolol3 + 
 Cadalene2 + 
 (+)-Cyclosativene3 8++++
 β-Farnesene3 + 
 β-Himachalene2  +
 γ-Humulene3  +
 β-Ionone3 + 
 α-Ylangene2 9++++
 4,8-Dimethyl-1,3,7-nonatriene (DMNT)3 5+++ 
 3E,7E-4,8,12-Trimethyltrideca-1,3,7,11-tetraene (TMTT)319++ 
N- or S-containing compounds
 2-s-Butyl-3-methoxypyrazine3 6+ 

Comparing the blends emitted after feeding by S. littoralis and T. urticae, some differences become apparent (Fig. 1; Table 1). Overall, fewer substances were released upon spider mite feeding. Remarkable is the total lack of homoterpenes, as well as alkanes and alkenes, in the spectrum. Similarly, no emission of methyl salicylate (MeSA) was induced by spider mite infestation. Moreover, the relative amounts of some of the sesquiterpenoids emitted differed, particularly for (+)-cyclosativene, α-ylangene and α-copaene. While larger amounts of (+)-cyclosativene and α-copaene were usually released after caterpillar feeding, emission of α-ylangene exceeded that of (+)-cyclosativene and α-copaene after spider mite infestation. Although the overall diversity of sesquiterpenoids induced by T. urticae was lower, β-himachalene and γ-humulene were found exclusively after attack by this herbivore and thus represented the only substances that could be detected after spider mite feeding, but not after caterpillar feeding.

Clear differences were also observed in salicylate and jasmonate concentrations after caterpillar and spider mite infestations. In the comparison drawn here, local and systemic responses, as well as early and late responses, were examined separately. As the time course of development of plant damage differs after attack by different herbivores, the comparison presented here is based on measurements at a given time for caterpillar feeding and at a certain stage of symptom development for spider mite infestation. Thus, the data sets were divided into early and late stages of infestation. In the case of caterpillar feeding, the early and late stages of infestation were taken to be 6 and 48 h after onset of feeding, respectively (larvae were allowed to feed on the plants for 4 h). After spider mite infestation, samples representing the early stages of damage were collected after the appearance of yellowish spots, while samples representing the late stages of damage were collected when initially infested leaves yellowed. This classification of samples ensured that the degree of damage was very similar for caterpillar feeding and spider mite infestation (in terms of leaves affected). However, secondary infections by opportunistic pathogens could not be fully excluded, particularly regarding the long incubation time. Thus we consequently screened the leaves for infections diagnosable by microscopic means, such as fungal infections, which could be excluded. A virus infection of the leaves via a herbivore vector, as described for whitefly Bemisia argentifolii (Bellows & Perring) (Mayer et al., 2002), is very unlikely as Ortlob (1968) showed that T. urticae is unable to transmit viruses. Samples for the determination of local SA and JA concentrations were taken from damaged leaves; systemic concentrations were measured using the uppermost undamaged leaves of the infested plants. Control samples were taken from undamaged, healthy plants.

At early stages of infestation, feeding by S. littoralis caused a marked increase of JA concentrations locally, up to c. 8.4-fold of the control values, while the JA concentrations after T. urticae infestation did not exceed baseline values. Conversely, the systemic concentrations of JA were only slightly increased early after caterpillar feeding (approx. 3.5 times), while the response to T. urticae attack was relatively strong, with an average rise up to 6.2-fold of the control values, although the result in the latter case was not significant (Fig. 2a). At later stages, the situation changed; there were increased JA concentrations (4.0–4.5-fold) after both types of herbivory in local tissue, indicating a transient increase after caterpillar feeding but a late increase after spider mite infestation. Regarding systemic tissue, concentrations rose strongly in plants damaged by caterpillar feeding (14.2-fold), but no changes could be observed in spider mite-infested plants with respect to the concentrations measured at early stages (Fig. 2b).

Figure 2.

Jasmonic acid (JA) concentrations in Medicago truncatula. (a) JA concentrations at early stages of infestation (for Spodoptera littoralis, 6 h after the onset of feeding; for Tetranychus urticae, after the appearance of yellowish spots). (b) JA concentrations at late stages of infestation (for S. littoralis, 48 h after the onset of feeding; for T. urticae, after yellowing of the initially infested leaves of M. truncatula). Data are the mean ± standard deviation for three independent experiments. Asterisks indicate a statistically significant increase of JA values for the particular treatment group (P < 0.05; Student's t-test) compared with the control. syst., systemic.

For SA concentrations, the alterations with time were notably different. In tissue local to the wounding site there was no accumulation of SA after S. littoralis feeding, either at early or at late stages. Conversely, concentrations were elevated in systemic tissue after S. littoralis and T. urticae infestation, as well as locally after T. urticae attack at early stages (2.0–2.8-fold; Fig. 3a). After 48 h, SA concentrations were very similar in caterpillar-damaged and control plants, while both local and systemic tissues of spider mite-infested plants at late stages showed large accumulation of SA (3.1- and 5.2-fold increases, respectively; Fig. 3b).

Figure 3.

Salicylic acid (SA) concentrations in Medicago truncatula. (a) SA concentrations at early stages of infestation (for Spodoptera littoralis, 6 h after the onset of feeding; for Tetranychus urticae, after the appearance of yellowish spots). (b) SA concentrations at late stages of infestation (for S. littoralis, 48 h after the onset of feeding; for T. urticae, after yellowing of the initially infested leaves of Medicago truncatula). Data are the mean ± standard deviation for three independent experiments. Asterisks indicate a statistically significant increase of SA values for the particular treatment group (P < 0.05; Student's t-test) compared with the control. syst., systemic.

Thus, it appears that feeding by S. littoralis produces a transient increase of JA locally in wounded tissue, while systemic concentrations are already elevated 6 h after the onset of feeding and continue to rise up to 48 h. In wounded tissue, no increase of SA concentrations could be detected, whereas a moderate increase took place in systemic tissue 6 h after wounding.

After spider mite infestation, the local response with regard to JA concentrations appeared late and was only moderate. In systemic tissue, there was a clear increase in JA concentration; the onset of this rise is likely to occur at the stage at which the leaves begin to yellow, thus the high standard deviation might indicate that samples were taken during the main period of JA accumulation. Regarding SA determination, a persistent increase in local as well as systemic tissues was found, although the response was not as strong in local tissue as in systemic tissue.

The detection of phenolic compounds after caterpillar feeding (Fig. 4a–c) revealed a two-phasic, time-dependent deposition of these autofluorescent compounds around the bite zone. Immediately after wounding, a bright yellow fluorescent edge at the bite zone could be seen. Within 6 h after wounding, the area adjacent to the wounding zone became blue fluorescent with excitation at c. 365 nm. This localized fluorescence subsequently spread around the wounding site. The widespread faint blue fluorescence after 48 h might also in part be attributable to the reduced autofluorescence of chlorophyll as a consequence of yellowing around the wounding site (Fig. 4c). Similarly, a weak yellow fluorescence was present after spider mite infestation at the stage of limited necrotic lesions (Fig. 4d). At the time of expansion of the lesions, yellow to blue fluorescence spread from the wounding site, which was mainly localized to the cell wall (Fig. 4e). At the final stage, in almost entirely yellowed leaves, a strictly localized blue fluorescence around the wounding site remained (Fig. 4f). As a control, leaves were mechanically damaged with a pin (Fig. 4g–i). Immediately after wounding and up to 24 h afterwards, no increased autofluorescence could be seen around the wounding site. Starting at 24 h, and increasing up to 48 h, blue autofluorescence was seen locally around the wounding sites (Fig. 4i) that resembled the reactions seen after feeding by S. littoralis after 6 h.

Figure 4.

Detection of phenolic compounds (a–i) and reactive oxygen species (ROS) (j–l) at wounding sites in Medicago truncatula. (a–c) Autofluorescence at the bite zone after feeding by Spodoptera littoralis: (a) immediately after wounding; (b) 6 h after wounding; (c) 48 h after wounding. (d–f) Autofluorescent phenolics around cells damaged by Tetranychus urticae: (d) at the stage of limited lesions; (e) at the onset of yellowing of the leaf; (f) after almost the entire leaf has yellowed. (g–i) Depostition of phenolic compounds after mechanical wounding (control): (g) immediately after wounding; (h) 4 h after wounding; (i) 48 h after wounding. (j–l) Detection of ROS using an iodine-starch stain: (j) 48 h after mechanical damage; (k) 48 h after feeding by S. littoralis; (l) after feeding by T. urticae (at the onset of yellowing of the leaf). vb, vascular bundle; t, basal cell of trichome (filled with autofluorescent compounds).

The principle of the iodine-starch stain used for the detection of ROS is the oxidation of iodide in the presence of ROS to iodine, which forms a coloured complex with starch. As can be seen in Fig. 4(j–l), mechanical wounding did not cause the accumulation of ROS, whereas both damage caused by S. littoralis and that caused by T. urticae induced the production of ROS around the wounding site. Interestingly, staining appeared only in the late stages of damage from herbivore feeding, with the first positive results not appearing before 24 h after the start of caterpillar feeding (weakly), or at the beginning of the yellowing of the leaves during spider mite infestation. In the case of caterpillar feeding, clear staining did not appear until 48 h.


Amongst the group of piercing-sucking arthropods there is a clear distinction, in terms of the responses of plants, between phloem-feeding herbivores and herbivores that feed on cellular contents; spider mites belong to the latter group (Walling, 2000). Phloem-feeding herbivores, such as aphids, leafhoppers and whiteflies, cause only minor tissue damage and seem to induce defence signalling pathways largely resembling those activated against pathogens. However, these herbivores can induce unique volatile blends (Du et al., 1998; Birkett et al., 2003). Mites that lacerate cells cause more extensive tissue damage, and thus the response of plants to these mites is more similar to their response to chewing herbivores (Walling, 2000). Nevertheless, there are some substantial differences in plant reactions to chewing and cell-content-feeding herbivores. Regarding volatile release in M. truncatula, the quantitative composition of the blends clearly differed depending on the type of herbivory (Fig. 1; Table 1). This is in accordance with previous findings that the blend emitted specifically attracts predators or parasitoids of the attacker (Takabayashi & Dicke, 1996; De Moraes et al., 1998; Du et al., 1998; Shimoda et al., 2002; Birkett et al., 2003; Horiuchi et al., 2003). There were relatively small qualitative differences, but it is remarkable that MeSA, TMTT (3E,7E-4,8,12-trimethyltrideca-1,3,7,11-tetraene) and E-nerolidol were not found in the chromatograms recorded after spider mite infestation, as MeSA and TMTT have been suggested to be spider mite-inducible volatiles in lima bean (Dicke et al., 1999) as well as in tomato (Kant et al., 2004). Moreover, all three compounds have been described as typical constituents of volatile blends produced by tomato in reaction to spider mites, only being absent in plants with impaired JA accumulation (Ament et al., 2004). In barrel medic, this is clearly not the situation, as elevated concentrations of JA were detected in this study, although the measured concentrations fluctuated greatly. This might have been a result of the sampling mode, in which symptoms were used as a measure. JA concentrations seem to rise during the expansion of localized necrotic lesions. Nevertheless, it is useful to establish comparisons in this way, because the time courses of damage development are very distinct after caterpillar and spider mite infestation. Furthermore, the time that elapsed before the first symptoms of disease became visible after the onset of spider mite feeding differed considerably in various infestations, and it therefore seemed preferable to use damage symptoms as criteria for sampling (in accordance with Kant et al., 2004).

However, clear differences were found in phytohormone concentrations depending on the type of herbivory (Figs 2 and 3). In contrast to results obtained for Helicoverpa zea Boddie larvae feeding on cotton (Bi et al., 1997b), the concentrations of SA were largely unaffected by caterpillar feeding on M. truncatula, while JA concentrations rose markedly. For spider mite infestation, enhanced production of both JA and SA was detected. Similar effects have recently been demonstrated for lima bean (Arimura et al., 2002), although no assessment of differences between local and systemic reactions was carried out. It is noteworthy that in M. truncatula a generally greater accumulation in systemic tissue than in local tissue at the feeding sites was observed. Furthermore, up-regulation of genes activated via the JA and SA pathways has been shown after spider mite infestation of tomato (Kant et al., 2004). Thus, it can be concluded that caterpillar feeding mainly activates JA-related signalling pathways, whereas spider mites induce reactions involving JA as well as SA. These differences might be connected to the specificities of the volatiles released, as JA treatment mimics the effect of caterpillar feeding with respect to the attraction of specific predators (Van Poecke & Dicke, 2002), whereas a combination of JA and MeSA is able to attract natural enemies of herbivorous mites (Shimoda et al., 2002).

The clear involvement of SA in the defence against spider mite infestation calls into question the hypothesis that SA signalling contributes to herbivore-induced pathways only if there is very limited tissue damage, as in the case of phloem feeders (Walling, 2000). These doubts are supported by another recent report showing that the puncture-feeding Tupiocoris notatus Distant not only induced defences similar to those induced by SA, but also repressed the expression of some JA-induced genes in Nicotiana attenuata Torrey ex Watson (Heidel & Baldwin, 2004).

In this context, it is intriguing to ask whether other parallels can be found in reactions known to be involved in defence against pathogens and protection against herbivory. Localized deposition of phenolic compounds has been described after pathogen attack (Bennett et al., 1996; Silva et al., 2002) and after the feeding of some sedentary herbivores (reviewed in Fernandes, 1990; Ollerstam et al., 2002) in species such as Lactuca spp. (in reaction to Bremia lactucae Regal, downy mildew fungus), Coffea spp. (in reaction to Hemileia vastatrix Berk. & Br., orange rust), Solanum dulcamara L. (in reaction to Eriophyes cladophthirus Nal., a gall mite) and Salix viminalis L. (in reaction to Dasineura marginemtorquens Bremi, the gall midge). Also, increased production of defensive compounds was demonstrated in reaction to caterpillar feeding (Bi et al., 1997a). We were able to show a localized accumulation of phenolic compounds in M. truncatula, surrounding the wounding site resulting from either type of herbivory, that was distinct from the reaction seen after mechanical wounding (Fig. 4a–i). This is consistent with findings that transcripts of an enzyme involved in phenylpropanoid phytoalexin biosynthesis, isoflavone-3′-hydroxylase, accumulate after Spodoptera exigua Hübner feeding on M. truncatula leaves (Liu et al., 2003). The progression of deposition of phenolic compounds appears to be two-phasic, similar to that reported for pathogen infection (Bennett et al., 1996) and treatment with pathogen-derived elicitors (M. Leitner, unpublished results). In contrast, mechanical damage only induces the accumulation of blue fluorescent compounds, which occurs later than that induced by herbivores. Thus, our data accord with a report on potato (Solanum tuberosum L.) demonstrating more rapid accumulation of the mRNAs of defence-related genes induced by herbivory compared with simple wounding (Korth & Dixon, 1997). It may be concluded that the second phase is a general reaction to wounding, perhaps with the purpose of preventing opportunistic microbes using wounds as penetration sites, while the first phase seems to be specific for biotic interactions. A further step in the confirmation of this assumption will certainly include the identification of the respective compounds.

As for the similarities of some components of plant defences against diverse attackers, it was intriguing to determine whether production of ROS, as is typical after pathogen attack, can also be found after herbivory (Fig. 4j–l). It has already been reported that wounding induces ROS production, at least in some species. For example, in Zinnia elegans L. (Olson & Varner, 1993), ROS production upon wounding has been shown using the same method as in this study. Yet another interesting report by Orozco-Cardenas & Ryan (1999) demonstrated that the production of ROS upon mechanical damage does not occur in all species. For example, the species tested belonging to the families Solanaceae, Cucurbitaceae and Poaceae showed wound-inducible H2O2 accumulation, whereas only one of the five legume species tested, Pisum sativum L., proved to be positive for this trait. This finding is consistent with our own results, which showed that mere mechanical damage did not induce the accumulation of ROS in M. truncatula, but that herbivore feeding did so. However, the late occurrence of ROS after arthropod feeding is an argument against their involvement in signal transduction. It might rather be supposed that they play a role in propagation of cell death or act against other potential invaders at the wounding site. Nevertheless, it cannot be ruled out that the method applied was not sensitive enough to detect an early, transient increase in ROS concentrations.

To summarize the results presented here, it can be stated that M. truncatula emitted a large variety of volatile substances in reaction to herbivory, differing in their quantitative and qualitative composition depending on the attacking organism. Furthermore, the response in terms of the phytohormones JA and SA clearly differed with type of herbivory, with a greater involvement of SA in the reaction to spider mite feeding and a different time course for the accumulation of JA. Both spider mite and caterpillar infestation induced the deposition of phenolic compounds around the wounding site in a seemingly two-phasic manner, in contrast to mechanical wounding, which caused only the respective second phase. ROS production occurred in the late stages of infestation, whereas wounding did not affect this defensive trait.

Defences against herbivores are distinct from reactions to mechanical wounding, and different types of herbivores are also recognized. While some responses seem to be more general, even being produced in defence against pathogens, others are clearly specific to the reaction against a particular attacker. The components of signal transduction at which the recognition and differentiation of the attacking organism take place remains to be determined.



We would like to thank Dr T. Huguet (INRA, Toulouse, France) for providing M. truncatula seeds, Dr R. Kaiser (Givaudan Company, Dübendorf, Switzerland) for terpenoid standards, and the Max-Planck-Gesellschaft for financial support. Thanks are also due to Angelika Berg for arthropod rearing.